Haematologica, Volume 106, Issue 1 by Haematologica

haematologica Journal of the Ferrata Storti Foundation

Editor-in-Chief Jacob M. Rowe (Haifa)

Deputy Editors Carlo Balduini (Pavia), Jerry Radich (Seattle)

Managing Director Antonio Majocchi (Pavia)

Associate Editors Hélène Cavé (Paris), Monika Engelhardt (Freiburg), Steve Lane (Brisbane), PierMannuccio Mannucci (Milan), Simon Mendez-Ferrer (Cambridge), Pavan Reddy (Ann Arbor), David C. Rees (London), Francesco Rodeghiero (Vicenza), Davide Rossi (Bellinzona), Gilles Salles (New York), Aaron Schimmer (Toronto), Richard F Schlenk (Heidelberg), Sonali Smith (Chicago)

Assistant Editors Anne Freckleton (English Editor), Britta Dorst (English Editor), Cristiana Pascutto (Statistical Consultant), Rachel Stenner (English Editor),

Editorial Board Jeremy Abramson (Boston); Paolo Arosio (Brescia); Raphael Bejar (San Diego); Erik Berntorp (Malmö); Dominique Bonnet (London); Jean-Pierre Bourquin (Zurich); Suzanne Cannegieter (Leiden); Francisco Cervantes (Barcelona); Nicholas Chiorazzi (Manhasset); Oliver Cornely (Köln); Michel Delforge (Leuven); Ruud Delwel (Rotterdam); Meletios A. Dimopoulos (Athens); Inderjeet Dokal (London); Hervé Dombret (Paris); Peter Dreger (Hamburg); Martin Dreyling (München); Kieron Dunleavy (Bethesda); Dimitar Efremov (Rome); Sabine Eichinger (Vienna); Jean Feuillard (Limoges); Carlo Gambacorti-Passerini (Monza); Guillermo Garcia Manero (Houston); Christian Geisler (Copenhagen); Piero Giordano (Leiden); Christian Gisselbrecht (Paris); Andreas Greinacher (Greifswals); Hildegard Greinix (Vienna); Paolo Gresele (Perugia); Thomas M. Habermann (Rochester); Claudia Haferlach (München); Oliver Hantschel (Lausanne); Christine Harrison (Southampton); Brian Huntly (Cambridge); Ulrich Jaeger (Vienna); Elaine Jaffe (Bethesda); Arnon Kater (Amsterdam); Gregory Kato (Pittsburg); Christoph Klein (Munich); Steven Knapper (Cardiff); Seiji Kojima (Nagoya); John Koreth (Boston); Robert Kralovics (Vienna); Ralf Küppers (Essen); Ola Landgren (New York); Peter Lenting (Le Kremlin-Bicetre); Per Ljungman (Stockholm); Henk M. Lokhorst (Utrecht); John Mascarenhas (New York); Maria-Victoria Mateos (Salamanca); Giampaolo Merlini (Pavia); Anna Rita Migliaccio (New York); Mohamad Mohty (Nantes); Martina Muckenthaler (Heidelberg); Ann Mullally (Boston); Stephen Mulligan (Sydney); German Ott (Stuttgart); Jakob Passweg (Basel); Melanie Percy (Ireland); Rob Pieters (Utrecht); Stefano Pileri (Milan); Miguel Piris (Madrid); Andreas Reiter (Mannheim); Jose-Maria Ribera (Barcelona); Stefano Rivella (New York); Francesco Rodeghiero (Vicenza); Richard Rosenquist (Uppsala); Simon Rule (Plymouth); Claudia Scholl (Heidelberg); Martin Schrappe (Kiel); Radek C. Skoda (Basel); Gérard Socié (Paris); Kostas Stamatopoulos (Thessaloniki); David P. Steensma (Rochester); Martin H. Steinberg (Boston); Ali Taher (Beirut); Evangelos Terpos (Athens); Takanori Teshima (Sapporo); Pieter Van Vlierberghe (Gent); Alessandro M. Vannucchi (Firenze); George Vassiliou (Cambridge); Edo Vellenga (Groningen); Umberto Vitolo (Torino); Guenter Weiss (Innsbruck).

Editorial Office Simona Giri (Production & Marketing Manager), Lorella Ripari (Peer Review Manager), Paola Cariati (Senior Graphic Designer), Igor Ebuli Poletti (Senior Graphic Designer), Marta Fossati (Peer Review), Diana Serena Ravera (Peer Review)

Affiliated Scientific Societies SIE (Italian Society of Hematology, www.siematologia.it) SIES (Italian Society of Experimental Hematology, www.siesonline.it)

haematologica Journal of the Ferrata Storti Foundation

Information for readers, authors and subscribers Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org). Haematologica publishes editorials, research articles, review articles, guideline articles and letters. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org). Manuscripts should be submitted online at http://www.haematologica.org/. Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail online (www.haematologica.org/content/policies). Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties. Detailed information about subscriptions is available online at www.haematologica.org. Haematologica is an open access journal. Access to the online journal is free. Use of the Haematologica App (available on the App Store and on Google Play) is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org). Rates of the International edition for the year 2020 are as following: Print edition

Institutional Euro 700

Personal Euro 170

Advertisements. Contact the Advertising Manager, Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, e-mail: marketing@haematologica.org). Disclaimer. Whilst every effort is made by the publishers and the editorial board to see that no inaccurate or misleading data, opinion or statement appears in this journal, they wish to make it clear that the data and opinions appearing in the articles or advertisements herein are the responsibility of the contributor or advisor concerned. Accordingly, the publisher, the editorial board and their respective employees, officers and agents accept no liability whatsoever for the consequences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage, and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published literature. Direttore responsabile: Prof. Carlo Balduini; Autorizzazione del Tribunale di Pavia n. 63 del 5 marzo 1955. Printing: Press Up, zona Via Cassia Km 36, 300 Zona Ind.le Settevene - 01036 Nepi (VT)

Associated with USPI, Unione Stampa Periodica Italiana. Premiato per l’alto valore culturale dal Ministero dei Beni Culturali ed Ambientali

haematologica Journal of the Ferrata Storti Foundation

Table of Contents Volume 106, Issue 1: January 2021 About the Cover 1

Images from the Haematologica Atlas of Hematologic Cytology: sea-blue histiocytosis Rosangela Invernizzi https://doi.org/10.3324/haematol.2020.273755

Editorials 2

Low-dose Btk inhibitors: an ‘aspirin’ of tomorrow? Bernard Payrastre and Agnès Ribes https://doi.org/10.3324/haematol.2020.265173

Bringing circulating tumor DNA to the clinic in Hodgkin lymphoma David M. Kurtz https://doi.org/10.3324/haematol.2020.265165

Genomic arrays for the identification of high-risk chronic lymphocytic leukemia: ready for prime time? Antonio Cuneo et al. https://doi.org/10.3324/haematol.2020.264689

Targeting the red cell enzyme pyruvate kinase with a small allosteric molecule AG-348 may correct underlying pathology of a glycolytic enzymopathy Abdu I. Alayash https://doi.org/10.3324/haematol.2020.266585

The arrival of personalized genomics in bone marrow failure Marcin W. Wlodarski https://doi.org/10.3324/haematol.2020.265124

Review Articles 14

Emerging epigenetic therapeutics for myeloid leukemia: modulating demethylase activity with ascorbate Andrew B. Das et al. https://doi.org/10.3324/haematol.2020.259283

Differentiation therapy of myeloid leukemia: four decades of development Vikas Madan and H. Phillip Koeffler https://doi.org/10.3324/haematol.2020.262121

Articles Acute Lymphoblastic Leukemia

Clinical significance of occult central nervous system disease in adult acute lymphoblastic leukemia: a multicenter report from the Campus ALL Network Maria Ilaria Del Principe et al. https://doi.org/10.3324/haematol.2019.231704

Outcomes after late bone marrow and very early central nervous system relapse of childhood B-acute lymphoblastic leukemia: a report from the Children’s Oncology Group phase III study AALL0433 Glen Lew et al. https://doi.org/10.3324/haematol.2019.237230

Haematologica 2021; vol. 106 no. 1 - January 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Acute Myeloid Leukemia

Measurable residual disease monitoring provides insufficient lead-time to prevent morphological relapse in the majority of patients with core-binding factor acute myeloid leukemia Robert Puckrin et al. https://doi.org/10.3324/haematol.2019.235721

Bone Marrow Failure

Cell Therapy & Immunotherapy

Expanded circulating hematopoietic stem/progenitor cells as novel cell source for the treatment of TCIRG1 osteopetrosis Valentina Capo et al. https://doi.org/10.3324/haematol.2019.238261

Chronic Lymphocytic Leukemia

Genomic arrays identify high-risk chronic lymphocytic leukemia with genomic complexity: a multicenter study Alexander C. Leeksma et al. https://doi.org/10.3324/haematol.2019.239947

Analysis of retrotransposon subfamily DNA methylation reveals novel early epigenetic changes in chronic lymphocytic leukemia Timothy M. Barrow et al. https://doi.org/10.3324/haematol.2019.228478

Chronic Myeloid Leukemia

111

The quiescent fraction of chronic myeloid leukemic stem cells depends on BMPR1B, Stat3 and BMP4-niche signals to persist in patients in remission Sandrine Jeanpierre et al. https://doi.org/10.3324/haematol.2019.232793

Coagulation & its Disorders

123

Inhibitor incidence in an unselected cohort of previously untreated patients with severe hemophilia B: a PedNet study Christoph Male et al. https://doi.org/10.3324/haematol.2019.239160

Hematopoiesis

130

HES1 and HES4 have non-redundant roles downstream of Notch during early human T-cell development Matthias De Decker et al. https://doi.org/10.3324/haematol.2019.226126

142

Cyba-deficient mice display an increase in hematopoietic stem cells and an overproduction of immunoglobulins Rodrigo Prieto-Bermejo et al. https://doi.org/10.3324/haematol.2019.233064

Hodgkin Lymphoma

154

Targeted genotyping of circulating tumor DNA for classical Hodgkin lymphoma monitoring: a prospective study Vincent Camus et al. https://doi.org/10.3324/haematol.2019.237719

Non-Hodgkin Lymphoma

163

Tandem autologous hematopoietic stem cell transplantation for treatment of adult T-cell lymphoblastic lymphoma: a multiple center prospective study in China Yao Liu et al. https://doi.org/10.3324/haematol.2019.226985

Haematologica 2021; vol. 106 no. 1 - January 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Plasma Cell Disorders

173

Preclinical development of a humanized chimeric antigen receptor against B-cell maturation antigen for multiple myeloma Lorena Perez-Amill et al. https://doi.org/10.3324/haematol.2019.228577

185

Exploiting MYC-induced PARPness to target genomic instability in multiple myeloma Daniele Caracciolo et al. https://doi.org/10.3324/haematol.2019.240713

Platelet Biology & its Disorders

196

Autoantibody-mediated desialylation impairs human thrombopoiesis and platelet lifespan Irene Marini et al. https://doi.org/10.3324/haematol.2019.236117

208

Low-dose Btk inhibitors selectively block platelet activation by CLEC-2 Phillip L.R. Nicolson et al. https://doi.org/10.3324/haematol.2019.218545

220

Desialylation of O-glycans on glycoprotein Iba drives receptor signaling and platelet clearance Yingchun Wang et al. https://doi.org/10.3324/haematol.2019.240440

230

238

Letters to the Editor 250

FcγRI and FcγRIII on splenic macrophages mediate phagocytosis of anti-glycoprotein IIb/IIIa autoantibody-opsonized platelets in immune thrombocytopenia Peter A. A. Norris et al. https://doi.org/10.3324/haematol.2020.248385

255

Impact and safety of chimeric antigen receptor T-cell therapy in older, vulnerable patients with relapsed/refractory large B-cell lymphoma Richard J. Lin, et al. https://doi.org/10.3324/haematol.2019.243246

259

Macrophage-HFE controls iron metabolism and immune responses in aged mice Naveen Kumar Tangudu et al. https://doi.org/10.3324/haematol.2019.235630

264

Factor IX alteration p.Arg338Gln (FIX Shanghai) potentiates FIX clotting activity and causes thrombosis Wenman Wu et al. https://doi.org/10.3324/haematol.2019.216713

269

Impact of graft composition on outcomes of haploidentical bone marrow stem cell transplantation Rima M. Saliba et al. https://doi.org/10.3324/haematol.2019.227371

275

T-cell immune response controls the high incidence of adenovirus infection in adult allogenic hematopoietic transplantation recipients Javier Sánchez-Céspedes et al. https://doi.org/10.3324/haematol.2019.240101

279

SIRPaFc treatment targets human acute myeloid leukemia stem cells Oleksandr Galkin et al. https://doi.org/10.3324/haematol.2019.245167

Haematologica 2021; vol. 106 no. 1 - January 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation 284

Comparative analysis of targeted novel therapies in relapsed, refractory chronic lymphocytic leukemia Toby A. Eyre et al. https://doi.org/10.3324/haematol.2019.241539

288

Platelets from patients with myocardial infarction can activate T cells Elena E. Solomou et al. https://doi.org/10.3324/haematol.2019.243402

291

295

High bias and low precision for estimated versus measured glomerular filtration rate in pediatric sickle cell anemia Jeffrey D. Lebensburger et al. https://doi.org/10.3324/haematol.2019.242156

299

Low incidence of EPOR mutations in idiopathic erythrocytosis Mathilde Filser et al. https://doi.org/10.3324/haematol.2019.244160

302

306

FcÎłRIIb-BCR coligation inhibits B-cell receptor signaling in chronic lymphocytic leukemia Rosa Bosch et al. https://doi.org/10.3324/haematol.2019.245795

Case Reports 310

Severe delayed hemolytic transfusion reaction due to anti-Fy3 in a patient with sickle cell disease undergoing red cell exchange prior to hematopoietic progenitor cell collection for gene therapy Elizabeth F. Stone et al. https://doi.org/10.3324/haematol.2020.253229

313

Clonal independence of JAK2 and CALR or MPL mutations in comutated myeloproliferative neoplasms demonstrated by single cell DNA sequencing Ella R. Thompson et al. https://doi.org/10.3324/haematol.2020.260448

316

Monoclonal gammopathy of clinical significance: in vivo demonstration of the anti-thrombotic effect of an acquired anti-thrombin antibody Berenice Schell et al. https://doi.org/10.3324/haematol.2019.242370

Errata Corrige 320

Low-dose X-rays leave scars on human hematopoietic stem and progenitor cells: the role of reactive oxygen species Masayuki Yamashita and Toshio Suda https://doi.org/10.3324/haematol.2020.272898

Haematologica 2021; vol. 106 no. 1 - January 2021 http://www.haematologica.org/

ABOUT THEâ&#x20AC;&#x2C6;COVER Images from the Haematologica Atlas of Hematologic Cytology: sea-blue histiocytosis Rosangela Invernizzi University of Pavia, Pavia, Italy E-mail: ROSANGELA INVERNIZZI - rosangela.invernizzi@unipv.it doi:10.3324/haematol.2020.273755

ea-blue histiocytosis, also called syndrome of the sea-blue histiocyte, is a very rare lysosomal storage disease, characterized by distinctive morphological features of storage cells in the bone marrow (Figure), liver, spleen, and other organs, due to the accumulation of ceroid or lipofuscin in their cytoplasm. Bone marrow sea-blue histiocytes are filled with granules that stain blue or blue-green with May-GrĂźnwald Giemsa stain (Figure A, B) or other panoptic stains, while they are brown in unstained films. Histiocyte granules are generally periodic acid Schiff positive (Figure C) and are sometimes iron positive. The histiocyte granules stain with Ziehl-Nielsen (acid fast) stain (Figure D), with phospholipids stained with Nile blue sulphate stain (Figure E) and lipids stained with Oil red O stain (Figure F, left). Moreover, under a fluorescence microscope, histiocytes show strong autofluorescence (Figure F, right).1

References 1. Invernizzi R. Storage diseases. Haematologica. 2020; 105(Suppl 1):S255-260.

haematologica | 2021; 106(1)

EDITORIALS Low-dose Btk inhibitors: an ‘aspirin’ of tomorrow? Bernard Payrastre1,2 and Agnès Ribes1,2 1

INSERM U1048 and Université Toulouse III Paul Sabatier and 2Laboratoire d’Hématologie, CHU de Toulouse, Toulouse Cedex 03, France.

E-mail: BERNARD PAYRASTRE - bernard.payrastre@inserm.fr doi:10.3324/haematol.2020.265173

ruton tyrosine kinase (Btk) is one of the five members of the Tec family of cytoplasmic non-receptor protein tyrosine kinases. Expressed in hematopoietic cells, with the exception of T cells, it is essential for B-cell maturation and signal transduction of the B-cell antigen receptor. Loss-of-function mutations in its gene lead to an immunodeficiency disease called X-linked agammaglobulinemia. Since the 2000s, Btk has emerged as a major therapeutic target in the treatment of B-cell malignancies including chronic lymphocytic leukemia, mantle-cell lymphoma and Waldenström macroglobulinemia.1 Ibrutinib is the first-in-class irreversible inhibitor of Btk acting by covalent modification of the cysteine residue C418 in the kinase ATP binding domain. In the long-term, daily intake of this drug has remarkable activity on these malignant blood disorders.1 There are, however, some side effects, including skin rashes, atrial fibrillation, hypertension, and increased risk of bleeding, in about 40% of treated patients.2 While most bleeding events are of low grade, some grade 3 or higher events are reported in 5% of patients.2 Interestingly, in this issue of Haematologica, Nicolson and colleagues propose a therapeutic application for Btk inhibitors at low doses in thrombosis.3 Two members of the Tec family expressed in platelets, Btk and Tec, are involved in signal transduction downstream of platelet receptors containing an immunoreceptor tyrosinebased activation motif (ITAM) or a closely related hemITAM.4 Glycoprotein VI (GPVI), a platelet-specific collagen receptor constitutively associated with the ITAM-containing FcRγchain, uses Btk to phosphorylate phospholipase Cγ2 (PLCγ2).5 PLCγ2-mediated calcium mobilization and protein kinase C activation are critical for collagen-evoked platelet activation and aggregation.4 Tec can partially compensate for the absence of Btk downstream of GPVI.6 The hemITAM-containing C-type lectin-like receptor 2 (CLEC-2), a podoplanin receptor highly expressed in platelets, also uses Btk and Tec.4,7 Finally, the platelet-specific receptor for von Willebrand factor, GPIb, which does not contain either ITAM or hemITAM, activates Btk.8 Following the observation of an increased bleeding risk in patients treated with ibrutinib, several studies have demonstrated that the drug affects platelet functions in vitro and ex vivo through both on- and off-target effects.2,9-12 Consistent with the fact that patients with X-linked agammaglobulinemia do not experience increased bleeding, ibrutinib-related hemorrhagic syndrome in patients with B-cell malignancies would be mainly due to off-target effects of the drug.12-14 Ibrutinib has, for instance, been shown to affect Src-family kinases (SFK) in platelets.12-14 The relatively high doses of ibrutinib used in the treatment of chronic lymphocytic leukemia (420 mg/day) and mantle-cell lymphoma (560 mg/day) likely contribute to the platelet effect via inhibition of Btk, Tec and also off-targets such as SFK. Second-generation Btk inhibitors with greater specificity for Btk, including acalabrutinib and tirabrutinib,1 have less impact on platelets and, compared to 2

ibrutinib, are associated with a lower risk of severe bleeding in treated patients.2,12-14 These drugs do not abolish the mild bleeding diathesis that may be influenced by intrinsic features of B-cell malignancies in addition to the anti-platelet effect of the drugs.2,12-14 Nicolson and colleagues report that low doses of Btk inhibitors selectively block platelet activation dependent on CLEC-2.3 Such low doses only delay platelet aggregation induced by GPVI triggering. The authors propose that the differential effect of Btk inhibition in CLEC-2, relative to GPVI signaling, is due to a positive feedback loop involving thromboxane A2 and adenosine diphosphate. This important Btkdependent positive feedback control from thromboxane A2 and adenosine diphosphate receptors observed in human platelets stimulated through CLEC-2 was not found in mouse platelets. Accordingly, mouse platelets require higher concentrations of ibrutinib to block CLEC-2-mediated activation. Furthermore, contrasting with the conclusion of a previous study,7 the authors convincingly demonstrated that the Syk tyrosine kinase lies upstream of Btk in the CLEC-2 signaling cascade. The mandatory role of Btk downstream of CLEC-2 was confirmed by using platelets from patients with X-linked agammaglobulinemia, which were unable to aggregate and phosphorylate PLCγ2 on the Btk phosphorylation site in response to CLEC-2 activation. Interestingly, Tec was not able to compensate for the lack of Btk in the CLEC-2 signaling. Nicolson et al.3 also showed that while Btk is important through both its kinase activity and its docking protein properties downstream of GPVI, only its kinase activity is critical for platelet activation by CLEC-2. As an application of their observation, Nicolson and colleagues3 proposed that low doses of ibrutinib could prevent deep vein thrombosis (DVT), as platelet CLEC-2 was previously shown to play a key role in this pathology.15 This hypothesis was tested in an experimental mouse model of DVT, known to be dependent on platelet CLEC-2. In the group of 13 mice treated with ibrutinib (35-70 mg/kg) a trend towards a reduction of the prevalence of inferior vena cava thrombus was observed compared to that in the control group. The authors suggest that the lack of statistical significance of this reduction may have resulted from an incomplete inhibition of CLEC-2 signaling through Btk pharmacological blockade within the 48 hours required for venous thrombosis in this experimental model, and from the fact that mouse platelet CLEC-2 is less critically dependent on Btk than is human platelet CLEC-2. Among the small number of mice that formed a thrombus despite ibrutinib treatment, the thrombus size was comparable to that in the control group, suggesting that CLEC-2 could primarily act in the phase of thrombus initiation. Considering the critical role of CLEC-2 in DVT15 as well as in thrombo-inflammation induced by bacterial infection,16 and its weak contribution to normal hemostasis,4 these results suggest that a low concentration of Btk inhibitors could be an haematologica | 2021; 106(1)

Editorials

Figure 1. Btk inhibition downstream of platelet GPVI and CLEC-2 as a new strategy to prevent deep vein thrombosis, thrombo-inflammation and atherothrombosis. This schematic and simplified representation highlights the role of Btk in the intracellular signaling pathways evoked by platelet receptors such as GPVI, GPIb-IX-V and CLEC-2. Low doses of Btk inhibitors, compared to the relatively high doses currently used for the treatment of B-cell malignancies, strongly prevent CLEC-2evoked platelet activation and may protect against deep vein thrombosis and infection-driven thrombosis. The Btk-dependent positive feedback loop involving adenosine diphosphate and thromboxane A2 in human platelets stimulated by CLEC-2 triggering is shown. Moreover, inhibiting Btk downstream of GPVI (strongly triggered by plaque collagen), and to a lesser extent GPIb, may protect against atherothrombosis. Low-dose irreversible inhibitors may be more effective on platelets that lack de novo synthesis of Btk than on nucleated cells such as B cells. It will, however, be important to determine the appropriate dose of Btk inhibitors to use, as some individuals are more sensitive than others.14 GP: glycoprotein; CLEC-2: C-type lectin-like receptor; Btk-i: Bruton tyrosine kinase inhibitor; Btk: Bruton tyrosine kinase; PLCÎł2: phospholipase CÎł2; ADP: adenosine diphosphate; TXA : thromboxane A2. 2

option to prevent thrombotic complications in these situations. The study by Nicolson and colleagues3 is the second example showing that low-dose Btk inhibitors could be considered as potential antithrombotic drugs. Recently, it was reported that low doses of Btk inhibitors, including ibrutinib and the novel irreversible inhibitors, prevent platelet thrombus formation, at arterial blood flow rates, on human atherosclerotic plaques obtained during carotid endarterectomy.17 The authors suggest that this atherosclerotic plaque-selective platelet activation was blocked due to the targeting of signal transduction of GPVI (strongly activated by plaque collagen), and to a lesser extent GPIb, by Btk inhibitors. Optimization of GPVI signaling inhibition by Btk inhibitors led to the demonstration that low blood concentrations of inhibitors are sufficient for GPVI-selective platelet inhibition relevant for atherothrombosis, while sparing primary hemostasis.18 This is consistent with the fact that, like CLEC-2, GPVI has a minimal role in normal hemostasis.4 Through a small pilot study, the authors could show that low-dose ibrutinib (140 mg/day or every 2 days for a week) was sufficient to prevent atherosclerotic plaque-induced platelet thrombus formation under flow ex vivo and was more effective than aspirin (100 mg/day).17-19 Whether CLEC-2-mediated Btk activation is also targeted in the prevention of atherothrombosis by Btk inhibitors haematologica | 2021; 106(1)

was not investigated in these studies. However, since podoplanin (which is not expressed in the normal vascular wall) has been found in atheromatous plaque, it is tempting to speculate that blockade of the CLEC-2/Btk axis also contributed to this effect.20 This could explain the pronounced efficacy of Btk inhibitors at relatively low doses in this condition. Finally, there is a recent report that ibrutinib treatment could reduce platelet-mediated inflammation, contributing to atherogenesis in an obese Rhesus macaque model of early atherosclerosis.21 In conclusion, Nicolson and colleagues bring important new information on the contribution of Btk in the platelet CLEC-2 signaling cascade and propose that targeting Btk downstream of CLEC-2 could protect against DVT and thrombo-inflammation.3 This study, alongside those from Busygina and colleagues17-19 and Kohs and colleagues,21 shed light on the potential application of Btk inhibitors in vascular and cardiovascular diseases (Figure 1). Considering the very active development of Btk inhibitors in cancer treatment, such tyrosine kinase inhibition-based antiplatelet therapy, rather than more conventional approaches targeting platelet-specific surface receptors, is attractive. One can, however, question whether the exciting results summarized above are a sufficient proof-ofconcept to open up a new therapeutic orientation for Btk inhibitors. The fact that the GPVI/Btk and CLEC-2/Btk axes are weakly involved in normal hemostasis and more 3

Editorials

strongly implicated in thrombotic complications such as DVT, thrombo-inflammation and atherothrombosis is a powerful argument. However, several issues remain to be resolved before undertaking clinical trials to evaluate the efficacy of low doses of Btk inhibitors compared to current antithrombotic treatments in these pathologies. For example, do we need to inhibit both Btk and Tec to obtain an antithrombotic effect in the different situations? Do low doses of Btk inhibitors preferentially inhibit Btk in platelets and spare B-cell functions? Do they prevent thrombus formation in vivo in other relevant animal models of DVT, thrombo-inflammation and atherothrombosis, such as rabbits, and do they spare homeostasis of the lymphatic system? It would also be useful to tackle more clinical issues, such as: (i) the potential relative protection of patients with X-linked agammaglobulinemia and oncology patients under Btk inhibitors against DVT and atherothrombotic complications; (ii) the impact of low doses of Btk inhibitors on the side effects observed with long-term treatment with high doses of ibrutinib, such as atrial fibrillation; or (iii) the potential efficacy of reversible Btk inhibitors in preventing thrombotic events with shortterm applications. Let us bet that future studies will resolve these issues as such a new treatment is of obvious appeal in the prevention of thrombosis in different pathological situations. Disclosures No Conflicts of interest to disclose. Contributions BP and AR wrote the manuscript and made the figure. Acknowledgments Thanks are due to our colleagues from INSERM-U1048 (I2MC) and the Hematology Laboratory of the CHU-Toulouse, France.

References 1. Singh SP, Dammeijer F, Hendriks R. Role of Bruton’s tyrosine kinase in B cells and malignancies. Mol Cancer. 2018;17(1):57. 2. Shatzel JJ, Olson SR, Tao DL, McCarty OJT, Danilov AV, Deloughery TG. Ibrutinib-associated bleeding: pathogenesis, management and risk reduction strategies. J Thromb Haemost. 2017;15(5):835-847. 3. Nicolson PLR, Nock SH, Hinds J, et al. Low-dose Btk inhibitors selec-

tively block platelet activation by CLEC-2. Haematologica. 2021;106(1):208-219. 4. Rayes J, Watson SP, Nieswandt B. Functional significance of the platelet immune receptors GPVI and CLEC-2. J Clin Invest. 2019;129(1):12-23. 5. Quek LS, Bolen J, Watson SP. A role for Bruton’s tyrosine kinase (Btk) in platelet activation by collagen. Cur Biol. 1998;8(20):1137-1140. 6. Atkinson BT, Ellmeier W, Watson SP. Tec regulates platelet activation by GPVI in the absence of Btk. Blood. 2003;102(10):3592-3599. 7. Manne BK, Badolia R, Dangelmaier C, et al. Distinct pathways regulate Syk protein activation downstream of immune tyrosine activation motif (ITAM) and hemITAM receptors in platelets. J Biol Chem. 2015;290(18):11557-11568. 8. Liu J, Fitzgerald ME, Berndt MC, Jackson CW, Gartner TK. Bruton tyrosine kinase is essential for botrocetin/VWF-induced signaling and GPIbdependent thrombus formation in vivo. Blood. 2006;108 (8):2596-2603. 9. Levade M, David E, Garcia C, et al. Ibrutinib treatment affects collagen and von Willebrand factor-dependent platelet functions. Blood. 2014;124(26):3991-3995. 10. Bye AP, Unsworth AJ, Vaiyapuri S, Stainer AR, Fry MJ, Gibbins JM. Ibrutinib inhibits platelet integrin αIIbb3 outside-in signaling and thrombus stability but not adhesion to collagen. Arterioscler Thromb Vasc Biol. 2015;35(11):2326-2335. 11. Kamel S, Horton L, Ysebaert L, et al. Ibrutinib inhibits collagen-mediated but not ADP-mediated platelet aggregation. Leukemia. 2015;29(4):783-787. 12. Nicolson PLR, Hughes CE, Watson S, et al. Inhibition of Btk-specific concentrations of ibrutinib and acalabrutinib delays but does not block platelet aggregation to GPVI. Haematologica. 2018;103(12): 2097-2108. 13. Bye AP, Unsworth AJ, Desborough MJ, et al. Severe platelet dysfunction in NHL patients receiving ibrutinib is absent in patients receiving acalabrutinib. Blood Adv. 2017;1(26):2610-2623. 14. Series J, Garcia C, Levade M, Viaud J, Sié P, Ysebaert L, Payrastre B. Differences and similarities in the effect of ibrutinib and acalabrutinib on platelet functions. Haematologica. 2019;104(11):2292-2299. 15. Payne H, Ponomaryov T, Watson SP, Brill A. Mice with a deficiency in CLEC-2 are protected against deep vein thrombosis. Blood. 2017;129(14):2013-2020. 16. Hitchcock JR, Cook CN, Bobat S, et al. Inflammation drives thrombosis after salmonella infection via CLEC-2 on platelets. J Clin Invest. 2015;125(12):4429-4446. 17. Busygina K, Jamasbi J, Seiler T, et al. Oral Bruton tyrosine kinase inhibitors selectively block atherosclerotic plaque–triggered thrombus formation in humans. Blood. 2018;131(24):2605-2616. 18. Denzinger V, Busygina K, Jamasbi J, et al. Optimizing platelet GPVI inhibition versus haemostatic impairment by the Btk inhibitors ibrutinib, acalabrutinib, ONO/GS-4059, BGB-3111 and evobrutinib. Thromb Haemost. 2019;119(3):397-406. 19. Busygina K, Denzinger V, Bernlochner I, Weber C, Lorenz R, Siess W. Btk inhibitors as first oral atherothrombosis-selective antiplatelet drugs? Thromb Haemost. 2019:119(8):1212-1221. 20. Inoue O, Hokamura K, Shirai T, Vascular smooth muscle cells stimulate platelets and facilitate thrombus formation through platelet CLEC-2: implications in atherothrombosis. PLoS One. 2015;10(9): e0139357. 21. Kohs T, Olson S, Xie A, et al. Effect of BTK inhibition on plateletmediated inflammation in an obese Rhesus macaque model of early atherosclerosis. ISTH 2020 Congress abstract. Res Pract Thromb Haemost 2020;4 (Suppl 1):PB0044.

haematologica | 2021; 106(1)

Editorials

Bringing circulating tumor DNA to the clinic in Hodgkin lymphoma David M. Kurtz Division of Oncology, Department of Medicine, Stanford University, Stanford, CA, USA. E-mail: DAVID M. KURTZ - dkurtz@stanford.edu doi:10.3324/haematol.2020.265165

ver the last decade, technologies to detect, genotype, and track cancers non-invasively through the blood have been developed,1-4 and are promising to revolutionize the way cancer is diagnosed and managed. These methods, termed â&#x20AC;&#x153;liquid biopsiesâ&#x20AC;?, largely rely on the detection of tumor-derived cell-free DNA from the blood plasma of patients, or so-called circulating tumor DNA (ctDNA). More recently, research efforts have focused not only on the development of such methods, but on the translation into the clinic and on defining the utility of ctDNA in various malignancies.5-7 In this issue of Haematologica, Camus and colleagues8 explore the utility of ctDNA detection by an amplicon-based next-generation sequencing (NGS) approach in classical Hodgkin lymphoma (cHL) in a prospective observational study. Detection of ctDNA largely has two major areas of clinic utility: i) genotyping of tumor-derived mutations non-invasively through the blood, and ii) disease quantification and detection of minimal residual disease (MRD) after therapy (Figure 1). In this article, the authors explore each of these potential uses in turn. They begin by assessing the performance of mutational genotyping from cell-free DNA in cHL,

finding variants in the majority of cases (70%). Importantly, in cases where both plasma and tumor DNA were available for sequencing, the authors found a high degree of concordance between mutations identified in the two compartments. Interestingly, they also found a higher median allele fraction in plasma than in tumor samples. This highlights a particular advantage of ctDNA for genotyping in cHL: due to the low abundance of malignant Reed-Sternberg (RS) cells, which are typically ~1% or less of all cells in a tumor, genotyping and NGS from tumor samples in cHL is particularly difficult (Figure 1). This suggests that future studies elucidating the genetic landscape of cHL could benefit from studying cell-free DNA rather than tumor DNA. The excellent performance of mutational genotyping from cell-free DNA in this study confirms prior reports examining genotyping in cHL using targeted hybrid-capture sequencing from cell-free DNA.9,10 These results are in-line with the performance of ctDNA genotyping for other B-cell lymphomas, most notably diffuse large B-cell lymphoma (DLBCL),6,11,12 suggesting a potential broad utility for this assay in many lymphoma subtypes. The authors go on to explore the utility of ctDNA quan-

Figure 1. Circulating tumor DNA in Hodgkin lymphoma. This figure shows an example of the challenges and opportunities for circulating tumor DNA in classical Hodgkin lymphoma. In this disease, the malignant Reed-Sternberg cells represent only a small fraction of the cellular material in the tumor. DNA from these cells (red), along with normal DNA (grey), are released into the circulation and can be collected and sequenced. Sequencing cell-free DNA largely has two use cases. First, identification of tumor-derived somatic mutations (top right), where cell-free DNA potentially contains a higher concentration of mutated molecules than DNA isolated from the tumor. Second, detection of low level minimal residual disease (MRD) during or after therapy (bottom right), where high sensitivity is critically important.

haematologica | 2021; 106(1)

Editorials

titation and MRD using their assay. When quantifying pretreatment ctDNA, they observed a number of striking correlations, including higher stage and International Prognostic Score (IPS) with higher ctDNA concentrations. Higher ctDNA was also correlated with higher tumor volume measured by positron emission tomography/computerized tomography (PET/CT) scans. Again, these results are remarkably consistent with results from other lymphomas including DLBCL,6,13 where higher stage, International Prognostic Index (IPI), and tumor volumes are associated with high ctDNA. The authors finally go on to explore the utility of their assay for the detection of MRD after two cycles of therapy. Interestingly, no patients had detectable MRD at this landmark, including patients who had eventual disease progression. While this suggests a high specificity of the assay, it also suggests improvements in sensitivity might be needed. Previous reports of ctDNA monitoring during therapy for cHL have suggested patients with suboptimal treatment response by PET scans (i.e., not in a complete response on interim imaging) can have residual detectable ctDNA;9,10 however, these studies have a limited number of cases. These data do suggest that sensitivity matters for detection of low-burden disease â&#x20AC;&#x201C; particularly during timepoints of radiographic response and remission. The performance of sensitive ctDNA assays for detection of MRD and molecular response in cHL, DLBCL,13 and other lymphomas should be validated in additional large prospective studies. It is likely that additional technical improvements in sensitivity will also improve the performance in these low disease burden use cases. While this report from Camus and colleagues, as well as reports from other groups, confirms the utility of ctDNA for mutational genotyping and disease detection in cHL and other lymphomas, a number of technical and logistical hurdles remain. While ctDNA-based mutational genotyping has shown utility for identifying single-nucleotide variants, other types of somatic alterations are more challenging to identify from plasma DNA, including copy number alterations, small insertions and deletions, and structural variants. New methods to detect these types of variants are therefore needed. Additionally, with multiple methods including amplicon and hybrid-capture based approaches finding their way into clinical laboratories, discrepancies in methodology and test characteristics are becoming more apparent, making comparisons between tests difficult. Efforts to harmonize and standardize ctDNA quantification and reporting are therefore needed for translation into the clinic. Furthermore, larger prospective studies are required to put the clinical utility of ctDNA quantification as a prognostic marker into context through multivariable analyses considering metabolic tumor volume and clinical risk factors such as the IPS for cHL and the IPI for DLBCL. Finally, and perhaps most significantly, questions remain regarding how to act on ctDNA assessments. While high pretreatment ctDNA level and failure to achieve a molecular response are markers of patients with a high risk of treatment failure, how to improve outcomes for these patients remains unclear. PET-adapted approaches have become common in cHL;14,15 however, these approaches have been difficult to implement for DLBCL, potentially due to the imperfect sensitivity and specificity in this dis6

ease. Similar risk-adapted trial designs in both cHL and DLBCL using ctDNA molecular response to guide therapy will therefore be needed to demonstrate superior outcomes and ultimately lead ctDNA assessment into routine clinical care. Finally, further research is needed to establish the utility of ctDNA in other lymphoma subtypes, particularly in low-grade lymphomas where clinical management strategies typically do not aim for curative endpoints. There still is therefore significant work ahead in translation of liquid biopsies into routine clinical management of B-cell lymphomas; however, early data from studies such as the work from Camus and colleagues provide promising evidence that ctDNA will change clinical management for B-cell lymphomas in the not-too-distant future. Disclosures Dr. Kurtz has served as a consultant for Roche and Genentech, has ownership equity in Foresight Diagnostics, and has patents pending related to methods for analysis of cell free nucleic acids and methods for treatment selection based on statistical frameworks of clinical outcome. Contributions DMK conceived of and wrote this manuscript.

References 1. Bettegowda C, Sausen M, Leary RJ, et al. Detection of circulating tumor DNA in early- and late-stage human malignancies. Sci Transl Med. 2014;6(224):224ra24. 2. Diehl F, Schmidt K, Choti MA, et al. Circulating mutant DNA to assess tumor dynamics. Nat Med. 2008;14(9):985-990. 3. Newman AM, Bratman SV, To J, et al. An ultrasensitive method for quantitating circulating tumor DNA with broad patient coverage. Nat Med. 2014;20(5):548-554. 4. Newman AM, Lovejoy AF, Klass DM, et al. Integrated digital error suppression for improved detection of circulating tumor DNA. Nat Biotechnol. 2016;34(5):547-555. 5. Abbosh C, Birkbak NJ, Wilson GA, et al. Phylogenetic ctDNA analysis depicts early-stage lung cancer evolution. Nature. 2017;545(7655):446451. 6. Scherer F, Kurtz DM, Newman AM, et al. Distinct biological subtypes and patterns of genome evolution in lymphoma revealed by circulating tumor DNA. Sci Transl Med. 2016;8(364):364ra155. 7. Tie J, Wang Y, Tomasetti C, et al. Circulating tumor DNA analysis detects minimal residual disease and predicts recurrence in patients with stage II colon cancer. Sci Transl Med. 2016;8(346):346ra92. 8. Camus V, Viennot M, Lequesne J, et al. Targeted genotyping of circulating tumor DNA for classical Hodgkin Lymphoma monitoring: a prospective study. Haematologica. 2020;106(1):154-162. 9. Desch AK, Hartung K, Botzen A, et al. Genotyping circulating tumor DNA of pediatric Hodgkin lymphoma. Leukemia. 2020;34(1):151-166. 10. Spina V, Bruscaggin A, Cuccaro A, et al. Circulating tumor DNA reveals genetics, clonal evolution, and residual disease in classical Hodgkin lymphoma. Blood. 2018;131(22):2413-2425. 11. Bohers E, Viailly PJ, Becker S, et al. Non-invasive monitoring of diffuse large B-cell lymphoma by cell-free DNA high-throughput targeted sequencing: analysis of a prospective cohort. Blood Cancer J. 2018;8(8):74. 12. Rossi D, Diop F, Spaccarotella E, et al. Diffuse large B-cell lymphoma genotyping on the liquid biopsy. Blood. 2017;129(14):1947-1957. 13. Kurtz DM, Scherer F, Jin MC, et al. Circulating tumor DNA measurements as early outcome predictors in diffuse large B-cell lymphoma. J Clin Oncol. 2018;36(28):2845-2853. 14. Andre MPE, Girinsky T, Federico M, et al. Early positron emission tomography response-adapted treatment in stage I and II Hodgkin lymphoma: final results of the randomized EORTC/LYSA/FIL H10 trial. J Clin Oncol. 2017;35(16):1786-1794. 15. Johnson P, Federico M, Kirkwood A, et al. Adapted treatment guided by Interim PET-CT scan in advanced Hodgkin's lymphoma. N Engl J Med. 2016;374(25):2419-2429.

haematologica | 2021; 106(1)

Editorials

Genomic arrays for the identification of high-risk chronic lymphocytic leukemia: ready for prime time? Antonio Cuneo,1 Gian Matteo Rigolin1 and Cristina Mecucci2 1

Hematology, Department of Medical Sciences, St. Anna University Hospital, Ferrara and 2Hematology, Department of Medicine, University of Perugia, Perugia, Italy E-mail: CUNEO ANTONIO - cut@unife.it doi:10.3324/haematol.2020.264689

ematopoietic tumors develop through the acquisition of multiple molecular cytogenetic lesions, which confer growth advantage to the neoplastic cells, favoring the emergence of clones resistant to treatment. Not surprisingly, a high number of chromosome aberrations represents an unfavorable prognostic factor in several hematologic neoplasias, including acute myeloid leukemia,1 myelodysplastic syndrome,2 myelofibrosis3 and chronic lymphocytic leukemia (CLL).4 Conventional banding analysis (CBA) is a well-established method allowing for a complete overview of the cell genome, with high specificity and low sensitivity. The presence of three or more chromosome lesions in the same clone, usually referred to as complex karyotype, was associated with a shorter overall survival in CLL as early as 1990.5 Since then, reports have documented the negative prognostic significance of a complex karyotype in patients treated with chemoimmunotherapy and with novel mechanism-based therapy.6-10 While prognostic factors associate with outcome inde-

pendently of the treatment, predictive factors are able to identify patients who are likely to obtain improved survival with a specific therapy and can only be identified in comparative trials.11 Because different types of treatment can be offered to CLL patients, biological factors predicting outcome with a specific drug are very useful in clinical practice and CBA was proposed as a desirable procedure in the diagnostic workup within prospective clinical studies.12 Interestingly, high efficacy of venetoclax plus obinutuzumab in patients with a complex karyotype and CLL was recently documented in the CLL14 trial,13 pointing to a possible role of complex karyotype as a predictive marker. Despite the introduction of effective mitogens, CBA is quite a cumbersome procedure requiring fresh dividing cells. For this reason, molecular methods such as array comparative genomic hybridization (CGH), single nucleotide insertional polymorphisms and next-generation sequencing have been applied in CLL, consistently showing that a high number of genetic lesions is associated with an inferior outcome.14-16

The long road of genetic advances with a clinical impact on chronic lymphocytic leukemia. CNV: copy number variation; cnLOH: copy neutral loss of heterozygosity; â&#x2030;Ľ3 and â&#x2030;Ľ5: three or more, or five or more aberrations in the same clone, respectively.

haematologica | 2021; 106(1)

Editorials

In this issue of Haematologica, Leeksma and co-workers17 report the results of high-resolution genome-wide detection of copy-number alterations (CNA) in a large multicenter cohort of CLL patients, showing that genomic arrays are more sensitive than molecular cytogenetic methods and that high genomic complexity (i.e., 5 or more CNA) represents an independent adverse prognosticator. In their analysis, 13 laboratories connected to the European Research Initiative in CLL (ERIC) group analyzed 2,293 patients by array-CGH and performed concurrent fluorescence in situ hybridization (FISH) in 260 patients using the classical four-probe panel detecting 17p-, 11q-, +12 and 13q- and concurrent CBA in 122 patients. Nine-hundred seventy-two patients, known to be untreated at the time of sampling, were included in the overall survival analyses. Overall, CNA were detected in 78.9% of the cases and, as expected, previously treated patients and patients with high-risk CLL (i.e., in advanced stage or with unfavorable genetic features) had a significantly higher number of CNA. In one-third of the cases genomic lesions detected by array-CGH were not detected by FISH. However, despite a high concordance between the two approaches, subclonal chromosome lesions detected by FISH, including 11q- and 17p-, were not detected by array-CGH. In previous analyses, similar results were observed when comparing FISH and CBA.18,19 Interestingly, Leeksma and co-workers suggest that CGH may be more sensitive than CBA, given that a direct comparison of results from 122 patients showed chromosomal abnormalities in 86.1% of patients by genomic arrays versus 73.8% by CBA (P=0.015). Among 972 untreated patients, 57 (5.8%) had high genomic complexity (i.e., ≥5 CNA). This parameter was an independent prognostic factor associated with shorter overall survival (hazard ratio = 2.19; 95% confidence interval: 1.35-3.54) and shorter time to first treatment (hazard ratio = 2.00; 95% confidence interval: 1.28-3.14), thus confirming the results of a recent international report on CBA.8 It is worth noting that, as the authors wisely pointed out, this series included patients treated with different chemoimmunotherapy regimens and that the role of genomic complexity in patients treated with novel agents is still unknown. Besides the classical biomarkers and molecular genetic lesions associated with inferior prognosis, array-CGH identified gains of 8q, and losses of 9p and 18p as candidate markers possibly associated with a shorter overall survival in univariable Cox regression analysis. Interestingly, a minority of patients (n=32) showed very complex rearrangements, consisting of ≥10 oscillating copy number alterations involving two or three copy number states on one chromosome, a pattern suggestive of catastrophic mitotic events, usually referred to as chromothripsis.20 Putative chromothripsis was associated with a dismal outcome in this series. It is worth noting that, despite heterogeneity of the patients and of the methods employed in the different laboratories (i.e., total blood or purified CD19-positive lymphocytes, different array-CGH platforms and proto8

cols) the study by Leeksma et al. has the following merits: (i) rigorous criteria were applied for the detection of CNA, with reference to the 5 Mb cut-off that is close to the chromosome band size; (ii) virtually all the results corresponded to established cytogenetic imbalances expected to occur in CLL; and (iii) the highly complex karyotype (i.e., ≥5 changes) was detected, with results in patients being superimposable to those with concurrent CBA. On the other hand, an intrinsic limit of genomic arrays, like that of other genomic techniques, is the lack of identification of subclones and translocations; furthermore, the major specific contributions generated in this analysis, i.e., chromothripsis and copy neutral loss of heterozygosity, could not be analyzed in detail. Although not yet validated in prospective studies, genomic arrays could become an accurate tool for CLL risk stratification, as they may represent a widely applicable method for the detection of DNA gains and losses. Indeed, while FISH remains the gold standard for the detection of 11q- and 17p-, the introduction of arrayCGH in clinical trials may be advantageous as it may allow for investigations on the role of genomic complexity as a prognostic/predictive factor in clinical trials, assisting clinicians in the choice of the most appropriate treatment for their patients. Disclosures No Conflicts of interest to disclose. Contributions AC, GMR and CM wrote and approved the editorial.

References 1. Döhner H, Estey E, Grimwade D, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood. 2017;129(4):424-447. 2. Greenberg PL, Tuechler H, Schanz J, et al. Revised international prognostic scoring system for myelodysplastic syndromes. Blood. 2012; 120(12):2454-2465. 3. Gangat N, Caramazza D, Vaidya R, et al. DIPSS plus: a refined Dynamic International Prognostic Scoring System for primary myelofibrosis that incorporates prognostic information from karyotype, platelet count, and transfusion status. J Clin Oncol. 2011;29(4):392-397. 4. Rigolin GM, del Giudice I, Formigaro L, et al. Chromosome aberrations detected by conventional karyotyping using novel mitogens in chronic lymphocytic leukemia: clinical and biologic correlations. Genes Chromosomes Cancer. 2015;54(12):818-826. 5. Juliusson G, Oscier DG, Fitchett M, et al. Prognostic subgroups in Bcell chronic lymphocytic leukemia defined by specific chromosomal abnormalities. N Engl J Med. 1990;323(11):720-724. 6. Cavallari M, Cavazzini F, Bardi A, et al. Biological significance and prognostic/predictive impact of complex karyotype in chronic lymphocytic leukemia. Oncotarget. 2018;9(76):34398-34412. 7. Rigolin GM, Saccenti E, Guardalben E, et al. In chronic lymphocytic leukaemia with complex karyotype, major structural abnormalities identify a subset of patients with inferior outcome and distinct biological characteristics. Br J Haematol. 2018;181(2):229-233. 8. Baliakas P, Jeromin S, Iskas M, et al. Cytogenetic complexity in chronic lymphocytic leukemia: definitions, associations, and clinical impact. Blood. 2019;133(11):1205-1216. 9. Thompson PA, O'Brien SM, Wierda WG, et al. Complex karyotype is a stronger predictor than del(17p) for an inferior outcome in relapsed or refractory chronic lymphocytic leukemia patients treated with ibrutinib-based regimens. Cancer. 2015;121(20):3612-3621. 10. Anderson MA, Tam C, Lew TE, et al. Clinicopathological features and outcomes of progression of CLL on the BCL2 inhibitor venetoclax. Blood. 2017;129(25):3362-3370.

haematologica | 2021; 106(1)

Editorials

11. Ballman KV. Biomarker: predictive or prognostic? J Clin Oncol. 2015;33(33):3968-3971. 12. Hallek M, Cheson BD, Catovsky D, et al. IwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760. 13. Al-Sawaf O, Lilienweiss E, Bahlo J, et al. High efficacy of venetoclax plus obinutuzumab in patients with complex karyotype and chronic lymphocytic leukemia. Blood. 2020;135(11):866-870. 14. Delgado J, Salaverria I, Baumann T, et al. Genomic complexity and IGHV mutational status are key predictors of outcome of chronic lymphocytic leukemia patients with TP53 disruption. Haematologica. 2014;99(11): e231-e234. 15. Gunnarsson R, Isaksson A, Mansouri M, et al. Large but not small copy-number alterations correlate to high-risk genomic aberrations and survival in chronic lymphocytic leukemia: a high-resolution genomic screening of newly diagnosed patients. Leukemia. 2010;24(1):211-215.

16. Rigolin GM, Saccenti E, Bassi C, et al. Extensive next-generation sequencing analysis in chronic lymphocytic leukemia at diagnosis: clinical and biological correlations. J Hematol Oncol. 2016;9(1):88. 17. Leeksma AC, Baliakas P, Moysiadis T, et al. Genomic arrays identify high-risk chronic lymphocytic leukemia with genomic complexity: a multicenter study. Haematologica. 2021;106(1):87-97. 18. Haferlach C, Dicker F, Schnittger S, et al. Comprehensive genetic characterization of CLL: a study on 506 cases analysed with chromosome banding analysis, interphase FISH, IgVH status and immunophenotyping. Leukemia. 2007; 21(12):2442-2451. 19. Rigolin GM, Cibien F, Martinelli S, et al. Chromosome aberrations detected by conventional karyotyping using novel mitogens in chronic lymphocytic leukemia with “normal” FISH: correlations with clinicobiologic parameters. Blood. 2012; 119(10):2310-2313. 20. Stephens PJ, Greenman CD, Fu B, et al. Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell. 2011;144(1):27-40.

Targeting the red cell enzyme pyruvate kinase with a small allosteric molecule AG-348 may correct underlying pathology of a glycolytic enzymopathy Abdu I. Alayash Laboratory of Biochemistry and Vascular Biology, Center for Biologics Evaluation and Research, Food and Drug Administration (FDA), Silver Spring, MD, USA E-mail: ABDU I. ALAYASH - abdu.alayash@fda.hhs.gov doi:10.3324/haematol.2020.266585

ature red blood cells (RBC) lack a nucleus and mitochondria, relying almost entirely on their own glycolytic degradative processes to generate energy. One of the major reaction pathways in RBC is the Embden-Meyerhof in which a series of enzymes convert glucose anaerobically into usable energy: adenosine triphosphate (ATP).1 The pathway utilizes two ATP to initiate the reaction, with ultimately two more ATP being produced. This pathway is essential to meeting the energy demands of RBC, including maintenance of red cell membrane flexibility and, therefore, impacting RBC shape.1 In the Embden-Meyerhof pathway, a pyruvate kinase (PK) isoform unique to RBC, PK-R, is a rate-limiting enzyme that plays a critical role in the formation of pyruvate from phosphoenolpyruvate (PEP) with the simultaneous generation of ATP from adenosine diphosphate (ADP)2 (Figure 1). Among the most common enzyme defects related to the Embden-Meyerhof pathway is an inherited disorder in which homozygote individuals display signs and symptoms of hemolytic anemia due to the deficiency of the PK-R enzyme. PK defects have been documented worldwide, although most cases have been identified in people of Northern European ancestry.3 This rare hereditary disorder is characterized by changes in RBC metabolism including manifestation of anemia and a compromised energetic profile (ATP production). RBC deficient in PK cannot produce enough energy to maintain normal membrane function. Potassium and water leak from the cell, while calcium concentrations increase. As a result, these cells become rigid, lose flexibility, and are more susceptible to premature hemolysis.4 In mammals, two PK genes are expressed depending on anatomical region and cell type: PK muscle (PKM) and PK liver and red blood cell (PKLR). PKLR controls the expreshaematologica | 2021; 106(1)

sion of the red blood cell (PK-R) or liver (PK-L) isoforms from tissue-specific promoters. Mutations in the PKLR gene cause PK deficiency with clinical symptoms apparently confined to RBC. PK-R of RBC is a tetrameric enzyme that exists in equilibrium between a less active T-state and a more active R-state that can be induced by binding to the glycolytic intermediate fructose bisphosphate (FBP). Therefore, intervention strategies designed to counter this condition included the stabilization of the active R state of PK-R, directly restoring PK activity above and beyond the endogenous activation by FBP.5 An early investigation that screened for drugs targeting the PK enzyme resulted in the discovery of a small molecule, AG-348, which allosterically activates wild type PK as well as the mutant form of the enzymes. The activities of this molecule were demonstrated in vitro, in mice, and ex vivo in human RBC.6 A partially resolved crystal structure of AG-348 bound to PK-R (2.75 A˚ resolution) showed that AG-348 is bound in the PK enzyme pocket at the dimer– dimer interface away from the FBP-binding domain and is buried in a cluster of apolar amino acids inducing post binding conformational change to the final R-state reminiscent of the classic R to T transition in another red cell protein, hemoglobin.7 In this issue of Haematologica, Rab et al.8 report on the effects of AG-348 on RBC from a small number of patients with PK deficiency. In this investigation, the group carried out ex vivo experiments on RBC from a broad range of patient phenotypes, including measuring several parameters such as activities of the intermediates in the glycolytic pathway and ATP levels. They showed that AG-348 affects thermostability of the PK-R, protein levels and the shape of RBC. They also reported a modest increase in ATP and improvement in PK thermostability. 9

Editorials

Figure 1. Schematic representation of the action of AG-348 on the enzyme pyruvate kinase in the final step of the glycolytic degradation of glucose in red blood cells (RBC). Glycolysis, the first step in cellular respiration begins with one glucose molecule (6 carbons) which is subsequently broken down in a series of enzymatic steps via ten enzymes into two pyruvic acid molecules (2 carbons) as well as releasing adenosine triphosphate and reduced nicotinamide adenine dinucleotide. The schemes on both sides (normal RBC metabolism and in pyruvate kinase deficiency) begin with the formation of 1,3 biphosphoglycerate after several enzymatic steps of phosphorylation and molecular arrangements. 2,3-bisphosphoglyceric acid (2,3-BPG), also known as 2,3-diphosphoglyceric acid (2,3-DPG) (allosteric affecter of hemoglobin), is a three-carbon isomer of the glycolytic intermediate 1,3-BPG. The final critical step in this reaction pathway involves the formation of pyruvate from phosphoenolpyruvate, a reaction catalyzed by the enzyme pyruvate kinase. The red cell in the center of the figure where these reactions occur is shown with two important receptors, Glut 1 and band 3, which play key roles in glucose transport and metabolism. The small molecule AG-348 penetrates the red cell and activates the enzyme pyruvate kinase.

Additionally, protein analyses using mass spectrometry suggest improvement in key glycolytic intermediates as a result of this treatment.8 Glycolytic pathways are also known to be regulated by band 3 in concert with hemoglobin-confirmation dependent transition from R to T states.9 However, western blot analysis in this manuscript surprisingly shows no differences in band 3 concentrations in samples from PK-deficient patients compared to healthy controls. These findings suggest that the effects of AG-348 may occur in a band 3-independent manner, although a more comprehensive proteomic analysis in the presence and absence of AG-348 may still reveal changes in band 3 and its associated proteins at different time points following treatment. Metabolic profiling of whole blood from some of the patients enrolled in this study from Rab et al. were carried out using liquid chromatography mass spectrometry (LCMS/MS). The data confirmed the decrease in PK activity along with reduced levels of PEP, and the modest increase in 2,3-diphosphoglycerate (DPG). Ex vivo incubation of PK-deficient RBC however, increased PK activity and ATP production in a dose-dependent manner. Although 2,3DPG levels were not systematically collected in this work, early in vivo experiments on RBC from mice treated with AG-348 showed modest (20%) decreases in 2,310

DPG.6 Because 2,3-DPG is quantitatively the most important organic phosphate in human RBC (being present in approximately four times the concentration of ATP), its measurement in those treated and untreated patients may better contribute to our understanding of the overall impact of this molecule on the metabolic pathways within RBC. Overall, in vitro assessment of AG-348 reported here in this manuscript by Rab et al. provides evidence that this molecule stimulates the activity of mutant PK-R enzymes, consistent with recently reported clinical responses to this drug in patients with PK deficiency.10 In an early investigation of a patient with a typical PK deficiency, elevated levels of 2,3-DPG in RBC with a corresponding right shifted oxygen equilibrium curve were found. This subject with a well-documented deficiency in PK had a marked increase in the concentration of 2,3DPG (2.5-fold) and a 2.0-fold decrease in the oxygen affinity of whole blood.11 In future studies it would be interesting to assess the role of AG-348 in another hemolytic condition, e.g., sickle cell anemia where both DPG and the RBC oxygen affinity are altered. Although no major glycolytic enzymatic abnormality is known to occur in this condition, several cytosolic and membrane alterations do occur, including changes in band 3 linked to glycolytic pathways which were recently reported in sickle cell disease mice.12 haematologica | 2021; 106(1)

Editorials

Disclosures No Conflicts of interest to disclose. Contributions None.

References 1. Beutler E. Energy metabolism and maintenance of erythrocytes. In: Williams WJ, Beutler E, Erslev A, Lichtman M, editors. Hematology. 4th ed. New York, NY, USA: McGraw-Hill; 1990. pp. 355-368. 2. Brown KA. Erythrocytes metabolism and defects. Lab Med. 1996;27(5):329-333. 3. Beutler E, Gelbart T. Estimating the prevalence of pyruvate kinase deficiency from the gene frequency in the general white population. Blood. 2000;95(11):3585-3588. 4. Zanella A, Fermo E, Bianchi P, Valentini G. Red cell pyruvate kinase deficiency: molecular and clinical aspects. Br J Haematol. 2005;130(1):11-25. 5. Jurica MS, Mesecar A, Heath PJ, Shi W, Nowak T, Stoddard BL. The

allosteric regulation of pyruvate kinase by fructose-1,6-bisphosphate. Structure. 1998;6(2):195-210. 6. Kung C, Hixon J, Kosinski PA, et al. AG-348 enhances pyruvate kinase activity in red blood cells from patients with pyruvate kinase deficiency. Blood. 2017;130(11):1347-1356. 7. Perutz MF. Stereochemistry of cooperative effects in haemoglobin. Nature. 1970;228(5273):726-739. 8. Rab MAE, van Oirschot BA, Kosinski PA, et al. AG-348 (Mitapivat), an allosteric activator of red blood cell pyruvate kinase, increases enzymatic activity, protein stability, and ATP levels over a broad range of PKLR genotypes. Haematologica. 2020;106(1):238-249. 9. Rogers, SC, Ross JGC, dâ&#x20AC;&#x2122;Avignon A, et al. Sickle hemoglobin disturbs normal coupling among erythrocyte O2 content, glycolysis, and antioxidant capacity. Blood. 2013;121(9):1651-1662. 10. Grace RF, Rose C, Layton M, et al. Safety and efficacy of mitapivat in pyruvate pinase deficiency. N Engl J Med. 2019;381(10):933-944. 11. Delivoria-Papadopoulos M, Oski F, Gottlib AJ. Oxygen-hemoglobin dissociation curves: effect of inherited enzyme defects in the red cell. Science. 1969;165(3893):601-602. 12. Jana S, Strader MB, Meng F, Hicks W, et al. Hemoglobin oxidationdependent reactions promote interactions with band 3 and oxidative changes in sickle cell-derived microparticles. JCI Insight. 2018;3(21): e120451.

The arrival of personalized genomics in bone marrow failure Marcin W. Wlodarski Department of Hematology, St. Jude Childrenâ&#x20AC;&#x2122;s Research Hospital, Memphis, TN, USA and Division of Pediatric Hematology and Oncology, Department of Pediatrics and Adolescent Medicine, Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany E-mail: MARCIN W WLODARSKI - marcin.wlodarski@stjude.org doi:10.3324/haematol.2020.265124

rom the description of a first genetic bone marrow failure (BMF) syndrome (dyskeratosis congenita) in the literature1 it took over 80 years to discover FANCA as the first gene causing BMF in humans.2 Using a laborious process of positional cloning, which starts by finding patients with similar phenotype patterns, performing linkage analysis using polymorphic markers to finally localize the affected cDNA, a number of additional BMF genes have been identified in the following years. These include DKC1 discovered in 1998,3 RPS19 in 19994 and ELANE in 1999,5 to name just a few. Fueled by mapping the human genome and rapid advances in genome-wide technologies, a large proportion of genes predisposing to BMF and myelodysplastic syndromes (MDS) were identified using next-generation sequencing (NGS) technologies in the 21st century. Overall, roughly 100 genes have, thus, so far been associated with BMF/MDS syndromes. Classical inherited BMF syndromes can be categorized into DiamondBlackfan anemia (DBA), Fanconi anemia (FA), severe congenital neutropenia (SCN), dyskeratosis congenita (DC), Shwachman-Diamond syndrome, and congenital thrombocytopenias (Figure 1A). From a biological perspective, all these entities are caused by loss of function of fundamental cellular pathways such as DNA repair, ribosome or telomere maintenance. In recent years we have witnessed a series of new discoveries on hereditary conditions predisposing to MDS.6-8 The majority of these MDS predisposing genes are essential hematopoietic transcription factors and result in heterogeneous phenotypes that can bridge over to haematologica | 2021; 106(1)

the classical BMF spectrum. The landscape of germline genetic changes in BMF/MDS syndromes includes mutations in coding regions or noncoding alterations in promoters, regulatory elements, synonymous mutations, small deletions spanning single exons, and whole gene deletions (Figure 1B). Somatic alterations are mostly point mutations in leukemia driver genes and chromosomal gains or losses, but revertant uniparental disomies are also frequently found. Most BMF experts perform a phenotype-driven approach by first obtaining a family and medical history, followed by detailed physical examination and analysis of blood counts and bone marrow cellularity, morphology and cytogenetics. This approach most often guides the selection of an appropriate genetic testing platform to identify the causative gene (Figure 1C). For example, deletions (whole gene or intragenic) are a common cause of DBA and FA and for their detection, a well-established clinical copy number method is required (such as high resolution comparative genomic hybridization/single nucleotide polymorphism [CGH/SNP] array or multiplex ligation-dependent probe amplification [MLPA]). On the other hand, intronic noncoding mutations in GATA2 might escape standard whole exome sequencing (WES) diagnostics because of coverage gaps, or simply due to missing expert knowledge about such lesions. The diagnostic genetic process for BMF and hereditary MDS is often challenging due to the continuous addition of novel genotypes and the evolving phenotype spectrum, but also the clinical need for a rapid turnaround 11

Editorials

Figure 1. Genomics of bone marrow failure and hereditary myelodysplastic syndromes. (A) Syndromes and associated genes. (B) Genetic changes encountered in bone marrow failure/ hereditary myelodysplastic syndromes (BMF/MDS). (C) Genetic methods required to identify causative genetics. (D) Outline of study by Blombery et al. 2020. FA: Fanconi anemia; DC: dyskeratosis congenita; SCN: severe congenital neutropenia; SDS: Shwachman-Diamond syndrome; DBA: Diamond Blackfan anemia; CAMT: congenital amegakaryocytic thrombocytopenia; RUSAT: radioulnar synostosis with amegakaryocytic thrombocytopenia; IBMF: inherited BMF; AA/hypoMDS: aplastic anemia/hypocellular MDS; Uncl. BMF: unclassified BMF; CGH: comparative genomic hybridization; SNP: single nucleotide polymorphism; MLPA: multiplex ligation-dependent probe amplification; NGS: next-generation sequencing; WGS: whole exome sequencing.

time to guide therapeutic decisions. Blombery and co-workers6 are now presenting a very innovative multi-institutional prospective study from the Melbourne Genomic Health Alliance BMF Flagship (Figure 1D). They analyzed a series of 115 pediatric and adult patients diagnosed consecutively with BMF at four institutions in Australia between May 2017 and August 2018. The main purpose was to assess the impact of comprehensive genomic evaluation in this heterogenous cohort. The authors performed a tour de force analysis using a genomic pipeline consisting of WES, targeted NGS, and in selected cases additional RNA sequencing (RNAseq) to confirm variant effect on RNA expression and droplet digital PCR (ddPCR) technique to validate small somatic clones. Every patient received personalized genetic counseling and the pathogenic variants identified through this study were used for clinical decision making. Before starting genetic analysis, patients were grouped into three diagnostic categories: inherited BMF (23 patients), acquired aplastic anemia (AA)/hypocellular MDS (47 patients) and unclassified BMF (45). The analysis resulted in a change of diagnostic category in 26% of the cohort. Most notably, the authors discovered germline causes in 3 of 47 and 16 of 45 of patients with pretest diagnoses of AA/MDS and clinically unclassifiable BMF, respectively. Additionally, somatic mutations were discovered in 51% of AA/MDS group, 24% of unclassified BMF, and 4% of inherited BMF (Figure 1D). Recent studies have demonstrated the utility of comprehensive genomic analysis in diagnosing BMF and hereditary MDS. Zhang and colleagues employed an NGS panel 12

of 85 genes to find that 8 of 71 patients with idiopathic BMF/MDS carry damaging germline mutations in GATA2, RUNX1, DKC1 or LIGIV.9 They also demonstrated the utility of NGS panels to detect genomic copy variants in BMF. Another study focused on finding copy number variants in a large cohort consisting of BMF patients and found that 16% of them carry pathogenic copy number variants.10 A landmark study by a French group in 2018 investigated 179 patients with suspected but genetically unresolved BMF. Using WES sequencing of fibroblast DNA, they found pathogenic mutations affecting 28 genes in 48% of this cohort.11 Finally, in a most recent study using WES performed in 86 families with MDS/AML, known genes were found mutated in 57% of families, and 65 new candidate loci were discovered in 43% of cases.12 An important finding of this work was that there is no single gene that accounts for a large number of unresolved cases with familial disease. However, aforementioned studies were of retrospective nature in cohorts collected over many years to decades. This leads to several questions: What is the immediate benefit to patient care? Should investigators approach families about pathogenic variants found in research only studies and if yes, how do we do this? In the current study, Blombery et al.13 used an accredited genomic pipeline where results can be returned immediately after testing to clinicians and patients. This approach had important implications for clinical management, including the choice to perform allogeneic stem cell transplantation and optimal sibling donor selection, treatment (such as corticosteroids in DBA), identification of at risk family members and implementation of disease-specific haematologica | 2021; 106(1)

Editorials

surveillance. Several examples showcased the importance of such a real-time prospective study for patients with BMF. In one family, the identification of a pathogenic TINF2 mutation permitted family screening and implementing DC-specific surveillance. In another family, diagnosis of homozygous RAD51C Fanconi-like syndrome had implications for breast and ovarian cancer screening in family members. Interestingly, in one case with a newly identified homozygous FANCA mutation (that resulted in exon skipping based on RNAseq), the chromosomal fragility test was ambiguous. This emphasizes the clinical impacts of personalized genomics in BMF patients, especially when the clinical features and functional testing do not completely align with a BMF syndrome, to obtain an accurate diagnosis and avoid missed or delayed diagnosis. NGS technologies have already fulfilled their promise as a diagnostic tool in BMF/MDS. Now, we should ride on this impressive momentum and take advantage of international consortia to implement a rapid diagnostic pipeline for these rare disorders in a prospective setting, as has already been demonstrated for pediatric cancers and seriously ill infants.14,15 The virtue will be to use rapid â&#x20AC;&#x153;all-inoneâ&#x20AC;? genomic platforms such as WGS complemented by RNAseq allowing simultaneous detection of all types of disease-relevant genomic lesions. Personalized genomics for patients with BMF has finally arrived. Disclosures No Conflicts of interest to disclose. Contributions None.

References 1. Zinsser F. Atrophia cutis reticularis cum pigmentatione dystrophia

haematologica | 2021; 106(1)

unguium et leukoplakia oris. Ikonographia Dermatologica. 1906; 1906;5:219-223. 2. Strathdee CA, Gavish H, Shannon WR, Buchwald M. Cloning of cDNAs for Fanconi's anaemia by functional complementation. Nature. 1992;356(6372):763-767. 3. Heiss NS, Knight SW, Vulliamy TJ, et al. X-linked dyskeratosis congenita is caused by mutations in a highly conserved gene with putative nucleolar functions. Nat Genet. 1998;19(1):32-38. 4. Draptchinskaia N, Gustavsson P, Andersson B, et al. The gene encoding ribosomal protein S19 is mutated in Diamond-Blackfan anaemia. Nat Genet. 1999;21(2):169-175. 5. Horwitz M, Benson KF, Person RE, Aprikyan AG, Dale DC. Mutations in ELA2, encoding neutrophil elastase, define a 21-day biological clock in cyclic haematopoiesis. Nat Genet. 1999;23(4):433436. 6. Khincha PP, Savage SA. Genomic characterization of the inherited bone marrow failure syndromes. Semin Hematol. 2013;50(4):333347. 7. Kennedy AL, Shimamura A. Genetic predisposition to MDS: clinical features and clonal evolution. Blood. 2019;133(10):1071-1085. 8. Sahoo SS, Kozyra EJ, Wlodarski MW. Germline predisposition in myeloid neoplasms: Unique genetic and clinical features of GATA2 deficiency and SAMD9/SAMD9L syndromes. Best Pract Res Clin Haematol. 2020;33(3):101197. 9. Zhang MY, Keel SB, Walsh T, et al. Genomic analysis of bone marrow failure and myelodysplastic syndromes reveals phenotypic and diagnostic complexity. Haematologica. 2015;100(1):42-48. 10. Waespe N, Dhanraj S, Wahala M, et al. The clinical impact of copy number variants in inherited bone marrow failure syndromes. NPJ Genom Med. 2017;2:18. 11. Bluteau O, Sebert M, Leblanc T, et al. A landscape of germ line mutations in a cohort of inherited bone marrow failure patients. Blood. 2018;131(7):717-732. 12. Rio-Machin A, Vulliamy T, Hug N, et al. The complex genetic landscape of familial MDS and AML reveals pathogenic germline variants. Nat Commun. 2020;11(1):1044. 13. Blombery P, Fox L, Ryland GL, et al. Utility of clinical comprehensive genomic characterisation for diagnostic categorisation in patients presenting with hypocellular bone marrow failure syndromes. Haematologica. 2021;106(1):64-73. 14. Rusch M, Nakitandwe J, Shurtleff S, et al. Clinical cancer genomic profiling by three-platform sequencing of whole genome, whole exome and transcriptome. Nat Commun. 2018;9(1):3962. 15. Clark MM, Hildreth A, Batalov S, et al. Diagnosis of genetic diseases in seriously ill children by rapid whole-genome sequencing and automated phenotyping and interpretation. Sci Transl Med. 2019;11(489):eaat6177.

REVIEW ARTICLE Ferrata Storti Foundation

Emerging epigenetic therapeutics for myeloid leukemia: modulating demethylase activity with ascorbate Andrew B. Das, Carlos C. Smith-Díaz and Margreet C.M. Vissers

Department of Pathology and Biomedical Science, University of Otago, Christchurch, New Zealand

Haematologica 2021 Volume 106(1):14-25

ABSTRACT

Correspondence: ANDREW B. DAS andrew.das@otago.ac.nz Received: June 26, 2020. Accepted: September 17, 2020. Pre-published: October 22, 2020. https://doi.org/10.3324/haematol.2020.259283

©2021 Ferrata Storti Foundation Material published in Haematologica is covered by copyright. All rights are reserved to the Ferrata Storti Foundation. Use of published material is allowed under the following terms and conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode. Copies of published material are allowed for personal or internal use. Sharing published material for non-commercial purposes is subject to the following conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode, sect. 3. Reproducing and sharing published material for commercial purposes is not allowed without permission in writing from the publisher.

he past decade has seen a proliferation of drugs that target epigenetic pathways. Many of these drugs were developed to treat acute myeloid leukemia, a condition in which dysregulation of the epigenetic landscape is well established. While these drugs have shown promise, critical issues persist. Specifically, patients with the same mutations respond quite differently to treatment. This is true even with highly specific drugs that are designed to target the underlying oncogenic driver mutations. Furthermore, patients who do respond may eventually develop resistance. There is now evidence that epigenetic heterogeneity contributes, in part, to these issues. Cancer cells also have a remarkable capacity to ‘rewire’ themselves at the epigenetic level in response to drug treatment, and thereby maintain expression of key oncogenes. This epigenetic plasticity is a promising new target for drug development. It is therefore important to consider combination therapy in cases in which both driver mutations and epigenetic plasticity are targeted. Using ascorbate as an example of an emerging epigenetic therapeutic, we review the evidence for its potential use in both of these modes. We provide an overview of 2-oxoglutarate dependent dioxygenases with DNA, histone and RNA demethylase activity, focusing on those which require ascorbate as a cofactor. We also evaluate their role in the development and maintenance of acute myeloid leukemia. Using this information, we highlight situations in which the use of ascorbate to restore 2-oxoglutarate dependent dioxygenase activity could prove beneficial, in contrast to contexts in which targeted inhibition of specific enzymes might be preferred. Finally, we discuss how these insights could be incorporated into the rational design of future clinical trials.

Introduction Acute myeloid leukemia (AML) is a cancer that arises from the stem and progenitor cells of the hematopoietic system. These abnormally differentiated, clonal cells infiltrate the bone marrow, blood and other tissues. Unless treatment is initiated early, the rapid onset of this proliferative disease is fatal. Over the past 50 years there have been remarkable improvements in the treatment of AML. Initially incurable, the 5-year survival for young patients (<60 years old) is now greater than 50%.1 This progress was initiated in the 1970s when the effectiveness of cytarabine and anthracyclines was discovered. Subsequently, various regimens involving combinations of the two drugs were studied to find the optimum dose, with bone marrow transplants introduced as an option in 1977. Prognosis has been even better for acute promyelocytic leukemia, a subset of AML in which the use of all-trans retinoic acid (ATRA) and arsenic trioxide has increased 5-year survival to 90%. However, for older patients, for whom intensive chemotherapy is not an option, outcomes remain bleak with a median survival of 5-11 months.2 Clearly, more effective treatments for AML are needed, with important clinical questions remaining, including why is there a variable response to drug therapy, and why do some patients develop resistance to therapy? When considering new treatments for AML, it is important to first evaluate our current understanding of the contributing haematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML

Figure 1. Levels of heterogeneity in acute myeloid leukemia and potential treatment strategies to address them. See text for further details.

variables. This will help to frame current and emerging treatment strategies for AML (Figure 1), and provide a framework for considering ascorbate as an epigenetic therapeutic.

Genetic heterogeneity in acute myeloid leukemia As evidenced by the variable response to chemotherapy, AML is a heterogeneous disease. As the technologies available to investigate AML have developed, our understanding of where this heterogeneity lies has expanded. Morphological and cytogenetic heterogeneity have been studied for over 30 years and formed the basis for the French-Amercian-British classification system.3 As genetic sequencing technologies improved, frequently mutated genes were added to classification and prognostic prediction by the World Health Organization4 and European LeukemiaNet.5 To date, next-generation sequencing technologies have shed the greatest light on heterogeneity in AML. In 2016, a landmark study that sequenced DNA samples from more than 1,540 AML patients found recurrent mutations in over 100 genes.6 Combining this information with cytogenetic and clinical data has shown that AML comprises at least 11 different subgroups with prognostic relevance.6,7 Therefore, substantial heterogeneity in molecular pathology underpins the apparently homogenous phenotype of patients presenting in the clinic. Singlecell sequencing has added another layer of complexity to this scenario: multiple AML clones may exist in the same patient with their relative proportions changing over time with treatment and relapse,8,9 with different co-occurring mutations contributing to different clinical outcomes.6,7 A further insight emerging from large studies using next-generation sequencing to investigate AML is that the haematologica | 2021; 106(1)

mutational burden is low in AML compared to other cancers, and that proteins involved in epigenetic processes are early drivers of the cancer phenotype.6,10 This information is of particular interest because the reversibility of epigenetic modifications makes these processes, and the proteins that mediate them, attractive drug targets.11-13 These data have also highlighted the potential for precision medicine, where treatments can be matched to the patient on a case-by-case basis. In the past 5 years, selective inhibitors for mutations in FLT3,14 IDH115 and IDH216 have gained approval from the Food and Drug Administration. There are also a growing number of clinical trials for drugs targeting epigenetic writers (DOT1L, PRMT5), readers (BRD2/3/4) and erasers (HDAC, LSD1).13 However, the major difficulties with targeted therapy to date, epigenetic or otherwise, are familiar adversaries: variable response to the drug, and drug resistance. At first glance this seems to be a repetition of the same issues experienced with chemotherapy. However, the nuance here is that patients with the same mutations respond quite differently to treatment. This is reported even with highly specific drugs that are designed to target the common oncogenic driver mutation.15,17 By uncovering deeper layers of heterogeneity in AML, this obstacle has provided an indication as to how it may be overcome.

Epigenetic heterogeneity, plasticity and drug resistance We have highlighted two reasons why patients respond quite differently to treatment; inter-patient heterogeneity (genetic) and intra-patient heterogeneity (genetic/clonal). However, these factors alone do not explain why some patients eventually develop resistance to a targeted drug. 15

A.B. Das et al.

The most commonly proffered reason is that the development of new mutations confers a survival advantage.15,18 This can be due the founding clone in the primary tumor gaining mutations and evolving into the relapse clone, or a subclone of the founding clone surviving initial therapy, gaining additional mutations and expanding at relapse.19 However, there is growing evidence that non-genetic resistance also plays a role.20 Two key variables that contribute towards this outcome are epigenetic heterogeneity and epigenetic plasticity. Epigenetic heterogeneity refers to the different epigenetic states across a population of genetically identical cells. This is a well-studied phenomenon in embryology, whereby multicellular organisms generate a vast array of cell phenotypes from a single genome. What is remarkable about normal development is the capacity of cells and organisms to produce consistent phenotypic outcomes despite being challenged by variable conditions.21,22 Waddington coined the term “canalization” to refer to this capacity of cells and used it interchangeably with the word “buffering”.23 Cancers in general and AML in particular also display buffering by producing a consistent phenotype. However, this apparent homogeneity can be unmasked by chemotherapy or targeted treatment that provides a large selection pressure. Epigenetic heterogeneity across a population of genetically identical cells means that some will be in a transcriptional state that confers resistance. Depending on the size of the fraction with resistance, this could manifest as an initial response to treatment, followed by relapse once the resistant clone has had time to regenerate the disease. Importantly, epigenetic heterogeneity has been correlated with poorer outcomes and a shorter time to relapse in AML.24 A closely related concept is epigenetic plasticity, whereby a cell can alter its epigenetic state in a heritable manner in response to stimuli. Those cells that are able to explore the epigenetic landscape more extensively will be more likely to discover resistant states. How epigenetic heterogeneity and plasticity enable non-genetic resistance is of great interest and we recommend the review by Waddington for further discussion.23 Of interest for this review is that: (i) both genetic and non-genetic evolution contribute to resistance in AML24-26 and (ii) epigenetic plasticity is a new target for therapy in AML.27

OGDD utilize non-heme iron (Fe2+), 2-oxoglutarate and oxygen to catalyze a wide range of hydroxylation reactions. Many enzymes function as DNA, histone and RNA demethylases and have been shown to require ascorbate to maintain optimal activity (Table 1). This biochemistry underpins ascorbate’s link to epigenetics and has received increasing attention in multiple contexts including cancer.32,33 The findings are particularly relevant for AML because many of these demethylase enzymes are involved in the development and maintenance of AML (Table 1). Of the 20 OGDD enzymes with demethylase activity listed in Table 1, 17 play a role in the development or maintenance of AML, 15 require ascorbate as a cofactor, and 12 fall into both categories (see Table 1 for details).

Targeting driver mutations: restoring tumor-suppressor activity The use of ascorbate as a drug differs from the typical scenario in which small molecules are designed to inhibit catalytic activity or interfere with protein-protein interactions. The proposed mechanism of action is that ascorbate could promote enzyme activity in cases in which decreased activity contributes to the AML phenotype. This requires residual functional enzyme to be present. We therefore propose two further criteria to select likely molecular contexts in which ascorbate might function to target oncogenic drivers. Firstly, evidence that the target demethylase functions as a tumor-suppressor in AML and secondly, that loss-of-function mutations are heterozygous to ensure there is residual functional enzyme. Using the information collated in Table 1 along with previously published next-generation sequencing data from 878 AML patients accessed through cBioPortal,34 we found three demethylases that fit these criteria: TET2, KDM3B and KDM6A (Figure 2). The rest of the demethylases from Table 1 are either oncogenes in AML, are tumor suppressor genes but have increased or mixed gene expression profiles, or there is no current evidence that ascorbate acts as a cofactor. TET2, KDM3B and KDM6A are discussed here first, with the more diverse group of enzymes considered in the subsequent sections.

Context 1. Heterozygous mutations in TET2 Ascorbate as an emerging epigenetic therapeutic Our increased understanding of heterogeneity in AML has opened up new strategies for therapy (Figure 1). Drugs that inhibit or activate epigenetic proteins can be used to target either oncogenic driver mutations, or epigenetic plasticity. With this in mind, we review the current evidence on whether ascorbate can be used as an epigenetic therapeutic in AML. Ascorbate is synthesized from glucose in the liver or kidneys of most animals. However, humans have acquired mutations in L-gulonolactone oxidase, the terminal synthetic enzyme, and must therefore obtain ascorbate from their diet.28 Ascorbate has an essential role in human physiology through its capacity to act as an electron donor and cofactor for a wide range of enzymes including many of the 2-oxoglutarate dependent dioxygenases (OGDD).29,30 This is a large family of enzymes that includes the first enzymes known to utilize ascorbate as a cofactor, the collagen hydroxylases.31 The 16

Mutations in TET2 arise in approximately 10% of cases of AML.35-37 Similar to other mutations that affect epigenetic regulation in AML, TET2 mutations arise early in the development of hematopoietic malignancies.6 Mutations in TET2 and other epigenetic regulators are commonly found in otherwise healthy individuals with clonal hematopoiesis, who are at risk of subsequently developing blood cancers such as AML.38-40 Mutations in TET2 increase hematopoietic stem cell (HSC) self-renewal and the expansion of myeloid lineage cells.41 The acquisition of subsequent mutations can lead to AML.42,43 These findings support a double-hit model in which premalignant HSC with TET2 mutations subsequently undergo further mutations leading to the development of cancer.44 TET2 has a fundamental role in hematopoiesis, and enables the appropriate differentiation of HSC.41 The TET enzymes are involved in active demethylation of DNA via oxidation of 5-methylcytosine to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine, and 5-carboxylcytosine. haematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML Table 1. Iron and 2-oxoglutarate-dependent epigenetic enzymes, their dependency on ascorbate and link to myeloid malignancies.

Gene

Synonyms

Protein

Substrate

Effect of ascorbatea

Role in myeloid malignanciesb

TET1

CXXC6, KIAA1676, LCX

Methylcytosine dioxygenase TET1

5mC, 5hmC and 5fC

Increased activity in vitro51,91 and in cell culture50,51,53,92â&#x20AC;&#x201C;94

TET2

KIAA1546, Nbla00191

Methylcytosine dioxygenase TET2

5mC, 5hmC and 5fC

TET3

KIAA0401

Methylcytosine dioxygenase TET3

5mC, 5hmC and 5fC

Increased activity in vitro,91 in cell culture,50,51,53,61,92 mouse models41,52 and clinical setting89 Increased activity in vitro91 in cell culture53,92 and a mouse model41

Oncogene in AML95 and MLL-rearranged leukemia96. Reduced expression in AML97 Increased expression in MLL-rearranged leukemia96 Tumor suppressor in AML, MDS, CMML35,36,41,52,97

KDM2A

CXXC8, FBL11, FBL7, FBXL11, JHDM1A, KIAA1004

Lysine-specific demethylase 2A

H3K36me1/me2

Increased activity in cell culture73,74

KDM2B

CXXC2, FBL10, FBXL10, JHDM1B, NDY1, PCCX2

Lysine-specific demethylase 2B

H3K36me1/me2 H3K4me3

Increased activity in cell culture73,74

KDM3A

JHDM2A, JMJD1, JMJD1A, KIAA0742, TSGA

Lysine-specific demethylase 3A

H3K9me2/me1

KDM3B

C5orf7, JHDM2B, JMJD1B, KIAA1082

Lysine-specific demethylase 3B

H3K9me2/me1

Increased activity in vitro65 and in cell culture66 Increased activity in cell culture66

KDM3C

JHDM2C, JMJD1C, KIAA1380, TRIP8

Lysine-specific demethylase 3C

H3K9me2/me1

Unknowna

KDM4A

JHDM3A, JMJD2, JMJD2A, KIAA0677

Lysine-specific demethylase 4A

KDM4B

JHDM3B, JMJD2B, KIAA0876

Lysine-specific demethylase 4B

KDM4C

GASC1, JHDM3C, JMJD2C, KIAA0780

Lysine-specific demethylase 4C

Increased activity in vitro103 and in cell culture104 Increased expression and activity in cell culture104,106 Increased activity in cell culture104

KDM5A

JARID1A, RBBP2, RBP2

Lysine-specific demethylase 5A

H3K9me3/me2 H3K36me3/me2 H1.4K26me3/me2 H3K9me3/me2 H3K36me3/me2 H1.4K26me3/me2 H3K9me3/me2 H3K36me3/me2 H1.4K26me3/me2 H3K4me3/me2/me1

KDM5B

JARID1B, PLU1, RBBP2H1

Lysine-specific demethylase 5B

H3K4me3/me2/me1

Decreased expression in cell culture109

KDM5C

DXS1272E, JARID1C, SMCX, XE169 HY, HYA, JARID1D, KIAA0234, SMCY UTX

Lysine-specific demethylase 5C Lysine-specific demethylase 5D Lysine-specific demethylase 6A

H3K4me3/me2/me1

Unknowna

H3K4me3/me2/me1

Increased activity in vitro112 Increased activity in vitro71 in cell culture70

KDM5D KDM6A

H3K27me3/me2

Increased expression in cell culture107

Increased expression in AML high expression of TET3 may be a positive prognostic marker.97 Possible tumor suppressor in AML98 Tumor suppressor in MLL-rearranged leukemia Over expression induces differentiation of MLL-AF10 transformed cells.72 Oncogene, required for AML cellular proliferation. Highly expressed in AML.99 Unknownb

Tumor suppressor, transcriptionally regulates HOXA1 through retinoic acid response elements in acute myeloid leukemia.63 Also commonly deleted, or underexpressed in AML/MDS.100,101 Required for the survival and proliferation of some AML cell lines.102 Required for MLL-AF9 translocated AML.105 Required for MLL-AF9 translocated AML.105 Required for MLL-AF9 translocated AML.105 Potentially oncogenic in AML.78 NUP98/JARID1A gene fusion linked to pediatric acute megakaryoblastic leukemia.108 Potential tumor suppressor MLL-AML.77 Potentially oncogenic â&#x20AC;&#x201C; CML,110 AML78,79 Associated with chemoresistance pediatric AML111 Unknownb Tumor suppressor. Low expression correlated with adverse outcome in male AML patients. Mutations in KDM6A associated with cytarabine resistance.68 Homozygous loss continued on the next page

haematologica | 2021; 106(1)

A.B. Das et al. continued from the previous page

KDM6B

JMJD3, KIAA0346

Lysine-specific demethylase 6B

H3K27me3/me2

Increased activity in cell culture70,94

KDM7A

JHDM1D, KDM7, KIAA1718

Lysine-specific demethylase 7A

Increased activity in vitro116

KDM7B

PHF8, KIAA1111, ZNF422, JHDM1F

Histone lysine demethylase PHF8

FTO

KIAA1752

Alpha-ketoglutaratedependent dioxygenase FTO

H3K9me2/me1 H3K27me2/me1 H4K20me1 H3K9me2/me1 H3K27me2 H4K20Me1 N6-mA mRNA

Unknowna

Increased activity in vitro118

of UTX in mice associated with AML113 Potential oncogene in AML – involved in the regulation of HSC self-renewal, knockout impairs self-renewal.114 Overexpressed in MDS CD34+ cells115 Unknownb

Regulates ATRA response in APL. Downregulated in ATRA-resistant APL cells117 Oncogene in AML. Inhibition of ATRA-induced AML cell differentiation119

‘Unknown’, we did not find evidence linking these enzymes to dependence on ascorbate as well as KDM4D, KDM4E, KDM6C, KDM7C, ALKBH1, ALKBH5, RIOX1, RIOX2, JMJD6, and JMJD8. in vitro, in which activity was tested using purified protein or cell lysate. b‘Unknown’, we did not find evidence linking these enzymes to the development or maintenance of myeloid leukemia. a

Figure 2. Ascorbate-dependent demethylases that function as tumor suppressors in acute myeloid leukemia. Top left panel. Relevant components of the DNA hydroxymethylation and active demethylation pathway are depicted. Top right panel. Heterozygous mutations seen in de novo acute myeloid leukemia (AML) that result in decreased TET2 activity. These mutations are mutually exclusive but collectively are found to be mutated in 30-50% of AML patients.6,59,60,122 Bottom left panel. Decreased KDM3B expression is seen in 13% of AML patients at first presentation. Bottom right panel. Heterozygous mutations in KDM3B are only seen in 1% of patients at first presentation. These data are based on next-generation sequencing of 878 AML patients accessed through cBioPortal34 (DNAseq n=878; RNAseq n=165). 2-OG, 2-oxoglutarate; 2-HG, 2-hydroxyglutarate.

These molecules are intermediates in the pathway of DNA demethylation, and 5hmC may be an epigenetic mark with its own reader.45-48 Loss-of-function mutations in TET2 cause a significant reduction in 5hmC levels, highlighting the importance of TET2 in hydroxymethylation and epigenetic regulation more broadly.49 TET2, as a member of the family of OGDD, requires Fe2+, oxygen and 2-oxoglutarate for activity. It is now well-established that ascorbate also increases TET 18

enzyme activity and leads to an increase in 5hmC in multiple cell culture and animal models.33,37,50,51 While still an area of active research, ascorbate has been shown to act as an electron donor to reduce inactive Fe3+ back to its catalytically active Fe2+ state, thereby increasing the Fe2+ necessary for TET activity.32,33,50 Furthermore, Yin et al. used fluorescence quenching to demonstrate that ascorbate uniquely interacts with the C-terminal catalytic domain of TET2, with other strong reducing chemicals not having haematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML

a similar effect.50 This suggests something specific about the interaction between ascorbate and the TET enzymes, although more structural evidence is required.50 A number of prominent studies have explored the specific relationship between TET activity, ascorbate and the development of hematologic malignancies. Agathocleous et al. used Gulo−/− mice (which lack the ability to synthesize ascorbate) and showed that ascorbate deprivation resulted in a higher level of HSC relative to body mass and significantly reduced 5hmC levels in HSC, findings that are consistent with reduced TET activity.52 The authors then transplanted Flt3ITD/Tet2Δ/+ AML cells into both Gulo−/− and control mice which resulted in accelerated development of leukemia in the Gulo−/− mice compared with the controls. Administration of ascorbate to the Flt3ITD/Tet2Δ/+/Gulo−/− mice suppressed leukemia development and prolonged survival. These results suggest that ascorbate could be beneficial in the context of leukemia with TET2 mutations. Interestingly, HSC also exhibited a 14-fold increased level of the ascorbate transporter Slc23a2 and a 6-fold higher concentration of ascorbate compared to the levels in mature immune cells, further indicating the importance of intracellular ascorbate availability to support cellular differentiation.52 In another study, Cimmino et al. used RNA interference to induce and reverse the knockdown of Tet2 in mice.41 Tet2 knockdown caused aberrant self-renewal of HSC, which was reversed upon the restoration of TET2 activity. The restoration of TET2 activity promoted myeloid differentiation, cell death and DNA demethylation. Additionally, ascorbate administration was shown to pharmacologically mimic the effect of Tet2 restoration.41 An important corollary of these findings is that ascorbate depletion mimics a loss of TET2. Indeed, ascorbate deficiency was able to cooperate with Flt3ITD to promote leukemogenesis in a manner similar to Tet2 loss.52 A number of studies have found that patients with hematologic cancers are ascorbate deficient,53-55 and these patients are likely to have decreased TET2 activity even in the absence of TET2 mutations. Collectively, these results indicate that the restoration of TET2 activity via ascorbate supplementation could provide an avenue for reversing disease progression in AML cases linked to heterozygous loss-of-function mutations in TET2. However, it is also important to address the issue of specificity. To what extent are the effects of ascorbate seen in the above studies mediated by mechanisms other than increased TET2 activity? Both Agathocleous et al. and Cimmino et al. found that ascorbate also had a beneficial effect when both copies of TET2 were abrogated.41,52 It is possible that ascorbate could be increasing the activity of other OGDD including histone demethylases. Agathocleous et al. did not find any evidence of increased histone methylation. Furthermore, there were no significant changes in the expression levels of Tet1-3, loss of bone collagen, depletion in carnitine or increased reactive oxygen species. Interestingly, Cimmino et al. found that knocking down Tet3 in addition to Tet2 resulted in a loss of the beneficial effect of ascorbate on both colony-forming potential as well as 5hmC. These results suggest that Tet3 was responsible for the residual effect of ascorbate in the absence of Tet2. Another mechanism proposed in the literature is that ascorbate can act as a pro-oxidant when used at high concentrations.28 Specifically, adding milhaematologica | 2021; 106(1)

limolar amounts of ascorbate to cell culture media can generate extracellular hydrogen peroxide via redox cycling, leading to increased cell death. Alternatively, cells can take up dehydroascorbate, an oxidation product of ascorbate, via GLUT1 which could lead to oxidative stress as the cells reduce it back to ascorbate. Agathocleous et al. did not investigate this possibility because they were not using supraphysiological concentrations of ascorbate. Cimmino et al. used catalase as a control and showed that this had no effect on ascorbate blocking the aberrant replating capacity of Tet2Δ/+ AML cells. Furthermore, using dichlorofluorescein fluorescence, they were not able to detect generation of intracellular reactive oxygen species, even though this method can detect as little as 50 mM hydrogen peroxide. Given these results and the low concentrations of ascorbate used, it is unlikely that the effects seen were mediated by the production of hydrogen peroxide.

Context 2. Mutations in IDH1, IDH2 and WT1 also decrease TET2 activity In addition to heterozygous TET2 mutations, decreased TET2 activity can result from mutations in IDH1, IDH2, and WT1, particularly in the context of AML (Figure 2).5658 Mutant isocitrate dehydrogenase (IDH) enzymes generate the oncometabolite 2-hydroxyglutarate, which acts as a competitor of TET2, in competition with its physiological substrate, 2-oxoglutarate.56 Wilms tumor protein 1 (WT1) is a transcription factor that recruits TET2 to DNA, enabling promoter demethylation.59 In both contexts, hypermethylation and decreased hydroxymethylation provide evidence of decreased TET2 activity. Data from cBioPortal and Wang et al.59 clearly show that mutations in the IDH1/2-TET2-WT1 pathway are mutually exclusive. Furthermore, they are collectively present in 30– 50% of AML cases.6,59,60 Together, they constitute a distinct subtype of AML characterized by dysregulated DNA (hydroxy)methylation (Figure 2). Ascorbate administration to mouse cells with IDH1 mutations has also been shown to reduce the rate of cellular proliferation and increase the expression of genes associated with hematopoietic differentiation. In this experiment, TET2 knockdown by short interfering RNA led to lower 5hmC increases in ascorbate-treated IDH1mutant cells, a finding consistent with a TET2-dependent mechanism.61 Importantly, this study excluded the generation of hydrogen peroxide by using 2-phosphoascorbate. This analogue of ascorbate does not redox cycle and therefore does not generate hydrogen peroxide but is converted to ascorbate during transport across the cell membrane. In one clinical study, a patient with AML experienced a full clinical remission for 2.5 years following treatment with intravenous ascorbate.62 The patient had mutations in a number of genes including DNMT3a, NPM1, TET2 and WT1. The TET2 and WT1 mutations were present in separate subclones, both mutations were heterozygous, and both involved truncation of the resultant protein leading to a loss of activity.62 Interestingly, the WT1 clone did not emerge at relapse, suggesting that it may have been more sensitive to treatment. Together with the evidence in Context 1, these findings support the hypothesis that the clinical outcome of patients suffering from AML involving reduced TET2 activity might be improved by treatment with ascorbate.28 19

A.B. Das et al.

Context 3. Decreased KDM3B expression An unexplored context for the potential benefit of ascorbate in AML is the altered expression of KDM3B. Decreased expression of this demethylase occurs in 13% of AML cases (Figure 2). This was determined by an in silico analysis of cancer patientsâ&#x20AC;&#x2122; data from the COSMIC database.63 This study showed that, in contrast to other cancers, decreased KDM3B expression is a feature of AML. On the other hand, KDM3B acts as an oncogene in acute lymphoblastic leukemia,64 which highlights the importance of determining disease- and subtype-specific roles for each demethylase. Although homozygous deletions of KDM3B are found in AML, these only account for 1% of AML cases, suggesting that most cases with decreased expression could also express some functional enzyme. KDM3B (originally known as JHDM2A) specifically demethylates mono- and di-methyl lysine at position 9 on H3 (H3K9).65 Knockdown of this demethylase results in a build-up of H3K9me2 at specific promotors, and decreased expression of genes at these locations. Importantly, KDM3B is also a known tumor-suppressor in AML, specifically in cell lines with MLL-AF6/9 or PMLRARA translocations.63 Overexpression of KDM3B in these cell lines resulted in repressed colony formation, suggesting that restoration of KDM3B activity is beneficial in specific AML contexts.63 Although pharmacological restoration of KDM3B has not been investigated in the context of AML, there is clear evidence that ascorbate activates KDM3B in embryonic stem cells.66 It was shown that ascorbate induced widespread demethylation of H3K9me2 in a KDM3A- and KDM3B-dependent manner. Given that H3K9 methylation is a repressive mark with regards to transcription, it is possible that demethylation due to KDM3B activation in the AML context reverses repression required for maintenance of AML. This hypothesis is supported by evidence that came from a screen of over 200,000 small molecules in which Xu et al. found that KA-7 was capable of upregulating KDM3B activity.67 KA-7 repressed proliferation and colony-forming assays in MLL-rearranged acute leukemia with a concomitant increase in KDM3B activity (demonstrated by increased demethylation of H3K9). These data are proof-of-principle that increasing the activity of KDM3B may be beneficial for AML patients with decreased KDM3B expression. Whether this increase in activity could be achieved with ascorbate should be explored in the preclinical setting as heterozygous mutations of TET2 have been tested.

Context 4. Heterozygous KDM6A mutations, and decreased expression at relapse Of the contexts discussed in this section, heterozygous KDM6A mutations constitute the smallest percentage of cases of de novo AML. Along with the seminal paper by Papaemmanuil et al.,6 COSMIC and cBioportal data34 show that only 1% of patients with de novo AML have a heterozygous loss-of-function mutation (Figure 2). However, KDM6A appears to be a tumor suppressor; low KDM6A expression is associated with poor survival outcomes and when KDM6A mutations arise at relapse they confer cytarabine resistance.68 Although cells used in follow-up cell culture experiments were either KDM6A-null or KDM6A-replete, AML patients are more likely to have heterozygous mutations than homozygous deletions.69 Therefore, there is likely to be functional KDM6A pres20

ent, with the potential for ascorbate to increase residual enzyme activity. Interestingly, the dependence of KDM6A on ascorbate has been demonstrated both using the purified protein as well as in a hematopoietic stem cell model.70,71 Ascorbate was required for KDM6A-mediated demethylation of H3K27, which is a repressive chromatin mark. KDM6A occupies the promotors of HOX gene clusters, and demethylation of H3K27 is critical for expression of some of the HOX genes. Zhang et al. found that ascorbate was critical for the differentiation of hemogenic-endothelial cells into hematopoietic cells in a KDM6A-dependent manner. Collectively this evidence suggests that ascorbate may be beneficial for a small percentage of patients (~1%) at presentation, and potentially a much larger proportion of patients at relapse (see Context 7). As with KDM3A, this hypothesis requires further investigation at the preclinical level.

Context-dependent roles for epigenetic 2-oxoglutarate-dependent dioxygenases in acute myeloid leukemia Thus far, we have considered contexts in which activation of tumor suppressor activity might be beneficial. However, our analysis of the information available has highlighted some other contexts with greater complexity. The remaining demethylases in Table 1 are either oncogenes in AML, are tumor suppressor genes with increased or mixed gene expression profiles, or there is no current evidence that ascorbate acts as a cofactor. Important contexts to consider are those in which the demethylase expression is heterogeneous across the patient group or those in which ascorbate-dependent demethylases are actually required for the maintenance of AML, i.e., they function as oncogenes (Figure 3). While an in-depth consideration is outside the remit of this review, it is important to be aware of contexts in which inhibition of the respective demethylase might be recommended on a mechanistic basis.

Context 5. Heterogeneous expression across patients with acute myeloid leukemia The demethylases KDM2A, KDM2B, KDM4C, KDM5A, KDM6B, KDM7A, and KDM7B, are neither consistently upregulated nor downregulated across AML patients. Of these, KDM2A is a tumor-suppressor in the context of MLL-rearranged leukemia, in which demethylation of H3K36 leads to inactivation of genes required for maintenance of leukemia.72 This enzyme requires ascorbate for optimal activity;73,74 however, this finding needs to be validated in the setting of leukemia before this can be considered as a context for treatment with ascorbate.

Context 6. Inhibiting oncogenic demethylase activity The demethylases TET1, KDM3C, KDM4A, KDM4B, KDM5B, KDM5C and FTO, are individually mutated in <1% of cases, whereas they each display increased expression in 5-7% of AML cases (Figure 3). There is evidence that TET1, KDM3C, KDM4A, KDM4B, KDM5B and FTO function as oncogenes in some AML contexts (see Table 1 for details). It is therefore interesting to consider that the inhibition of demethylases may provide clinical benefit. This possibility has been considered since the initial discovery of histone demethylases, and we sughaematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML

Figure 3. Context-dependent roles for demethylases in acute myeloid leukemia. Top left panel. Heterogeneous expression of histone KDM in acute myeloid leukemia (AML) including some for which there is no clear evidence that ascorbate is required for optimal activity. Top right panel. A number of histone KDM are oncogenes and have upregulated expression in 5-7% of AML patients. Targeted inhibition of these enzymes might be beneficial. Bottom left panel. Decreased KDM6A expression is seen in 45% of AML patients at relapse and confers resistance to cytarabine.69 Bottom right panel. TET2 can demethylate enhancers and thereby play a role in epigenetic plasticity. Data are based on studies of 878 AML patients and were accessed through cBioPortal34 (DNAseq n= 878; RNAseq n= 165). 2-OG, 2-oxoglutarate.

gest the review by Thinnes et al. for further reading on this subject.75 It is interesting to note that histone deacetylase inhibitors are being investigated in clinical trials,13 whereas histone demethylase inhibitors have lagged behind.76 This is likely due to the multiple roles that histone demethylases play in AML disease progression. For example, there are conflicting reports in the literature as to whether KDM5B inhibition suppresses or promotes leukemogenesis.77-79 Clearly, further investigation is required to elucidate the specific contexts in which inhibition of these demethylases might have clinical benefit.

Targeting epigenetic plasticity Cancer stem cells can evade therapy by exploring the epigenetic landscape and finding transcriptional states that confer resistance.20,27,80 This is a form of Lamarckian induction that is enabled by epigenetic plasticity. For example, drug treatment may result in enhancer switching81 to maintain the expression of key survival genes.27 While targeting this capacity of cancer cells is a relatively new concept, there is some evidence that KDM6A and TET2 may enable plasticity in some contexts.

Context 7. Increasing KDM6A activity to overcome drug resistance Epigenetic proteins that mediate plasticity have been identified in different models of drug resistance in AML. Using overexpressed MLL-AF9 fusion protein to generate cell culture and mouse models of AML, BET inhibitor therhaematologica | 2021; 106(1)

apy was employed to create a robust model of drug resistance.25 Using this model, targeting the histone demethylase Lsd1(Kdm1a) overcame non-genetic, acquired resistance and re-sensitized the cells to BET inhibition.27 Interestingly, the CRISPR screen that identified Lsd1 as a target also showed that targeting Utx (Kdm6a) and Mll4 prevented differentiation of the resistant population. Lsd1 and Mll4 counter each other via demethylation/methylation of H3K4 at enhancers, suggesting that enhancer switching was mediating the resistant phenotype.27 Consistent with the emergence of new enhancers, there were also increased H3K27 acetylation and related markers at regions of increased chromatin accessibility. Importantly, KDM6A demethylates H3K27 in an ascorbate-dependent manner,70,71 which must occur prior to H3K27 acetylation.82 Given that knocking down Kdm6a prevented differentiation and resensitization, it is possible that ascorbate could increase Kdm6a activity and thereby augment the effect of Lsd1 inhibition in restoring sensitivity to BET inhibition. Further evidence in support of this hypothesis comes from clinical studies: when KDM6A mutations arise at relapse they confer cytarabine resistance.68 A follow-up study found that 45% of relapsed patients have lower KDM6A expression at relapse than at initial presentation.69 The same study found that KDM6A null cells were more resistant to cytarabine and daunorubicin, with re-expression of KDM6A restoring sensitivity to cytarabine. Determining whether treatment with ascorbate phenocopies KDM6A overexpression is a critical next step. Given that there are relatively few options for relapsed AML patients, investigating this possibility is rel21

A.B. Das et al.

Table 2. Clinical trials involving treatment with ascorbate for acute myeloid leukemia and myelodysplastic syndromes.

Trial number

Conditions

Specific mutations

Drug combinations

Control

Phase

Status

NCT00184054

AML

N/A

Single-arm

Completed

NCT03397173

MDS, AML

Single-arm

Recruiting

NCT00329498

MDS, AML

Heterozygous TET2 mutations required for patient enrolment N/A

Arsenic trioxide + ascorbate (IV) Azacitidine + ascorbate

Single-arm

Completed120

NCT03999723

MDS, AML, CMML

N/A

Recruiting

MDS, AML

N/A

Azacitidine + placebo Single-arm

NCT00671697

Completed121

NCT02877277

MDS, AML

Azacitidine + placebo

N/A

Completed89

Not registered

AML

TET2, DNMT3A, IDH2 mutations identified N/A

Manipulation of ascorbate levels Azacitidine + ascorbate Decitabine, arsenic trioxide + ascorbate Azacitidine + ascorbate

DCAGa

N/A

Completed86

Low-dose DCAG + ascorbate (IV)

AML: acute myeloid leukemia, N/A: not applicable; IV: intravenous; MDS: myelodysplastic syndrome; aDCAG: low-dose decitabine prior to aclarubicin and cytarabine.

evant for a large proportion of patients who are resistant to therapy.

Context 8. Heterogeneous roles for TET2 at enhancers A number of studies have found that TET2 has the capacity to demethylate DNA at enhancers. This has been demonstrated in embryonic stem cells, hematopoietic stem cells, AML and breast cancer.43,83,84 In the context of AML, loss of TET2 promotes leukemogenesis via hypermethylation of enhancers.43 Therefore, in those cases in which heterozygous TET2 mutations are part of the natural history of AML, DNA demethylation at key enhancers would be an additional mechanism in favor of ascorbate treatment. However, the specific enhancers that TET2 binds to are cell-specific and likely to be influenced by AML subtype.83 It is possible that resistance to targeted treatment in TET2 wild-type AML could arise by TET2dependent demethylation of enhancers leading to the expression of key survival genes. In this scenario direct inhibition of TET2 with small molecules might be more effective.85

Clinical trials and conclusions By providing an overview of the ascorbate-dependent OGDD with demethylase activity, we have highlighted a range of contexts in which ascorbate could be used as an epigenetic therapeutic in AML (Contexts 1-4 and 7). Undoubtedly, the greatest weight of preclinical data supporting this is in AML with decreased TET2 activity (Contexts 1 and 2). For Contexts 3-8 much pre-clinical work remains to be done to determine whether ascorbate or, conversely, OGDD inhibition may be of clinical benefit. The question remaining is, how can this be translated into clinical benefit? It is important to note that no single-agent therapy has been effective in AML; combination therapy has been the foundation of all successful treatments. Ascorbate will therefore need to be combined with chemotherapy or other targeted therapies in randomized clinical trials in 22

order to gauge its utility in the clinic. Thus far, only one randomized trial has reported outcomes from such a comparison: ascorbate plus decitabine prior to aclarubicin and cytarabine (A-DCAG, n=39) was compared to DCAG alone (n=34).86 Zhao et al. found that A-DCAG significantly increased the chance of clinical remission after first induction, as well as extending median overall survival by 6 months when compared to DCAG alone in patients over 60 years of age. This result is promising, although the findings need to be validated in a larger, randomized cohort. In addition to lack of randomization, we have noted that clinical trials looking at the benefit of ascorbate (Table 2) do not systematically evaluate the potential proposed in Contexts 1 and 2. Because most trials did not investigate TET2/IDH/WT1 mutational status, there was no way to infer the underlying mechanism of action. Future clinical trials involving ascorbate will need to stratify response by mutation status if epigenetic mechanisms of action are to be validated in patient cohorts. Another consideration is the route of ascorbate administration. This is important because the maximum steadystate plasma concentration achievable by oral dosing is approximately 100 mM.87 Preclinical and clinical data suggest that oral administration may be adequate for the maintenance of TET2 function.52,88 Specifically, Agathocleous et al. found that oral supplementation was sufficient to prevent the onset of leukemia when Gulo-/mice were transplanted with Tet2Î&#x201D;/+;Flt3ITD/+ leukemic cells. Although they did not measure 5hmC changes with supplementation, they did show that the 5hmC:mC ratio was lower in hematopoietic progenitor cells from Gulo-/mice transplanted with Tet2Î&#x201D;/+;Flt3ITD/+ leukemic cells than in Gulo+/+ mice transplanted with the same cells. 52 Furthermore, Gillberg et al. found that oral administration of ascorbate was able to increase the 5hmC:mC ratio in patients undergoing azacytidine treatment for myeloid neoplasms.89 On the other hand, Cimmino et al. used 250 mM ascorbate in cell culture, and intraperitoneal injections in experiments with mice to achieve similar results.41 Whether or not increasing plasma concentrations beyond the physiological maximum of 100 mM haematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML

increases TET2 activity further and provides additional clinical benefit in humans needs to be investigated. Although achieving supraphysiological levels would require administration via an intravenous route, this should be distinguished from high-dose intravenous trials in other cancers in which plasma ascorbate concentrations can be manipulated to exceed 10 mM transiently. Interestingly, modeling of ascorbate distribution through tissues suggests that 1 mM ascorbate is sufficient for saturation of tissues distant from the blood vessel supply.28 Increasing the plasma concentration above this level may not have an additional impact via the epigenetic mechanisms that we have described in this paper. Therefore, for clinicians and hematology researchers looking to investigate the use of ascorbate as an epigenetic therapeutic for AML, we recommend the following: (i) at a minimum, consider mutation, copy number, and gene expression profiling for TET2, IDH1, IDH2, WT1, KDM3B, and KDM6A and (ii) determine an appropriate dosing regimen. As part of this, whether or not supraphysiological concentrations by intravenous administration can increase OGDD activity beyond that achieved by oral supplementation needs to be investigated. In order to do this, levels of ascorbate and 5hmC (or histone methylation) in myeloid cells should be measured. Furthermore, the frequency of

References 1. Kantarjian H. Acute myeloid leukemia major progress over four decades and glimpses into the future. Am J Hematol. 2016;91(1):131-145. 2. Döhner H, Weisdorf DJ, Bloomfield CD. Acute myeloid leukemia. N Engl J Med. 2015;373(12):1136-1152. 3. Bennett JM, Catovsky D, Daniel M-T, et al. Proposals for the classification of the acute leukaemias. French-American-British (FAB) co-operative group. Br J Haematol. 1976;33 (4):451-458. 4. Arber DA, Orazi A, Hasserjian R, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood. 2016;127 (20):2391-2406. 5. Döhner H, Estey E, Grimwade D, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood. 2017;129 (4):424-447. 6. Papaemmanuil E, Gerstung M, Bullinger L, et al. Genomic classification and prognosis in acute myeloid leukemia. N Engl J Med. 2016;374(23):2209-2221. 7. Gerstung M, Papaemmanuil E, Martincorena I, et al. Precision oncology for acute myeloid leukemia using a knowledge bank approach. Nat Genet. 2017;49(3):332340. 8. Potter N, Miraki-Moud F, Ermini L, et al. Single cell analysis of clonal architecture in acute myeloid leukaemia. Leukemia 2019;33 (5):1113-1123. 9. Ferrando AA, López-Otín C. Clonal evolution in leukemia. Nat Med. 2017;23(10):1135-1145. 10. Kandoth C, McLellan MD, Vandin F, et al. Mutational landscape and significance across 12 major cancer types. Nature. 2013;502(7471):333-339. 11. Dawson MA, Kouzarides T. Cancer epigenetics: from mechanism to therapy. Cell.

haematologica | 2021; 106(1)

dosing and temporal relationship to standard-of-care treatment will need to be established. Where possible, deep sequencing90 to obtain mutational, copy number and gene expression profiling details should be carried out at diagnosis, remission and relapse to provide information on the sensitivity of subclones to treatment. This will provide ample data to determine whether epigenetic mechanisms of action for ascorbate can be validated in patient cohorts. Disclosures We are supported by the University of Otago, and have received funding from Canterbury Medical Research Fund, Bone Marrow Cancer Research Trust, and a NZSO Roche Translational Cancer Research Fellowship. Contributions ABD, CCSD and MCMV wrote the manuscript and approved the final version. Acknowledgments We are supported by the University of Otago, and have received funding from Canterbury Medical Research Foundation, Bone Marrow Cancer Research Trust, and a NZSO Roche Translational Cancer Research Fellowship. Biorender was used in the making of figures for this publication.

2012;150(1):12-27. 12. Wouters BJ, Delwel R. Epigenetics and approaches to targeted epigenetic therapy in acute myeloid leukemia. Blood. 2015;127 (1):42-53. 13. Fennell KA, Bell CC, Dawson MA. Epigenetic therapies in acute myeloid leukaemia: where to from here? Blood. 2019;134(22):1891-1901. 14. Stone RM, Mandrekar SJ, Sanford BL, et al. Midostaurin plus chemotherapy for acute myeloid leukemia with a FLT3 mutation. N Engl J Med. 2017;377(5):454-464. 15. Quek L, David MD, Kennedy A, et al. Clonal heterogeneity of acute myeloid leukemia treated with the IDH2 inhibitor enasidenib. Nat Med. 2018;24(8):1167-1177. 16. Stein EM, DiNardo CD, Pollyea DA, et al. Enasidenib in mutant IDH2 relapsed or refractory acute myeloid leukemia. Blood. 2017;130(6):722-731. 17. Stein EM, DiNardo CD, Fathi AT, et al. Molecular remission and response patterns in patients with mutant- IDH2 acute myeloid leukemia treated with enasidenib. Blood. 2019;133(7):676-687. 18. Martincorena I, Campbell PJ. Somatic mutation in cancer and normal cells. Science. 2015;349(6255):1483-1489. 19. Ding L, Ley TJ, Larson DE, et al. Clonal evolution in relapsed acute myeloid leukaemia revealed by whole-genome sequencing. Nature. 2012;481(7382):506-510. 20. Bell CC, Gilan O. Principles and mechanisms of non-genetic resistance in cancer. Br J Cancer. 2020;122(4):465-472. 21. Waddington CH. Canalization of development and the inheritance of acquired characters. Nature. 1942;150(3811):563-565. 22. Pujadas E, Feinberg AP. Regulated noise in the epigenetic landscape of development and disease. Cell. 2012;148(6):1123-1131. 23. Waddington CH. The Strategy Of The Genes: A Discussion Of Some Aspects Of Theoretical Biology. 1st edition. London: Allen and Unwin. 1957.

24. Li S, Garrett-Bakelman FE, Chung SS, et al. Distinct evolution and dynamics of epigenetic and genetic heterogeneity in acute myeloid leukemia. Nat Med. 2016;22(7): 792-799. 25. Fong CY, Gilan O, Lam EYN, et al. BET inhibitor resistance emerges from leukaemia stem cells. Nature. 2015;525(7570):538-542. 26. Shlush LI, Mitchell A, Heisler L, et al. Tracing the origins of relapse in acute myeloid leukaemia to stem cells. Nature. 2017;547(7661):104-108. 27. Bell CC, Fennell KA, Chan Y-C, et al. Targeting enhancer switching overcomes non-genetic drug resistance in acute myeloid leukaemia. Nat Commun. 2019;10(1):2723. 28. Vissers MCM, Das AB. Potential mechanisms of action for vitamin C in cancer: reviewing the evidence. Front Physiol. 2018;9:809. 29. Kuiper C, Vissers MCM. Ascorbate as a cofactor for Fe- and 2-oxoglutarate dependent dioxygenases: physiological activity in tumor growth and progression. Front Oncol. 2014;4:359. 30. Vissers MCM, Das AB. Ascorbate as an enzyme cofactor. In: Chen Q, Vissers MCM, editors. Vitamin C: New Biochemical and Functional Insights. Boca Raton: CRC Press. p28. 31. Murad S, Grove D, Lindberg KA, Reynolds G, Sivarajah A, Pinnell SR. Regulation of collagen synthesis by ascorbic acid. Proc Natl Acad Sci U S A. 1981;78(5):2879-2882. 32. Monfort A, Wutz A. Breathing-in epigenetic change with vitamin C. EMBO Rep. 2013;14(4):337-346. 33. Young JI, Züchner S, Wang G. Regulation of the epigenome by vitamin C. Annu Rev Nutr. 2015;35(1):545-564. 34. Cerami E, Gao J, Dogrusoz U, et al. The cBio Cancer Genomics Portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2(5): 401-404. 35. Delhommeau F, Dupont S, Della Valle V, et

A.B. Das et al. al. Mutation in TET2 in myeloid cancers. N Engl J Med. 2009;360(22):2289-2301. 36. Ferrone CK, Blydt-Hansen M, Rauh MJ. Age-associated TET2 mutations: common drivers of myeloid dysfunction, cancer and cardiovascular disease. Int J Mol Sci. 2020;21(2):626. 37. Solary E, Bernard OA, Tefferi A, Fuks F, Vainchenker W. The ten-eleven translocation-2 (TET2) gene in hematopoiesis and hematopoietic diseases. Leukemia. 2014;28 (3):485-496. 38. Genovese G, Kähler AK, Handsaker RE, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med. 2014;371(26):2477-2487. 39. Busque L, Patel JP, Figueroa ME, et al. Recurrent somatic TET2 mutations in normal elderly individuals with clonal hematopoiesis. Nat Genet. 2012;44(11): 1179-1181. 40. Xie M, Lu C, Wang J, et al. Age-related mutations associated with clonal hematopoietic expansion and malignancies. Nat Med. 2014;20(12):1472-1478. 41. Cimmino L, Dolgalev I, Wang Y, et al. Restoration of TET2 function blocks aberrant self-renewal and leukemia progression. Cell. 2017;170(6):1079-1095. 42. Shih AH, Jiang Y, Meydan C, et al. Mutational cooperativity linked to combinatorial epigenetic gain of function in acute myeloid leukemia. Cancer Cell. 2015;27(4): 502-515. 43. Rasmussen KD, Jia G, Johansen JV, et al. Loss of TET2 in hematopoietic cells leads to DNA hypermethylation of active enhancers and induction of leukemogenesis. Genes Dev. 2015;29(9):910-922. 44. Rasmussen KD, Helin K. Role of TET enzymes in DNA methylation, development, and cancer. Genes Dev. 2016;30(7): 733-750. 45. Crawford DJ, Liu MY, Nabel CS, Cao X-J, Garcia BA, Kohli RM. Tet2 catalyzes stepwise 5-methylcytosine oxidation by an iterative and de novo mechanism. J Am Chem Soc. 2016;138(3):730-733. 46. He YF, Li BZ, Li Z, et al. Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science. 2011;333(6047):1303-1307. 47. Kohli RM, Zhang Y. TET enzymes, TDG and the dynamics of DNA demethylation. Nature. 2013;502(7472):472-479. 48. Spruijt CG, Gnerlich F, Smits AH, et al. Dynamic readers for 5-(hydroxy)methylcytosine and its oxidized derivatives. Cell. 2013;152(5):1146-1159. 49. Ko M, Huang Y, Jankowska AM, et al. Impaired hydroxylation of 5-methylcytosine in myeloid cancers with mutant TET2. Nature. 2010;468(7325):839-843. 50. Yin R, Mao SQ, Zhao B, et al. Ascorbic acid enhances Tet-mediated 5-methylcytosine oxidation and promotes DNA demethylation in mammals. J Am Chem Soc. 2013;135(28):10396-10403. 51. Blaschke K, Ebata KT, Karimi MM, et al. Vitamin C induces Tet-dependent DNA demethylation and a blastocyst-like state in ES cells. Nature. 2013;500(7461):222-226. 52. Agathocleous M, Meacham CE, Burgess RJ, et al. Ascorbate regulates haematopoietic stem cell function and leukaemogenesis. Nature. 2017;549(7673):476-481. 53. Liu M, Ohtani H, Zhou W, et al. Vitamin C increases viral mimicry induced by 5-aza-2′deoxycytidine. Proc Natl Acad Sci. 2016;113(37):10238-10244. 54. Huijskens MJAJ, Wodzig WKWH, Walczak

M, Germeraad WTV, Bos GMJ. Ascorbic acid serum levels are reduced in patients with hematological malignancies. Results Immunol. 2016;6:8-10. 55. Carr AC, Spencer E, Das A, et al. Patients undergoing myeloablative chemotherapy and hematopoietic stem cell transplantation exhibit depleted vitamin C status in association with febrile neutropenia. Nutrients. 2020;12(6):1879. 56. Figueroa ME, Abdel-Wahab O, Lu C, et al. Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell. 2010;18(6):553-567. 57. Rampal R, Alkalin A, Madzo J, et al. DNA hydroxymethylation profiling reveals that WT1 mutations result in loss of TET2 function in acute myeloid leukemia. Cell Rep. 2014;9(5):1841-1856. 58. Magotra M, Sakhdari A, Lee PJ, et al. Immunohistochemical loss of 5-hydroxymethylcytosine expression in acute myeloid leukaemia: relationship to somatic gene mutations affecting epigenetic pathways. Histopathology. 2016;69(6):10551065. 59. Wang Y, Xiao M, Chen X, et al. WT1 recruits TET2 to regulate its target gene expression and suppress leukemia cell proliferation. Mol Cell. 2015;57(4):662-673. 60. Cancer Genome Atlas Research Network. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 2013;368(22):2059-2074. 61. Mingay M, Chaturvedi A, Bilenky M, et al. Vitamin C-induced epigenomic remodelling in IDH1 mutant acute myeloid leukaemia. Leukemia. 2018;32(1):11-20. 62. Das AB, Kakadia PM, Wojcik D, et al. Clinical remission following ascorbate treatment in a case of acute myeloid leukemia with mutations in TET2 and WT1. Blood Cancer J. 2019;9(10):82. 63. Xu X, Nagel S, Quentmeier H, et al. KDM3B shows tumor-suppressive activity and transcriptionally regulates HOXA1 through retinoic acid response elements in acute myeloid leukemia. Leuk Lymphoma. 2018;59(1):204-213. 64. Kim J-Y, Kim K-B, Eom GH, et al. KDM3B is the H3K9 demethylase involved in transcriptional activation of lmo2 in leukemia. Mol Cell Biol. 2012;32(14):2917-2933. 65. Yamane K, Toumazou C, Tsukada Y ichi, et al. JHDM2A, a JmjC-containing H3K9 demethylase, facilitates transcription activation by androgen receptor. Cell. 2006;125(3):483-495. 66. Ebata KT, Mesh K, Liu S, et al. Vitamin C induces specific demethylation of H3K9me2 in mouse embryonic stem cells via Kdm3a/b. Epigenetics Chromatin. 2017;10 (1):36. 67. Xu X, Dirks WG, Drexler HG, Hu Z. A small molecular agonist of histone demethylase KDM3B selectively represses MLLrearranged acute Leukemia. Blood. 2018;132 (Suppl_1):4682-4682. 68. Greif PA, Hartmann L, Vosberg S, et al. Evolution of cytogenetically normal acute myeloid leukemia during therapy and relapse: an exome sequencing study of 50 patients. Clin Cancer Res. 2018;24(7):17161726. 69. Stief SM, Hanneforth A-L, Weser S, et al. Loss of KDM6A confers drug resistance in acute myeloid leukemia. Leukemia. 2020;34(1):50-62. 70. Zhang T, Huang K, Zhu Y, et al. Vitamin C– dependent lysine demethylase 6 (KDM6)-

mediated demethylation promotes a chromatin state that supports the endothelial-tohematopoietic transition. J Biol Chem. 2019;294(37):13657-13670. 71. Lee MG, Villa R, Trojer P, et al. Demethylation of H3K27 regulates polycomb recruitment and H2A ubiquitination. Science. 2007;318(5849):447-450. 72. Zhu L, Li Q, Wong SHK, et al. ASH1L links histone H3 lysine 36 dimethylation to MLL Leukemia. Cancer Discov. 2016;6(7):770783. 73. Wang T, Chen K, Zeng X, et al. The histone demethylases Jhdm1a/1b enhance somatic cell reprogramming in a vitamin-C-dependent manner. Cell Stem Cell. 2011;9(6):575587. 74. Tsukada Y, Fang J, Erdjument-Bromage H, et al. Histone demethylation by a family of JmjC domain-containing proteins. Nature. 2006;439(7078):811-816. 75. Thinnes CC, England KS, Kawamura A, Chowdhury R, Schofield CJ, Hopkinson RJ. Targeting histone lysine demethylases progress, challenges, and the future. Biochim Biophys Acta. 2014;1839(12):1416-1432. 76. Tsai CT, So CWE. Epigenetic therapies by targeting aberrant histone methylome in AML: molecular mechanisms, current preclinical and clinical development. Oncogene. 2017;36(13):1753-1759. 77. Wong SHK, Goode DL, Iwasaki M, et al. The H3K4-methyl epigenome regulates leukemia stem cell oncogenic potential. Cancer Cell. 2015;28(2):198-209. 78. Shokri G, Doudi S, Fathi-Roudsari M, Kouhkan F, Sanati M-H. Targeting histone demethylases KDM5A and KDM5B in AML cancer cells: a comparative view. Leuk Res. 2018;68:105-111. 79. Su H, Ma X, Huang Y, Han H, Zou Y, Huang W. JARID1B deletion induced apoptosis in Jeko-1 and HL-60 cell lines. Int J Clin Exp Pathol. 2015;8(1):171-183. 80. Shaffer SM, Dunagin MC, Torborg SR, et al. Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance. Nature. 2017;546(7658):431-435. 81. Godfrey L, Crump NT, Thorne R, et al. DOT1L inhibition reveals a distinct subset of enhancers dependent on H3K79 methylation. Nat Commun. 2019;10(1):2803. 82. Lavarone E, Barbieri CM, Pasini D. Dissecting the role of H3K27 acetylation and methylation in PRC2 mediated control of cellular identity. Nat Commun. 2019;10 (1):1679. 83. Rasmussen KD, Berest I, Kebler S, et al. TET2 binding to enhancers facilitates transcription factor recruitment in hematopoietic cells. Genome Res. 2019;29(4):564-575. 84. Wang L, Ozark PA, Smith ER, et al. TET2 coactivates gene expression through demethylation of enhancers. Sci Adv. 2018;4(11):eaau6986. 85. Chua GNL, Wassarman KL, Sun H, et al. Cytosine-based TET enzyme inhibitors. ACS Med Chem Lett. 2019;10(2):180-185. 86. Zhao H, Zhu H, Huang J, et al. The synergy of vitamin C with decitabine activates TET2 in leukemic cells and significantly improves overall survival in elderly patients with acute myeloid leukemia. Leuk Res. 2018;66:1-7. 87. Lykkesfeldt J, Tveden-Nyborg P. The pharmacokinetics of vitamin C. Nutrients. 2019;11(10):2412. 88. Gillberg L, Ørskov AD, Liu M, Harsløf LBS, Jones PA, Grønbæk K. Vitamin C - a new player in regulation of the cancer epigenome. Semin Cancer Biol. 2018;51:59-67. 89. Gillberg L, Ørskov AD, Nasif A, et al. Oral

haematologica | 2021; 106(1)

Emerging epigenetic therapeutics for AML vitamin C supplementation to patients with myeloid cancer on azacitidine treatment: normalization of plasma vitamin C induces epigenetic changes. Clin Epigenetics. 2019;11(1):143. 90. Griffith M, Miller CA, Griffith OL, et al. Optimizing cancer genome sequencing and analysis. Cell Syst. 2015;1(3):210-223. 91. Hore TA, von Meyenn F, Ravichandran M, et al. Retinol and ascorbate drive erasure of epigenetic memory and enhance reprogramming to naĂŻve pluripotency by complementary mechanisms. Proc Natl Acad Sci U S A. 2016;113(43):12202-12207. 92. Minor EA, Court BL, Young JI, Wang G. Ascorbate induces ten-eleven translocation (Tet) methylcytosine dioxygenase-mediated generation of 5-hydroxymethylcytosine. J Biol Chem. 2013;288(19):13669-13674. 93. Chen J, Guo L, Zhang L, et al. Vitamin C modulates TET1 function during somatic cell reprogramming. Nat Genet. 2013;45(12): 1504-1509. 94. He X, Kim M, Kim S, et al. Vitamin C facilitates dopamine neuron differentiation in fetal midbrain through TET1- and JMJD3dependent epigenetic control manner. Stem Cells. 2015;33(4):1320-1332. 95. Jiang X, Hu C, Ferchen K, et al. Targeted inhibition of STAT/TET1 axis as a therapeutic strategy for acute myeloid leukemia. Nat Commun. 2017;8(1):2099. 96. Huang H, Jiang X, Li Z, et al. TET1 plays an essential oncogenic role in MLL-rearranged leukemia. Proc Natl Acad Sci U S A. 2013;110(29):11994-11999. 97. Zhang T, Zhao Y, Zhao Y, Zhou J. Expression and prognosis analysis of TET family in acute myeloid leukemia. Aging (Albany NY). 2020;12(6):5031-5047. 98. Shrestha R, Sakata-Yanagimoto M, Maie K, et al. Molecular pathogenesis of progression to myeloid leukemia from TET-insufficient status. Blood Adv. 2020;4(5):845-854. 99. He J, Nguyen AT, Zhang Y. KDM2b/JHDM1b, an H3K36me2-specific demethylase, is required for initiation and maintenance of acute myeloid leukemia. Blood. 2011;117(14):3869-3880. 100. Dâ&#x20AC;&#x2122;Oto A, Tian Q, Davidoff AM, Yang J. Histone demethylases and their roles in cancer epigenetics. J Med Oncol Ther. 2016;1(2):34-40. 101. MacKinnon RN, Kannourakis G, Wall M,

haematologica | 2021; 106(1)

Campbell LJ. A cryptic deletion in 5q31.2 provides further evidence for a minimally deleted region in myelodysplastic syndromes. Cancer Genet. 2011;204(4):187-194. 102. Chen M, Zhu N, Liu X, et al. JMJD1C is required for the survival of acute myeloid leukemia by functioning as a coactivator for key transcription factors. Genes Dev. 2015;29(20):2123-2139. 103. Klose RJ, Yamane K, Bae Y, et al. The transcriptional repressor JHDM3A demethylates trimethyl histone H3 lysine 9 and lysine 36. Nature. 2006;442(7100):312-316. 104. Song MH, Nair VS, Oh KI. Vitamin C enhances the expression of IL17 in a Jmjd2dependent manner. BMB Rep. 2017;50 (1):49-54. 105. Agger K, Miyagi S, Pedersen MT, Kooistra SM, Johansen JV, Helin K. Jmjd2/Kdm4 demethylases are required for expression of Il3ra and survival of acute myeloid leukemia cells. Genes Dev. 2016;30(11):1278-1288. 106. Liu TM, Yildirim ED, Li P, et al. Ascorbate and iron are required for the specification and long-term self-renewal of human skeletal mesenchymal stromal cells. Stem Cell Rep. 2020;14(2):210-225. 107. Eid W, Abdel-Rehim W. Vitamin C promotes pluripotency of human induced pluripotent stem cells via the histone demethylase JARID1A. Biol Chem. 2016;397(11):1205-1213. 108. De Rooij JDE, Hollink IHIM, ArentsenPeters STCJM, et al. NUP98/JARID1A is a novel recurrent abnormality in pediatric acute megakaryoblastic leukemia with a distinct HOX gene expression pattern. Leukemia. 2013;27(12):2280-2288. 109. Yu X-X, Liu Y-H, Liu X-M, et al. Ascorbic acid induces global epigenetic reprogramming to promote meiotic maturation and developmental competence of porcine oocytes. Sci Rep. 2018;8(1):6132. 110. Xue S, Lam YM, He Z, et al. Histone lysine demethylase KDM5B maintains chronic myeloid leukemia via multiple epigenetic actions. Exp Hematol. 2020;82:53-65. 111. Zhan D, Zhang Y, Xiao P, et al. Whole exome sequencing identifies novel mutations of epigenetic regulators in chemorefractory pediatric acute myeloid leukemia. Leukemia Res. 2018;65:20-24. 112. Lee MG, Norman J, Shilatifard A, Shiekhattar R. Physical and functional asso-

ciation of a trimethyl H3K4 demethylase and Ring6a/MBLR, a polycomb-like protein. Cell. 2007;128(5):877-887. 113. Gozdecka M, Meduri E, Mazan M, et al. UTX-mediated enhancer and chromatin remodeling suppresses myeloid leukemogenesis through noncatalytic inverse regulation of ETS and GATA programs. Nat Genet. 2018;50(6):888-894. 114. Mallaney C, Ostrander EL, Celik H, et al. Kdm6b regulates context-dependent hematopoietic stem cell self-renewal and leukemogenesis. Leukemia. 2019;33(10): 2506-2521. 115. Wei Y, Chen R, Dimicoli S, et al. Global H3K4me3 genome mapping reveals alterations of innate immunity signaling and overexpression of JMJD3 in human myelodysplastic syndrome CD34+ cells. Leukemia. 2013;27(11):2177-2186. 116. Tsukada Y -i., Ishitani T, Nakayama KI. KDM7 is a dual demethylase for histone H3 Lys 9 and Lys 27 and functions in brain development. Genes Dev. 2010;24(5):432-437. 117. Arteaga MF, Mikesch J-H, Qiu J, et al. The histone demethylase PHF8 governs retinoic acid response in acute promyelocytic leukemia. Cancer Cell. 2013;23(3):376-389. 118. Gerken T, Girard CA, Tung Y-CL, et al. The obesity-associated FTO gene encodes a 2oxoglutarate-dependent nucleic acid demethylase. Science. 2007;318(5855):14691472. 119. Li Z, Weng H, Su R, et al. FTO plays an oncogenic role in acute myeloid leukemia as a N6-methyladenosine RNA demethylase. Cancer Cell. 2017;31(1):127-141. 120. Park CH, Kimler BF, Yi SY, et al. Depletion of L-ascorbic acid alternating with its supplementation in the treatment of patients with acute myeloid leukemia or myelodysplastic syndromes. Eur J Haematol. 2009;83(2):108118. 121. Welch JS, Klco JM, Gao F, et al. Combination decitabine, arsenic trioxide, and ascorbic acid for the treatment of myelodysplastic syndrome and acute myeloid leukemia: a phase I study. Am J Hematol. 2011;86(9): 796-800. 122. Metzeler KH, Herold T, RothenbergThurley M, et al. Spectrum and prognostic relevance of driver gene mutations in acute myeloid leukemia. Blood. 2016;128(5):686698.

ARTICLEARTICLE REVIEW Ferrata Storti Foundation

Differentiation therapy of myeloid leukemia: four decades of development Vikas Madan1 and H. Phillip Koeffler1,2,3

Cancer Science Institute of Singapore, National University of Singapore, Singapore; Cedars-Sinai Medical Center, Division of Hematology/Oncology, UCLA School of Medicine, Los Angeles, CA, USA and 3Department of Hematology-Oncology, National University Cancer Institute of Singapore (NCIS), National University Hospital, Singapore.

1 2

Haematologica 2021 Volume 106(1):26-38

ABSTRACT

Correspondence: VIKAS MADAN vikasmadan@aol.com H. PHILLIP KOEFFLER H.Koeffler@cshs.org Received: June 10, 2020. Accepted: September 15, 2020. Pre-published: October 9, 2020. https://doi.org/10.3324/haematol.2020.262121

ÂŠ2021 Ferrata Storti Foundation Material published in Haematologica is covered by copyright. All rights are reserved to the Ferrata Storti Foundation. Use of published material is allowed under the following terms and conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode. Copies of published material are allowed for personal or internal use. Sharing published material for non-commercial purposes is subject to the following conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode, sect. 3. Reproducing and sharing published material for commercial purposes is not allowed without permission in writing from the publisher.

cute myeloid leukemia is characterized by arrested differentiation, and agents that overcome this block are therapeutically useful, as shown by the efficacy of all-trans retinoic acid in acute promyelocytic leukemia. However, the early promise of differentiation therapy did not translate into clinical benefit for other subtypes of acute myeloid leukemia, in which cytotoxic chemotherapeutic regimens remained the standard of care. Recent advances, including insights from sequencing of acute myeloid leukemia genomes, have led to the development of targeted therapies, comprising agents that induce differentiation of leukemic cells in preclinical models and clinical trials, thus rejuvenating interest in differentiation therapy. These agents act on various cellular processes including dysregulated metabolic programs, signaling pathways, epigenetic machinery and the cell cycle. In particular, inhibitors of mutant IDH1/2 and FLT3 have shown clinical benefit, leading to approval by regulatory bodies of their use. Besides the focus on recently approved differentiation therapies, this review also provides an overview of differentiation-inducing agents being tested in clinical trials or investigated in preclinical research. Combinatorial strategies are currently being tested for several agents (inhibitors of KDM1A, DOT1L, BET proteins, histone deacetylases), which were not effective in clinical studies as single agents, despite encouraging anti-leukemic activity observed in preclinical models. Overall, recently approved drugs and new investigational agents being developed highlight the merits of differentiation therapy; and ongoing studies promise further advances in the treatment of acute myeloid leukemia in the near future.

Introduction Acute myeloid leukemia (AML) is a heterogeneous malignancy characterized by impaired differentiation and uncontrolled clonal expansion of myeloid precursors. A varied range of genetic and epigenetic alterations disrupt normal differentiation of myeloid precursors leading to maturation arrest coupled with self-renewal capacity. Despite extensive research and remarkable advances in our understanding of the molecular mechanisms governing AML pathogenesis, chemotherapy-based treatment regimens developed in the 1970s constitute the standard of care.1 All-trans retinoic acid (ATRA) for acute promyelocytic leukemia (APL) is a notable exception, which not only revolutionized the clinical outcome of APL, but also demonstrated that inducing terminal differentiation is an effective mode of treating AML.2 This therapeutic approach of using pharmacological agents to stimulate differentiation of immature cancer cells into more mature forms is termed differentiation therapy. Apart from its overt effect on APL blasts, ATRA also differentiated some AML cell lines lacking PML-RARA and the drug showed some activity in a few clinical trials;36 however, in general, ATRA has been ineffective in non-APL subtypes of AML.7-13 Initial studies with the HL-60 cell line also suggested that many compounds could induce terminal differentiation, including cytokines, vitamin D3 analogs and ligands of PPARÎł.14 However, their therapeutic utility has been limited.15, 16 More recently, novel oncogenic pathways in AML have been identified, facilitathaematologica | 2021; 106(1)

Advances in differentiation therapy of AML

ing the development of targeted therapies, and heralding a new era in AML treatment. New differentiation-inducing agents that target mutant isocitrate dehydrogenase (IDH)1, IDH2 or FLT3 have gained regulatory approval. Other agents directed against distinct targets are being developed to induce differentiation of leukemic blasts. This review updates our previous description of AML differentiation therapy,14,15 with a focus on new developments in the field, especially in the last decade.

Newly-approved targeted therapies for acute myeloid leukemia Agents targeting two distinct oncogenic events, IDH1/2 mutations and activating FLT3 mutations, have been approved for AML therapy since 2017. These agents exert anti-leukemic activity by inducing differentiation of leukemic blasts and are exciting additions to differentiation therapy of AML.

Targeting mutant IDH1/2 enzymatic activity induces differentiation IDH1 and IDH2 are NADP+-dependent enzymes that normally catalyze the oxidative decarboxylation of isocitrate to produce α-ketoglutarate in the tricarboxylic acid cycle.17 Somatic mutations in the active site of IDH1 and IDH2 are observed in about 20% cases of AML and 5% of myelodysplastic syndromes.17 Mutations of IDH1 and IDH2, which largely occur in a mutually exclusive manner, result in accumulation of the oncometabolite D-2hydroxyglutatrate (2-HG).18 2-HG disrupts activity of αketoglutarate-dependent enzymes including members of the ten-eleven-translocation (TET) family of 5-methylcytosine hydroxylases and the jumonji-domain-containing group of histone lysine demethylases. This leads to a block of normal differentiation and promotes oncogenic transformation.19-21 Ivosidenib (AG-120; Tibsovo, Agios) is a first-in-class inhibitor of mutant IDH1 developed from the prototype compound AGI-5198.22 Ivosidenib is effective against AML cells harboring mutant IDH1 by lowering their 2-HG levels and causing cellular differentiation.22 A phase I dose escalation and expansion study of this drug in patients with relapsed/refractory AML harboring IDH1 mutations resulted in about 30% of cases achieving either complete remission (CR) or CR with partial hematologic recovery (CRh), with a median overall survival (OS) of 9 months (NCT02074839) (Table 1).23 Ivosidenib induced myeloid differentiation of AML blasts, and differentiation syndrome (DS) occurred in about 11% of patients23 (see description of ‘Differentiation syndrome’ below). A subset of patients, who achieved clearance of mutant IDH1, showed longer CR (11.1 vs. 6.5 months) and OS (14.5 vs. 10.2 months) compared with those without clearance of the mutant IDH1. Notably, patients who did not respond to ivosidenib had significantly higher rates of mutations of the receptor tyrosine kinase pathway.23 In a recent study, ivosidenib monotherapy was shown to induce durable remissions in patients with newly diagnosed IDH1-mutant AML ineligible for standard chemotherapy.24 Preliminary results from an ongoing phase Ib/II trial (NCT02677922) showed that the combination of ivosidenib with 5-azacytidine was well tolerated with CR and overall response rates exceeding those achieved with 5-azacytidine alone.25 haematologica | 2021; 106(1)

In 2018, the Food and Drug Administration (FDA) approved the use of ivosidenib for patients with IDH1mutant relapsed/refractory AML; and subsequently the FDA approval was granted for newly diagnosed IDH1mutant patients who are ≥75 years old or who have comorbidities precluding the use of intensive induction chemotherapy. Other inhibitors of mutant IDH1, which were effective in preclinical models include AG-881 (Agios Pharmaceuticals), IDH-305 (Novartis Pharmaceuticals), FT2102 (Forma Therapeutics) and BAY1436032 (Bayer); however, only FT2102 is being developed further for hematologic cancers (Table 1).26 The first selective inhibitor of mutant IDH2 was AGI6780, which potently inhibited its aberrant enzyme activity in AML cells and induced differentiation of these cells in vitro.27 Subsequently, a clinically useful inhibitor, enasidenib (AG-221; IDHIFA), was developed by Agios Pharmaceuticals/Celgene Corporation. This drug reduced 2-HG levels and induced differentiation of IDH2-mutant AML blasts ex vivo and in xenograft mouse models, conferring these mice with a survival advantage.28 In a phase I/II study of relapsed/refractory AML with IDH2 mutation (either R140 or R172), the drug was well-tolerated and induced hematologic responses associated with maturation in 40% of patients, half of whom achieved CR (NCT01915498).29 The median OS among patients with relapsed/refractory AML was 9.3 months, while it was 19.7 months in those who attained CR with enasidenib.29 Moreover, older patients who were not candidates for standard cytotoxic therapy had a 30% response rate to enasidenib, including an 18% CR rate.30 The differentiated neutrophils retained the IDH2 mutation, supporting differentiation as the mechanism of action of enasidenib.31,32 While enasidenib potently suppressed 2-HG in most patients, levels of 2-HG were not predictive of clinical response. Lack of clinical response despite suppression of 2-HG levels suggests the need for additional biomarkers of drug response. An absence of response to enasidenib could also occur because of other genetic or epigenetic abnormalities in the leukemic clones. For example, co-occurring mutations of genes in the RAS pathway are significantly associated with the lack of response to enasidenib.32 In 2017, enasidenib received FDA approval for use in adult relapsed/refractory AML patients with IDH2 mutations. Ongoing clinical studies are evaluating the efficacy of combining mutant IDH1/2 inhibitors with either 5-azacytidine or other therapeutic agents (Table 1). Interestingly, an interim analysis of an ongoing study (NCT02677922) showed significantly higher response and CR rates with a combination of enasidenib and 5-azacytidine than with 5-azacytidine monotherapy in patients with newly diagnosed AML, who were not candidates for intensive chemotherapy.33 Encouraging response rates noted in a preliminary analysis of an ongoing trial also support the feasibility of combining either ivosidenib or enasidenib with standard AML induction therapy in frontline treatment of the newly diagnosed patients harboring mutations of IDH1 and IDH2, respectively.34 Despite the clinical benefit of selective inhibitors of mutant IDH1/2, acquired resistance to this targeted therapy poses new challenges in treating AML. For instance, IDH2-mutant patients developed resistance to enasidenib by acquisition of a second mutation at residues located at the interface where enasidenib binds to the IDH2 dimer.35 Resistance to enasidenib can also occur through acquisition 27

V. Madan and H.P. Koeffler

of additional oncogenic mutations in other genes.36 Similarly, secondary resistance to ivosidenib was associated with acquisition of either IDH1-second site mutations or mutations of IDH2 or RTK pathway genes.37

Differentiation syndrome following mutant IDH1/2 therapy DS, originally described as retinoic acid syndrome, is a common side effect observed in APL patients treated with

ATRA or arsenic trioxide. It is a serious complication that develops within 1-3 weeks of initiation of differentiation therapy in APL.38 The pathogenesis of DS involves release of pro-inflammatory cytokines from differentiating blast cells, leading to a systemic inflammatory state. This, combined with increased vascular permeability, leads to a myriad of clinical symptoms that include unexplained fevers, weight gain, hypotension, rash, hypoxia, renal failure, dyspnea with pulmonary infiltrates and pleuropericardial effu-

Table 1. Clinical trials of IDH1 and IDH2 inhibitors.

Target

Inhibitor

Clinical trial identifier

Phase

Recruitment status

Combination

Study population

Mutant IDH1

Ivosidenib (AG-120)

NCT03245424

IDH1-mutated relapsed/refractory AML

NCT04044209

Approved for marketing Phase II

Nivolumab

NCT04493164 NCT03471260

Phase II Phase Ib/II

Temporarily suspended Not yet recruiting Recruiting

NCT04250051 NCT02632708

Phase I Phase I

NCT02677922

Phase Ib/II

NCT02074839

Phase I

Not yet recruiting Active, not recruiting Active, not recruiting Recruiting

IDH1-mutated relapsed/refractory AML and high-risk MDS IDH1-mutated AML and high-risk MDS Advanced hematologic malignancies with an IDH1 mutation IDH1-mutated relapsed/refractory AML IDH1-mutated newly diagnosed AML

NCT03173248 NCT03839771

Phase III Phase III

Recruiting Recruiting

NCT02719574

Phase I/II

Recruiting

NCT04013880

Phase Ib/II

NCT01915498

Phase I/II

NCT04092179 NCT02632708

Phase Ib/II Phase I

NCT02677922

Phase Ib/II

NCT03683433 NCT03825796 NCT02577406

Phase II Phase II Phase III

Withdrawn (loss of funding) Active, not recruiting Not yet recruiting Active, not recruiting Active, not recruiting Recruiting Recruiting Active, not recruiting

NCT03881735 NCT03839771

Phase II Phase III

Recruiting Recruiting

NCT03515512

Phase I

Recruiting

Standard chemotherapy -

NCT03728335

Phase I

Recruiting

NCT02492737

Phase I

Completed

Olutasidenib (FT 2102)

Mutant IDH2

Enasidenib (AG-221)

Vorasidenib (AG-881)

CPX-351 Venetoclax + 5-Aza FLAG chemotherapy Standard chemotherapy 5-Aza 5-Aza Standard chemotherapy 5-Aza or cytarabine ASTX727 Venetoclax Standard chemotherapy 5-Aza 5-Aza CPX-351 -

IDH1-mutated newly diagnosed AML Advanced hematologic malignancies with an IDH1 mutation IDH1-mutated newly diagnosed AML IDH1-mutated newly diagnosed AML IDH1-mutated relapsed/refractory AML and MDS IDH1-mutated relapsed/refractory AML and MDS Advanced hematologic malignancies with an IDH2 mutation IDH2-mutated relapsed/refractory AML IDH2-mutated newly diagnosed AML IDH2-mutated newly diagnosed AML IDH2-mutated relapsed/refractory AML IDH2-mutated relapsed AML Older subjects (â&#x2030;Ľ60 years old) with IDH2-mutated relapsed/refractory AML (compared to conventional therapy) IDH2-mutated relapsed/refractory AML IDH2-mutated newly diagnosed AML Maintenance therapy for IDH2-mutant AML or CMML following allogeneic stem cell transplantation Maintenance therapy for IDH2-mutant AML following allogeneic stem cell transplantation Advanced hematologic malignancies with either IDH1 or IDH2 mutation

AML: acute myeloid leukemia; 5-Aza: 5-azacytidine; ASTX727: decitabine + cedazuridine (E7727; cytidine deaminase inhibitor; Astex Pharmaceuticals); CMML: chronic myelomonocytic leukemia; CPX-351: liposomal cytarabine and daunorubicin; FLAG: fludarabine, cytarabine, plus granulocyte colony-stimulating factor; MDS: myelodysplastic syndrome.

haematologica | 2021; 106(1)

Advances in differentiation therapy of AML

sion.38 Although potentially life-threatening, clinical symptoms of DS are routinely managed using corticosteroids, while discontinuation of differentiation therapy might be required in cases with severe DS.38,39 DS, similar to that observed in APL following ATRA/arsenic trioxide therapy, has also been reported in patients treated with inhibitors of mutant IDH1/2. The most common clinical features of IDH-DS were dyspnea and pulmonary infiltrates or pleuropericardial effusion, while hypotension was the least common symptom.40 IDH-DS was effectively managed by dose interruption and treatment with glucocorticoids and hydroxyurea.23,29,41 In contrast to a rapid occurrence of DS in APL (median of 12 days),38,42,43 the onset of DS was delayed in patients treated with IDH inhibitors.23,40,41 A recent comprehensive analysis concluded that the incidence of DS in IDH inhibitor-treated patients (19%)40 is similar to the rate observed for ATRA-treated APL (â&#x2C6;ź25%).39,42,43 This necessitates routine monitoring for DS in patients on IDH inhibitor therapy. Further mechanistic studies are necessary to establish how inhibition of mutant IDH1 and IDH2 relieves the block in myeloid differentiation. Inhibitors of mutant IDH1/2 potently suppress 2-HG, and this is likely to restore the activity of Îą-ketoglutarate-dependent enzymes that regulate DNA and histone methylation. Therefore, studies need to focus on investigating the function of TET enzymes and jumonji histone lysine demethylases, and whether restoration of their function correlates with clinical outcome in subjects treated with inhibitors of mutant IDH1/2. A better understanding is also required of why the clinical response to IDH inhibitors is variable despite robust inhibition of 2-HG levels. In addition, better predictive biomarkers are needed to enhance the therapeutic utility of IDH inhibitors.

FLT3 inhibitors promote differentiation of acute myeloid leukemia blasts FMS-like tyrosine kinase 3 (FLT3) is one of the most frequently mutated genes in AML (~30%), with two distinct alterations, either an internal tandem duplication (FLT3ITD) in the juxtamembrane domain or point mutations in the tyrosine kinase domain (TKD). These activating mutations lead to ligand-independent, constitutive FLT3 signaling which promotes proliferation of leukemic cells. Activation of FLT3 also suppresses myeloid differentiation by inhibiting the function of CEBPA via ERK1/2-mediated phosphorylation, while pharmacological inhibition of FLT3 causes granulocytic differentiation of AML cells.44 Early studies with FLT3 inhibitors demonstrated clinical manifestation of differentiation with evidence of neutrophilic dermatoses.45-47 Since then, several inhibitors of FLT3 have been investigated in clinical trials and two agents, gilteritinib and midostaurin, have received regulatory approval. In clinical studies, the differentiation-inducing effect of FLT3 inhibition is clearly evident with selective FLT3 inhibitors, gilteritinib and quizartinib. Gilteritinib (Xospata, Astellas Pharma) is a type I FLT3 inhibitor active against both ITD and TKD mutations. In a murine xenograft model, gilteritinib induced regression of tumors expressing mutant FLT3.48 In an initial clinical study, 49% of FLT3-mutant AML patients responded to gilteritinib. Of note, 37% of the patients who had received prior treatment with another FLT3 inhibitor responded to gilteritinib [NCT02014558 (CHRYSALIS study)] (Table 2).49 Significantly, gilteritinib stimulated differentiation towards haematologica | 2021; 106(1)

granulocytic and monocytic lineages in about half of the FLT3-mutant patients, who exhibited significant reduction in marrow blast percentage, even though the mutation burden of FLT3 was unchanged.50 In this subgroup of patients, DS was reported in two patients, in whom it was resolved with glucocorticoid therapy.50 Intriguingly, the authors noted no apparent signs of myeloid differentiation in the other half of patients treated with gilteritinib although reductions in AML blasts and bone marrow cellularity were observed. This alternate pattern of response to gilteritinib was associated with a reduction in allele frequency of mutant FLT3 and indicated that mechanisms other than induction of myeloid differentiation, might contribute to the anti-leukemic effect of gilteritinib. Promising data emerging from an interim analysis of a phase III clinical study (NCT02421939 [ADMIRAL study]) comparing gilteritinib monotherapy to salvage chemotherapy in FLT3-mutant relapsed/refractory AML patients led to the FDA (November 2018) and European Medicines Agency (EMA) (September 2019) approving single-agent gilteritinib therapy in adult patients with FLT3-mutated relapsed/refractory AML. This trial demonstrated significantly longer OS (9.3 months vs. 5.6 months) and higher response rates (CR/CRh rates of 34% vs. 15%) in gilteritinib-treated patients than in those treated with chemotherapy.51 The efficacy of gilteritinib is also being explored in a variety of other clinical settings, including in combination with either 5-azacytidine or other chemotherapeutic agents, as well as for maintenance therapy in patients with FLT3-mutated AML (Table 2). Quizartinib is a type II inhibitor that binds adjacent to the ATP-binding domain when the FLT3 protein is in its inactive conformation and, therefore, lacks activity against TKD-mutant FLT3.52 It is a highly potent and selective inhibitor of FLT3 and showed comparatively greater efficacy as a single agent than chemotherapy in a larger phase III trial (QuANTUM-R study; NCT02039726). In this study of relapsed/refractory AML with FLT3-ITD, quizartinib monotherapy prolonged OS compared with salvage chemotherapy.53 Quizartinib caused terminal myeloid differentiation of leukemic blasts with a surge in the number of FLT3-ITD-retaining neutrophils, and development of clinical DS.54 Superior response rates were obtained in relapsed/refractory AML harboring FLT3-ITD when quizartinib was combined with either 5-azacytidine or low-dose cytarabine.55 A larger, placebo-controlled trial (QuANTUM-First; NCT02668653) combining quizartinib with standard induction and consolidation chemotherapy in newly-diagnosed FLT3-ITD-positive AML patients is ongoing (Table 2). Although gilteritinib and quizartinib act as differentiating agents, differentiation is not universally observed in all patients treated with FLT3 inhibitors, suggesting alternative mechanisms of clearing AML blasts. Midostaurin (Rydapt, Novartis) is a first-generation FLT3 inhibitor that received FDA and EMA approval in 2017 for use in adults with FLT3-mutated newly diagnosed AML when combined with intensive chemotherapy; the EMA also approved midostaurin for single-agent maintenance therapy. As a single-agent, midostaurin causes a reduction in leukemic blasts in FLT3-mutant relapsed/refractory AML patients, without achieving CR.56 Combining midostaurin with chemotherapy significantly improved its efficacy, with a 92% CR rate being observed in FLT3-mutated patients (Table 2).57 Moreover, in a large, phase III, double29

V. Madan and H.P. Koeffler Table 2. An overview of acute myeloid leukemia clinical trials of FLT3 inhibitors.

Inhibitor Gilteritinib

Quizartinib

Clinical trial identifier

Phase

Recruitment status

Combination

Study population

NCT02014558 (CHRYSALIS) NCT03730012 NCT04293562

Phase I/II

Completed

Relapsed/refractory AML

Phase I/II Phase III

Recruiting Recruiting

Atezolizumab CPX-351

NCT04336982

Phase I/II

CC-90009

NCT04240002

Phase I/II

Not yet recruiting Recruiting

FLT3-mutated relapsed/refractory AML Newly diagnosed AML with or without FLT3 mutations (standard chemotherapy compared to therapy with CPX-351 and/or gilteritinib) FLT3-ITD positive relapsed/refractory AML

NCT02927262

Phase II

Active, not recruiting

NCT02421939 (ADMIRAL) NCT03625505 NCT02997202

Phase III Phase Ib Phase III

Active, not recruiting Recruiting Recruiting

Venetoclax -

NCT03182244

Phase III

Recruiting

NCT02752035

Phase II/III

Recruiting

NCT03836209 NCT02310321

Phase II Phase I/II

NCT02236013 NCT04027309

Phase I Phase III

NCT04140487

Phase I/II

NCT03135054

Phase II

Recruiting Active, not recruiting Recruiting Not yet recruiting Not yet recruiting Recruiting

Single agent or combined with 5-Aza Standard chemotherapy Standard chemotherapy

NCT02039726 (QuANTUM-R) NCT04107727

Phase III

NCT03989713

Phase II

NCT03793478

Phase I/II

NCT03552029

Phase I

NCT04112589 NCT02668653 (QuANTUM-First) NCT03735875 NCT04128748

Phase I/II Phase III Phase Ib/II Phase I/II

NCT04209725 NCT03661307

Phase II Phase I/II

NCT04047641

Phase II

Recruiting

NCT01892371

Phase I/II

Active, not recruiting

Phase II

Active, not recruiting Recruiting

FLAG chemotherapy -

Standard chemotherapy Standard chemotherapy 5-Aza and venetoclax Omacetaxine mepesuccinate -

Sandard chemotherapy with or without quizartinib Not high-dose cytarabine, yet recruiting mitoxantrone Recruiting Re-Induction Chemotherapy Recruiting Milademetan

Recruiting FLAG-IDA chemotherapy Active, Standard chemotherapy not recruiting Recruiting Venetoclax Not CPX-351 yet recruiting Recruiting CPX-351 Recruiting Decitabine and venetoclax Cladribine, idarubicin and cytarabine 5-Aza or cytarabine

Children, adolescents and young adults with FLT3-mutated relapsed/refractory AML Maintenance therapy following induction/ consolidation therapy for FLT3-ITD positive AML in CR FLT3-mutated relapsed/refractory AML (gilteritinib compared to salvage chemotherapy) Relapsed/refractory AML Maintenance therapy following allogeneic transplant for FLT3-ITD positive AML in CR FLT3-mutated relapsed/refractory AML (gilteritinib compared to salvage chemotherapy) FLT3-mutated newly diagnosed AML not eligible for intensive induction chemotherapy FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML Newly diagnosed AML FLT3-mutated newly diagnosed AML (also for one-year maintenance therapy) FLT3-mutated relapsed/refractory AML and high risk MDS Newly diagnosed or relapsed/refractory AML carrying FLT3-ITD Monotherapy vs salvage chemotherapy in FLT3-ITD positive relapsed/refractory AML Newly diagnosed AML wild-type for FLT3 Relapsed/refractory AML with FLT3-ITD Pediatric relapsed/refractory AML with FLT3-ITD (also for maintenance therapy) FLT3-ITD positive relapsed/refractory AML FLT3-ITD positive newly diagnosed AML patients unfit for intensive chemotherapy Relapsed/refractory AML Newly diagnosed AML with FLT3-ITD (also for continuation therapy) FLT3-mutated relapsed/refractory AML AML and high risk MDS FLT3-ITD positive relapsed/refractory AML Newly diagnosed or relapsed AML and high risk MDS Newly diagnosed or relapsed/refractory AML and high risk MDS Relapsed/refractory AML and MDS continued on the next page

haematologica | 2021; 106(1)

Advances in differentiation therapy of AML continued from the previous page

Inhibitor

Midostaurin

Crenolanib

Clinical trial identifier

Phase

Recruitment status

Combination

Study population

NCT02984995 NCT01565668 NCT00651261 (RATIFY) NCT04385290

Phase II Phase II Phase III Phase I/II Phase I

Standard chemotherapy Gemtuzumab Ozogamicin HDM201

FLT3-ITD positive relapsed/refractory AML FLT3-ITD positive relapsed/refractory AML FLT3-mutated newly diagnosed AML (< 60 years old) FLT3-mutated newly diagnosed AML

NCT04496999 NCT03512197

Phase III

Completed Completed Active, not recruiting Not yet recruiting Not yet recruiting Recruiting

NCT04097470

Phase II

NCT03280030

Phase II

NCT03591510

Phase II

Not yet recruiting Active, not recruiting Suspended

NCT03258931 NCT03900949

Phase III Phase I

Recruiting Recruiting

NCT03379727

Phase IIIb

NCT03836209 NCT01477606

Phase II Phase II

Active, not recruiting Recruiting Completed

NCT03951961 (MAURITIUS) NCT04027309

Phase II

Standard chemotherapy Decitabine Standard chemotherapy Standard chemotherapy Standard chemotherapy Gemtuzumab ozogamicin plus chemotherapy standard chemotherapy Standard chemotherapy Standard chemotherapy

Phase III

Not yet recruiting Recruiting

Standard chemotherapy

NCT03686345

Phase II

Recruiting

Standard chemotherapy

NCT01202877 NCT03258931 NCT02283177 NCT03250338

Phase II Phase III Phase II Phase III

Completed Recruiting Completed Recruiting

NCT02400255

Phase II

Recruiting

5-Aza Standard chemotherapy Standard chemotherapy Standard chemotherapy with or without crenolanib -

NCT01657682 NCT02400281

Phase II Phase I/II

Completed Completed

Standard salvage chemotherapy or 5-Aza

Relapsed/refractory AML with FLT3 mutation and wildtype TP53 Newly diagnosed AML negative for FLT3 mutation AML patients ineligible for intensive chemotherapy FLT3-mutated newly diagnosed AML Pediatric patients with FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML Newly diagnosed AML carrying FLT3-ITD (also for maintenance therapy) MRD-positive FLT3-mutant AML after allogeneic stem cell transplantation FLT3-mutated newly diagnosed AML (also for maintenance therapy) CBF-AML (also for 1-year maintenance therapy) Relapsed/refractory AML and MDS FLT3-mutated newly diagnosed AML FLT3-mutated newly diagnosed AML FLT3-mutated relapsed/refractory AML subjects â&#x2030;¤ 75 years of age Maintenance therapy following allogeneic transplant for FLT3-mutant AML FLT3-mutated relapsed/refractory AML FLT3-mutated relapsed/refractory AML

AML: acute myeloid leukemia; MDS: myelodysplastic syndrome; 5-Aza: 5-azacytidine; CBF: core binding factor; CPX-351: liposomal cytarabine and daunorubicin; FLAG-IDA: fludarabine, cytarabine, idarubicin and glycosylated granulocyte colony-stimulating factor (G-CSF); MRD: minimal residual disease, FLAG: fludarabine, cytarabine plus G-CSF.

blind, placebo-controlled trial in newly diagnosed FLT3mutated AML, addition of midostaurin to chemotherapy resulted in significant improvements in OS and event-free survival.58 However, none of these studies demonstrated that midostaurin exerted its anti-leukemic activity via differentiation of AML blasts. Midostaurin is a broad-range multikinase inhibitor, which targets KIT, PDGFRÎą/b, VEGFR2, and members of the serine/threonine kinase PKC family, apart from inhibiting FLT3. Thus, it is possible that its clinical efficacy results from blocking additional pro-survival and proliferation signals. Indeed, midostaurin also caused clinical responses in FLT3-negative AML and further trials are carefully investigating its clinical activity in FLT3 wild-type AML (Table 2). Crenolanib is another selechaematologica | 2021; 106(1)

tive FLT3 inhibitor, which produced high response rates in newly-diagnosed and relapsed FLT3-mutant AML patients, as well as those refractory to other tyrosine kinase inhibitors.52,59 It is being investigated in several late-stage clinical studies, primarily in combination with chemotherapy (Table 2). Several preclinical studies have demonstrated that FLT3 inhibitors broadly induce apoptosis of AML cells.60,61 Hence, FLT3 inhibitors possibly function through multiple pathways, among which induction of apoptotic cell death and terminal differentiation of leukemic blasts are prominent. Secondary resistance frequently develops in AML patients treated with FLT3 inhibitors, and complex patterns of clonal selection and evolution have been report31

V. Madan and H.P. Koeffler

ed.52,59,62-64 Acquired resistance to type II FLT3 inhibitors often involves gain of secondary mutations either in the TKD or the gatekeeper residues of the FLT3 gene,65 and type I inhibitors, such as gilteritinib and crenolanib, which have activity against both ITD and TKD mutations are effective in these patients. Overall, rational combination therapies with agents targeting known resistance pathways are needed to eradicate leukemic clones.

Other differentiation-inducing agents evaluated in clinical trials The approval of new targeted drugs bodes well for AML differentiation therapy, but improving therapeutic options for all subgroups remains a major challenge. Epigenetic enzymes are attractive targets for AML therapy and inhibitors of histone deacetylases (HDAC), KDM1A, DOT1L and BET proteins have been extensively studied in AML. We discuss below several of these drugs, which stimulated differentiation of leukemic blasts in preclinical models, but their single-agent activity in clinical trials has been disappointing.

Inhibition of KDM1A (LSD1) promotes myeloid differentiation KDM1A is a demethylase enzyme with specificity for monomethyl- and dimethyl-histone H3 lysine-4 (H3K4) and lysine-9 (H3K9). It is highly expressed in AML cells and is required to maintain the oncogenic program downstream of MLL-AF9.66 Pharmacological inhibition of this enzyme potentiates ATRA-induced differentiation of AML cells and diminishes their engraftment in immune-deficient mice.67 Combined treatment with ATRA and tranylcypromine (an inhibitor of KDM1A) caused a gene-specific increase in H3K4me2 levels and upregulation of myeloiddifferentiation associated gene expression.67 MLL-AF9expressing cells treated with KDM1A inhibitors also exhibit gain in chromatin accessibility at sites bound by PU.1 and CEBPA; and depletion of either of these transcription factors confers resistance of MLL-AF9-expressing cells to KDM1A inhibition.68 Preclinical results have also demonstrated sensitivity of other AML subtypes (e.g., AML with RUNX1-RUNX1T1) to KDM1A inhibitors.69 The efficacy of tranylcypromine combined with ATRA is being evaluated in clinical trials (Table 3). ORY-1001 (Iadademstat), a more potent and selective derivative of tranylcypromine, effectively reduced growth of AML cell lines, induced expression of differentiation markers and increased survival of murine xenograft models.70 The drug is synergistic with ATRA as well as with other epigenetic inhibitors targeting DOT1L, DNMT1 or HDAC.70 Because of favorable pharmacological characteristics and tolerability in a phase I/IIa study in relapsed/refractory AML,71 ORY-1001 is now being investigated in combination with 5-azacytidine in elderly AML individuals who are not eligible for intensive chemotherapy (Table 3). Other KDM1A inhibitors are currently being investigated as either monotherapy or in combination with ATRA in early phase clinical trials (Table 3).

DOT1L inhibition in MLL-rearranged acute myeloid leukemia Rearrangements of the MLL gene occur in 5-10% of AML cases and are associated with an adverse prognosis. 32

DOT1L, a H3K79 histone methyltransferase, is recruited by MLL-fusion proteins resulting in increased H3K79 methylation at target gene loci.72,73 This mis-targeting of DOT1L activity is essential for the transforming ability of MLL fusion proteins, and enhances expression of HOXA9 and MEIS1.74-76 Depletion of DOT1L in MLL-AF9 leukemic cells induces differentiation and apoptosis.75 EPZ004777 was the first selective inhibitor of DOT1L, which inhibited H3K79 methylation and reversed the gene expression signature of MLL fusion leukemias, including downregulation of HOXA cluster genes.74,75,77 This drug stimulated myeloid differentiation in murine and human cells harboring MLL fusions.76-78 Pinometostat (EPZ-5676) is closely related to EPZ004777 but is more potent and has better pharmacokinetic properties.79 However, in a phase I trial of pinometostat in advanced stage MLL-rearranged leukemia, only two of 51 patients achieved a CR, while seven non-responding patients also exhibited morphological changes consistent with myeloid differentiation (NCT01684150).80 Despite apparent biological activity (inhibition of H3K79 methylation) in patients, pinometostat as a stand-alone therapy was not sufficient to produce a clinical benefit. This lack of therapeutic activity may be attributed to heterogeneity of MLL fusion proteins, which might result in differential sensitivity to DOT1L inhibitors. Given their insufficient clinical efficacy as single agents, combinatorial strategies incorporating DOT1L inhibitors are being tested. Preclinical studies have shown synergistic anti-leukemic effects of combining DOT1L inhibition with either cytotoxic chemotherapy, hypomethylating agents, or inhibition of KDM1A or menin (a MLL cofactor),81-84 suggesting that dual targeting of the MLL-fusion complex may prove superior to DOT1L inhibitor monotherapy. Phase Ib/II trials of pinometostat plus either standard chemotherapy or 5-azacytidine are ongoing in AML with MLL gene rearrangements (Table 3).

Myeloid differentiation induced by inhibition of BET proteins Bromodomain and extra terminal (BET) family of proteins bind to acetyl-modified lysine residues of histone tails and activate transcription. A screen of RNA interference in a mouse AML model expressing the MLL-AF9 fusion and oncogenic NRAS identified BRD4, a BET family member, as indispensable for disease progression.85 Genetic and pharmacological inactivation (JQ1 and I-BET151) of BRD4 led to downregulation of MYC, BCL2 and CDK6 in AML blasts, inhibited growth of MLL-rearranged AML cell lines and induced differentiation of primary leukemic samples and murine hematopoietic cells transformed with MLL fusions.8587 Another BET bromodomain inhibitor, AZD5153, which ligates two bromodomains of BRD4 simultaneously, has prominent anti-proliferative activity against AML cell lines in vitro and in xenograft models.88 In other preclinical studies, BRD4 inhibitors exhibited anti-leukemic effects in AML driven by mutations of either IDH2, NPM1 or FLT3-ITD.89-91 Several BET inhibitors are currently under clinical investigation (Table 3); however, emerging results from these early phase trials are disappointing. In a phase I trial of OTX015 (MK-8628), this BET inhibitor elicited a clinical response in only five of 36 AML patients, with the response lasting 2â&#x20AC;&#x201C;5 months.92 The clinical response to mivebresib (ABBV-075) was similarly inadequate and frequently associated with adverse effects including cytopenia.93 A number of trials of haematologica | 2021; 106(1)

Advances in differentiation therapy of AML

BET inhibitors have been terminated because of a combination of issues including low response rates, dose-limiting toxicities and inter-patient variability in pharmacokinetic properties (Table 3).

Improved drug design and combination strategies might establish the therapeutic utility of BET inhibitors. For instance, hetero-bifunctional compounds, BET-PROTAC (proteolysis-targeting chimera, ARV-825 and ARV-771), in

Table 3. Clinical trials on drugs targeting epigenetic regulators.

Target

Inhibitor

Clinical trial identifier

Phase

Recruitment status

Combination

Study population

KDM1A

ORY-1001 (Iadademstat)

EudraCT 2013-002447-29 EudraCT 2018-000482-36 NCT02717884

Phase I

Completed

Relapsed/refractory AML

Phase I

Ongoing

5-Aza

Phase I/II

NCT02273102 NCT02261779

Phase I Phase I/II

IMG-7289 INCB059872

NCT02842827 NCT02712905

Phase I Phase I/II

GSK2879552

NCT02177812

Phase I

Pinometostat (EPZ-5676)

NCT01684150

Phase I

Completed

NCT02141828

Phase I

Completed

NCT03724084

Phase Ib/II

Recruiting

NCT03701295

Phase Ib/II

Recruiting

NCT02303782

Phase Ib/II

Withdrawn

5-Aza

NCT01713582 NCT02698189 NCT01943851

Phase I Phase Ib Phase I/II

Completed Active, not recruiting Completed

Newly diagnosed AML not eligible for intensive treatment De novo and secondary AML De novo and secondary AML Relapsed/refractory AML

NCT02308761

Phase I

Completed

Relapsed/refractory AML

NCT02543879

Phase I/Ib

Completed

5-Aza

INCB057643

NCT02711137

Phase I/II

INCB054329

NCT02431260

Phase I/II

Mivebresib (ABBV-075) PLX51107

NCT02391480

Phase I

Terminated (safety issues) Terminated (PK variability) Completed

Relapsed/refractory hematologic malignancies Advanced solid tumors and hematologic malignancies Advanced malignancies including AML

Relapsed/refractory AML

NCT02683395

Phase 1b/IIa

Advanced hematologic malignancies

NCT04022785

Phase I

Terminated (business decision) Recruiting

5-Aza

AML and MDS

Phase II Phase II

Completed Recruiting Recruiting

DNMTi 5-Aza 5-Aza

AML and MDS AML and MDS with TET2 mutations MDS, CMML and AML

Tranylcypromine

DOT1L

BET proteins

OTX015 (MK-8628)

Molibresib (I-BET 762, GSK525762) RO6870810 (TEN-010) FT-1101

Aberrant DNA demethylation

Vitamin C

NCT02877277 NCT03397173 NCT03999723

Older AML patients ineligible for standard treatment Recruiting ATRA and AraC Non-APL AML or MDS who failed azanucleoside treatment Completed ATRA Adult AML and MDS Unknown ATRA Relapsed/refractory AML or patients not eligible for intensive treatment Completed ATRA AML and MDS Terminated 5-Aza Newly diagnosed AML (business decision) ATRA Relapsed/refractory AML Terminated (unfavorable ATRA Relapsed/refractory AML risk-benefit ratio) Relapsed/refractory AML with MLL-rearrangements or PTD Pediatric relapsed/refractory AML with MLL-rearrangements Standard Newly diagnosed chemotherapy AML with MLL-rearrangements 5-Aza Relapsed, refractory or newly diagnosed AML with MLLrearrangements

AML: acute myeloid leukemia; MDS: myelodysplastic syndrome; APL: acute promyelocytic leukemia; ATRA: all-trans retinoic acid; 5-Aza: 5-azacytidine; AraC: cytarabine; CMML: chronic myelomonocytic leukemia; DNMTi: DNA methyltransferase inhibitor; PTD: partial tandem duplication.

haematologica | 2021; 106(1)

V. Madan and H.P. Koeffler

which a BET inhibitor (OTX015/JQ1) is linked to a ligand for an E3 ubiquitin ligase (either cereblon or von Hippelâ&#x20AC;&#x201C; Lindau),94,95 have exhibited high potency in attenuating MYC and inducing apoptosis of AML cells.96 Moreover, concomitant inhibition of both BRD4 and DOT1L had significant synergistic activity against MLL-rearranged leukemia cell lines, primary human AML cells and murine leukemia models.97 Clearly, additional studies are needed to establish the clinical efficacy of drug combinations involving BET inhibitors.

cells.107,108 Alisertib, an Aurora kinase A inhibitor, is currently under clinical evaluation for use in the treatment of AMKL (NCT02530619). Several trials of either specific or dual inhibitors of Aurora kinase A and B showed moderate efficacy for treating other types of AML. In a phase I study, a combination of alisertib and conventional induction therapy (cytarabine and idarubicin) resulted in an 86% remission rate in newly diagnosed AML patients (NCT01779843),109 suggesting enhanced activity of this drug combination. Thus, Aurora kinase inhibitors may hold promise in AML therapy.

Histone deacetylase inhibitors Recruitment of HDAC is a major pathogenic event in AML subtypes driven by PML-RARA, inv(16) (CBFAMYH11) and RUNX1-RUNX1T1 chimeric proteins, indicating that inhibiting the activity of HDAC might be therapeutic. In preclinical studies, a combination of a HDAC inhibitor (trichostatin A) with ATRA relieved the differentiation block in ATRA-resistant APL cells. Panobinostat, another HDAC inhibitor, caused proteosomal degradation of the RUNX1-RUNX1T1 oncoprotein, triggering myeloid differentiation of leukemic cells in a murine model of human t(8;21) AML.98 Although several classes of HDAC inhibitors showed promise in preclinical models, their clinical efficacy as monotherapy was negligible in early phase clinical trials.99102 Combinations of HDAC inhibitors with hypomethylating agents, chemotherapeutic agents or immunotherapy and targeted therapies are currently in different stages of clinical testing. Despite evidence of synergism between HDAC inhibitors and hypomethylating agents in preclinical and early clinical studies, large multi-arm clinical studies showed increased toxicity and decreased OS in patients treated with a HDAC inhibitor plus 5-azacytidine compared with 5-azacytidine alone.103-105 Collectively, the data suggest that HDAC inhibitors have limited clinical benefit, illustrating the potential problem of off-target effects arising from a lack of specificity of this approach.

Differentiation therapies in early clinical development Differentiation of acute myeloid leukemia blasts using Aurora kinase inhibitors Aurora kinases are serine/threonine kinases that regulate chromosomal alignment and segregation during mitosis and meiosis. Expression of Aurora kinase A and B is higher in leukemic blasts than in normal CD34+ cells.106 Inhibition of Aurora kinase A in acute megakaryocytic leukemia (AMKL) increases polyploidization of blasts and induces their differentiation.107 Furthermore, these inhibitors significantly prolonged survival of mice in an AMKL model, as well as mice engrafted with primary human AMKL

Restoration of TET activity with vitamin C promotes differentiation TET proteins are dioxygenase enzymes that catalyze oxidation of 5-methylcytosine to 5-hydroxymethylcytosine, leading to DNA demethylation. The TET2 gene is frequently mutated in AML and other hematological malignancies, as well as in clonal hematopoiesis of indeterminate potential.110 Stimulation of the remaining wild-type allele of TET2 is a plausible therapeutic approach. In a murine model of reversible silencing of TET2, reexpression of Tet2 in hematopoietic cells primed the cells towards myeloid differentiation and promoted cell death.111 Vitamin C is a co-factor that activates TET proteins; and its depletion impairs the function of TET2 and accelerates tumorigenesis in a murine model of AML.112 Importantly, vitamin C treatment mimics pharmacological restoration of TET2 activity as evidenced by increased 5-hydroxymethylcytosine levels in AML cells and Tet2-deficient mice.111 In AML with biallelic TET2 mutations or copy number neutral loss of heterozygosity of the TET2 locus, vitamin C might still enhance 5-hydroxymethylcytosine levels by potentiating TET3 activity, which can potentially compensate sufficiently for lack of TET2.111 Serum ascorbic acid (vitamin C) levels are significantly reduced in patients with diverse hematologic malignancies,113,114 and high doses of oral vitamin C might improve their outcome. The clinical efficacy of vitamin C combined with 5-azacytidine in patients with AML or myelodysplastic syndrome with or without TET2 mutations is currently being investigated (Table 3). Conceivably, dietary supplements of vitamin C may also help to augment TET2 activity and impede clonal hematopoietic progression in individuals with clonal hematopoiesis of indeterminate potential, especially those harboring mutant TET2.

Inhibition of DHODH relieves the block in myeloid differentiation in acute myeloid leukemia HOXA9 is suppressed during myeloid differentiation, but its upregulation occurs in a large proportion of AML driven by different oncogenic events. In a murine model of inducible Hoxa9 expression, inhibition of dihydroorotate

Table 4. Clinical trials of DHODH inhibitors in acute myeloid leukemia.

Inhibitor

Clinical trial identifier

Phase

Recruitment status

Combination

Study population

ASLAN003

NCT03451084

Phase IIa

Recruiting

BAY2402234 PTC299 Brequinar

NCT03404726 NCT03761069 NCT03760666

Phase I Phase Ib Phase Ib/IIa

Recruiting Recruiting Recruiting

5-Aza (â&#x2030;Ľ60-years old) -

AML patients ineligible for standard treatment and relapsed/refractory AML AML and MDS Relapsed/refractory AML Relapsed/refractory AML

AML: acute myeloid leukemia; MDS: myelodysplastic syndrome; 5-Aza: 5-azacytidine; DHODH: dihydroorotate dehydrogenase.

haematologica | 2021; 106(1)

Advances in differentiation therapy of AML

dehydrogenase (DHODH), a key enzyme in uridine synthesis that catalyzes conversion of dihydroorotate to orotate, induced terminal differentiation of leukemic cells.115 In an independent CRISPR/Cas9 screen of AML cell lines, loss of DHODH also decreased growth of AML cells.116 This suggests that targeting the de novo pyrimidine synthesis pathway may induce AML differentiation. DHODH is neither mutated nor overexpressed in cancers; but cancer cells that are metabolically reprogrammed may have an increased dependency on de novo pyrimidine production,117 thus creating a potential therapeutic window. Earlier attempts to target DHODH with inhibitors (e.g., teriflunomide, brequinar, and leflunomide) in several cancers were unsuccessful because of either low potency or off-target effects.118 Subsequently, more potent and selective DHODH inhibitors (ASLAN003, BAY2402234 and PTC299) (Table 4) have been developed, which show antileukemic activity in various preclinical models;117,119-121 and are in early phase clinical trials. The mechanism of action of DHODH inhibitors is unclear; but these drugs appear to have potential therapeutic value for the treatment of AML. However, by analogy with ATRA and the differentiating agents discussed above, it is unlikely that monotherapy will achieve durable responses.

pression (by PD-0332991) induced myeloid differentiation in human and murine AML cells expressing the MLL fusion.124 Another cell cycle regulator, CDK1 phosphorylates CEBPA, thereby inhibiting its differentiation-inducing activity. In preclinical studies, pharmacological and genetic inhibition of CDK1 relieved the differentiation block in AML cells harboring mutations of FLT3.125 Inhibition of CDC25 using IRC-083864 can also overcome the differentiation block in FLT3-ITD AML cells..126 Overall, cell cycle regulators may be potential therapeutic targets in AML.

Combination therapies using differentiation-inducing agents

Recent studies have brought to light previously uncharacterized modulators of myeloid differentiation which can be targeted to overcome the differentiation block in AML. Pharmacological agents targeting these candidate modulators need to be developed to explore their clinical utility.

Combinatorial therapeutic approaches may be needed to increase efficacy and reduce development of resistance in AML. In APL, ATRA is effective in achieving remission by inducing terminal differentiation, but does not eradicate the leukemic clone; however, high cure rates are attained when ATRA is combined with arsenic or anthracycline. It is postulated that ATRA differentiation therapy in APL is unable to eliminate the leukemic stem cell, and this may also be relevant to differentiation therapy of AML more widely. In fact, a recent study revealed that mature APL cells become leukemogenic upon termination of ATRA treatment, suggesting that withdrawal of differentiation therapy can lead to reacquisition of clonogenic and leukemogenic properties through de-differentiation.127 Hence, combination regimens involving newly approved differentiation therapies (mutant IDH1/2 and FLT3 inhibitors) should be optimized for efficient clearing of the leukemic clones. Similarly, rational combination strategies might enhance the clinical utility of epigenetic inhibitors that have been largely ineffective as single agents.

The CAF1 complex as a novel target for differentiation therapy of acute myeloid leukemia

Concluding remarks

Potential new targets for acute myeloid leukemia differentiation therapy

CHAF1B is a subunit of the chromatin assembly factor 1 complex (CAF1) that facilitates assembly of H3-H4 tetramers at replication forks during the S phase. High CHAF1B levels are observed in leukemia and are associated with a poor prognosis.122 Overexpression of CHAF1B enhances the leukemic potential of the MLL-AF9 oncoprotein, and its inhibition promotes differentiation of leukemic cells.122 CHAF1B protein occupies discrete genomic loci and interferes with the CEBPA-mediated differentiation of leukemic cells;122 and is a potential therapeutic target.

Inhibiting transcriptional repressors to re-activate the myeloid differentiation program A common feature of AML driven by different oncogenic events is increased recruitment of transcriptional co-repressors, which affects transcription of differentiation-promoting genes. Two co-repressors (SMARCA5 and CHD4) are enriched in the PU.1/RUNX1/CEBPA master transcription factor hub specifically in AML cells.123 An inhibitor of SMARCA5/CHD4 suppresses growth of AML cell lines via their terminal differentiation while not affecting growth of normal hematopoietic progenitors.123

Cell cycle regulators as therapeutic targets in acute myeloid leukemia CDK6 is required for the growth and maintenance of an immature state of MLL-rearranged AML cells, and its suphaematologica | 2021; 106(1)

Since 2017, four drugs (inhibitors of mutant IDH1/2 and FLT3) that induce differentiation of AML cells have been approved by the FDA. These novel therapeutic agents mark significant strides forward in differentiation therapy of AML. However, for other differentiation-inducing agents (e.g., inhibitors of HDAC, BET proteins, DOT1L), which showed promise in preclinical studies, early reports from clinical trials have not been encouraging. This highlights the limitations of preclinical efficacy models usually performed in immunocompromised mice lacking the immune component, as well as species-specific differences in drug pharmacokinetics and metabolism. Animal models are also limited in their ability to adequately replicate the molecular heterogeneity of AML. In addition, there is a pressing need to identify molecular subgroups which might be responsive to the new agents, as illustrated by the success of the RATIFY trial in assessing the benefit of midostaurin in FLT3-mutant AML. Another concern is that the new drugs may have been tested on limited cohorts of patients. For example, DOT1L inhibitors have been clinically evaluated in AML patients with rearranged MLL, but their differentiation-promoting activity has also been demonstrated in DNMT3Amutant AML cells in preclinical studies,128,129 suggesting that investigation of DOT1L inhibitors could be expanded to other cohorts of patients. In addition, initial trials usually involve patients with relapsed/refractory disease which can 35

V. Madan and H.P. Koeffler

be biologically and clinically distinct from newly diagnosed AML. Overall, this illustrates that numerous difficulties need to be overcome for translation of preclinical candidates into clinically useful differentiation therapy. APL is arguably the least molecularly complex type of AML, which may be relevant to the early success of differentiation therapy in APL. In summary, new insights into disease pathogenesis acquired from mutational and gene expression analyses have directed AML differentiation therapy to several novel targets. For many targets, pre-clinical potential has not translated into clinical success; nonetheless, additional promising targets (DHODH, Aurora kinases) have emerged and are being actively explored. Overall, we are in an exciting era of new therapeutic targets and drugs not imagined 10 years ago. Disclosures No Conflicts of interest to disclose

References 1. Dombret H, Gardin C. An update of current treatments for adult acute myeloid leukemia. Blood. 2016;127(1):53-61. 2. Cicconi L, Lo-Coco F. Current management of newly diagnosed acute promyelocytic leukemia. Ann Oncol. 2016;27(8):14741481. 3. Breitman TR, Selonick SE, Collins SJ. Induction of differentiation of the human promyelocytic leukemia cell line (HL-60) by retinoic acid. Proc Natl Acad Sci U S A. 1980;77(5):2936-2940. 4. Drach J, Lopez-Berestein G, McQueen T, Andreeff M, Mehta K. Induction of differentiation in myeloid leukemia cell lines and acute promyelocytic leukemia cells by liposomal all-trans-retinoic acid. Cancer Res. 1993;53(9):2100-2104. 5. Schlenk RF, Frohling S, Hartmann F, et al. Phase III study of all-trans retinoic acid in previously untreated patients 61 years or older with acute myeloid leukemia. Leukemia. 2004;18(11):1798-1803. 6. Schlenk RF, Döhner K, Krauter Jr, et al. Alltrans retinoic acid improves outcome in younger adult patients with nucleophosmin-1 mutated acute myeloid leukemia – results of the AMLSG 07-04 randomized treatment trial. Blood. 2011;118(21):80. 7. Coombs CC, Tallman MS, Levine RL. Molecular therapy for acute myeloid leukaemia. Nat Rev Clin Oncol. 2016;13(5): 305-318. 8. Petrie K, Zelent A, Waxman S. Differentiation therapy of acute myeloid leukemia: past, present and future. Curr Opin Hematol. 2009;16(2):84-91. 9. Burnett AK, Hills RK, Green C, et al. The impact on outcome of the addition of alltrans retinoic acid to intensive chemotherapy in younger patients with nonacute promyelocytic acute myeloid leukemia: overall results and results in genotypic subgroups defined by mutations in NPM1, FLT3, and CEBPA. Blood. 2010;115(5):948956. 10. Estey EH, Thall PF, Pierce S, et al. Randomized phase II study of fludarabine + cytosine arabinoside + idarubicin +/- alltrans retinoic acid +/- granulocyte colonystimulating factor in poor prognosis newly

Contributions VM performed literature search and wrote the review, HPK conceived and wrote the manuscript. Acknowledgments We thank Dr Wendy A Stewart, London, England, UK, for valuable discussions and critical reading of the manuscript. This work was funded by the Leukemia and Lymphoma Society, the Singapore Ministry of Health’s National Medical Research Council (NMRC) under its Singapore Translational Research (STaR) Investigator Award to HPK (NMRC/STaR/0021/2014), the NMRC Centre Grant awarded to the National University Cancer Institute of Singapore (NMRC/CG/012/2013) and the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centres of Excellence initiatives. We also thank the Melamed Family for their generous support (2019 Melamed Discovery Fund in Hematologic Malignancies and Cell-Based Therapy).

diagnosed acute myeloid leukemia and myelodysplastic syndrome. Blood. 1999;93 (8):2478-2484. 11. Belhabri A, Thomas X, Wattel E, et al. All trans retinoic acid in combination with intermediate-dose cytarabine and idarubicin in patients with relapsed or refractory non promyelocytic acute myeloid leukemia: a phase II randomized trial. Hematol J. 2002;3(1):49-55. 12. Milligan DW, Wheatley K, Littlewood T, Craig JI, Burnett AK, Group NHOCS. Fludarabine and cytosine are less effective than standard ADE chemotherapy in highrisk acute myeloid leukemia, and addition of G-CSF and ATRA are not beneficial: results of the MRC AML-HR randomized trial. Blood. 2006;107(12):4614-4622. 13. de The H. Differentiation therapy revisited. Nat Rev Cancer. 2018;18(2):117-127. 14. Koeffler HP. Induction of differentiation of human acute myelogenous leukemia cells: therapeutic implications. Blood. 1983;62(4): 709-721. 15. Nowak D, Stewart D, Koeffler HP. Differentiation therapy of leukemia: 3 decades of development. Blood. 2009;113 (16):3655-3665. 16. Sachs L. The control of hematopoiesis and leukemia: from basic biology to the clinic. Proc Natl Acad Sci U S A. 1996;93(10):47424749. 17. Medeiros BC, Fathi AT, DiNardo CD, Pollyea DA, Chan SM, Swords R. Isocitrate dehydrogenase mutations in myeloid malignancies. Leukemia. 2017;31(2):272-281. 18. Ward PS, Patel J, Wise DR, et al. The common feature of leukemia-associated IDH1 and IDH2 mutations is a neomorphic enzyme activity converting alpha-ketoglutarate to 2-hydroxyglutarate. Cancer Cell. 2010;17(3):225-234. 19. Figueroa ME, Abdel-Wahab O, Lu C, et al. Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell. 2010;18(6):553567. 20. Lu C, Ward PS, Kapoor GS, et al. IDH mutation impairs histone demethylation and results in a block to cell differentiation. Nature. 2012;483(7390):474-478. 21. Xu W, Yang H, Liu Y, et al. Oncometabolite 2-hydroxyglutarate is a competitive

inhibitor of alpha-ketoglutarate-dependent dioxygenases. Cancer Cell. 2011;19(1):1730. 22. Popovici-Muller J, Lemieux RM, Artin E, et al. Discovery of AG-120 (ivosidenib): a firstin-class mutant IDH1 inhibitor for the treatment of IDH1 mutant cancers. ACS Med Chem Lett. 2018;9(4):300-305. 23. DiNardo CD, Stein EM, de Botton S, et al. Durable remissions with ivosidenib in IDH1-mutated relapsed or refractory AML. N Engl J Med. 2018;378(25):2386-2398. 24. Roboz GJ, DiNardo CD, Stein EM, et al. Ivosidenib induces deep durable remissions in patients with newly diagnosed IDH1mutant acute myeloid leukemia. Blood. 2020;135(7):463-471. 25. Dinardo CD, Stein AS, Stein EM, et al. Mutant IDH1 inhibitor ivosidenib (IVO; AG-120) in combination with azacitidine (AZA) for newly diagnosed acute myeloid leukemia (ND AML). J Clin Oncol. 2019;37 (15_suppl):7011. 26. Montalban-Bravo G, DiNardo CD. The role of IDH mutations in acute myeloid leukemia. Future Oncol. 2018;14(10):979993. 27. Wang F, Travins J, DeLaBarre B, et al. Targeted inhibition of mutant IDH2 in leukemia cells induces cellular differentiation. Science. 2013;340(6132):622-626. 28. Yen K, Travins J, Wang F, et al. AG-221, a first-in-class therapy targeting acute myeloid leukemia harboring oncogenic IDH2 mutations. Cancer Discov. 2017;7(5): 478-493. 29. Stein EM, DiNardo CD, Pollyea DA, et al. Enasidenib in mutant IDH2 relapsed or refractory acute myeloid leukemia. Blood. 2017;130(6):722-731. 30. Pollyea DA, Tallman MS, de Botton S, et al. Enasidenib, an inhibitor of mutant IDH2 proteins, induces durable remissions in older patients with newly diagnosed acute myeloid leukemia. Leukemia. 2019;33(11): 2575-2584. 31. Amatangelo MD, Quek L, Shih A, et al. Enasidenib induces acute myeloid leukemia cell differentiation to promote clinical response. Blood. 2017;130(6):732-741. 32. Stein EM, DiNardo CD, Fathi AT, et al. Molecular remission and response patterns in patients with mutant-IDH2 acute myeloid leukemia treated with enasidenib.

haematologica | 2021; 106(1)

Advances in differentiation therapy of AML Blood. 2019;133(7):676-687. 33. DiNardo CD, Schuh AC, Stein EM, et al. Enasidenib plus azacitidine significantly improves complete remission and overall response compared with azacitidine alone in patients with newly diagnosed acute myeloid leukemia (AML) with isocitrate dehydrogenase 2 (IDH2) mutations: interim phase II results from an ongoing, randomized study. Blood. 2019;134 (Suppl_1):643. 34. Stein EM, DiNardo CD, Mims AS, et al. Ivosidenib or enasidenib combined with standard induction chemotherapy is well tolerated and active in patients with newly diagnosed AML with an IDH1 or IDH2 mutation: initial results from a phase 1 trial. Blood. 2017;130(Supplement_1):726. 35. Intlekofer AM, Shih AH, Wang B, et al. Acquired resistance to IDH inhibition through trans or cis dimer-interface mutations. Nature. 2018;559(7712):125-129. 36. Quek L, David MD, Kennedy A, et al. Clonal heterogeneity of acute myeloid leukemia treated with the IDH2 inhibitor enasidenib. Nat Med. 2018;24(8):1167-1177. 37. Choe S, Wang H, DiNardo CD, et al. Molecular mechanisms mediating relapse following ivosidenib monotherapy in IDH1mutant relapsed or refractory AML. Blood Adv. 2020;4(9):1894-1905. 38. Montesinos P, Bergua JM, Vellenga E, et al. Differentiation syndrome in patients with acute promyelocytic leukemia treated with all-trans retinoic acid and anthracycline chemotherapy: characteristics, outcome, and prognostic factors. Blood. 2009;113(4): 775-783. 39. Zuckerman T, Ganzel C, Tallman MS, Rowe JM. How I treat hematologic emergencies in adults with acute leukemia. Blood. 2012;120(10):1993-2002. 40. Norsworthy KJ, Mulkey F, Scott EC, et al. Differentiation syndrome with ivosidenib and enasidenib treatment in patients with relapsed or refractory IDH-mutated AML: a U.S. Food and Drug Administration systematic analysis. Clin Cancer Res. 2020;26(16): 4280-4288. 41. Fathi AT, DiNardo CD, Kline I, et al. Differentiation syndrome associated with enasidenib, a selective inhibitor of mutant isocitrate dehydrogenase 2: analysis of a phase 1/2 study. JAMA Oncol. 2018;4(8): 1106-1110. 42. Tallman MS, Andersen JW, Schiffer CA, et al. Clinical description of 44 patients with acute promyelocytic leukemia who developed the retinoic acid syndrome. Blood. 2000;95(1):90-95. 43. Tallman MS, Andersen JW, Schiffer CA, et al. All-trans-retinoic acid in acute promyelocytic leukemia. N Engl J Med. 1997;337 (15):1021-1028. 44. Radomska HS, Basseres DS, Zheng R, et al. Block of C/EBP alpha function by phosphorylation in acute myeloid leukemia with FLT3 activating mutations. J Exp Med. 2006;203(2):371-381. 45. Fathi AT, Le L, Hasserjian RP, Sadrzadeh H, Levis M, Chen YB. FLT3 inhibitor-induced neutrophilic dermatosis. Blood. 2013;122(2): 239-242. 46. Man CH, Fung TK, Ho C, et al. Sorafenib treatment of FLT3-ITD(+) acute myeloid leukemia: favorable initial outcome and mechanisms of subsequent nonresponsiveness associated with the emergence of a D835 mutation. Blood. 2012;119(22):51335143. 47. Varadarajan N, Boni A, Elder DE, et al. FLT3 inhibitor-associated neutrophilic dermatoses. JAMA Dermatol. 2016;152(4): 480-

haematologica | 2021; 106(1)

482. 48. Mori M, Kaneko N, Ueno Y, et al. Gilteritinib, a FLT3/AXL inhibitor, shows antileukemic activity in mouse models of FLT3 mutated acute myeloid leukemia. Invest New Drugs. 2017;35(5):556-565. 49. Perl AE, Altman JK, Cortes J, et al. Selective inhibition of FLT3 by gilteritinib in relapsed or refractory acute myeloid leukaemia: a multicentre, first-in-human, open-label, phase 1-2 study. Lancet Oncol. 2017;18(8):1061-1075. 50. McMahon CM, Canaani J, Rea B, et al. Gilteritinib induces differentiation in relapsed and refractory FLT3-mutated acute myeloid leukemia. Blood Adv. 2019;3(10): 1581-1585. 51. Perl AE, Martinelli G, Cortes JE, et al. Gilteritinib significantly prolongs overall survival in patients with FLT3-mutated (FLT3mut+) relapsed/refractory (R/R) acute myeloid leukemia (AML): results from the phase III ADMIRAL trial. Cancer Res. 2019;79:13. 52. Short NJ, Kantarjian H, Ravandi F, Daver N. Emerging treatment paradigms with FLT3 inhibitors in acute myeloid leukemia. Ther Adv Hematol. 2019;10:2040620719827310. 53. Cortes JE, Khaled S, Martinelli G, et al. Quizartinib versus salvage chemotherapy in relapsed or refractory FLT3-ITD acute myeloid leukaemia (QuANTUM-R): a multicentre, randomised, controlled, open-label, phase 3 trial. Lancet Oncol. 2019;20(7):984997. 54. Sexauer A, Perl A, Yang X, et al. Terminal myeloid differentiation in vivo is induced by FLT3 inhibition in FLT3/ITD AML. Blood. 2012;120(20):4205-4214. 55. Swaminathan M, Kantarjian HM, Daver N, et al. The combination of quizartinib with azacitidine or low dose cytarabine is highly active in patients (Pts) with FLT3-ITD mutated myeloid leukemias: interim report of a phase I/II trial. Blood. 2017;130 (Supplement 1):723. 56. Fischer T, Stone RM, Deangelo DJ, et al. Phase IIB trial of oral midostaurin (PKC412), the FMS-like tyrosine kinase 3 receptor (FLT3) and multi-targeted kinase inhibitor, in patients with acute myeloid leukemia and high-risk myelodysplastic syndrome with either wild-type or mutated FLT3. J Clin Oncol. 2010;28(28):4339-4345. 57. Stone RM, Fischer T, Paquette R, et al. Phase IB study of the FLT3 kinase inhibitor midostaurin with chemotherapy in younger newly diagnosed adult patients with acute myeloid leukemia. Leukemia. 2012;26(9): 2061-2068. 58. Stone RM, Mandrekar SJ, Sanford BL, et al. Midostaurin plus chemotherapy for acute myeloid leukemia with a FLT3 mutation. N Engl J Med. 2017;377(5):454-464. 59. Daver N, Schlenk RF, Russell NH, Levis MJ. Targeting FLT3 mutations in AML: review of current knowledge and evidence. Leukemia. 2019;33(2):299-312. 60. Larrosa-Garcia M, Baer MR. FLT3 inhibitors in acute myeloid leukemia: current status and future directions. Mol Cancer Ther. 2017;16(6):991-1001. 61. Ma J, Zhao S, Qiao X, et al. Inhibition of Bcl2 synergistically enhances the antileukemic activity of midostaurin and gilteritinib in preclinical models of FLT3-mutated acute myeloid leukemia. Clin Cancer Res. 2019;25(22):6815-6826. 62. Zhang H, Savage S, Schultz AR, et al. Clinical resistance to crenolanib in acute myeloid leukemia due to diverse molecular mechanisms. Nat Commun. 2019;10(1):244.

63. McMahon CM, Ferng T, Canaani J, et al. Clonal selection with RAS pathway activation mediates secondary clinical resistance to selective FLT3 inhibition in acute myeloid leukemia. Cancer Discov. 2019;9(8):10501063. 64. Smith CC, Paguirigan A, Jeschke GR, et al. Heterogeneous resistance to quizartinib in acute myeloid leukemia revealed by singlecell analysis. Blood. 2017;130(1):48-58. 65. Smith CC, Wang Q, Chin CS, et al. Validation of ITD mutations in FLT3 as a therapeutic target in human acute myeloid leukaemia. Nature. 2012;485(7397):260-263. 66. Harris WJ, Huang X, Lynch JT, et al. The histone demethylase KDM1A sustains the oncogenic potential of MLL-AF9 leukemia stem cells. Cancer Cell. 2012;21(4):473-487. 67. Schenk T, Chen WC, Gollner S, et al. Inhibition of the LSD1 (KDM1A) demethylase reactivates the all-trans-retinoic acid differentiation pathway in acute myeloid leukemia. Nat Med. 2012;18(4):605-611. 68. Cusan M, Cai SF, Mohammad HP, et al. LSD1 inhibition exerts its antileukemic effect by recommissioning PU.1- and C/EBPalpha-dependent enhancers in AML. Blood. 2018;131(15):1730-1742. 69. McGrath JP, Williamson KE, Balasubramanian S, et al. Pharmacological inhibition of the histone lysine demethylase KDM1A suppresses the growth of multiple acute myeloid leukemia subtypes. Cancer Res. 2016;76(7): 1975-1988. 70. Maes T, Mascaro C, Tirapu I, et al. ORY1001, a potent and selective covalent KDM1A inhibitor, for the treatment of acute leukemia. Cancer Cell. 2018;33(3): 495-511. 71. Somervaille TC, Salamero O, Montesinos P, et al. Safety, phamacokinetics (PK), pharmacodynamics (PD) and preliminary activity in acute leukemia of Ory-1001, a first-in-class inhibitor of lysine-specific histone demethylase 1A (LSD1/KDM1A): initial results from a first-in-human phase 1 study. Blood. 2016;128(22):4060. 72. Guenther MG, Lawton LN, Rozovskaia T, et al. Aberrant chromatin at genes encoding stem cell regulators in human mixed-lineage leukemia. Genes Dev. 2008;22(24):34033408. 73. Milne TA, Martin ME, Brock HW, Slany RK, Hess JL. Leukemogenic MLL fusion proteins bind across a broad region of the Hox a9 locus, promoting transcription and multiple histone modifications. Cancer Res. 2005;65 (24):11367-11374. 74. Nguyen AT, Taranova O, He J, Zhang Y. DOT1L, the H3K79 methyltransferase, is required for MLL-AF9-mediated leukemogenesis. Blood. 2011;117(25):6912-6922. 75. Bernt KM, Zhu N, Sinha AU, et al. MLLrearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. Cancer Cell. 2011;20(1):66-78. 76. Deshpande AJ, Chen L, Fazio M, et al. Leukemic transformation by the MLL-AF6 fusion oncogene requires the H3K79 methyltransferase Dot1l. Blood. 2013;121 (13):2533-2541. 77. Daigle SR, Olhava EJ, Therkelsen CA, et al. Selective killing of mixed lineage leukemia cells by a potent small-molecule DOT1L inhibitor. Cancer Cell. 2011;20(1):53-65. 78. Chen L, Deshpande AJ, Banka D, et al. Abrogation of MLL-AF10 and CALM-AF10mediated transformation through genetic inactivation or pharmacological inhibition of the H3K79 methyltransferase Dot1l. Leukemia. 2013;27(4):813-822. 79. Daigle SR, Olhava EJ, Therkelsen CA, et al. Potent inhibition of DOT1L as treatment of

V. Madan and H.P. Koeffler MLL-fusion leukemia. Blood. 2013;122(6): 1017-1025. 80. Stein EM, Garcia-Manero G, Rizzieri DA, et al. The DOT1L inhibitor pinometostat reduces H3K79 methylation and has modest clinical activity in adult acute leukemia. Blood. 2018;131(24):2661-2669. 81. Klaus CR, Iwanowicz D, Johnston D, et al. DOT1L inhibitor EPZ-5676 displays synergistic antiproliferative activity in combination with standard of care drugs and hypomethylating agents in MLL-rearranged leukemia cells. J Pharmacol Exp Ther. 2014;350(3):646-656. 82. Kuhn MW, Song E, Feng Z, et al. Targeting chromatin regulators inhibits leukemogenic gene expression in NPM1 mutant leukemia. Cancer Discov. 2016;6(10):1166-1181. 83. Dafflon C, Craig VJ, Mereau H, et al. Complementary activities of DOT1L and Menin inhibitors in MLL-rearranged leukemia. Leukemia. 2017;31(6):1269-1277. 84. Feng Z, Yao Y, Zhou C, et al. Pharmacological inhibition of LSD1 for the treatment of MLL-rearranged leukemia. J Hematol Oncol. 2016;9:24. 85. Zuber J, Shi J, Wang E, et al. RNAi screen identifies Brd4 as a therapeutic target in acute myeloid leukaemia. Nature. 2011;478(7370):524-528. 86. Dawson MA, Prinjha RK, Dittmann A, et al. Inhibition of BET recruitment to chromatin as an effective treatment for MLL-fusion leukaemia. Nature. 2011;478(7370):529-533. 87. Mertz JA, Conery AR, Bryant BM, et al. Targeting MYC dependence in cancer by inhibiting BET bromodomains. Proc Natl Acad Sci U S A. 2011;108(40):16669-16674. 88. Rhyasen GW, Hattersley MM, Yao Y, et al. AZD5153: a novel bivalent BET bromodomain inhibitor highly active against hematologic malignancies. Mol Cancer Ther. 2016;15(11):2563-2574. 89. Fiskus W, Sharma S, Qi J, et al. BET protein antagonist JQ1 is synergistically lethal with FLT3 tyrosine kinase inhibitor (TKI) and overcomes resistance to FLT3-TKI in AML cells expressing FLT-ITD. Mol Cancer Ther. 2014;13(10):2315-2327. 90. Chen C, Liu Y, Lu C, et al. Cancer-associated IDH2 mutants drive an acute myeloid leukemia that is susceptible to Brd4 inhibition. Genes Dev. 2013;27(18):1974-1985. 91. Dawson MA, Gudgin EJ, Horton SJ, et al. Recurrent mutations, including NPM1c, activate a BRD4-dependent core transcriptional program in acute myeloid leukemia. Leukemia. 2014;28(2):311-320. 92. Berthon C, Raffoux E, Thomas X, et al. Bromodomain inhibitor OTX015 in patients with acute leukaemia: a dose-escalation, phase 1 study. Lancet Haematol. 2016;3(4): e186-195. 93. Odenike O, Wolff JE, Borthakur G, et al. Results from the first-in-human study of mivebresib (ABBV-075), a pan-inhibitor of bromodomain and extra terminal proteins, in patients with relapsed/refractory acute myeloid leukemia. J Clin Oncol. 2019;37 (15_suppl):7030. 94. Lu J, Qian Y, Altieri M, et al. Hijacking the E3 ubiquitin ligase cereblon to efficiently target BRD4. Chem Biol. 2015;22(6):755-763. 95. Raina K, Lu J, Qian Y, et al. PROTACinduced BET protein degradation as a therapy for castration-resistant prostate cancer. Proc Natl Acad Sci U S A. 2016;113(26): 7124-7129. 96. Saenz DT, Fiskus W, Qian Y, et al. Novel BET protein proteolysis-targeting chimera

exerts superior lethal activity than bromodomain inhibitor (BETi) against post-myeloproliferative neoplasm secondary (s) AML cells. Leukemia. 2017;31(9):1951-1961. 97. Gilan O, Lam EY, Becher I, et al. Functional interdependence of BRD4 and DOT1L in MLL leukemia. Nat Struct Mol Biol. 2016;23(7):673-681. 98. Bots M, Verbrugge I, Martin BP, et al. Differentiation therapy for the treatment of t(8;21) acute myeloid leukemia using histone deacetylase inhibitors. Blood. 2014;123 (9):1341-1352. 99. Quintas-Cardama A, Santos FP, GarciaManero G. Histone deacetylase inhibitors for the treatment of myelodysplastic syndrome and acute myeloid leukemia. Leukemia. 2011;25(2):226-235. 100. Schaefer EW, Loaiza-Bonilla A, Juckett M, et al. A phase 2 study of vorinostat in acute myeloid leukemia. Haematologica. 2009;94(10):1375-1382. 101. Kirschbaum MH, Foon KA, Frankel P, et al. A phase 2 study of belinostat (PXD101) in patients with relapsed or refractory acute myeloid leukemia or patients over the age of 60 with newly diagnosed acute myeloid leukemia: a California Cancer Consortium study. Leuk Lymphoma. 2014;55(10):23012304. 102. Schlenk RF, Krauter J, Raffoux E, et al. Panobinostat monotherapy and combination therapy in patients with acute myeloid leukemia: results from two clinical trials. Haematologica. 2018;103(1):e25-e28. 103. Bewersdorf JP, Shallis R, Stahl M, Zeidan AM. Epigenetic therapy combinations in acute myeloid leukemia: what are the options? Ther Adv Hematol. 2019;10: 2040620718816698. 104. Prebet T, Sun Z, Ketterling RP, et al. Azacitidine with or without entinostat for the treatment of therapy-related myeloid neoplasm: further results of the E1905 North American Leukemia Intergroup study. Br J Haematol. 2016;172(3):384-391. 105. Garcia-Manero G, Sekeres MA, Egyed M, et al. A phase 1b/2b multicenter study of oral panobinostat plus azacitidine in adults with MDS, CMML or AML with 30% blasts. Leukemia. 2017;31(12):2799-2806. 106. Goldenson B, Crispino JD. The aurora kinases in cell cycle and leukemia. Oncogene. 2015;34(5):537-545. 107. Wen Q, Goldenson B, Silver SJ, et al. Identification of regulators of polyploidization presents therapeutic targets for treatment of AMKL. Cell. 2012;150(3):575-589. 108. Thiollier C, Lopez CK, Gerby B, et al. Characterization of novel genomic alterations and therapeutic approaches using acute megakaryoblastic leukemia xenograft models. J Exp Med. 2012;209(11):2017-2031. 109. Fathi AT, Wander SA, Blonquist TM, et al. Phase I study of the aurora A kinase inhibitor alisertib with induction chemotherapy in patients with acute myeloid leukemia. Haematologica. 2017;102(4):719-727. 110. Bowman RL, Levine RL. TET2 in normal and malignant hematopoiesis. Cold Spring Harb Perspect Med. 2017;7(8):a026518. 111. Cimmino L, Dolgalev I, Wang Y, et al. Restoration of TET2 function blocks aberrant self-renewal and leukemia progression. Cell. 2017;170(6):1079-1095. 112. Agathocleous M, Meacham CE, Burgess RJ, et al. Ascorbate regulates haematopoietic stem cell function and leukaemogenesis. Nature. 2017;549(7673):476-481.

113. Liu M, Ohtani H, Zhou W, et al. Vitamin C increases viral mimicry induced by 5-aza-2'deoxycytidine. Proc Natl Acad Sci U S A. 2016;113(37):10238-10244. 114. Waldo AL, Zipf RE. Ascorbic acid level in leukemic patients. Cancer. 1955;8(1):187190. 115. Sykes DB, Kfoury YS, Mercier FE, et al. Inhibition of dihydroorotate dehydrogenase overcomes differentiation blockade in acute myeloid leukemia. Cell. 2016;167(1):171-186. 116. Tzelepis K, Koike-Yusa H, De Braekeleer E, et al. A CRISPR dropout screen identifies genetic vulnerabilities and therapeutic targets in acute myeloid leukemia. Cell Rep. 2016;17(4):1193-1205. 117. Sykes DB. The emergence of dihydroorotate dehydrogenase (DHODH) as a therapeutic target in acute myeloid leukemia. Expert Opin Ther Targets. 2018;22(11):893-898. 118. Madak JT, Bankhead A, 3rd, Cuthbertson CR, Showalter HD, Neamati N. Revisiting the role of dihydroorotate dehydrogenase as a therapeutic target for cancer. Pharmacol Ther. 2019;195:111-131. 119. Cao L, Weetall M, Trotta C, et al. Targeting of hematologic malignancies with PTC299, a novel potent inhibitor of dihydroorotate dehydrogenase with favorable pharmaceutical properties. Mol Cancer Ther. 2019;18 (1):3-16. 120. Zhou J, Quah JY, Chooi JY, et al. ASLAN003, a novel and potent dihydroorotate dehydrogenase (DHODH) inhibitor, induces differentiation of acute myeloid leukemia. Blood. 2018;132 (Suppl_1):4047. 121. Christian S, Merz C, Evans L, et al. The novel dihydroorotate dehydrogenase (DHODH) inhibitor BAY 2402234 triggers differentiation and is effective in the treatment of myeloid malignancies. Leukemia. 2019;33(10):2403-2415. 122. Volk A, Liang K, Suraneni P, et al. A CHAF1B-dependent molecular switch in hematopoiesis and leukemia pathogenesis. Cancer Cell. 2018;34(5):707-723. 123. Kishtagari A, Ng KP, Jarman C, et al. A firstin-class inhibitor of ISWI-mediated (ATPDependent) transcription repression releases terminal-differentiation in AML cells while sparing normal hematopoiesis. Blood. 2018;132(Suppl_1):216. 124. Placke T, Faber K, Nonami A, et al. Requirement for CDK6 in MLL-rearranged acute myeloid leukemia. Blood. 2014;124 (1):13-23. 125. Radomska HS, Alberich-Jorda M, Will B, Gonzalez D, Delwel R, Tenen DG. Targeting CDK1 promotes FLT3-activated acute myeloid leukemia differentiation through C/EBPalpha. J Clin Invest. 2012;122(8):2955-2966. 126. Bertoli S, Boutzen H, David L, et al. CDC25A governs proliferation and differentiation of FLT3-ITD acute myeloid leukemia. Oncotarget. 2015;6(35):38061-38078. 127. McKenzie MD, Ghisi M, Oxley EP, et al. Interconversion between tumorigenic and differentiated states in acute myeloid leukemia. Cell Stem Cell. 2019;25(2):258272. 128. Rau RE, Rodriguez BA, Luo M, et al. DOT1L as a therapeutic target for the treatment of DNMT3A-mutant acute myeloid leukemia. Blood. 2016;128(7):971-981. 129. Lu R, Wang P, Parton T, et al. Epigenetic perturbations by Arg882-mutated DNMT3A potentiate aberrant stem cell gene-expression program and acute leukemia development. Cancer Cell. 2016;30(1):92-107.

haematologica | 2021; 106(1)

ARTICLE

Acute Lymphoblastic Leukemia

Clinical significance of occult central nervous system disease in adult acute lymphoblastic leukemia: a multicenter report from the Campus ALL Network

Maria Ilaria Del Principe,1 Elisa Buzzatti,1 Alfonso Piciocchi,2 Fabio Forghieri,3 Massimiliano Bonifacio,4 Federica Lessi,5 Silvia Imbergamo,5 Enrico Orciuolo,6 Giovanni Rossi,7 Nicola Fracchiolla,8 Silvia Trappolini,9 Benedetta Neri,10 Chiara Sarlo,11 Patrizia Zappasodi,12 Michelina Dargenio,13 Mariagiovanna Cefalo,10 Maria Antonietta Irno-Consalvo,1 Consuelo Conti,14 Giovangiacinto Paterno,1 Gottardo De Angelis,1 Mariarita Sciumè,8 Irene Della Starza,15 Adriano Venditti,1 Robin Foà15 and Anna Rita Guarini16

Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):39-45

Ematologia Dipartimento di Biomedicina e Prevenzione, Università degli Studi di Roma “Tor Vergata”, Roma; 2GIMEMA Data Center, Roma; 3Ematologia Dipartimento di Scienze Mediche e Chirurgiche, Università degli Studi di Modena e Reggio Emilia, Azienda Ospedaliera di Modena, Modena; 4Dipartimento di Medicina, Sezione di Ematologia, Universitày di Verona, Verona; 5Ematologia ed Immunologia Clinica, Azienda Ospedaliera di Padova, Padova; 6UO Ematologia, Azienda Ospedaliero-Universitaria Pisana, Pisa; 7U.O. di Ematologia e Trapianto di Cellule Staminali, IRCCS “Casa Sollievo della Sofferenza”, San Giovanni Rotondo, Foggia; 8UOC di Ematologia, Fondazione IRCCS Ca' Granda, Ospedale Maggiore Policlinico, Milano; 9Clinica di Ematologia, AOU Ospedali Riuniti di Ancona, Ancona; 10Ematologia, Ospedale Sant’Eugenio, Dipartimento di Biomedicina e Prevenzione, Università degli Studi di Roma “Tor Vergata”, Roma; 11Ematologia, Policlinico Universitario-Campus Biomedico, Roma; 12Divisione di Ematologica, Fondazione IRCCS Policlinico San Matteo, Università di Pavia, Pavia; 13Ematologia e Trapianto di Cellule Staminali, Ospedale Vito Fazzi, Lecce; 14Ematologia, Dipartimento di Onco-Ematologia, Fondazione Policlinico Tor Vergata, Roma; 15Ematologia, Dipartimento di Medicina Traslazionale e di Precisione, Università “Sapienza”, Roma and 16Dipartimento di Medicina Molecolare, Università “Sapienza”, Roma, Italy 1

ABSTRACT

n acute lymphoblastic leukemia (ALL), flow cytometry (FCM) detects leukemic cells in patients’ cerebrospinal fluid (CSF) more accurately than conventional cytology (CC). However, the clinical significance of FCM positivity with a negative cytology (i.e., occult central nervous system [CNS] disease) is not clear. In the framework of the national Campus ALL program, we retrospectively evaluated the incidence of occult CNS disease and its impact on outcome in 240 adult patients with newly diagnosed ALL. All CSF samples were investigated by CC and FCM. The presence of ≥10 phenotypically abnormal events, forming a cluster, was considered to be FCM positivity. No CNS involvement was documented in 179 patients, while 18 were positive by modified conventional morphology with CC and 43 were occult CNS disease positive. The relapse rate was significantly lower in CNS disease negative patients and the disease-free and overall survival (OS) were significantly longer in CNS disease negative patients than in those with manifest or occult CNS disease positivity. In multivariate analysis, the status of manifest and occult CNS disease positivity was independently associated with a worse OS. In conclusion, we demonstrate that in adult ALL patients at diagnosis FCM can detect occult CNS disease at high sensitivity and that the status of occult CNS disease positivity is associated with an adverse outcome. (Registered at clinicaltrials.gov identifier: NCT03803670).

Introduction Over the last two decades, improved response rates have been reported in adult patients with acute lymphoblastic leukemia (ALL).1-3 In this context of a superior systemic disease control, central nervous system (CNS) involvement has become haematologica | 2021; 106(1)

Correspondence: MARIA ILARIA DEL PRINCIPE del.principe@med.uniroma2.it Received: July 29, 2019. Accepted: December 20, 2019. Pre-published: December 26, 2019. https://doi.org/10.3324/haematol.2019.231704

M.I. Del Principe et al.

an ever more influential limitation to the achievement of a long-term cure and a main cause of mortality. At diagnosis, about 5-10% of adult ALL patients have CNS involvement,4-6 which translates into a shorter overall survival (OS) compared to that of patients without CNS involvement.4 Conventional cytology (CC) examination of the cerebrospinal fluid (CSF) remains the gold standard for the diagnosis of CNS involvement in ALL; CC is estimated to have a >95% specificity. However, it has a relatively low sensitivity (<50%), resulting in frequent false negative determinations. Such a low sensitivity is due to the poor cellularity of CSF and to the difficulties in distinguishing benign from malignant cells on morphologic grounds only.7,8 Flow cytometric (FCM) immunophenotyping is a valuable tool for the diagnosis and staging of hematologic disorders involving lymph nodes, blood, bone marrow and other body fluids. Current FCM assays allow detection of phenotypically abnormal cells up to the limit of at least 0.01% (1 target cell in 104 events), representing, therefore, a very effective tool for minimal residual disease monitoring in acute leukemia.9 Indeed, several recently published experiences have demonstrated the superior sensitivity of FCM over CC for the detection of CNS disease in patients with ALL and non-Hodgkin lymphoma.10-13 These studies have also contributed to establish a new standard that is the so-called “occult CNS disease” (OCNSD), namely the status of FCM positivity and CC negativity. None of these reports has, however, clarified whether a condition of OCNSD has an additional prognostic role compared to the well-established negative impact of CC positivity. We therefore conducted a multicenter, retrospective study in the framework of the national Campus ALL program aimed at improving the management of adult ALL patients. The aims of the present study were: (i) to evaluate the incidence of OCNSD in a large series of adult patients with ALL; and (ii) to assess the impact of OCNSD on the clinical outcome of these patients.

combined with steroids once a week for a total of 3-4 administrations during the induction and consolidation cycles, respectively. In LAL0904, cranio-spinal irradiation (CI) was dispensed after the consolidation phase,14 while in the other GIMEMA/NILG protocols CI was omitted and all patients received a CNS-crossing agent-based chemotherapy. According to the NILG-ALL10/07 protocol, 12 triple agent (methotrexate 12.5 mg, cytarabine 50 mg, dexamethasone 4 mg) IT injections were given as CNS prophylaxis. Finally, in the Hyper-CVAD/MTX-ARAC program, 16 prophylactic IT were planned.17,18 CNS therapy for patients with a CCpositive LP consisted of IT injections of 12 mg methotrexate, 50 mg cytarabine and 10 mg methylprednisolone twice weekly until CSF blast clearance, and then once weekly for two administrations.

Cell counts and conventional cytology Cytospins for CC examination were prepared as previously described in detail.19,20 CC positivity was defined as unequivocal, morphological evidence of leukemic blast in the CSF and/or a CSF WBC count ≥5/mL with less than 10 erythrocytes/mL.3,21 Traumatic LP were excluded from the analysis.

Flow cytometry analysis All centers involved were selected on the basis of a strict adherence to a standardized approach relying on the same procedures (time elapsed from collection to processing, number of fluorochromes, number of acquired events and analysis). Samples for FCM analysis were locally processed within 60 minutes from harvest, as described elsewhere.19 A cocktail of 6-8 monoclonal antibodies was used (Online Supplementary Table S1). On average, 1,080 events were acquired (range 0-210,000). In agreement with the recommendations for the analysis of rare events, a cluster of at least 10 phenotypically abnormal events was regarded as proof of CSF infiltration10 (Figure 1). Traumatic LP were excluded from the analysis.

Statistical analysis The statistical analysis is described in the Online Supplementary Appendix.

Ethical considerations Methods Study design and patients Our retrospective analysis included patients seen between January 2007 and December 2017 at 13 Italian hematology centers. Cases were documented using a case report form. Variables included the following data: age, sex, ALL onset, genetic/cytogenetic features, B/T phenotype, white blood cell count (WBC) at diagnosis and at the time of lumbar puncture (LP), lactate dehydrogenase (LDH), chemotherapy, date of complete remission (CR), CSF cell count and chemistry, CC and FCM results, date of systemic and/or CNS relapse, allogeneic stem cell transplant (ASCT), date of death or the last follow-up. Personal information was treated in a confidential manner and all sensitive data were analyzed anonymously. Samples were collected at diagnosis. In patients with a high WBC count, which might be due to the traumatic procedure and confound the CSF picture, the explorative LP was performed once the WBC count was reduced below 10x109/L by administering steroids. Patients were treated within or according to GIMEMA (LAL0904, LAL1308, LAL1913, LAL1104)14 or NILG (NILGALL10/07)15,16 protocols or the Hyper-CVAD/MTX-ARAC regimen.17,18 In the GIMEMA protocols, CNS prophylaxis consisted in intrathecal injection (IT) of methotrexate (12.5 or 15 mg) alone or 40

Approval of the local institutional review board and ethics committee was obtained at all participating sites. The trial was registered at clinicaltrials.gov identifier: NCT03803670.

Results Patients' characteristics The clinical and laboratory characteristics of the 240 patients are summarized in Table 1. At diagnosis, 179 (75%) CSF samples were negative by both FCM and CC (CNSneg), while 43 (18%) were OCNSD positive (positive by FCM and negative by CC=OCNSDpos) and 18 (7%) were positive by both FCM and CC (manifest CNS disease positive = MCNSDpos) (Table 1). No case proved to be FCM-negative and CC-positive. The characteristics of patients belonging to the three groups are listed in Table 1. There was an equal male:female ratio among CNSneg, OCNSDpos and MCNSDpos patients. There was no significant difference in median age, median WBC count, B/T lineage, LDH levels between the three patient categories. Cytogenetic/genetic data were available in 178 of 240 cases (74%) and no difference in distribution among the three categories was haematologica | 2021; 106(1)

CNS involvement in adult ALL patients

Figure 1. Flow cytometry detection of occult central nervous system (CNS) involvement in a patient with B-lineage acute lymphoblastic leukemia (ALL). The blast population is shown in gray which denotes a cluster of few cells CD19 (C) and CD10 (D) positive, and CD34, CD22 negative (E and F) and CD20 weak (F).

observed. On the other hand, the status of OCNSDpos and MCNSDpos was significantly associated with a high CSF cellularity (P<0.001) (Table 1) and the levels of CSF proteins (P=0.023) (Table 1). One hundred and seventy-one patients (71%) were treated within or according to GIMEMA protocols, 37 (15%) with the HyperCVAD/MTX-ARAC regimen, and 32 (14%) according to the NILG ALL10/07 protocol. Considering the heterogeneity of the chemotherapy regimens utilized, we analyzed our series dividing the patients into three groups on the basis of the intensity of the treatment received. Accordingly, 91 patients (37.9%) underwent a conventional treatment, 120 (50%) an intensified pediatric-inspired regimen, and 29 (12.1%), qualified as unfit or frail, were treated with a reduced intensity schedule (Table 1).

Outcome Of the 232 evaluable patients, 198 (85%) achieved a CR with no significant differences between the three CNS status-based groups (P=0.3). Of these 198 patients, 116 (59%) experienced a relapse; in 18 of 116 (15%), disease recurrence occurred in the CNS alone or was combined with a hematologic relapse. The relapse rate was significantly higher in OCNSDpos and MCNSDpos patients than in CNSneg patients (P=0.001) (Table 2). The 3-year disease-free survival (DFS) was also significantly longer in CNSneg patients compared to OCNSDpos or MCNSDpos patients: 39% (95% confidence interval [CI]: 31-48) versus 21% (95%CI: 4.5haematologica | 2021; 106(1)

33.9) versus 21% (95%CI: 7.9-58.4), respectively (P=0.005) (Table 2). On the contrary, there was no difference in 3year DFS between OCNSDpos and MCNSDpos patients (P=0.3) (Figure 2). The 3-year overall survival (OS) in CNSneg, OCNSDpos and MCNSDpos patients was 53% (95%CI: 45.5-61.5), 31% (95%CI: 19.2-50.5) and 22% (95%CI: 9.4-52.7), respectively (P<0.0001) (Table 2). There was no difference in 3-year OS between OCNSDpos and MCNSDpos patients (P=0.2) (Figure 3).

Multivariate analysis The clinical impact of the CNS status on OS was also challenged in the multivariate Cox proportional hazard analysis applied to models including age, transplant, sex, WBC count and treatment received. Multivariate analysis confirmed that the OCNSDpos (HR=1.82, 95%CI: 1.155.92; P=0.01) or MCNSDpos status (HR=3.23, 95%CI: 1.762.89; P<0.0001), defined at the time of diagnosis, were factors that independently impacted on OS together with the treatment regimens (intensified vs. conventional vs. reduced intensity for age) (Table 3).

Discussion This retrospective study shows that FCM offers better technical support than CC in detecting leukemic cells in the CFS of adult patients with ALL, and documents the 41

M.I. Del Principe et al.

clinical impact of OCNSD on the outcome of these patients. By introducing FCM analysis, the detection power improved to such an extent that evidence of CNS involvement increased from 7% to 25% of ALL cases at diagnosis. This analysis confirms previous reports that demonstrated the superior sensitivity of FCM over CC.10,12,13,22,23 In a large retrospective study of 326 CSF sam-

ples collected from patients affected by diffuse large B-cell and Burkitt lymphomas, a CSF involvement was detected by FCM in 33 (13%) diffuse large B-cell lymphomas and in 9 (11%) Burkitt lymphomas.24 FCM allows detection of a hematologic disease in CSF specimens even when the cellularity is very low.9,25 This peculiarity has been confirmed in pediatric ALL patients where FCM was able to

Table 1. Clinical characteristics of patients according to the central nervous system (CNS) status.

Level N Sex, N (%)

F M

Age, years, median (range) Lineage, N (%)

B T

WBC x 109/L, median (%) Cytogenetic, N (%) Treatment, N (%)

LDH, U/L, median (range) CSF-WBC per mm3, median (%) CSF protein, mg/dL, median (range)

Abnormal Normal Conventional Intensified Reduced

ALL

CNSneg

OCNSDpos

MCNSDpos

240 103 (42.9) 137 (57.1) 45 (17-80) 184 (76.7) 56 (23.3) 11 (0.140- 573) 118 (64.5) 65 (35.5) 91 (37.9) 120 (50.0) 29 (12.1) 482 (21-8,332) 1 (0-3,000) 36 (5.9-326)

179 76 (42.5) 103 (57.5) 45 (17-80) 140 (78.2) 39 (21.8) 11 (0.140- 573) 91 (63.6) 52 (36.4) 70 (39.1) 85 (47.5) 24 (13.4) 478 (21-8,332) 1 (0-17) 35 (5-94)

43 20 (46.5) 23 (53.5) 46 (17-72) 34 (79.1) 9 (20.9) 10 (1.44-291) 20 (69.0) 9 (31.0) 15 (34.9) 23 (53.5) 5 (11.6) 555 (55-5,532) 1 (0-7) 38 (16-161)

18 7 (38.9) 11 (61.1) 36.50 (18-73) 10 (55.6) 8 (44.4) 9 (0.4-133,84) 7 (63.6) 4 (36.4) 6 (33.3) 12 (66.7) 0 (0.0) 372 (180-4,086) 39 (7-3,000) 51 (23-326)

P 0.835 0.302 0.088 0.799 0.860 0.400

0.806 <0.001 0.023

ALL: acute lymphoblastic leukemia; N: number; F: female; M: male; CNSneg: cerebrospinal fluid (CSF) samples negative by both flow cytometry (FCM) and conventional cytology (CC); OCNSDpos: CSF samples positive by FCM and negative by CC; MCNSDpos: CSF positive by both FCM and CC; WBC: white blood cells; LDH: lactate dehydrogenase.

CNSneg OCNSDpos MCNSDpos

Overall P value=0.005

OCNSDpos vs. MCNSDpos

P=0.3

Figure 2. Disease-free survival (DFS) based on central nervous system (CNS) status. Kaplan-Meier plot comparing DFS of cerebrospinal fluid (CSF) samples negative by both flow cytometry (FCM) and conventional cytology (CC) (CNSneg), occult CSF samples positive by FCM and negative by CC (OCNSDpos), and CSF positive by both FCM and CC (MCNSDpos).

haematologica | 2021; 106(1)

CNS involvement in adult ALL patients

substantially improve recognition of occult CSF involvement.26-28 In agreement with pediatric reports,27 the CNS status of our adults did not correlate with risk factors associated with the risk of relapse, such as WBC count at onset, B/T phenotype or cytogenetic/genetic features. In pediatric ALL, FCM positivity alone in the absence of a positive CC seems to affect clinical outcome.27-29 Similar observations have been made in patients with high-risk non-Hodgkin lymphomas and Burkitt lymphomas, in whom FCM positivity of CSF was associated with a significantly higher risk of CNS relapse and a worse prognosis.24,30

In adult ALL patients, the role of OCNSD is less clear because of the limited number of studies based on small series of patients. By analyzing 168 CSF samples collected from 31 patients with ALL, Subira et al.31 reported a concordance between FCM and CC except for ten samples. All patients found to be FCM negative remained free from CNS disease. In a small population of 38 adults with ALL or lymphoblastic lymphoma, we previously observed that the median OS of patients with FCM single positivity was intermediate between double positive or negative patients.19 The uncertain clinical significance of the FCM analysis

Table 2. Correlation between central nervous system (CNS) status and outcome. N Hematologic response, N (%) ASCT, N (%) Relapse, N (%) Relapse site, N (%) OS 3 years DFS 3 years

Level

ALL

CNSneg

OCNSDpos

MCNSDpos

CR No CR No Yes No Yes CNS BM Estimate % (95%CI) Estimate % (95%CI)

240 198 (85.3) 34 (14.7) 88 (44.9) 108 (55.1) 78 (40.2) 116 (59.8) 16 (16.8) 79 (83.2) 46.4 (40.1-53.8) 34.3 (27.9-42.2)

179 152 (87.4) 22 (12.6) 65 (44.2) 82 (55.8) 70 (47.0) 79 (53.0) 8 (12.7) 55 (87.3) 52.9 (45.5-61.5) 38.6 (31-48)

43 32 (80.0) 8 (20.0) 17 (47.2) 19 (52.8) 7 (22.6) 24 (77.4) 7 (31.8) 15 (68.2) 31.1 (19.2-50.5) 20.6 (10.2-41.9)

18 14 (77.8) 4 (22.2) 6 (46.2) 7 (53.8) 1 (7.1) 13 (92.9) 1 (10.0) 9 (90.0) 22.2 (9.4-52.7) 21.4 (7.9-58.4)

P 0.317 0.944 0.001 0.099 <0.001 0.005

ALL: acute lymphoblastic leukemia; N: number; CNSneg: cerebrospinal fluid (CSF) samples negative by both flow cytometry (FCM) and conventional cytology (CC); OCNSDpos: CSF samples positive by FCM and negative by CC; MCNSDpos: CSF positive by both flow cytometry and CC; ASCT: allogeneic stem cell transplant; CR: complete remission; BM: bone marrow; OS: overall survival; DFS: disease-free survival; CI: confidence interval.

CNSneg OCNSDpos MCNSDpos

Overall P value <0.0001

OCNSDpos vs. MCNSDpos P=0.2

Figure 3. Overall survival (OS) based on the central nervous system (CNS) status. Kaplan-Meier plot comparing OS of patientsâ&#x20AC;&#x2122; cerebrospinal fluid (CSF) samples negative by both flow cytometry (FCM) and conventional cytology (CC) (CNSneg), occult CSF samples positive by FCM and negative by CC (OCNSDpos), and CSF positive by both FCM and CC (MCNSDpos).

haematologica | 2021; 106(1)

M.I. Del Principe et al. Table 3. Univariate and multivariate analysis of all variables associated with survival.

HR Age Sex: M vs. F Lineage T vs. B WBC Cytogenetic: normal vs. abnormal Treatment: conventional vs. intensified Treatment: conventional vs. reduced LDH CSF_WBC CSF_proteins ASCT yes vs. no OCNSDpos vs. CNSneg MCNSDpos vs. CNSneg

1.01 1.044 0.957 1 1.15 0.61 1.47 1 1 1.008 0.564 1.915 2.887

Univariate analysis Lower Higher 95%CI 95%CI 1 0.739 0.6367 1 0.76 0.42 0.89 1 1 1.002 0.387 1.259 1.73

1.03 1.474 14.375 1 1.74 0.88 2.44 1 1.001 1.013 0.823 2.913 4.817

P 0.0062 0.8076 0.8313 0.3764 0.5101 0.0089 0.1295 0.908 0.5719 0.0071 0.0029 0.0024 <0.0001

Multivariate analysis Lower Higher 95%CI 95%CI

0.584 1.708

0.403 1.027

0.848 2.842

0.0047 0.0393

2.03 3.392

1.333 2.015

3.093 5.71

0.001 <0.0001

FCM: confidence flow cytometry; CC: conventional cytology; M: male; F: female; HR: hazard ratio; CI: confidence interval; CSF: cerebrospinal fluid; CNSneg: CSF samples negative by both FCM and CC; OCNSDpos: CSF samples positive by FCM and negative by CC; MCNSDpos: CSF positive by both FCM and CC; WBC: white blood cell count; ASCT: allogeneic stem cell transplant; LDH: lactate dehydrogenase.

of CSF is confirmed by the discordant position of the current guidelines. In fact, while the National Comprehensive Cancer Network (NCCN) guidelines32 do not indicate that FCM analysis of the CSF should be part of the initial work-up, the more recent American pocket guide for the clinician recommends (although not strongly) performing this examination at diagnosis.33 Based on our large multicenter report, occult CNS status does indeed have a significant impact on outcome. In fact, patients with OCNSD had a worse DFS and OS compared to those who were OCNSD negative. The superimposable duration of OS of OCNSD and MCNSD patients indicates that even the presence of a few cells in the CNS sanctuary has a clinical impact; these few cells can only be detected by using approaches more sensitive than CC, such as FCM. The pronounced neurotropism of ALL34-36 can be responsible for disease recurrence once the leukemic cells, having survived systemic chemotherapy within the CNS sanctuary, migrate to the circulation.37,38 Thus, the availability of highly sensitive methods capable of accurately defining whether or not the CSF is colonized by leukemic cells not only offers a refined diagnostic/prognostic work-up, but also helps to personalize CNS prophylaxis through the early identification of patients who may benefit from more aggressive approaches. With the limitations of its retrospective nature, the results of our study demonstrate that, in adult ALL patients, FCM can more precisely identify and quantify the number of patients with CNS involvement at diagno-

References 1. Thomas X, Boiron JM, Huguet F, et al. Outcome of treatment in adults with acute lymphoblastic leukemia: analysis of LALA94 trial. J Clin Oncol. 2004;22(20):40754086. 2. Kantarjian HM, Oâ&#x20AC;&#x2122;Brien S, Smith TL, et al. Results of treatment with Hyper-CVAD, a dose-intensive regimen in adult acute lymphocytic leukemia. J Clin Oncol. 2000; 18(3):547-561.

sis and that this impacts significantly on the clinical course and outcome of the disease, thus enabling a further refinement of the current diagnostic risk-stratification process. This refined CNS evaluation should become a routine tool for the work-up of ALL patients at presentation. Further and larger prospective studies are needed to further standardize the procedures and promote optimal clinical application of this technique. Disclosures This study was carried out as part of the routine clinical workup of patients. The authors declare no competing financial interests. Contributions MIDP, AV and AG designed the study, interpreted data, wrote the manuscript; AP analyzed data and performed statistical analysis; EB, FF, MB, FL, SI, EO, GR, NF, ST, BN, CS, PZ, MD, MC, GD, MS, GP provided patients information, collected clinical data and contributed to data analysis; MAC and CC obtained flow cytometry data; IDS interpreted data and contributed to data analysis; RF wrote and revised the manuscript. All authors approved the manuscript. Acknowledgments The authors would like to thank all participants of the Campus ALL program. Partly supported by Associazione Italiana Ricerca sul Cancro (AIRC), Special 5x1000 Program Metastases (21198), Milan (Italy) to RF.

3. Thomas X, Le QH. Central nervous system involvement in adult acute lymphoblastic leukemia. Hematology. 2008;13(5):293-302. 4. Lazarus HM, Richards SM, Chopra R, et al. Medical Research Council (MRC)/National Cancer Research Institute (NCRI) Adult Leukaemia Working Party of the United Kingdom and the Eastern Cooperative Oncology Group. Central nervous system involvement in adult acute lymphoblastic leukemia at diagnosis: results from international ALL trial MRC UKALL XII/ECOG E2993. Blood. 2006;108(2):65-72.

5. Jabbour E, Thomas D, Cortes J, Kantarjian H, Oâ&#x20AC;&#x2122;Brien S. Central nervous system prophylaxis in adults with acute lymphoblastic leukemia. Cancer. 2010;116(10):2290-2300. 6. Larson RA. Managing CNS disease in adults with acute lymphoblastic leukemia. Leuk Lymphoma. 2018;59:3-13. 7. Bromberg JE, Breems DA, Kraan J, et al. CSF flow cytometry greatly improves diagnostic accuracy in CNS hematologic malignancies. Neurology. 2007;68(20):1674-1679. 8. Kaplan JG, DeSouza TG, Farkash A, et al. Leptomeningeal metastases: comparison of

haematologica | 2021; 106(1)

CNS involvement in adult ALL patients clinical features and laboratory data of solid tumors, lymphomas and leukemias. J Neurooncol. 1990;9(3):225-229. 9. Craig F, Foon KA. Flow cytometric immunophenotyping for hematologic neoplasms. Blood. 2008;111(8):3941-3967. 10. Quijano S, Lopez A, Manuel Sancho J, et al. Spanish Group for the Study of CNS disease in NHL. Identification of leptomeningeal disease in aggressive B-cell non Hogkin’s lymphoma: improved sensitivity of flow cytometry. J Clin Oncol. 2009;27(9):14621469. 11. de Graaf MT, de Jongste AH, Kraan J, Boonstra JG, Sillevis Smitt PA, Gratama JW. Flow cytometric characterization of cerebrospinal fluid cells. Cytometry B Clin Cytom. 2011;80(5):271-281. 12. Zeiser R, Burger JA, Bley TA, WindfuhrBlum M, Schulte-Monting J, Behringer DM. Clinical follow-up indicates differential accuracy of magnetic resonance imaging and immunocytology of the cerebral spinal fluid for the diagnosis of neoplastic meningitis-a single centre experience. Br J Haematol. 2004;124(6):762-768. 13. Di Noto R, Scalia G, Abate G, et al. Critical role of multidimensional flow cytometry in detecting occult leptomeningeal disease in newly diagnosed aggressive B-cell lymphomas. Leuk Res. 2008:32(8):1196-1199. 14. Annino L, Vignetti M, Paoloni FP, et al. Treatment of adolescents and young adults with acute lymphoblastic leukemia (ALL): an update of the GIMEMA experience. Blood. 2009;114(22):3097. 15. Bassan R, Masciulli A, Intermesoli T, et al. Randomizad trial of radiation-free central nervous system prophylaxis comparing intrathecal triple therapy with liposomal cytarabine in acute lymphoblastic leukemia. Haematologica. 2015;100(6):786-793. 16. Bassan R, Masciulli A, Intermesoli T, et al. Final results of Northern Italy Leukemia Group (NILG) trial 10/07 combining pediatric-type therapy with minimal residual disease study and risk-oriented hematopoietic cell transplantation in adult acute lymphoblastic leukemia (ALL). Blood. 2016; 128(22):176. 17. Kantarjian HM, O'Brien S, Smith TL, et al. Results of treatment with hyper-CVAD, a dose-intensive regimen, in adult acute lymphocytic leukemia. J Clin Oncol. 2000; 18(3):547-561. 18. Kantarjian H, Thomas D, O'Brien S, et al. Long-term follow-up results of hyperfrac-

haematologica | 2021; 106(1)

tionated cyclophosphamide, vincristine, doxorubicin, and dexamethasone (HyperCVAD), a dose-intensive regimen, in adult acute lymphocytic leukemia. Cancer. 2004; 101(12):2788-2801. 19. Del Principe MI, Buccisano F, Cefalo M, et al. High sensitivity of flow cytometry improves detection of occult leptomeningeal disease in acute lymphoblastic leukemia and lymphoblastic lymphoma. Ann Hematol. 2014; 93(9):1509-1513. 20. Del Principe MI, Buccisano F, Soddu S, et al. Involvement of central nervous system in adult patients with acute myeloid leukemia: incidence and impact on outcome. Semin Hematol. 2018;55(4):209-214. 21. Mahmoud HH, Rivera GK, Hancock ML, et al. Low leukocyte counts with blast cells in cerebrospinal fluid of children with newly diagnosed acute lymphoblastic leukemia. N Engl J Med. 1993;329(5):314-319. 22. Roma A, Garcia A, Avagnina A, Rescia C, Elsner B. Lymphoid and myeloid neoplasms involving cerebrospinal fluid: comparison of morphologic examination and imunophenotyping by flow cytometry. Diagn Cytopathol. 2002;27(5):271-275. 23. Mitri Z, Siddiqui MT, El Rassi F, et al. Sensitivity and specificity of cerebral fluid flow cytometry for the diagnosis of leukemic meningitis in acute lymphoblastic leukemia/lymphoma. Leuk Lymphoma. 2014;55(7):1498-1500. 24. Wilson W, Bromberg J, Stetler-Stevenson M, et al. Detection and outcome of occult leptomeningeal disease in diffuse large B-cell lymphoma and Burkitt lymphoma. Haematologica. 2014;99(7):1228-1235. 25. Nuckel H, Novotny JR, Noppeney R, Savidou I, Duhrsen U. Detection of malignant haematopoietic cells in the cerebrospinal fluid by conventional cytology and flow cytometry. Clin Lab Haem. 2006; 28(1):22-29. 26. Sayed D, Badrawy H, Ali AM, Shaker S. Immunophenotyping and immunoglobulin heavy chain gene rearrangement analysis in cerebrospinal fluid of pediatric patients with acute lymphoblastic leukemia. Leuk Res. 2009;33(5):655-661. 27. Cancela CSP, Murao M, Assumpção JG, et al. Immunophenotyping of the cerebrospinal fluid as a prognostic factor at diagnosis of acute lymphoblastic leukemia in children and adolescents. Pediatr Hematol Oncol. 2017;34(2):53-65. 28. Ranta S, Nilsson F, Harila-Saari A, et al.

Detection of central nervous system involvement in childhood acute lymphoblastic leukemia by cytomorphology and flow cytometry of the cerebrospinal fluid. Pediatr Blood Cancer. 2015;62(6):951956. 29. Martinez-Laperche C, Gomez-Garcia AM, Lassaletta A, et al. Detection of occult cerebrospinal fluid involvement during maintenance therapy identifies a group of children with acute lymphoblastic leukemia at high risk for relapse. Am J Hematol. 2013; 88(5):360-365. 30. Benevolo G, Stacchini A, Spina M, et al. Final results of a multicenter trial addressing role of CSF flow cytometric analysis in NHL patients at high risk for CNS dissemination. Blood. 2012;120(16):3222-3228. 31. Subira D, Castanon S, Roman A, et al. Flow cytometry and the study of central nervous disease in patients with acute leukaemia. Br J Haematol. 2001;112(2):381-384. 32. SAlvarnas JC , Brown PA, Aoun P, et al. Acute lymphoblastic leukemia, version 2.2015. J Natl Compr Canc Netw. 2015;13(10):1240-1279. 33. SArber DA, Borowitz MJ, Cessna M, et al. Initial diagnostic workup of acute leukemia: guideline from the College of American Pathologists and the American Society of Hematology. Arch Pathol Lab Med. 2017; 141(10):1342-1393. 34. SAkers SM, O'Leary HA, Minnear FL, et al. VE-cadherin and PECAM-1 enhance ALL migration across brain microvascular endothelial cell monolayers. Exp Hematol. 2010;38(9):733-743. 35. SAkers SM, Rellick SL, Fortney JE, Gibson LF. Cellular elements of the subarachnoid space promote ALL survival during chemotherapy. Leuk Res. 2011;35(6):705711. 36. Svan der Velden VH, de Launaij D, de Vries JF, et al. New cellular markers at diagnosis are associated with isolated central nervous system relapse in paediatric B-cell precursor acute lymphoblastic leukaemia. Br J Haematol. 2016;172(5):769-781. 37. SFrishman-Levy L, Izraeli S. Advances in understanding the pathogenesis of CNS acute lymphoblastic leukaemia and potential for therapy. Br J Haematol. 2017; 176(2):157-167. 38. SPui CH, Howard SC. Current management and challenges of malignant disease in the CN in paediatric leukaemia. Lancet Oncol. 2008;9(3):257-268.

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):46-55

Acute Lymphoblastic Leukemia

Outcomes after late bone marrow and very early central nervous system relapse of childhood B-acute lymphoblastic leukemia: a report from the Children’s Oncology Group phase III study AALL0433

Glen Lew,1 Yichen Chen,2 Xiaomin Lu,3 Susan R. Rheingold,4 James A. Whitlock,5 Meenakshi Devidas,2 Caroline A. Hastings,6 Naomi J. Winick,7 William L. Carroll,8 Brent L. Wood,9 Michael J. Borowitz,10 Michael A. Pulsipher11# and Stephen P. Hunger4#

Aflac Cancer & Blood Disorders Center, Children’s Healthcare of Atlanta, Department of Pediatrics, Emory University School of Medicine, Atlanta, GA, USA; 2Department of Global Pediatric Medicine, St Jude Children’s Research Hospital, Memphis, TN, USA; 3 Gilead Sciences, Inc., Foster City, CA, USA; 4Department of Pediatrics and the Center for Childhood Cancer Research, Children’s Hospital of Philadelphia and the Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, USA; 5Division of Hematology/Oncology, Department of Pediatrics, The Hospital for Sick Children, University of Toronto, Toronto, Ontario, Canada; 6UCSF Benioff Children's Hospital Oakland, Oakland, CA, USA; 7University of Texas Southwestern Medical Center, Dallas, TX, USA; 8Perlmutter Cancer Center and Department of Pediatrics, NYU Langone Medical Center, New York, NY, USA; 9Department of Laboratory Medicine, University of Washington, Seattle, WA, USA; 10Department of Pathology, Johns Hopkins Medical Institutions, Baltimore, MD, USA and 11Division of Hematology, Oncology, and Blood and Marrow Transplantation, Children's Hospital Los Angeles, USC Keck School of Medicine, Los Angeles, CA, USA 1

MAP and SPH contributed equally as co-senior authors.

ABSTRACT

Correspondence: GLEN LEW glen.lew@choa.org Received: September 3, 2019. Accepted: January 24, 2020. Pre-published: January 30, 2020. https://doi.org/10.3324/haematol.2019.237230

utcomes after relapse of childhood B-acute lymphoblastic leukemia (B-ALL) are poor, and optimal therapy is unclear. The children’s Oncology Group study AALL0433 evaluated a new platform for relapsed ALL. Between March 2007 and October 2013 AALL0433 enrolled 275 participants with late bone marrow or very early isolated central nervous system (iCNS) relapse of childhood B-ALL. Patients were randomized to receive standard versus intensive vincristine dosing; this randomization was closed due to excess peripheral neuropathy in 2010. Patients with matched sibling donors received allogeneic hematopoietic cell transplantation (HCT) after the first three blocks of therapy. The prognostic value of minimal residual disease (MRD) was also evaluated in this study. The 3-year event free and overall survival (EFS/OS) for the 271 eligible patients were 63.6±3.0% and 72.3±2.8% respectively. MRD at the end of Induction-1 was highly predictive of outcome, with 3-year EFS/OS of 84.9±4.0% and 93.8±2.7% for patients with MRD <0.1%, versus 53.7±7.8% and 60.6± 7.8% for patients with MRD ≥0.1% (P<0.0001). Patients who received HCT versus chemotherapy alone had an improved 3-year disease-free survival (77.5±6.2% vs. 66.9 + 4.5%, P=0.03) but not OS (81.5±5.8% for HCT vs. 85.8±3.4% for chemotherapy, P=0.46). Patients with early iCNS relapse fared poorly, with a 3-year EFS/OS of 41.4±9.2% and 51.7±9.3%, respectively. Infectious toxicities of the chemotherapy platform were significant. The AALL0433 chemotherapy platform is efficacious for late bone marrow relapse of B-ALL, but with significant toxicities. The MRD threshold of 0.1% at the end of Induction-1 was highly predictive of the outcome. The optimal role for HCT for this patient population remains uncertain. This trial is registered at clinicaltrials.gov (NCT# 00381680). haematologica | 2021; 106(1)

Outcomes for relapsed B-ALL from Children’s Oncology Group study AALL0433

Introduction Childhood acute lymphoblastic leukemia (ALL) is highly curable,1,2 but 10% of patients will not survive, primarily because of relapse. After relapse, B-lymphoblastic phenotype, longer first remission, and isolated extramedullary relapse have been associated with improved chances of salvage.3-6 Minimal residual disease (MRD) post re-Induction has also been shown to be predictive of outcome.7,8 Previous studies have shown that up to one-half of patients with late marrow or very early isolated extramedullary (IEM) relapse can be cured.5,9,10,11,12 However, this group is heterogeneous, and it is challenging to distinguish which patients require therapies beyond chemotherapy, particularly hematopoietic cell transplantation (HCT). The Children’s Oncology Group (COG)11-14 and other international cooperative groups9,15,16,17,18 have used highly varied chemotherapy approaches for relapsed ALL with largely similar outcomes. Although HCT has been successfully utilized in relapsed ALL,19-21 there are limited and conflicting data whether HCT is superior to conventional chemotherapy for late marrow or IEM relapse.22 Patients with HLA-matched sibling donors were previously reported to have better outcomes than those from unrela-ted/ alternative donors due to decreased toxicities. However, more recent reports suggest similar HCT outcomes regardless of the donor type.20,23 COG AALL0433 (NCT# 00381680) for intermediaterisk first relapse of B-ALL was instituted to evaluate the efficacy of a hybrid platform, combining elements from prior COG studies for both bone marrow and IEM relapse.24-26 Eligible patients overlap considerably with the S2 group from the Berlin, Franfurt, Muenster (BFM) relapse trials.15 This report details results from AALL0433.

S1). The duration of therapy was 2 years. iCNS and combined marrow/CNS patients who did not proceed to HCT received 18 Gy cranial radiotherapy at the beginning of maintenance. Testicular relapse patients with residual involvement (biopsyproven) after Induction-1 were to receive 24 Gy radiotherapy. Patients were randomized at enrollment to either standard (Arm A, 1.5 mg/m2, maximum 2 mg) or intensive (Arm B, 2 mg/m2, maximum 2.5 mg) vincristine dosing. Randomization was suspended in June 2010 after scheduled interim monitoring revealed excess peripheral neuropathy in Arm B meeting predefined stopping rules. The amended study (without randomization) re-opened in November 2010. Since there were no differences in outcomes for Arm A pre- and post- randomization closure, outcome data for the standard arm were pooled All patients with an HLA matched sibling donor were recommended to proceed to HCT after completing three treatment blocks. HCT regimens were not prescribed, and were given at the discretion of treating centers; patients were encouraged to enroll on concurrent COG HCT study ASCT0431.19 Unrelated/alternative donor HCT was not included in AALL0433, but outcome data were collected for those who underwent unrelated/alternative donor HCT per investigator discretion.

MRD assessment MRD was performed at two COG reference laboratories using previously described flow cytometric methods.27-29 MRD was measured at the end of Induction-1 and Induction-3. Results were blinded to investigators, as one study objective was to establish prognostic MRD thresholds for patients with relapse.

Toxicity assessment Data on adverse events and clinically significant laboratory findings are summarized using NCI Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Adverse event reporting was supplemented with NCI’s Adverse Event Expedited Reporting System (AdEERS) and MedWatch reports.

Methods Statistical analysis Patient characteristics AALL0433 enrolled subjects age 1-29.99 years with intermediate-risk relapse of B-ALL between March 2007 and October 2013. Eligible patients had late bone marrow or combined marrow/ EM relapse ≥36 months from diagnosis, or very early IEM (CNS/testicular) relapse <18 months from diagnosis. Bone marrow relapse was defined as an M3 marrow (>25% blasts) occurring after attaining complete remission (CR) without concurrent CNS/testicular relapse. Isolated CNS (iCNS) relapse was defined as cerebrospinal fluid having ≥5 cells/microliter with blasts or clinical or radiologic signs of CNS leukemia. Biopsy was required to establish the diagnosis of isolated testicular relapse. Combined relapse was defined as a CNS and/or testicular relapse with concurrent M2 (≥5% blasts) or M3 marrow. AALL0433 was approved by the National Cancer Institute (NCI) and Institutional Review Boards of participating institutions. Informed consent was obtained in accordance with Department of Health and Human Services Guidelines.

Treatment AALL0433 utilized chemotherapy beginning with three intensive blocks from COG AALL01P2 [24], followed by intensification, reinduction, and maintenance from COG 9412/AALL02P225, 26 (Figure 1 and Online Supplementary Table haematologica | 2021; 106(1)

Primary outcomes were 3-year event-free survival (EFS) and overall survival (OS) rates. EFS was defined as time from enrollment to first event (induction failure, induction death, death in remission, relapse, or second malignant neoplasm [SMN] or date of last contact for those who were event-free). OS was calculated as time from enrollment to death or last contact for those alive. Disease-free survival (DFS) and OS for patients on the standard arm were compared between those receiving chemotherapy versus HCT (matched sibling or other donor) post achieving second complete remission (CR2), where CR2 was defined as having <5% blasts in the bone marrow morphologically. For analyses involving HCT in CR2, DFS/OS were defined from date of CR2; and DFS times for chemotherapy patients were adjusted by median time to HCT for the cohort. Survival rates were estimated using the Kaplan-Meier method with standard errors of Peto et al.30,31 The two-sided log-rank test was used for the comparison of the survival curves. The cumulative incidence function for competing risks was used to calculate cumulative incidence rates, and comparisons were made using the K-sample test.32 Comparison of proportions was done using the c2 test or Fisher’s Exact test. P-values <0.05 were considered statistically significant for all comparisons. All analyses were performed using SAS® version 9.4. Graphics were generated using R (http://www.R-project.org, version 3.4.0). Data frozen on 06/30/2018 are included in this report. 47

G. Lew et al.

Results Participants AALL0433 enrolled 275 subjects between March 2007 and October 2013; four patients were ineligible, not meeting inclusion criteria. The median follow-up time was 4.13 years (range: 0.05â&#x20AC;&#x201C;9.91). The patient flow and characteristics are given in Figure 2 and Table 1. Of 271 eligible patients, 185 had bone marrow, 57 had combined marrow/ EM, and 29 had very early iCNS relapse. No patients with isolated or combined testicular relapse were enrolled. Nine patients died in Induction-1. Five patients went off protocol therapy due to induction failure (three with M3 marrow

after Induction-1, two failing to achieve an M1 marrow [< 5% blasts] after Induction-2). Three patients who achieved CR2 died in Induction-3. Two hundred and fifty-seven patients (95%) achieved CR2, and 254 completed the threeblock induction in CR2. One hundred and eighty-three patients (161 late marrow/combined, 22 iCNS) chemotherapy/radiotherapy per protocol, and seventy-four patients underwent HCT in CR2 (51 late marrow, 16 combined, and seven iCNS relapses). Fourty-seven patients received matched sibling donor HCT as outlined in the protocol but 27 underwent alternative donor HCT. The median time to transplant from CR2 was 102 days (range: 52-379) for late marrow and 88 days (range: 27-253) for iCNS relapse.

Figure 1. Schema. CNS: central nervous system; VCR: Vincristine; CSF: cerebrospinal fluid; ITT: intention-to-treat ; HLA: human leukocyte antigen; SCT: stem cell transplant.

haematologica | 2021; 106(1)

Outcomes for relapsed B-ALL from Children’s Oncology Group study AALL0433

Treatment outcome The EFS/OS for all 271 eligible patients were, 63.6±3.0% and 72.3±2.8% at 3 years, and 51.0±3.5% and 62.9±3.3% at 5 years, respectively (Figure 3). The overall EFS/OS for patients with bone marrow/combined relapse were 66.3±3.1% and 74.8±2.9% at 3 years, and 52.5±3.7% and 65.1±3.5% at 5 years. Younger patients at the time of enrollment had superior outcomes compared to older patients, with 3-year EFS and OS of 68.7±3.6% and 76.2±3.3% respectively for patients less than 13 years old versus 54.1±5.3% and 65.1±5.1% for patients greater than 13 years (P=0.0037 for EFS, P=0.0009 for OS). The 3-year cumulative incidence of isolated bone marrow second relapse was 14.7±2.5%, considering other sites of relapse, induction failure, induction death, death in remission and SMN as competing risks. The 3-year cumulative incidence of remission death was 6.0±1.7%,

considering induction events (induction deaths and induction failure), relapse and SMN as competing events. Six patients developed SMN (all therapy-related myeloid leukemia or myelodysplastic syndrome).

Establishment of MRD threshold The outcomes of patients with bone marrow or combined marrow/ EM relapse treated on the standard arm were stratified by MRD levels after Induction-1. Of 122 patients with data available, 62 (51%) had MRD <0.01%, 41 (34%) had MRD >0.1%, and 19 (16%) had levels between 0.01-0.1%. The 3-year EFS/OS were 84.9±4.0% and 93.8±2.7% for patients with MRD <0.1%, versus 53.7 ±7.8% and 60.6±7.8% for patients with MRD >0.1% (P<0.0001) (Figure 4). The most prognostic MRD threshold was 0.1%, as there were no significant differences in outcome between patients with MRD <0.01%

Figure 2. CONSORT Diagram. BM: bone marrow; HCT: hematopoietic cell transplantation; CR2: second complete remission; CNS: central nervous system; Ph-ALL: Philadelphia chromosome–like acute lymphoblastic leukemia; MPAL: mixed phenotype acute leukemia.

haematologica | 2021; 106(1)

G. Lew et al. Table 1. Patient characteristics.

Characteristics

Age at enrollment in years Median Range Sex – Male Race* White Black Other Unknown Ethnicity* Hispanic or Latino Neither Hispanic nor Latino Unknown AALL0433 Stratum Late isolated bone marrow relapse Early isolated CNS relapse Early isolated testicular relapse Early CNS + testicular relapse Late combined bone marrow + EM relapse +/- CNS, +/- testicular disease Chemotherapy Late pre-B relapse Early isolated CNS relapse HCT Matched sibling donor Late pre-B relapse Early isolated CNS relapse Other donor Late pre-B relapse Early isolated CNS relapse

All eligible patients n=271 N (%)

Regimen A Standard VCR n=203 N (%)

Regimen B Intensive VCR n=68 N (%)

11 3-29 152 (56.1)

11 3-29 109 (53.7)

11 3-24 43 (63.2)

208 (76.8) 15 (5.5) 11 (4.1) 37 (13.7)

154 (75.9) 12 (5.9) 9 (4.4) 28 (13.8)

54 (79.4) 3 (4.4) 2 (2.9) 9 (13.2)

62 (22.9) 193 (71.2) 16 (5.9)

55 (27.1) 142 (70.0) 6 (3.0)

7 (10.3) 51 (75.0) 10 (14.7)

185 (68.3) 29 (10.7) 0 (0.0) 0 (0.0) 57 (21.0) 183 (67.5) 161 22 74 (27.3) 47 42 5 27 24 3

138 (68.0) 22 (10.8) 0 (0.0) 0 (0.0) 43 (21.2) 139 (68.5) 123 16 55 (27.1) 34 30 4 21 19 2

47 (69.1) 7 (10.3) 0 (0.0) 0 (0.0) 14 (20.6) 44 (64.7) 38 6 19 (27.9) 13 12 1 6 5 1

*Self-reported race/ethnic categories per NIH criteria; CNS: central nervous system; EM relapse: extramedullary relapse; Late pre-B relapse: late isolated bone marrow and combined bone marrow + EM relapse in precursor B-ALL; B-ALL: B-acute lymphoblastic leukemia; VCR: Vincristine.

versus those with intermediate MRD levels (0.01-0.1%) (Online Supplementary Figure S1A). Seventy of the above patients had end of Induction-3 MRD data; 62 (88.6%) were <0.1% and 8 (11.4%) were ≥0.1%. The eight patients with MRD >0.1% had inferior outcomes with 3-year EFS 37.5±17.1% and OS 50.0±17.7% compared to 85.1±4.6% and 95.1±2.8% for those <0.1% (P=0.0007 for EFS; P=0.0105 for OS) (Online Supplementary Figure S1B).

Hematopoietic cell transplantation The DFS/OS for patients with bone marrow/combined relapse who received chemotherapy alone versus HCT were compared. The DFS analyses for patients receiving chemotherapy were adjusted by median time to transplant, resulting in 10 chemotherapy patients (six relapses, three deaths, one withdrawal of consent) being excluded from these analyses. There were no significant differences in the DFS/OS between patients receiving matched sibling donor HCT versus other HCT (Online Supplementary Figure S2); further analyses combined all HCT patients regardless 50

of the donor source. Analyzing patients on the standard arm alone, a significant improvement in DFS was seen for HCT over chemotherapy, with a 3-year DFS of 77.5±6.2% versus 66.9±4.5% (P=0.03) (Figure 5). However, this did not correspond with improved OS (81.5±5.8% for HCT vs. 85.8±3.4% for chemotherapy; P=0.46). When patients on both vincristine dose arms were included, the 3-year DFS maintained a trend towards significance (73.1±5.6% for HCT vs. 65.6±4.0%, P=0.08). The HCT outcomes at different MRD thresholds were also examined (Figure 6). In patients with MRD <0.1% at the end of Induction-1, HCT showed a trend of improved 3-year DFS of 90.7±6.5% versus 74.1±5.8% for chemotherapy (P=0.07). The difference in 3-year OS was not significant, 95.5±4.7% for HCT versus 93.1±3.4% for chemotherapy (P=0.16). In patients with MRD >0.1% after Induction-1, no benefit was seen for HCT over chemotherapy for 3-year DFS (HCT 56.3±13.2%, chemotherapy 55.0±11.1, P=0.23) or 3-year OS (HCT 62.5±12.8%, chemotherapy 70.0±10.3%, P=1.00). haematologica | 2021; 106(1)

Outcomes for relapsed B-ALL from Children’s Oncology Group study AALL0433

B B

Figure 3. Overall outcomes for all eligible patients. (A) Event-free survival (EFS) curves, 3-year and 5-year EFS: 63.6±3.0%, and EFS 51.0%±3.5%, respectively. (B) Overall survival (OS) , 3-year and 5-year OS 72.3±2.8%, and OS 62.9±3.3%, respectively.

Isolated CNS relapse The 3-year EFS/OS for eligible patients with very early iCNS relapse were 41.4±9.2% and 51.7±9.3%, respectively. No deaths or induction failures were observed during the three induction blocks for iCNS patients. After completion of Induction-3, seven patients were treated with HCT, and 22 continued chemotherapy/radiotherapy. After adjusting DFS/OS times for the chemotherapy cohort by median time to transplant from CR2 (88 days), one chemotherapy patient relapsed before that and was excluded from the DFS/OS analyses. The 3-year DFS/OS for transplanted patients (n=7) were 71.4±17.0% and 71.4±17.1%, compared to 28.6±9.9% and 42.9±10.8% for those who received chemotherapy/radiotherapy (n=21), respectively (P=0.12 for DFS, P=0.18 for OS) (Online Supplementary Figure S3). Three patients of the 14 of 28 patients who relapsed, recurred in the CNS, 10 in bone marrow (BM) alone, and one patient in both BM/CNS.

Vincristine randomization Planned interim toxicity analysis in 2010 showed an increase in grade 3 or higher peripheral neuropathy in the higher-dose vincristine arm 14 of 68 (21%) compared to 3 of 69 (4%) in the control arm (P=0.004). The difference was primarily reported in patients >13 years old, with an haematologica | 2021; 106(1)

Figure 4 Survival curves by 0.1% minimal residual disease (MRD) threshold after induction-1 for late bone marrow relapse patients on standard arm (Arm A), with available MRD data. (A) Event-free survival (EFS) curves by 0.1% MRD threshold, 3-year EFS: 84.9±4.0% for patients with MRD <0.1% versus 53.7±7.8% for patients with MRD ≥0.1% (P<0.0001). (B) Overall survival (OS) curves by 0.1% MRD threshold, 3-year OS: 93.8±2.7% for patients with MRD <0.1% and 60.6±7.8% for patients with MRD ≥0.1% (P<0.0001).

incidence of 9 of 26 (25%) in the higher-dose arm compared to 1 of 21 (5%) in the control arm. There was no significant difference in patients <13, with 2 of 48 (4.2%) in the standard versus 5 of 42 (11.9%) in the higher-dose arm, P=0.245. These results led to the suspension, followed by closure of the vincristine randomization as there was no evidence of an improved MRD response at the time of interim analysis (data not shown). At final analysis, there was a trend toward inferiority of the higher dose arm, with a 3-year EFS of 70.1±3.5% versus 55.1±6.5% for standard versus intensive dosing (P=0.091), and OS of 78.6 ±3.1% versus 63.6±6.3% (P=0.071) (Online Supplementary Figure S4).

Infection/deaths Eighteen on-study deaths (6.6%) due to toxicity were reported in patients receiving chemotherapy. All were due to infection and their sequelae (Online Supplementary Table S2). Nine occurred in Induction-1, giving an induction death rate of 3.3%. An additional nine deaths occurred in patients achieving CR2 – all followed highdose cytarabine in the Induction-3 (n=3) or Intensification-2 phases (n=6). Grade 3 or higher bloodstream infections were also common in these phases, with incidences of 23.6%, 29%, and 30.6% in Induction1, Induction-3, and Intensification-2, respectively. 51

G. Lew et al.

Discussion Despite the high success rate of treatment for newly diagnosed childhood B-ALL, survival for those relapsing after frontline treatment remains poor. In contrast to newly-diagnosed B-ALL, where therapy is similar across international study groups, therapies used for relapse are highly variable. Divergent therapeutic approaches, smaller patient numbers for clinical trials, and varying utilization of HCT make it difficult to compare trial results from different cooperative study groups. We report the outcomes from the first Phase 3 trial for BM and very early IEM relapse of B-ALL from the Children’s Oncology Group. The 3-year EFS and OS of 63.6% and 72.3% from AALL0433 are similar to the overall results reported from the UK ALLR3 trial, which revealed 3-year progression free survival of 60% and OS of 72%.18 However, 57% of patients in the R3 study underwent HCT in the sescond remission as opposed to continued chemotherapy. In contrast, less than one third of patients enrolled on AALL0433 received HCT. While the overall results are broadly similar between studies, the significant differences in the percentage of patients recei-ving HCT in the second remission makes it challenging to compare the relative efficacy of the different chemotherapy regimens. Although the overall outcomes from AALL0433 are similar to those reported by other groups, significant infectious toxicities occurred even after Induction-1, with half of treatment-related deaths occurring in CR2. The toxicity death rate from the current study is similar to the 5% (4 of 76) death rate from the 9412 protocol which served as the framework for AALL0433,25 and also within the range reported in recent trials for relapsed ALL.10,17,24 Higher vincristine doses were hypothesized to improve outcomes. Capping of vincristine dose was adopted from clinical observations from the 1960’s,33 but results in a relatively lower dose (based on body surface area) for many older patients with ALL, who are known to have a higher risk of relapse. However, vincristine intensification was not feasible in this non-blinded randomization due to an unacceptably high incidence of peripheral neuropathy, particularly among patients older than 13 years. Patients on this trial had significant prior vincristine exposure during primary therapy, which may have been a cofactor. Although the randomization was stopped before meeting statistical endpoints, the higher dose arm trended toward inferior outcomes. Genetic predictors of vincristine neuropathy were explored as a secondary study objective, and polymorphisms in CEP72 were discovered to be associa-ted with neuropathy incidence.34 MRD has been shown to be one of the most important predictors of outcome in newly diagnosed ALL, but prognostic threshold values after relapse are less clear. The end-induction MRD threshold of 0.01% is highly prognostic in frontline ALL trials from COG and European cooperative groups27,29,35,36 and was also highly prognostic in the AIEOP REC 2003 relapse trial.8 However, the UKALL R3 relapse protocol used the same 0.01% threshold, and showed improved progression-free survival for those receiving mitoxantrone (66%) versus idarubicin (46%) despite no difference in MRD clearance between arms.17,18 A higher MRD threshold of 0.1% was highly predictive of outcome in BFM REZ P95/96/2002.7,37 The same 0.1% threshold was the strongest predictor of out52

Figure 5 Survival curves by hematopoietic cell transplantation versus chemotherapy alone, for late bone marrow relapse patients on Arm A. (A) Adjusted disease-free survival (DFS) curves by hematopoietic cell transplantation (HCT) versus chemotherapy alone, 3-year DFS: 77.5±6.2% for HCT and 66.9±4.5% for chemotherapy alone (P=0.03). One relapse occurred after 6 years, where only two patients had more than 6 years of follow-up, which resulted in DFS for HCT dropping from 77.5% to 38.8%. (B) Adjusted overall survival (OS) curves by HCT versus chemotherapy alone, 3-year OS: 81.5±5.8% for HCT and 85.8±3.4% for chemotherapy alone (P=0.46).

come on AALL0433, and is currently being used to riskstratify patients on the successor COG relapse study AALL1331 (NCT02101853). It is feasible that the optimal MRD threshold used to stratify risk groups may be higher in relapse than in newly diagnosed ALL. The indications for HCT after late BM or early IEM relapse of B-ALL are unclear, as randomized trials have not been feasible in this population. Although there was an improved 3-year DFS for patients receiving HCT over chemotherapy in AALL0433, this did not result in improved OS. Sub-analysis showed improved DFS with HCT in patients with MRD <0.1% after Induction-1, but no improvement in those with MRD ≥0.1%, and no improvement in OS for HCT over chemotherapy regardless of the MRD level. This is contrary to other reports where HCT improved outcomes for patients with poor MRD response.38,39 Lower pre-transplant MRD levels correlate with improved HCT outcomes,19,40 but it is less clear how MRD kinetics immediately after induction (typically several months before HCT) relate to eventual outcomes. The emergence of immunotherapeutic options (such as blinatumomab and CAR T-cell therapies) during haematologica | 2021; 106(1)

Outcomes for relapsed B-ALL from Children’s Oncology Group study AALL0433

Figure 6. Survival curves by hematopoietic cell transplantation versus chemotherapy alone, for late bone marrow relapse patients on Arm A, using a 0.1% minimal residual disease threshold. (A) Adjusted disease-free survival (DFS) curves for minimal residual disease (MRD) <0.1%, by hematopoietic cell transplantation (HCT) versus chemotherapy alone, 3-year DFS: 90.7±6.5% for HCT and 74.1±5.8% for chemotherapy alone (P=0.07). (B) Adjusted overall survival (OS) curves for MRD <0.1%, by HCT versus chemotherapy alone, 3-year OS: 95.5±4.7% for HCT and 93.1±3.4% for chemotherapy alone (P=0.16). (C) Adjusted disease-free survival (DFS) curves for MRD ≥0.1%, by HCT versus chemotherapy alone, 3-year DFS: 56.3±13.2% for HCT and 55.0± 11.1% for chemotherapy alone (P=0.23). (D) Adjusted OS curves for MRD ≥0.1%, by HCT versus chemotherapy alone, 3-year OS: 62.5±12.8% for HCT and 70.0±10.3% for chemotherapy alone (P=1.00).

AALL043341-43 may have improved the chances of salvage and prolonged OS for patients with a second relapse. The AALL0433 regimen has significant toxicity. AALL0433 includes 45,000 mg/m2 of high-dose methotrexate, 12,200 mg/m2 of cyclophosphamide, 2,900 mg/m2 of etoposide, and five courses of high-dose cytarabine, which were associated with nine deaths in second remission. As broadly similar results have been seen in other recent trials for this patient population with regimens with lower toxicity,5,6 further intensification of traditional chemotherapeutic agents is unlikely to improve outcomes for relapsed ALL. This study has several limitations. Measurement of MRD at the time AALL0433 opened was performed as a research test (with blinded results), but became common clinical practice by the time of study closure, leading some investigators to pursue HCT off study. AALL0433 was not powered to formally compare HCT and chemotherapy, especially for iCNS patients where numhaematologica | 2021; 106(1)

bers were small (although a similar trend favoring HCT for iCNS patients was also noted in ALL R3).44 Additionally, vincristine do-sing was not blinded to investigators or families, so reporting bias is feasible in the incidence of neuropathy reported on the higher dose arm. Newer therapies for relapsed ALL including immunotherapies and small molecule inhibitors, are being actively investigated.45 Many of these therapies have shown impressive early results in patients with a second or later relapse and merit testing in patients at the time of the first relapse. The COG AALL1331 relapse trial includes a randomization to intensive post-induction chemotherapy blocks (based upon UKALL R3) versus replacement with blinatumomab. Immunotherapies will also be tested in upcoming COG trials for patients with newly diagnosed ALL (NCT #’s 03914625, 03959085, 03876769). Ultimately, the goal of curing more patients will result from preventing relapses, and not treating them. 53

G. Lew et al.

Disclosures No conflicts of interest to disclose. Contributions GL wrote the primary manuscript and chaired the study; YC, XL, and MD performed data analysis and contributed to the manuscript; BLW and MJB performed minimal residual disease testing for the study; GL, SRR, JAW, CAH, NJW, WLC, MAP and SPH developed the study, supervised the conduct of the study, interpreted the data, and contributed to the manuscript.

Elizabeth Raetz, Mignon Loh, Susan Conway, Catherine Shannon, and the other members of the AALL0433 committee for their contributions to this work. SPH is the Jeffrey E. Perelman Distinguished Chair in the Department of Pediatrics at the Children’s Hospital of Philadelphia. Funding This study was supported by grants U10 CA98543, U10 CA98413, U10 CA180886, 1U24-CA196173 and U10 CA180899 from the National Institutes of Health, and by St. Baldrick’s Foundation.

Acknowledgments The authors would like to acknowledge Rochelle Yanofsky,

References 1. Hunger SP, Lu X, Devidas M, et al. Improved survival for children and adolescents with acute lymphoblastic leukemia between 1990 and 2005: a report from the Children's Oncology Group. J Clin Oncol. 2012; 30(14):1663-1669. 2. Pui CH, Yang JJ, Hunger SP, et al. Childhood acute lymphoblastic leukemia: progress through collaboration. J Clin Oncol. 2015; 33(27):2938-2948. 3. Chessells JM. Relapsed lymphoblastic leukaemia in children: a continuing challenge. Br J Haematol. 1998;102(2):423-438. 4. Nguyen K, Devidas M, Sheng SC, et al. Factors influencing survival after relapse from acute lymphoblastic leukemia: a Children's Oncology Group study. Leukemia. 2008;22(12):2142-2150. 5. Bhojwani D, Pui CH. Relapsed childhood acute lymphoblastic leukaemia. Lancet Oncol. 2013;14(6):e205-217. 6. Locatelli F, Moretta F, Rutella S. Management of relapsed acute lymphoblastic leukemia in childhood with conventional and innovative approaches. Curr Opin Oncol. 2013:25(6):707-715. 7. Eckert C, von Stackelberg A, Seeger K, et al. Minimal residual disease after induction is the strongest predictor of prognosis in intermediate risk relapsed acute lymphoblastic leukaemia - long-term results of trial ALLREZ BFM P95/96. Eur J Cancer. 2013;49(6):1346-1355. 8. Paganin M, Zecca M, Fabbri G, et al. Minimal residual disease is an important predictive factor of outcome in children with relapsed ‘high-risk’ acute lymphoblastic leukemia. Leukemia. 2008;22(12):21932200. 9. Lawson SE, Harrison G, Richards S, et al. The UK experience in treating relapsed childhood acute lymphoblastic leukaemia: a report on the medical research council UKALLR1 study. Br J Haematol. 2000; 108(3):531-543. 10. Tallen G, Ratei R, Mann G, et al. Long-term outcome in children with relapsed acute lymphoblastic leukemia after time-point and site-of-relapse stratification and intensified short-course multidrug chemotherapy: results of trial ALL-REZ BFM 90. J Clin Oncol. 2009;28(14):2339-2347. 11. Buchanan GR, Rivera GK, Pollock BH, et al. Alternating drug pairs with or without periodic reinduction in children with acute lymphoblastic leukemia in second bone marrow remission: a Pediatric Oncology Group study. Cancer. 2000;88(5):1166-1174.

12. Kelly ME, Lu X, Devidas M, et al. Treatment of relapsed precursor-B acute lymphoblastic leukemia with intensive chemotherapy: POG (Pediatric Oncology Group) study 9411 (SIMAL 9). J Pediatr Hematol Oncol. 2013;35(7):509-513. 13. Abshire TC, Pollock BH, Billett AL, et al. Weekly polyethylene glycol conjugated Lasparaginase compared with biweekly dosing produces superior induction remission rates in childhood relapsed acute lymphoblastic leukemia: a Pediatric Oncology Group study. Blood. 2000;96(5):1709-1715. 14. Gaynon PS, Harris RE, Altman AJ, et al. Bone marrow transplantation versus prolonged intensive chemotherapy for children with acute lymphoblastic leukemia and an initial bone marrow relapse within 12 months of the completion of primary therapy: Children's Oncology Group study CCG1941. J Clin Oncol. 2006;24(19):3150-3156. 15. Henze G, von Stackelberg A, Eckert C. ALLREZ BFM--the consecutive trials for children with relapsed acute lymphoblastic leukemia. Klin Padiatr. 2013;225(Suppl 1): S73-78 . 16. Roy A, Cargill A, Love S, et al. Outcome after first relapse in childhood acute lymphoblastic leukaemia - lessons from the United Kingdom R2 trial. Br J Haematol. 2005;130(1):67-75. 17. Parker C, Waters R, Leighton C, et al. Effect of mitoxantrone on outcome of children with first relapse of acute lymphoblastic leukaemia (ALL R3): an open-label randomised trial. Lancet. 2010;376(9757):20092017. 18. Parker C, Krishnan S, Hamadeh L, et al. Outcomes of patients with childhood B-cell precursor acute lymphoblastic leukaemia with late bone marrow relapses: long-term follow-up of the ALLR3 open-label randomised trial. Lancet Haematol. 2019; 6(4):e204-e216. 19. Pulsipher MA, Langholz B, Wall DA, et al. The addition of sirolimus to tacrolimus/methotrexate GVHD prophylaxis in children with ALL: A phase 3 Children's Oncology Group / Pediatric Blood and Marrow Transplant Consortium trial. Blood. 2014;123(13):2017-2025. 20. Peters C, Schrappe M, von Stackelberg A, et al. Stem-cell transplantation in children with acute lymphoblastic leukemia: a prospective international multicenter trial comparing sibling donors with matched unrelated donors - the ALL-SCT-BFM-2003 trial. J Clin Oncol. 2015;33(11):1265-1274. 21. Uderzo C, Valsecchi MG, Bacigalupo A, et al. Treatment of childhood acute lymphoblastic leukemia in second remission

with allogeneic bone marrow transplantation and chemotherapy: ten-year experience of the Italian Bone Marrow Transplantation Group and the Italian Pediatric Hematology Oncology Association. J Clin Oncol. 1995; 13(2):352-358. 22. Pulsipher MA, Peters C, Pui CH. High risk pediatric acute lymphoblastic leukemia: to transplant or not to transplant? Biol Blood Marrow Transplant. 2011;17(Suppl 1):S137148. 23. MacMillan M, Davies SM, Nelson GO, et al. Twenty years of unrelated donor bone marrow transplantation for pediatric acute leukemia facilitated by the National Marrow Donor Program. Biol Blood Marrow Transplant. 2008;14(Suppl 9):S16-22. 24. Raetz EA, Borowitz MJ, Devidas M, et al. Reinduction platform for children with first marrow relapse of acute lymphoblastic leukemia: a Children's Oncology Group study. J Clin Oncol. 2008;26(24):3971-3978. 25. Barredo JC, Devidas M, Lauer SJ, et al. Isolated CNS relapse of acute lymphoblastic leukemia treated with intensive systemic chemotherapy and delayed CNS radiation: a Pediatric Oncology Group study. J Clin Oncol. 2006;24(19):3142-3149. 26. Barredo JC, Hastings C, Lu X, et al. Isolated late testicular relapse of B-cell acute lymphoblastic leukemia treated with intensive systemic chemotherapy and response-based testicular radiation: a Children's Oncology Group study. J Pediatr Blood Cancer. 2018; 65(5):e26928. 27. Borowitz MJ, Devidas M, Hunger SP, et al. Clinical significance of minimal residual disease in childhood acute lymphoblastic leukemia and its relationship to other prognostic factors: a Children's Oncology Group study. Blood. 2008;111(12):54775485. 28. Borowitz MJ, Pullen DJ, Shuster JJ, et al. Minimal residual disease detection in childhood precursor-B-cell acute lymphoblastic leukemia: relation to other risk factors: a Children's Oncology Group study. Leukemia. 2003;17(8):1566-1572. 29. Borowitz MJ, Wood BL, Devidas M, et al. Prognostic significance of minimal residual disease in high risk B-ALL: a report from Children's Oncology Group study AALL0232. Blood. 2015;126(8):964-971. 30. Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481. 31. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. Br J Cancer. 1977;35(1):1-39. 32. Gray RJ. A class of K-sample tests for com-

haematologica | 2021; 106(1)

Outcomes for relapsed B-ALL from Childrenâ&#x20AC;&#x2122;s Oncology Group study AALL0433

paring the cumulative incidence of a competing risk. Ann Statistics, 1988;16(3):11411154. 33. McCune JS, Lindley C. Appropriateness of maximum-dose guidelines for vincristine. Am J Health Syst Pharm. 1997;54(15):17551758. 34. Diouf B, Crews KR, Lew G, et al. Association of an inherited genetic variant with vincristine-related peripheral neuropathy in children with acute lymphoblastic leukemia. JAMA. 2015;313(8):815-823. 35. Conter V, Bartram CR, Valsecchi MG, et al. Molecular response to treatment redefines all prognostic factors in children and adolescents with B-cell precursor acute lymphoblastic leukemia: results in 3184 patients of the AIEOP-BFM ALL 2000 study. Blood. 2010;115(16):3206-3214. 36. Vora, A, Goulden N, Mitchell C, et al. Augmented post-remission therapy for a minimal residual disease-defined high-risk subgroup of children and young people with clinical standard-risk and intermediate-risk acute lymphoblastic leukaemia (UKALL 2003): a randomised controlled trial. Lancet

haematologica | 2021; 106(1)

Oncol. 2014;15(8):809-818. 37. Eckert C, Groeneveld-Krentz S, KirschnerSchwabe R, et al. Improving stratification for children with late bone marrow B-cell acute lymphoblastic leukemia relapses with refined response classification and integration of genetics. J Clin Oncol. 2019; 37(36):3493-3506. 38. Eckert C, Henze G, Seeger K, et al. Use of allogeneic hematopoietic stem-cell transplantation based on minimal residual disease response improves outcomes for children with relapsed acute lymphoblastic leukemia in the intermediate-risk group. J Clin Oncol. 2013;31(21):2736-2742. 39. Bader P, Kreyenberg H, Henze GH, et al. Prognostic value of minimal residual disease quantification before allogeneic stem-cell transplantation in relapsed childhood acute lymphoblastic leukemia: the ALL-REZ BFM Study Group. J Clin Oncol. 2009;27(3):377384. 40. Pulsipher MA, Langholz B, Wall DA, et al. Risk factors and timing of relapse after allogeneic transplantation in pediatric ALL: for whom and when should interventions be

tested? Bone Marrow Transplant. 2015; 50(9):1173-1179. 41. von Stackelberg A, Locatelli F, Zugmaier G, et al. Phase I/phase II study of blinatumomab in pediatric patients with relapsed/refractory acute lymphoblastic leukemia. J Clin Oncol. 2016; 34(36):43814389. 42. Maude SL, Frey N, Shaw PA, et al. Chimeric antigen receptor T cells for sustained remissions in leukemia. N Engl J Med. 2014; 371(16):1507-1517. 43. Maude SL, Laetsch TW, Buechner J, et al. Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia. N Engl J Med. 2018;378(5):439-448. 44. Masurekar AN, Parker CA, Shanyinde M, et al. Outcome of central nervous system relapses in childhood acute lymphoblastic leukaemia--prospective open cohort analyses of the ALLR3 trial. PLoS One. 2014; 9(10):e108107. 45. Pierro J, Hogan LE, Bhatla T, Carroll WL. New targeted therapies for relapsed pediatric acute lymphoblastic leukemia. Expert Rev Anticancer Ther. 2017;17(8):725-736.

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):56-63

Acute Myeloid Leukemia

Measurable residual disease monitoring provides insufficient lead-time to prevent morphological relapse in the majority of patients with core-binding factor acute myeloid leukemia

Robert Puckrin,1,2 Eshetu G. Atenafu,3 Jaime O. Claudio,1 Steven Chan,1,2 Vikas Gupta,1,2 Dawn Maze,1,2 Caroline McNamara,1,2 Tracy Murphy,1,2 Andre C. Schuh,1,2 Karen Yee,1,2 Hassan Sibai,1,2 Mark D. Minden,1,2 Cuihong Wei,1,4 Tracy Stockley,1,4 Suzanne Kamel-Reid1,4 and Aaron D. Schimmer1,2

Princess Margaret Cancer Centre, University Health Network; 2Department of Medicine, University of Toronto; 3Department of Biostatistics, Princess Margaret Cancer Centre, University Health Network and 4Department of Clinical Laboratory Genetics, Laboratory Medicine Program, University Health Network, Toronto, Canada 1

ABSTRACT

Correspondence: AARON SCHIMMER aaron.schimmer@uhn.ca Received: August 20, 2019 Accepted: January 2, 2020. Pre-published: January 2, 2020. https://doi.org/10.3324/haematol.2019.235721

ore-binding factor acute myeloid leukemia is characterized by t(8;21) or inv(16) and the fusion proteins RUNX1-RUNX1T1 and CBFB-MYH11. International guidelines recommend monitoring for measurable residual disease every 3 months for 2 years after treatment. However, it is not known whether serial molecular monitoring can predict and prevent morphological relapse. We conducted a retrospective singlecenter study of 114 patients in complete remission who underwent molecular monitoring with real-time quantitative polymerase chain reaction analysis of RUNX1-RUNX1T1 or CBFB-MYH11 transcripts every 3 months. Morphological relapse was defined as re-emergence of >5% blasts and molecular relapse as ≥1 log increase in transcript level between two samples. Over a median follow-up time of 3.7 years (range, 0.2-14.3), remission persisted in 71 (62.3%) patients but 43 (37.7%) developed molecular or morphological relapse. Patients who achieved <3 log reduction in RUNX1RUNX1T1 or CBFB-MYH11 transcripts at the end of chemotherapy had a significantly higher risk of relapse compared to patients who achieved ≥3 log reduction (61.1% vs. 33.7%, P=0.004). The majority of relapses (74.4%, n=32) were not predicted by molecular monitoring and occurred rapidly with <100 days from molecular to morphological relapse. Molecular monitoring enabled the detection of impending relapse and permitted pre-emptive intervention prior to morphological relapse in only 11 (25.6%) patients. The current practice of molecular monitoring every 3 months provided insufficient lead-time to identify molecular relapses and prevent morphological relapse in the majority of patients with core-binding factor acute myeloid leukemia treated at our institution. Further research is necessary to determine the optimal monitoring strategies for these patients.

Introduction Core-binding factor (CBF) acute myeloid leukemia (AML) is recognized by the World Health Organization as a subtype of AML with recurrent genetic abnormalities characterized by the chromosomal abnormalities t(8;21) or inv(16)/t(16;16).1 At the molecular level, these chromosomal alterations result in production of the fusion proteins RUNX1-RUNX1T1 in cases with t(8;21) and CBFB-MYH11 in those with inv(16)/t(16;16). CBF-AML accounts for 10-15% of adult AML and tends to be associated with younger patient age and higher sensitivity to induction and consolidation chemotherapy.2,3 Despite a relatively favorable prognosis, up to 40% of patients with CBF-AML relapse.2,4,5 haematologica | 2021; 106(1)

MRD monitoring in CBF-AML

The presence of minimal or measurable residual disease (MRD), defined as the presence of leukemic cells after treatment at levels of 1:104 to 1:106, predicts poor outcomes and impending relapse in AML.6-8 Molecular monitoring for MRD is well-established in hematologic malignancies with clearly defined disease markers, such as BCR-ABL1 in chronic myeloid leukemia and NPM1 in NPM1-mutated AML. With the ability of real-time quantitative polymerase chain reaction (RT-qPCR) to quantify transcript levels of RUNX1-RUNX1T1 and CBFB-MYH11, CBF-AML represents one of the other most well-established targets for MRD monitoring in AML.8 Multiple studies have shown that the results of MRD monitoring can predict relapse in CBF-AML.4,9-18 Given this capacity to identify patients at risk of imminent relapse, the recently published consensus statement from the European LeukemiaNet recommends molecular MRD monitoring for CBF-AML in the peripheral blood (PB) and bone marrow (BM) at diagnosis, after two cycles of induction/consolidation chemotherapy, at the end of treatment, and every 3 months for 24 months of follow-up.6 Presumably, by monitoring patients for rising molecular transcripts, those at risk of impending relapse can be identified and treated with allogeneic bone marrow transplantation (BMT) prior to the emergence of overt disease and the need for re-induction chemotherapy. However, the relapse kinetics of CBF-AML are poorly understood, and it is not known whether serial MRD monitoring can detect molecular relapses with sufficient lead-time to intervene and prevent morphological relapse. Furthermore, it is uncertain if patients’ outcomes can be improved by using MRD status to risk-stratify patients and guide treatment decisions such as escalation to allogeneic BMT. We therefore conducted a study to determine the relapse kinetics and clinical utility of serial MRD monitoring in CBF-AML.

Methods

MRD measurements were generally performed at diagnosis, after induction and the final cycle of consolidation, and then every 3 months during follow-up for 24 months. MRD assessment was routinely performed on BM aspirates, but PB was tested in patients who could not tolerate repeat BM aspirates. A molecular relapse was typically confirmed with a repeat sample 4 weeks later, with referral for allogeneic BMT or administration of chemotherapy if confirmed. Patients who developed morphological relapse were treated with re-induction chemotherapy to achieve second complete remission and referred for allogeneic BMT if a suitable donor was available.19

Study protocol We determined the risk of relapse among patients with CBFAML and the median time from molecular relapse to morphological relapse. We empirically defined rapid relapse as <100 days from molecular to morphological relapse as we considered this timeframe insufficient to directly administer allogeneic BMT to prevent morphological relapse. Morphological complete remission was defined as <5% blasts in BM with recovery of peripheral cell counts and no evidence of extramedullary disease. Morphological relapse was defined as the re-emergence of leukemic blasts in PB, ≥5% blasts in BM, or development of extramedullary disease.20 Complete molecular remission was defined as two successive MRD-negative samples >4 weeks apart, and molecular relapse as an increase of MRD transcript level ≥1 log between two successive samples.6

Statistical analysis

We performed Student t-tests for continuous variables and a c2 test/Fisher exact test for categorical variables to analyze differences between patients with remission or relapse and between those with rapid or slow relapse kinetics. A log-rank test was used to compare relapse-free survival between patients who achieved ≥3 or <3 log reduction in MRD at the end of treatment. Results with a P value <0.05 were considered statistically significant. All analyses were performed using GraphPad statistical software and version 9.4 of the SAS System for Windows (© 2002-2012 SAS Institute, Inc., Cary, NC, USA).

Study design and population

This retrospective study included all patients ≥18 years old with CBF-AML at Princess Margaret Cancer Centre in Toronto, Canada between January 2000 and 2017. Patients were excluded if they did not have MRD monitoring performed. This study was approved by the institutional Research Ethics Board.

Treatment and measurable residual disease monitoring protocol The standard treatment protocol for eligible patients with CBFAML at our institution during the study period was induction chemotherapy with daunorubicin for 3 days and continuous infusion of cytarabine for 7 days. Patients who achieved complete remission received three cycles of consolidation with high-dose cytarabine along with daunorubicin during the first cycle of consolidation (see Online Supplementary Methods). Allogeneic BMT was not routinely recommended unless there was molecular progression during follow-up or if risk factors were present such as t(8;21) with C-KIT mutation.19 Patients underwent MRD monitoring using RT-qPCR of RUNX1-RUNX1T1 or CBFB-MYH11 fusion transcripts in an accredited (ISO15189) laboratory at the University Health Network (see Online Supplementary Methods). Each patient’s own diagnostic level was used as baseline for calculating log reduction. The detection limit for these assays was 1 in 10,000 throughout the study period. haematologica | 2021; 106(1)

Results Study population We identified 206 patients with CBF-AML who were evaluated at our institution during the study period. Of these 206 patients, 114 had MRD monitoring performed and were included in our analysis. We excluded 92 patients who did not have MRD monitoring available because of death during treatment (n=19), administration of non-intensive or supportive care only (n=12), allogeneic BMT performed in first complete remission (n=17), MRD monitoring conducted at an outside institution (n=24), or for unspecified reasons (n=20) (Figure 1). The median age of the 114 patients was 46.5 years (range, 18-79) and 44.7% were female. The t(8;21) was present in 58.8% and inv(16)/t(16;16) in 41.2% of patients. Over a median follow-up time of 3.7 years (range, 0.214.3), MRD measurements were performed a mean (± standard deviation, SD) 4.9±3.2 times per patient with a median (range) sampling interval of 98 (0-378) days. A total of 564 MRD measurements were performed, of which 77.5% within 120 days of the previous MRD sample. Most MRD measurements were performed on BM samples and 13.7% were performed on PB. 57

R. Puckrin et al.

Figure 1. Flow diagram of patients with core-binding factor acute myeloid leukemia. CBF-AML: core-binding factor acute myeloid leukemia; MRD: measurable residual disease, BMT: bone marrow transplantation; CR1: first complete remission.

Risk of relapse All patients except one (99.1%) achieved complete remission with standard induction chemotherapy; the single patient with refractory disease achieved complete remission with re-induction chemotherapy. Long-term remission was maintained in 71 (62.3%) patients but 43 (37.7%) relapsed, of whom 34 developed morphological relapse and nine had isolated molecular relapse but did not progress to morphological relapse because of pre-emptive treatment with allogeneic BMT or death from complications of treatment or disease. Most relapses occurred early with a median time from complete remission to molecular or morphological relapse of 9.1 months (range, 1.6-38.6); only 2/43 (4.7%) relapses occurred >2 years after the achievement of complete remission. The risk of relapse was significantly higher in patients who achieved <3 log reduction in RUNX1-RUNX1T1 or CBFB-MYH11 transcripts at the end of consolidation chemotherapy compared to patients who achieved â&#x2030;Ľ3 log reduction (61.1% vs. 33.7%, P=0.004) (Figure 2). Patients with <4 log reduction in MRD transcripts at the end of treatment also had a higher risk of relapse compared to patients with â&#x2030;Ľ4 log reduction (51.2% vs. 29.3%, P=0.026). Furthermore, relapsed patients were significantly less likely to have achieved MRD-negative status at the end of chemotherapy (23.1% vs. 45.2%, P=0.025) or complete molecular remission during follow-up (7.0% vs. 74.6%, P<0.0001), compared to patients who did not relapse (Table 1). There was no significant association between risk of relapse and MRD log reduction at the end of induction or with other clinical, cytogenetic, or genetic variables. There was a 58

trend, which bordered on statistical significance, towards more extramedullary disease in patients with relapse (Table 1). Of the 71 patients who did not relapse, 28 (39.4%) had persistent molecular disease during follow-up with detectable MRD at low transcript levels but with <1 log increase between two successive positive samples. Persistent molecular disease continued for a median of 1.5 years and up to a maximum of 3.4 years from the time of complete remission. Of these 28 patients, 13 eventually achieved MRD-negative status and 15 had persistent molecular disease to the end of follow-up MRD monitoring.

Relapse kinetics Of the 43 patients who relapsed, the majority (74.4%, n=32) had rapid relapse kinetics with <100 days from molecular to morphological relapse. Of the 32 patients with rapid relapse kinetics, 17 (53.1%) had simultaneous molecular and morphological relapse, meaning that they progressed from a MRD-negative status (n=3) or stable levels of MRD (n=14) on one sample to morphological relapse on the subsequent sample. The remaining 15/32 (46.9%) developed molecular relapse prior to morphological relapse, but progression occurred rapidly over a median of 48 days (range, 11-95) such that no pre-emptive therapy, including allogeneic BMT, could be administered. Of the patients with rapid relapse kinetics, 16 (50.0%) received re-induction chemotherapy at the time of morphological relapse followed by allogeneic BMT. Eight patients (25.0%) received additional chemotherapy but haematologica | 2021; 106(1)

MRD monitoring in CBF-AML

Figure 2. Kaplan-Meier analysis of relapse-free survival, according to reduction in measurable residual disease at end of chemotherapy. MRD: measurable residual disease.

did not proceed to allogeneic BMT and died from complications of chemotherapy or disease. Five (15.6%) patients died without further chemotherapy and three (9.4%) were lost to follow-up. MRD monitoring enabled timely detection of molecular relapse and permitted intervention prior to morphological relapse in only 11 patients (25.6%). Of these patients with slower relapse kinetics, eight (72.7%) went on to receive allogeneic BMT and three (27.3%) received additional chemotherapy but died from complications of chemotherapy or disease prior to allogeneic BMT. Pre-emptive intervention with allogeneic BMT at the time of molecular relapse prevented the development of morphological relapse in 6/11 (54.5%) patients. The median overall survival was 33.6 months for patients with slower relapse kinetics compared to 21.7 months for patients with rapid relapse kinetics (P=0.070). There were no clinical features, molecular mutations, or cytogenetic abnormalities that significantly predicted patients with rapid versus slower relapse kinetics (Table 2). Furthermore, our inability to predict most relapses did not appear to be due to differences in MRD sampling technique, as there was no significant difference between patients with rapid and slower relapse kinetics in terms of the median time interval between MRD samples (70.5 vs. 70.0 days, P=0.74), the proportion of delayed MRD measurements (7.1% vs. 12.5%, P=0.35), or the number of MRD tests performed on PB (10.0% vs. 8.3%, P=1.0). Among patients who had a rise in MRD transcripts detected prior to the emergence of morphological relapse, a confirmatory sample was obtained within 4 weeks in 8/17 (47.1%) cases in the rapid relapse group (median 31 days; range, 3-91) and in 5/8 (62.5%) cases in the slower relapse group (median 28 days; range, 8-185), with no significant difference in time to confirmatory sample between the two groups (P=0.68). The time period during which relapses occurred was also similar between the two haematologica | 2021; 106(1)

groups, with 20/32 (62.5%) relapses occurring after 2010 in the rapid relapse group compared to 5/11 (45.5%) in the slower relapse group (P=0.48). Finally, a secondary analysis restricted to patients who achieved an optimal molecular response (defined here as â&#x2030;Ľ3 log reduction in MRD at the end of chemotherapy) yielded similar findings as our primary analysis. The majority of these patients (67.9%, n=19) experienced rapid progression from molecular to morphological relapse (median 11 days; range, 0-95), and MRD monitoring enabled the timely detection of impending relapse in only 9/28 (32.1%) of patients with an optimal molecular response.

Discussion Although CBF-AML is associated with a higher rate of complete remission and a relatively favorable prognosis, up to 40% of patients develop relapse. Current international guidelines for CBF-AML recommend MRD monitoring every 3 months for 2 years after remission. This practice is intended to identify patients at risk of recurrence by detecting impending relapse and, presumably, to enable early intervention to prevent morphological relapse. However, our study demonstrates that this monitoring strategy fails to detect the majority of relapsing patients in a timely manner, as 74.4% of patients at our institution developed rapid morphological relapses such that no pre-emptive therapy, including allogeneic BMT, could be administered. In almost half these patients, molecular relapse occurred simultaneously with morphological relapse, while in the remaining cases the progression from molecular to morphological relapse occurred rapidly over a median of only 48 days. The current practice of measuring MRD every 3 months during follow-up therefore appears to provide insufficient lead-time to iden59

R. Puckrin et al.

tify molecular relapses and intervene prior to morphological relapse. As a consequence, 114 patients in our study underwent serial MRD monitoring with BM sampling every 3 months for 24 months, but only 11 patients (9.6%) experienced a direct change in clinical management from this practice by receiving pre-emptive treatment prior to morphological relapse. This suggests that MRD monitoring may be of rare clinical utility in the follow-up of patients with CBF-AML. The results of our study are significant given the lack of large longitudinal studies examining the clinical utility of serial MRD monitoring in CBF-AML. Our results are consistent with the long-term follow-up reported from the UK MRC AML-15 trial, in which serial MRD monitoring failed to predict impending relapse in 42/71 (59.2%) patients with morphological relapse.11 However, the authors of that study attributed the inability to predict impending relapse to infrequent MRD measurements, as

sampling intervals were >3 months in many patients. This contrasts with the generally good MRD monitoring adherence in our population of patients, among whom the median MRD sampling interval was 98 days. However, a limitation of our study is its real-world retrospective design, and we cannot exclude that delays in MRD sampling or variations in specimen quality may have limited our ability to predict impending relapse in some patients. Furthermore, the majority of relapses in our study occurred early at a median of 9 months after remission, which may have further limited the ability of serial MRD monitoring to predict relapse. Other studies have found serial MRD monitoring to be more strongly predictive of impending relapse. In a follow-up analysis of 94 patients enrolled in the French CBF-2006 trial who underwent MRD monitoring of PB every 3 months after remission, morphological relapse was predicted by persistent MRD positivity or molecular relapse in 21/29

Table 1. Analysis of patients with long-term remission versus relapse.

Variable Clinical variables Sex, n (%) Age at diagnosis, years WBC count at diagnosis, x109/L BM blasts at diagnosis; % AML diagnosis, n (%)

Female Median (range) Median (Range) Median (Range) De novo Secondary Yes Yes

CNS disease, n (%) Extramedullary disease, n (%) Cytogenetic abnormalities, n (%) inv(16) Yes t(8;21) Yes del(X) Yes del(Y) Yes del(9) Yes tri(8) Yes tri(9) Yes tri(21) Yes tri(22) Yes ≥3 cytogenetic abnormalities Yes Genetic mutations, n (%) C-KIT mutation Yes NPM1 mutation Yes FLT3-ITD mutation Yes FLT3-TKD mutation Yes Outcomes CR to induction, n (%) Yes CR to consolidation, n (%) Yes Morphological relapse, n (%) Yes Duration of CR, months Median (range) Death Yes MRD status End of induction MRD ≥3 log reduction, n (%) Yes End of consolidation MRD ≥3 log reduction, n (%) Yes End of consolidation MRD ≥4 log reduction, n (%) Yes End of consolidation MRD negative, n (%) Yes Complete molecular remission during follow-up, n (%) Yes Number of MRD tests per patient, n (%) Mean (SD) Number of MRD tests performed on PB, n (%) Sum Time between MRD tests, days Median (range) Number of MRD tests performed >120 days apart, n (%) Sum

Remission (n=71)

Relapse (n=43)

P value

31 (43.7%) 47 (18-72) 8.0 (0.90-360) 53.5 (12-95) 65 (91.5%) 5 (7.1%) 1 (1.4%) 4 (5.6%)

20 (46.5%) 45 (19-79) 9.9 (1.3-341) 67 (15-100) 38 (88.4%) 5 (11.6%) 3 (7.0%) 8 (18.6%)

0.77 0.83 0.24 0.35 0.50 0.15 0.055

32 (45.1%) 39 (54.9%) 4 (5.6%) 14 (19.7%) 5 (7.0%) 8 (11.3%) 3 (4.2%) 3 (4.2%) 6 (8.5%) 14 (19.7%)

15 (34.9%) 28 (65.1%) 6 (14%) 8 (18.6%) 2 (4.7%) 2 (4.7%) 0 (0%) 1 (2.3%) 0 (0%) 5 (11.6%)

0.28 0.28 0.17 0.88 0.71 0.32 0.29 1.00 0.082 0.26

7/39 (17.9%) 0/2 (0%) 1/2 (50%) 0/2 (0%)

7/30 (23.3%) 0/4 (0%) 0/4 (0%) 1/4 (25%)

0.58 NA 0.33 1.00

71 (100%) 71 (100%) 0 (0) 87.1 (10.2-173.4) 3 (4.2%)

42 (97.7%) 39 (90.7%) 34 (79.1%) 9.1 (1.6-38.6) 30 (69.8%)

0.38 0.02 <0.0001 <0.0001 <0.0001

25/42 (59.5%) 55/62 (88.7%) 41/62 (66.1%) 28/62 (45.2%) 53/71 (74.6%) 6.3 (3.1) 66/446 (14.8%) 109 (26.3-378) 116/446 (26.0%)

15/33 (45.5%) 28/39 (71.8%) 17/39 (43.6%) 9/39 (23.1%) 3/43 (7.0%) 2.7 (2.2) 11/118 (9.3%) 70 (0-164) 11/118 (9.3%)

0.23 0.031 0.038 0.025 <0.0001 <0.0001 0.13 <0.0001 <0.0001

WBC: white blood cell count; BM: bone marrow; AML: acute myeloid leukemia; CNS: central nervous system; CR: complete remission; NA: not available; ITD: internal tandem duplication; TKD: tyrosine kinase domain; MRD: measurable residual disease; PB: peripheral blood; SD: standard deviation.

haematologica | 2021; 106(1)

MRD monitoring in CBF-AML

(72.4%) cases.21 However, it is unclear if the timelier detection of impending relapse in that study translated into early intervention and improved clinical outcomes, as the cumulative incidence of morphological relapse remained high at 74.5% among the 23 patients with molecular relapse. This issue is further complicated by incomplete understanding of the relapse kinetics and optimal sampling interval of MRD measurements in CBF-AML. The current recommendation of MRD monitoring every 3 months is derived from the UK MRC AML-15 trial which reported a median time from molecular relapse in the BM to mor-

phological relapse of 4.9 months in cases with t(8;21) and 3 months in those with inv(16).11 Similarly, the CBF-2006 trial demonstrated a median time from molecular to morphological relapse of 3.9 months.21 However, other studies have described slower relapse kinetics with a median time from molecular to morphological relapse of up to 6 months or even 1 year in patients with inv(16).22,23 Alternatively, other analyses are consistent with the rapid relapse kinetics observed in our study, with reported median times from molecular to morphological relapse of 35 to 60 days in small studies.24,25 The discordant relapse kinetics across studies may be attributable to differences

Table 2. Analysis of patients with rapid versus slow relapse kinetics.

Variable Clinical variables Sex, n (%) Age at diagnosis, years WBC count at diagnosis, x109/L BM blasts at diagnosis, % Diagnosis, n (%) CNS disease, n (%) Extramedullary disease, n (%) Cytogenetic abnormalities, n (%) inv(16) t(8;21) del(X) del(Y) del(9) tri(8) tri(9) tri(21) tri(22) ≥3 cytogenetic abnormalities Genetic mutations, n (%) CKIT mutation NPM1 mutation FLT3-ITD mutation FLT3-TKD mutation Treatment Induction chemotherapy, n (%) Consolidation chemotherapy, median (range) Allogeneic BMT Outcomes CR to induction, n (%) CR to consolidation, n (%) Morphologic relapse, n (%) Molecular relapse before morphological relapse, n (%) Duration of CR, months Death, n (%) MRD status End of induction MRD ≥3 log reduction, n (%) End of consolidation MRD ≥3 log reduction, n (%) End of consolidation MRD negative, n (%) Number of MRD tests/patient Number of MRD tests performed on PB, n (%) Sampling interval between MRD testing, days Number of MRD tests performed >120 days apart, n (%)

Level

Rapid relapse (n=32)

Slow relapse (n=11)

P value

Female Median (Range) Median (Range) Median (Range) De novo AML Secondary AML Yes Yes

15 (46.9%) 45 (19-79) 9.8 (1.3-198) 67 (15-93) 27 (84.4%) 5 (15.6%) 1 (3.1%) 5 (15.6%)

5 (45.5%) 41 (24-59) 10.1 (3.7-341) 68 (20-100) 11 (100%) 0 (0%) 2 (18.2%) 3 (27.3%)

0.94 0.64 0.27 0.35 0.31

Present Present Present Present Present Present Present Present Present Present

9 (28.1%) 23 (71.9%) 4 (12.5%) 7 (21.9%) 1 (3.1%) 2 (6.3%) 0 (0%) 1 (3.1%) 0 (0%) 4 (12.5%)

6 (54.6%) 5 (45.5%) 2 (18.2%) 1 (9.1%) 1 (9.1%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (9.1%)

0.15 0.15 0.64 0.66 0.45 1.00 NA 1.00 NA 1.00

Present Present Present Present

6/24 (25%) 0/3 (0%) 0/3 (0%) 0/3 (0%)

1/6 (16.7%) 0/1 (0%) 0/1 (0%) 1/1 (100%)

1.00 NA NA 0.25

7+3 Cycles Performed

32 (100%) 3 (1-4) 17 (53.1%)

11 (100%) 3 (2-4) 8 (72.7%)

NA 0.41 0.31

Yes Yes Yes

31 (96.9%) 28 (87.5%) 32 (100%)

11 (100%) 11 (100%) 2 (18.2%)

1.00 0.56 <0.0001

Yes Median (range) Yes

15 (46.9%) 9.0 (1.6-34.7) 24 (75.0%)

11 (100%) 9.8 (7.6-38.6) 6 (54.5%)

0.001 0.22 0.26

Yes Yes Yes Mean (SD) Sum Median (range)

11/25 (44.0%) 19/29 (65.5%) 6/29 (20.7%) 2.2 (1.7) 7/70 (10.0%) 70.5 (0-126)

4/8 (50.0%) 9/10 (90.0%) 3/10 (30.0%) 4.3 (2.7) 4/48 (8.3%) 70 (46-164)

1.00 0.23 0.67 0.004 1.00 0.74

Sum

5/70 (7.1%)

6/48 (12.5%)

0.35

0.16 0.40

haematologica | 2021; 106(1)

R. Puckrin et al.

in sampling source (i.e., PB or BM), variable sensitivity of individual RT-qPCR assays, or differing MRD sampling intervals.26 Individual patient-level factors may also contribute to differences in relapse kinetics, although our study did not identify any clinical variables significantly predictive of rapid relapse kinetics. Additional research is therefore warranted to better characterize the relapse kinetics of CBF-AML and to determine whether more frequent MRD sampling (e.g., PB every 4-8 weeks) might be more predictive of impending relapse. An additional area of uncertainty and interest is whether serial MRD monitoring can be used to inform treatment decisions. Currently, the European LeukemiaNet does not formally recommend changing therapy based on MRD status in CBF-AML due to lack of supportive data.6 However, the AML05 trial demonstrated that utilizing MRD status to identify high-risk patients with t(8;21) (i.e., those with <3 log reduction in MRD after the second cycle of consolidation or loss of molecular response within 6 months) as candidates for escalation of therapy to allogeneic BMT in first complete remission may lead to improved clinical outcomes.27 The GIMEMA AML1310 study also provides support for a risk-adapted, MRD-directed treatment strategy, as patients with intermediate-risk AML who were MRDpositive at the end of the first cycle of consolidation chemotherapy were referred for allogeneic BMT and achieved similar disease-free and overall survival rates as those with favorable-risk AML.28 In the present study, preemptive intervention with allogeneic BMT at the time of molecular relapse prevented the development of morphological relapse in 6/11 patients with slower relapse kinetics. However, this finding requires confirmation in other populations of patients given the relatively small size and retrospective design of our study. The role for pre-emptive chemotherapy or allogeneic BMT in CBF-AML is also uncertain because most patients with morphological relapse respond well to salvage treatments.26 Future studies should examine the clinical outcomes of different treatment strategies (e.g., observation or pre-emptive chemotherapy, immunotherapy, or BMT) for patients with CBF-AML in molecular relapse. Although our study raises questions about the utility of serial MRD monitoring during follow-up after remission, we did identify the prognostic importance of measuring MRD at the end of treatment. In our study population, achievement of â&#x2030;Ľ3 log reduction in RUNX1-RUNX1T1 or CBFB-MYH11 transcripts at this time-point was associated with a significantly lower risk of relapse. This is consistent with the findings of other trials, including the UK MRC AML-15, AML05, and CBF-2006 trials, in which

References 1. Arber DA, Orazi A, Hasserjian R, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood. 2016;127(20):2391-2405. 2. Grimwade D, Hills RK, Moorman AV, et al. Refinement of cytogenetic classification in acute myeloid leukemia: determination of prognostic significance of rare recurring chromosomal abnormalities among 5876 younger adult patients treated in the

achievement of deep reductions in MRD transcript levels after induction or consolidation chemotherapy was associated with improved relapse-free survival.11,12,27,29 Our study did not analyze the prognostic significance of measuring MRD after the second cycle of chemotherapy as recommended by the European LeukemiaNet, as the local institutional policy during this study period recommended measuring MRD after induction and the final cycle of consolidation chemotherapy. Other studies have identified additional potential prognostic factors that may predict relapse and poor outcomes in CBF-AML, including older age, elevated white blood cell count at diagnosis, extramedullary disease, secondary AML, additional cytogenetic changes (e.g., loss of chromosome X or Y, trisomy 22, deletion of 9q), and genetic mutations (e.g., KIT, FLT3, RAS, ASXL).3 In conclusion, measuring MRD at the end of treatment is predictive of relapse risk in CBF-AML. However, the current guideline recommendation of MRD monitoring every 3 months during follow-up failed to detect the majority of molecular relapses with sufficient lead-time to intervene and prevent morphological relapse in our study population. Further research is warranted to characterize the relapse kinetics of CBF-AML and to identify the patients at highest risk of relapse and the optimal strategies to monitor these patients over time. Disclosures ADS has received honorarium from Novartis, Jazz, and Otsuka Pharmaceuticals and research support from Medivir AB and Takeda. A.D.S. is named as an inventor on a patent application related to the use of DNT cells in AML. A.D.S. owns stock in AbbVie Pharmaceuticals. VG received honorarium, research funding through institution, and served on advisory board of Novartis and Incyte. CM has received honorarium from Novartis. Contributions RP performed research, performed data analysis, and wrote the manuscript. EA performed data analysis. JO, SC, VG, DM, CM, TM, AS, KY, HS, MM, CW, TS, and SK contributed to performing research and writing the manuscript. AS supervised the study and contributed to writing the manuscript. Acknowledgments We thank Jill Flewelling (Princess Margaret Cancer Center) for administrative assistance. This work was supported by the Canadian Institutes of Health Research, the Princess Margaret Cancer Centre, The Princess Margaret Cancer Foundation, and the Ontario Ministry of Health. ADS holds the Ronald N. Buick Chair in Oncology Research.

United Kingdom Medical Research Council trials. Blood. 2010;116(3):354-365. 3. Duployez N, Willekens C, Marceau-Renaut A, et al. Prognosis and monitoring of corebinding factor acute myeloid leukemia: current and emerging factors. Expert Rev Hematol. 2015;8(1):43-56. 4. Ustun C and Marcucci G. Emerging diagnostic and therapeutic approaches in core binding factor acute myeloid leukaemia. Curr Opin Hematol. 2015;22(2):85-91. 5. Schlenk RF, Benner A, Krauter J, et al. Individual patient data-based meta-analysis of patients aged 16 to 60 years with core

binding factor acute myeloid leukemia: a survey of the German Acute Myeloid Leukemia Intergroup. J Clin Oncol. 2004;22(18):3741-3750. 6. Schuurhuis GJ, Heuser M, Freeman S, et al. Minimal/measurable residual disease in AML: a consensus document from the European LeukemiaNet MRD Working Party. Blood. 2018;131(12): 1275-1291. 7. Paietta E. Minimal residual disease in acute myeloid leukemia: coming of age. Hematology Am Soc Hematol Educ Program. 2012;2012:35-42. 8. Ofran Y, Rowe JM. Introducing minimal

haematologica | 2021; 106(1)

MRD monitoring in CBF-AML

residual disease in acute myeloid leukemia. Curr Opin Hematol. 2015;22(2): 139-145. 9. Deotare U, Shaheen M, Brandwein JM, et al. Predictive value of molecular remissions postconsolidation chemotherapy in patients with core binding factor acute myeloid leukemia (CBF-AML) - a single center analysis. Hematol Oncol. 2016;35(4): 810-813. 10. Hoyos M, Nomdedeu JF, Esteve J, et al. Core binding factor acute myeloid leukemia: the impact of age, leukocyte count, molecular findings, and minimal residual disease. Eur J Haematol. 2013;91(3):209-218. 11. Yin JA, Oâ&#x20AC;&#x2122;Brien MA, Hills RK, et al. Minimal residual disease monitoring by quantitative RT-PCR in core binding factor AML allows risk stratification and predicts relapse: results of the United Kingdom MRC AML-15 trial. Blood. 2012;120(14): 2826-2835. 12. Jourdan E, Boissel N, Chevret S, et al. Prospective evaluation of gene mutations and minimal residual disease in patients with core binding factor acute myeloid leukemia. Blood. 2013;121(12):2213-2223. 13. Zhang L, Li Q, Li W, et al. Monitoring of minimal residual disease in acute myeloid leukemia with t(8;21)(q22;q22). Int J Hematol. 2013;97(6):786-792. 14. Guieze R, Renneville A, Cayuela JM, et al. Prognostic value of minimal residual disease by real-time quantitative PCR in acute myeloid leukemia with CBFB-MYH11 rearrangement: the French experience. Leukemia. 2010;24(7):1386-1388. 15. Corbacioglu A, Scholl C, Schlenk RF, et al. Prognostic impact of minimal residual disease in CBFB-MYH11-positive acute myeloid leukemia. J Clin Oncol. 2010;28(23):3724-3729. 16. Markova J, Markova J, Trnkova Z, et al. Monitoring of minimal residual disease in

haematologica | 2021; 106(1)

patients with core binding factor acute myeloid leukemia and the impact of C-KIT, FLT3, and JAK2 mutations on clinical outcome. Leuk Lymphoma. 2009;50(9):14481460. 17. Weisser M, Haferlach C, Hiddeman W, et al. The quality of molecular response to chemotherapy is predictive for the outcome of AML1-ETO-positive AML and is independent of pretreatment risk factors. Leukemia. 2007;21(6):1177-1182. 18. Krauter J, Gorlich K, Ottmann O, et al. Prognostic value of minimal residual disease quantification by real-time reverse transcriptase polymerase chain reaction in patients with core binding factor leukemias. J Clin Oncol. 2003;21(23):4413-4422. 19. Leukemia Site Group. Acute myeloid leukemia. Princess Margaret Cancer Centre Clinical Practice Guidelines. 2015; https://www.uhn.ca/PrincessMargaret/He alth_Professionals/Programs_Departments /Leukemia/Documents/CPG_Leukemia_A cuteMyeloidLeukemia.pdf [Accessed September 8, 2020] 20. Cheson BD, Bennet JM, Kopecky KJ, et al. Revised Recommendations of the International Working Group for Diagnosis, Standardization of Response Criteria, Treatment Outcomes, and Reporting Standards for Therapeutic Trials in Acute Myeloid Leukemia. J Clin Oncol. 2003;21 (24):4642-4649. 21. Willekens C, Blanchet O, Renneville A, et al. Prospective long-term minimal residual disease monitoring using RQ-PCR in RUNX1RUNX1T1-positive acute myeloid leukemia: results of the French CBF-2006 trial. Haematologica. 2016;101(3):328-335. 22. Ommen HB, Schnittger S, Jovanovic JV, et al. Strikingly different molecular relapse kinetics in NPM1c, PML-RARA, RUNX1RUNX1T1, and CBFB-MYH11 acute myeloid leukemias. Blood. 2010;115(2):198-

205. 23. Stentoft J, Hokland P, Ostergaard M, et al. Minimal residual core binding factor AMLs by real time quantitative PCR--initial response to chemotherapy predicts event free survival and close monitoring of peripheral blood unravels the kinetics of relapse. Leuk Res. 2006;30(4):389-395. 24. Doubek M, Palasek I, Pospisil Z, et al. Detection and treatment of molecular relapse in acute myeloid leukemia with RUNX1 (AML1), CBFB, or MLL gene translocations: frequent quantitative monitoring of molecular markers in different compartments and correlation with WT1 gene expression. Exp Hematol. 2009;37(6): 659-672. 25. Lane S, Saal R, Mollee P, et al. A >or=1 log rise in RQ-PCR transcript levels defines molecular relapse in core binding factor acute myeloid leukemia and predicts subsequent morphologic relapse. Leuk Lymphoma. 2008;49(3):517-523. 26. Kayser S, Walter RB, Stock W, et al. Minimal residual disease in acute myeloid leukemia-current status and future perspectives. Curr Hematol Malig Rep. 2015;10(2):132-44. 27. Zhu HH, Zhang XH, Qin YZ, et al. MRDdirected risk stratification treatment may improve outcomes of t(8;21) AML in the first complete remission: results from the AML05 multicenter trial. Blood. 2013;121 (20):4056-4062. 28. Venditti A, Piciocchi A, Candoni A, et al. GIMEMA AML1310 trial of risk-adapted, MRD-directed therapy for young adults with newly diagnosed acute myeloid leukemia. Blood. 2019;132(12):935-945. 29. Agrawal M, Corbacioglu A, Paschka P, et al. Minimal residual disease monitoring in acute myeloid leukemia (AML) with translocation t(8;21)(q22;q22): results of the AML Study Group (AMLSG) [Abstract]. Blood. 2016;128(22):1207.

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):64-73

Bone Marrow Failure

Utility of clinical comprehensive genomic characterization for diagnostic categorization in patients presenting with hypocellular bone marrow failure syndromes Piers Blombery,1,2,3 Lucy C. Fox,3,4,5* Georgina L. Ryland,3* Ella R. Thompson,2,3 Jennifer Lickiss,3 Michelle McBean,3 Satwica Yerneni,3 David Hughes,6 Anthea Greenway,6 Francoise Mechinaud,6 Erica M. Wood,5 Graham J. Lieschke,1,7 Jeff Szer,1 Pasquale Barbaro,8 John Roy,8 Joel Wight,9 Elly Lynch,10,11,12 Melissa Martyn,10,11,12 Clara Gaff2,10,12 and David Ritchie1

Clinical Hematology, Peter MacCallum Cancer Center/Royal Melbourne Hospital, Melbourne, Victoria; 2University of Melbourne, Melbourne, Victoria; 3Department of Pathology, Peter MacCallum Cancer Center, Melbourne, Victoria; 4Epworth Healthcare, Melbourne, Victoria; 5Transfusion Research Unit, School of Public Health & Preventive Medicine, Monash University, Melbourne, Victoria; 6Royal Children’s Hospital, Melbourne, Victoria; 7Australian Regenerative Medicine Institute, Monash University, Melbourne, Victoria; 8 Children’s Health Queensland and University of Queensland, South Brisbane, Queensland; 9 Department of Hematology, Austin Health, Melbourne, Victoria; 10Melbourne Genomics Health Alliance, Melbourne, Victoria; 11Victorian Clinical Genetics Service, Melbourne, Victoria and 12Murdoch Children’s Research Institute, Melbourne, Victoria, Australia 1

*LCF and GLR contributed equally as co-second authors

ABSTRACT

Correspondence: PIERS BLOMBERY piers.blombery@petermac.org Received: September 6, 2019. Accepted: February 7, 2020. Pre-published: February 13, 2020. https://doi.org/10.3324/haematol.2019.237693

one marrow failure (BMF) related to hypoplasia of hematopoietic elements in the bone marrow is a heterogeneous clinical entity with a broad differential diagnosis including both inherited and acquired causes. Accurate diagnostic categorization is critical to optimal patient care and detection of genomic variants in these patients may provide this important diagnostic and prognostic information. We performed real-time, accredited (ISO15189) comprehensive genomic characterization including targeted sequencing and whole exome sequencing in 115 patients with BMF syndromes (median age 24 years, range: 3 months - 81 years). In patients with clinical diagnoses of inherited BMF syndromes, acquired BMF syndromes or clinically unclassifiable BMF we detected variants in 52% (12 of 23), 53% (25 of 47) and 56% (25 of 45) respectively. Genomic characterization resulted in a change of diagnosis in 30 of 115 (26%) including the identification of germline causes for 3 of 47 and 16 of 45 cases with pre-test diagnoses of acquired and clinically unclassifiable BMF respectively. The observed clinical impact of accurate diagnostic categorization included choice to perform allogeneic stem cell transplantation, disease-specific targeted treatments, identification of at-risk family members and influence of sibling allogeneic stem cell donor choice. Multiple novel pathogenic variants and copy number changes were identified in our cohort including in TERT, FANCA, RPS7 and SAMD9. Whole exome sequence analysis facilitated the identification of variants in two genes not typically associated with a primary clinical manifestation of BMF but also demonstrated reduced sensitivity for detecting low level acquired variants. In conclusion, genomic characterization can improve diagnostic categorization of patients presenting with hypoplastic BMF syndromes and should be routinely performed in this group of patients.

Introduction Bone marrow failure (BMF) related to hypoplasia of hematopoietic elements in the bone marrow is a heterogeneous clinical entity with a broad differential diagnosis. BMF is most commonly an acquired disorder related to an exogenous marhaematologica | 2021; 106(1)

Genomic characterization of BMF syndrome patients

row insult such as drugs (e.g., cytotoxic chemotherapy), toxins or immune destruction of hemopoietic progenitors (acquired aplastic anemia [aAA]). Less commonly, BMF may be the result of an inherited/genetic cause such as Fanconi anemia (FA), Shwachman-Diamond syndrome (SDS), Diamond-Blackfan anemia (DBA) and other rare genetic disorders. Moreover, myelodysplastic syndrome (MDS) can present in the minority of cases (approximately 10%) with marrow hypoplasia (hMDS) rather than the typical hypercellular form with the notable exception that most children and adolescents with MDS will be categorized as refractory cytopenia of childhood (RCC) with the majority of these cases being hypocellular.1 An accurate diagnosis of the precise etiology and subtype of BMF is critical to providing appropriate clinical care. However, the clinical and morphological features of the bone marrow biopsy of the various BMF entities (both acquired and genetic) are non-specific with significant overlap limiting the ability to differentiate the underlying type of BMF. Whilst investigations such as telomere length assessment and chromosomal fragility testing can be helpful in suggesting an underlying diagnosis (of dyskeratosis congenita [DKC] and FA respectively), these techniques have significant technical limitations.2,3 In addition, they do not provide the clinical utility of knowing a specific causative genetic variant that can provide extra information regarding risk stratification, prediction of inheritance patterns and screening of asymptomatic/pre-symptomatic family members. Genomic evaluation using next generation sequencing (NGS) is a powerful technique to identify variants in a large number of genes in order to resolve the diagnostic and therapeutic challenges faced by physicians treating BMF. Indeed, literature to date suggests a significant contribution to diagnostic rate and reclassification using NGS in patients with inherited/genetic BMF syndromes (IBMFS).4 Moreover, in the acquired setting, detection of variants in patients with aAA may inform of the risk of transformation to hematological malignancy and response to immunosuppression5,6 as well as aiding the diagnostic categorization between aAA and hMDS.4 In some diagnostic laboratories, detection of germline variants is now being performed by whole exome sequencing (WES) as a single approach applicable to a broad range of phenotypes, whereas targeted panel testing is more typically used in the acquired/hematological malignancy setting due to a requirement for greater sensitivity to detect subclonal variants. Here we describe the results of the Melbourne Genomic Health Alliance Bone Marrow Failure Flagship a multi-institutional study of 115 patients which aimed to prospectively assess the diagnostic impact of comprehensive genomic evaluation including both WES and targeted NGS panel testing in patients presenting with BMF characterized by bone marrow hypoplasia.

Methods Recruitment and ethics Eligible patients were recruited from four participating institutions in Victoria, Australia. Inclusion criteria were (i) age ≥3 months, (ii) a clinicopathological diagnosis (based on current guidelines)1,8 of either aAA, IBMFS, hMDS or a BMF syndrome characterized by marrow hypoplasia/aplasia but not able to be haematologica | 2021; 106(1)

definitively categorized. All patients received pre-test counselling and assessment. Standardized pre-test counselling was provided by the treating hematologist or participants were referred to the study hematologist for the purpose of the study. Participants received written information in the form of the Patient Information Consent Form (PICF). Genetic counsellors and clinical geneticists were available at each of the study sites for referral if required. Research was conducted after Institutional Review Board ethics approval (HREC/13/MH/326 and HREC/17/PMCC/163) and all research was conducted in accordance with the Declaration of Helsinki.

Genomic characterization All patients underwent baseline testing with (i) WES and (ii) targeted panel sequencing with a hybridization-based NGS BMF panel (a subset of the Peter MacCallum Cancer Center PanHaem panel).9 Sequence variants alone were analyzed from WES and both sequence variants as well as genome-wide copy number changes were assessed from the targeted BMF panel. Genes analyzed by either WES or targeted panel sequencing are shown in Figure 1 and listed in the Online Supplementary Table S1. Details of bioinformatic processing of sequence data are provided in the Online Supplementary Appendix. Briefly, both aligned WES and targeted panel data underwent variant calling using a typical germline variant caller (GATK HaplotypeCaller10) and GATK recommended best practices.11,12 In addition the targeted panel aligned sequence data was processed through a dedicated bioinformatics pipeline which included variant calling with GATK4/Mutect2 (https://software.broadinstitute.org/gatk/) in order to improve detection of low level acquired variants. Copy number variation (CNV) was estimated by comparing read counts from targeted panel data to a reference pool comprising normal samples with similar technical artefacts and visualized using CNspector.13 Depending on the clinical context and genomic findings, patients underwent further characterization with RNA-studies (including whole transcriptome RNA sequencing), Sanger sequencing and/or droplet digital PCR (ddPCR) (see the Online Supplementary Appendix). Sequencing was performed on peripheral blood, bone marrow, buccal epithelial cells or cultured skin fibroblasts depending on the clinical context. All variants were described according to Human Genome Variation Society (HGVS) nomenclature and germline variants classified according to the American College of Medical Genetics guidelines for the interpretation of sequence variants.14 Accredited (International Organization for Standardization [ISO] 15189) diagnostic reports were issued to referrers for all patients in real-time. In addition, genomic results were reviewed in centralized multidisciplinary case conferences attended by the treating clinicians, molecular hematopathologists, medical scientists, clinical geneticists and genetic counsellors.

Results Cohort description 115 patients were recruited from May 2017 to August 2018. The median age of the overall cohort was 24 years (range: 3 months - 81 years) with a male to female (M:F) ratio of 1:1.4. Upon enrolment, the referring clinical team were asked to assign a BMF subtype based on the available clinicopathological features to capture the “pre-test diagnosis” for each patient, noting that these had been assigned on varied degrees of evidence. Any genomic evaluation completed before enrolment was considered part of “standard-of-care” investigations and were performed 65

P. Blombery et al.

Table 1. Causative genomic variants in patients with the clinical diagnosis of an inherited bone marrow failure syndrome.

Patient ID Age/ Pre-test Post-test Germline variants Sex diagnosis diagnosis 14# 25# 31# 76# 44# 85#

4M 14F 16M 33M 35M 33F

DBA DBA DBA DKC FA FA

SDS

102

32F

SDS

107#

18F

SCN

18 32# 104#

18F 5F 20F

SCN SCN SCN

Acquired Comment variants

RPS19 c.251_252del; p.(Arg84Lysfs*69) [het] RPL35A whole gene copy loss [het] RPS19 exon 2-3 copy loss [het] TINF2 c.844C>T; p.(Arg282Cys) [het] RAD51C c.773G>A; p.(Arg258His) [hom] FANCI c.3184C>T; p.(Gln1062*) [het] FANCI c.3041G>A; p.(Cys1014Tyr) [het] SBDS c.183_184delinsCT; p.(Lys62*) [het] SBDS c.258+2T>C; p.? [het] SBDS c.183_184delinsCT; p.(Lys62*) [het] SBDS c.258+2T>C; p.? [het] ELANE c.684C>A; p.(Tyr228*) [het] CSF3R c.2308C>T; p.(Gln770*) HAX1 c.91del; p.(Glu31Lysfs*54) [hom] SRP54 c.349_351del; p.(Thr117del) [het] SRP54 c.349_351del; p.(Thr117del) [het] -

Variant previously described in DBA25 RPL35A deletion previously described in DBA26 RPS19 copy number losses previously described in DBA27 Variant previously described in DKC28,29 Variant previously described in FA-like syndrome30-32 Variant previously described in FA33,34 Established pathogenic variants in SDS Established pathogenic variants in SDS Variant previously described in SCN35,36 Acquired mutation detected at VAF of 2.9% Variant previously described in SCN37,38 Variant previously described in SCN39,40 Variant previously described in SCN39,40

#Case description provided in the Online Supplementary Appendix. DBA: Diamond-Blackfan anemia; DKC: Dyskeratosis congenita; FA: Fanconi anemia; SCN: severe congenital neutropenia; SDS: Shwachman-Diamond syndrome; VAF: variant allele frequency; het: heterozygous; hom: homozygous. Reference transcripts – CSF3R, NM_156039.3; ELANE, NM_001972.2; FANCI, NM_001113378.1; HAX1, NM_006118.3; RAD51C, NM_058216.2; RPS19, NM_001022.3; RPL35A, NM_000996.2; SBDS, NM_016038.2; SRP54, NM_003136.3; TINF2, NM_001099274.1.

according to individual treating teams/institutional guidelines. For analysis purposes, patients were divided into three different pre-test diagnosis cohorts (i) patients with a pre-diagnosis of an IBMFS (n=23; median age 10 [range: 2-36 years]; M:F ratio 1:1.88), (ii) patients with a pre-test diagnosis of aAA or hMDS (n=47; median age 32 [range: 7-80 years]; M:F ratio 1:1.24) and (iii) patients with a pretest diagnosis of clinically unclassifiable BMF (UBMF) (n=45; median age 25 [range: 1-81 years]; M:F ratio 1:1.37) (Figure 2). If the referring clinical team could not definitively assign a pre-test diagnosis of IBMFS (including subtype) or aAA/hMDS, the patient was assigned to the UBMF category. After comprehensive genomic characterization was completed, multidisciplinary re-review of all clinical and pathological features incorporating newly acquired genomic data was performed and a “post-test diagnosis” was assigned. Clinical management of patients after assignment of post-test diagnosis was captured in regard to (i) proceeding to allogeneic stem cell transplant, (ii) screening of relatives for identified causative variants and (iii) disease-specific interventions.

Genomic characterization identifies causative variants in the majority of patients with a pre-test diagnosis of an IBMFS In patients with a pre-test diagnosis of an IBMFS, genomic characterization confirmed an IBMFS by detecting causative germline variants in 52.2% (12 of 23). The genomic variants detected in this cohort are listed in Table 1. In all cases where a causative germline variant was identified, the post-test diagnosis was consistent with the pre-test diagnosis. One acquired CSF3R variant was detected in this group in a patient with severe congenital neutropenia (SCN) and a germline ELANE variant who had life-long treatment with G-CSF. A causative 66

germline variant was not identified in patients with a pretest diagnosis of SCN in 71.4% (10 of 14) of cases despite genomic characterization. Despite no change in the pre-test diagnosis in this group, the observed clinical impacts of rendering a precise genomic diagnosis to this group of patients with established clinical diagnoses included one patient (#76) where the identification of a specific variant in TINF2 permitted screening of his pre-symptomatic children for DKC. In addition, this group included an adult patient (#44) with the observation of a homozygous variant in RAD51C (Arg258His) resulting in a Fanconi-like syndrome (associated with positive chromosomal fragility studies) with subsequent implications for breast and ovarian cancer risk screening in siblings. Further clinicopathological details for cases are presented in the Online Supplementary Appendix.

Both acquired and inherited variants were detected in patients with a pre-test diagnosis of aAA/hMDS In patients with a pre-test diagnosis of aAA/hMDS, acquired variants were detected in 53.2% of patients (25 of 47) (see Table 2). Forty-one of 47 patients had a pretest diagnosis of aAA (22 of 41 [53.7%] with variants) and 6 of 47 had a pre-test diagnosis of hMDS (3 of 6 [50%] with variants). Variant allele frequency (VAF) of acquired variants in this group ranged from 0.35%56.1%. Acquired variants were detected in PIGA (n=15), ASXL1 (n=6), BCOR (n=4), RUNX1 (n=4), CBL (n=3), CSMD1 (n=3), DNMT3A (n=2), TET2 (n=2), U2AF1 (n=2), SRSF2 (n=1), SF3B1 (n=1), TERT (n=1), IKZF1 (n=1), IDH2 (n=1) and BCORL1 (n=1). Three patients in this group had multiple subclonal PIGA variants detectable, a phenomenon previously described in patients with paroxysmal nocturnal hemoglobinuria haematologica | 2021; 106(1)

Genomic characterization of BMF syndrome patients

Table 2. Variants detected in patients with a clinicopathological diagnosis of acquired aplastic anemia or hypoplastic myelodysplastic syndrome.

Patient ID Age/ Sex

Pre-test diagnosis

Post-test diagnosis

Germline variants

Acquired variants ASXL1 c.3785del; p.(Ser1262Thrfs*18) ASXL1 c.1864_1865del; p.(Leu622Alafs*12) CSMD1 c.1124G>A; p.(Cys375Tyr) BCORL1 c.1429dup; p.(Thr477Asnfs*58) CSMD1 c.4377del; p.(Gly1460Valfs*22) -

2.8% 1.6% 1.8% 3.3% 1.7% 100%

PIGA c.986dup; p.(Ser330Lysfs*9) PIGA c.849-2A>G; p.? CBL c.1259G>C; p.(Arg420Pro)

1.6% 3.9% 42.9% 7.5% 52.4% 16% 18.1% 6.9% 4.3% 1.0% 41.7% 6.0% 4.2% 2.4% 13.0% 1.9% 8.6% 1.8% 56.1% 10.6% 8.0% 8.3% 24.3% 22.5% 2.0% 2.6% 1.9% 8.2% 1.4% 16.6% 9.9% 40.2% 0.52% 8.3% 1.8% 0.35% 0.47% 0.89% 19.8% 7.1% 32.5% 50.3% 1.5%

22 30

19F 17M

aAA aAA

28M

aAA

39#

27F

aAA

IBMFS (FA)

31F

aAA

47#

64M

aAA

IBMFS (APS)

87#

49M

aAA

IBMFS (Telomeropathy)

50F

aAA

47F

aAA

43 49 55 57 58 65

24M 22M 71F 36F 49F 73M

aAA aAA aAA aAA aAA aAA

aAA aAA aAA aAA aAA MDS

66 71 78 89

36F 38M 37F 37F

aAA aAA aAA aAA

aAA aAA aAA MDS

101 108

38F 24F

aAA aAA

aAA MDS

114 45 9 111

40F 58M 19M 75M

aAA hMDS hMDS hMDS

FANCA c.2980A>G; p.(Ser994Gly) [hom] SAMD9L c.2956C>T; p.(Arg986Cys) [het] TERT c.1223T>C; p.(Leu408Pro) [het]

BCOR c.3549_3561dup; p.(Val1188Metfs*27) TERT c.-124C>T PIGA c.749T>C; p.(Leu250Pro) PIGA c.684del; p.(Ile228Metfs*19) PIGA c.1347dup; p.(Ile450Tyrfs*3) PIGA c.105_106del; p.(Ile35Metfs*5) RUNX1 c.317G>T; p.(Trp106Leu) ASXL1 c.2329G>T; p.(Glu777*) ASXL1 c.2053G>T; p.(Gly685*) PIGA c.491C>T; p.(Ser164Leu) PIGA c.715+2T>G; p.? PIGA c.944G>A; p.(Cys315Tyr) DNMT3A c.2645G>A; p.(Arg882His) BCOR c.4537C>T; p.(Arg1513*) PIGA c.715+1G>T; p.? TET2 c.488del; p.(Phe163Serfs*20) IDH2 c.419G>A; p.(Arg140Gln) SF3B1 c.1997A>G; p.(Lys666Arg) U2AF1 c.470A>C; p.(Gln157Pro) ASXL1 c.1934dup; p.(Gly646Trpfs*12) BCOR c.1971C>A; p.(Tyr657*) PIGA c.715+1G>A; p.? PIGA c.811_812del; p.(Leu271Glyfs*11) RUNX1 c.790_791del; p.(Gln264Glufs*335) RUNX1 c.602G>A; p.(Arg201Gln) CBL c.1259G>A; p.(Arg420Gln) CBL c.616C>T; p.(Arg206*) ASXL1 c.2555C>G; p.(Ser852*) TET2 c.4910_4911insTC; p.(Ser1638Hisfs*58) U2AF1 c.101C>T; p.(Ser34Phe) PIGA c.548G>T; p.(Cys183Phe) PIGA c.525_526del; p.(Cys176*) PIGA c.227_229delinsTT; p.(Asn76Ilefs*19) BCOR c.4504G>A; p.(Asp1502Asn) DNMT3A c.2525A>G; p.(Gln842Arg) IKZF1 c.476A>G; p.(Asn159Ser) RUNX1 c.602G>A; p.(Arg201Gln) SRSF2 c.284_304del; p.(Pro95_Ser101del) CSMD1 c.2257C>T; p.(Arg753Cys)

Acquired VAF

# Case description provided in the Online Supplementary Appendix. aAA: acquired aplastic anemia; FA: Fanconi anemia; APS: ataxia-pancytopenia syndrome; MDS: myelodysplastic syndrome; hMDS, hypocellular MDS; VAF: variant allele frequency; het: heterozygous; hom: homozygous. Reference transcripts â&#x20AC;&#x201C; ASXL1, NM_015338.5; BCOR, NM_017745.5; BCORL1, NM_021946.4; CBL, NM_005188.3; CSMD1, NM_033225.5; DNMT3A, NM_022552.4; FANCA, NM_000135.2; IDH2, NM_002168.2; IKZF1, NM_006060.4; PIGA, NM_002641.3; RUNX1, NM_001754.4; SAMD9L, NM_152703.2; SF3B1, NM_012433.2; SRSF2, NM_003016.4; TERT, NM_198253.2; TET2, NM_001127208.2; U2AF1, NM_006758.2.

haematologica | 2021; 106(1)

P. Blombery et al.

Figure 1. Genes involved in inherited and acquired bone marrow failure syndromes. Genes causative of an inherited bone marrow failure syndrome (IBMFS), and genes associated with a risk of transformation to hematological malignancy or response to immunosuppression in patients with acquired aplastic anemia (aAA) and hypoplastic myelodysplastic syndrome (hMDS) were assessed in this study. Four genes, CSF3R, GATA2, MPL and RUNX1, were assessed in both germline and acquired settings. Genes targeted by whole exome sequencing (WES) only (i.e., not included on targeted panel) are denoted with an asterisk. Genes KMT2D and PSTPIP1 are not represented in this figure; variants detected in these two genes were considered ‘off-target’ findings.

(PNH).15 No patients in this group had a clinical diagnosis of overt PNH. Over the limited follow-up available, 3 of 25 patients with acquired variants progressed with an aggressive hematological malignancy (acute myeloid leukemia or high-grade MDS) during the study period compared to 0 of 22 patients with no acquired variant detected (P=0.2368, Fisher’s exact test). Six of 47 (12.8%) patients in this group had a change of diagnostic categorization as a result of genomic characterization. Three of these six patients had a pre-test diagnosis of aAA and were re-categorized as having a myelodysplastic syndrome (transformed from underlying aAA) after genomic characterization. In addition to the genomic testing results themselves, the re-categorization of these patients was also the result of a re-assessment of the patient’s disease that was prompted by the availability of genomic testing (including repeat of investigations such as morphological assessment of bone marrow biopsy). The other three patients had a change in pre-test diagnosis from aAA/hMDS to an IBMFS. These cases include a 64-year-old male (#47) who presented with pancytopenia and was found to have bone marrow hypoplasia and morphological dysplasia as well as an acquired pathogenic CBL variant. A pathogenic SAMD9L variant was identified during germline testing. Detailed clinical review revealed a family history of early onset ataxia in a number of family members and subsequently the patient was diagnosed with ataxia-pancytopenia syndrome – one presentation of the SAMD9L group of inherited disor68

ders16 (see the Online Supplementary Appendix). In addition, a 49-year-old male (#87) presented with pancytopenia and severe bone marrow hypoplasia and was diagnosed and treated for aAA. Genomic testing identified a germline variant in TERT (Leu408Pro). Identification of the TERT variant prompted telomere length testing and the patient was found to have telomere lengths <1st percentile. In addition, an acquired TERT promoter variant was detected in this patient adding further evidence to the pathogenicity of the germline variant. The patient had no other clinical features of DKC (see the Online Supplementary Appendix). Finally, a 27-year-old female (#39) was diagnosed with aAA after presenting with pancytopenia. She was commenced on eltrombopag but had no improvement in her blood counts. Genomic characterization revealed a homozygous FANCA variant (Ser994Gly) that resulted in exon skipping (demonstrated by RNA sequencing) with further testing demonstrating equivocal chromosomal fragility but failure of monoubiquitination in FANCD2 consistent with a diagnosis of FA (see the Online Supplementary Appendix).

Genomic characterization clarifies diagnostic categorization in patients with clinically unclassifiable BMF In patients with a pre-test diagnosis of UBMF, genomic characterization identified a causative germline or acquired variant in 25 of 45 (55.6%). In 24 of 25 patients this resulted in a change in the diagnostic categorization haematologica | 2021; 106(1)

Genomic characterization of BMF syndrome patients

Table 3. Variants detected in patients with a clinical diagnosis of undifferentiated bone marrow failure syndromes. Patient ID Age/ Pre-test Post-test Germline Acquired Sex diagnosis diagnosis variants variants 13# 67# 106#

1F 2F 4F

UBMF UBMF UBMF

IBMFS (DBA) IBMFS (DBA) IBMFS (DBA)

RPS19 c.184C>T; p.(Arg62Trp) [het] RPS19 c.184C>T; p.(Arg62Trp) [het] RPS7 c.75+1G>T; p.? [het]

60#

UBMF

IBMFS (DKC)

TERT c.3148A>G; p.(Lys1050Glu) [het] TERT c.1670T>C; p.(Leu557Pro) [het]

96#

34M

UBMF

IBMFS (FA)

118#

21F

UBMF

53#

11M

UBMF

20#

UBMF

62#

UBMF

23#

UBMF

IBMFS (GATA2 syndrome) IBMFS (FA-like syndrome) IBMFS (SAMD9 syndrome) IBMFS (SAMD9 syndrome) IBMFS (SAMD9 syndrome) IBMFS (SDS)

FANCA c.2852G>A; p.(Arg951Gln) [het] FANCA c.3971C>T; p.(Pro1324Leu) [het] GATA2 c.351C>G; p.(Thr117=) [het]

69#

UBMF

81#

13M

82#

69F

103

40F

4# 79

5M 53F

81M

UBMF

hMDS

73F

UBMF

hMDS

52#

41F

UBMF

98 112

53F 65M

UBMF UBMF

MDS with isolated del(5q) UBMF MDS

63 88

69M 68F

UBMF UBMF

FANCM c.1972C>T; p.(Arg658*) [het] FANCM c.5101C>T; p.(Gln1701*) [het] SAMD9 c.4057A>G; p.(Lys1353Glu) [het]

IBMFS (Kabuki Syndrome) UBMF IBMFS (Autoimmune cytopenias associated with PSTPIP1 syndrome) UBMF RCC UBMF Therapy-related myeloid neoplasm

aAA Therapy-related myeloid neoplasm

SETBP1 c.2602G>A; p.(Asp868Asn) -

Variant previously described in DBA25,41 Variant previously described in DBA25,41 Novel variant predicted to cause abnormal splicing Variant previously described in telomeropathy42 Novel variant. Impairment of TERT function demonstrated in vitro. Variants previously described in FA43-47 Variant previously described in GATA2 syndrome48

Confirmed de novo

SAMD9 c.2318T>C; p.(Ile773Thr) [het]

Confirmed de novo

SAMD9 c.3821A>G; p.(Tyr1274Cys) [het]

Confirmed de novo

Established pathogenic variants

SBDS c.183_184delinsCT; p.(Lys62*) [het] SBDS c.258+2T>C; p.? [het] DNAJC21 c.972_983+1414del [het]

IBMFS (SDS-like syndrome) UBMF IBMFS DDX41 c.435-2_435-1delinsCA; p.? [het] (DDX41 syndrome) UBMF IBMFS DDX41 c.517G>A; p.(Gly173Arg) [het] (DDX41 syndrome)

UBMF

Comment

KMT2D c.9218del; p.(Val3073Glufs*46) [het] PSTPIP1 c.748G>A; p.(Glu250Lys) [het]

DDX41 c.1574G>A; p.(Arg525His) DNMT3A c.1903C>T; p.(Arg635Trp) TET2 c.2366_2402del; p.(Asn789Metfs*12) -

Novel variant resulting in exon 7 skipping on RNA sequencing. Variant previously described in DDX41 syndrome21 Variant previously described in DDX41 syndrome18

Novel variant. Cytopenias observed as a result of autoimmune destruction. Variant previously described in Hyperzincaemia/hypercalprotectinemia49

RUNX1 c.496C>T; p.(Arg166*) TP53 c.833C>T; p.(Pro278Leu) TP53 c.818G>A; p.(Arg273His) TP53 c.592del; p.(Glu198Lysfs*49) TET2 c.5437C>T; p.(Gln1813*) DNMT3A c.2634dup; p.(Asn879Glnfs*42) TET2 c.4145A>G; p.(His1382Arg) TET2 c.3415del; p.(Ile1139Leufs*13) TET2 c.4072T>G; p.(Cys1358Gly) 5q deletion (5q14.3-33.2) DNMT3A c.2191T>A; p.(Phe731Ile) BCORL1 c.2699del; p.(Asp900Alafs*25) DNMT3A c.2141C>G; p.(Ser714Cys) BCOR c.4848T>G; p.(Tyr1616*) TET2 c.4075C>T; p.(Arg1359Cys)

# Case description provided in the Online Supplementary Appendix. aAA: acquired aplastic anemia; DBA: Diamond-Blackfan anemia; DKC: Dyskeratosis congenita; FA: Fanconi anemia; hMDS: hypoplastic myelodysplastic syndrome; MDS: myelodysplastic syndrome; SDS: Shwachman-Diamond syndrome; UBMF: undifferentiated bone marrow failure; het: heterozygous; RCC: Refractory cytopenia of childhood. Reference transcripts â&#x20AC;&#x201C; BCOR, NM_017745.5; BCORL1, NM_021946.4; DDX41, NM_016222.2; DNAJC21, NM_194283.3; DNMT3A, NM_022552.4; FANCA, NM_000135.2; FANCM, NM_020937.2; GATA2, NM_032638.4; KMT2D, NM_003482.3; PSTPIP1, NM_003978.4; RPS19, NM_001022.3; RPS7, NM_001011.3; RUNX1, NM_001754.4; SAMD9, NM_017654.3; SBDS, NM_016038.2; SETBP1, NM_015559.2; TERT, NM_198253.2; TET2, NM_001127208.2; TP53, NM_000546.5.

haematologica | 2021; 106(1)

P. Blombery et al.

Figure 2. Pre- and post-test diagnosis classification of patients. Numbers of patients classified into each diagnostic group, (i) inherited bone marrow failure syndrome (IBMFS), (ii) acquired aplastic anemia (aAA) or hypoplastic MDS (hMDS) or (iii) clinically unclassifiable bone marrow failure (UBMF) as evaluated prior to genomic testing (pre-test diagnosis) and again following genomic testing (posttest diagnosis) with transitions indicated by the connecting lines (in addition three patients changed diagnosis within the aAA/hMDS category from aAA to MDS). RCC: refractory cytopenia of childhood; tMN: therapy related myeloid neoplasm.

from a pre-test diagnosis of UBMF to a specific post-test diagnosis. Therefore genomic characterization resulted in diagnostic categorization in 24 of 45 (53.3%) of patients in this diverse group of patients. Variants detected in this cohort are listed in Table 3. Causative germline variants were detected in 16 patients and acquired variants were detected in nine patients. The diverse inherited and acquired diagnoses included FA, SDS, DKC, DBA, aAA, hMDS, therapy-related myeloid neoplasm (tMN) and MDS with isolated del(5q). Three patients in this group received a post-test diagnosis of hMDS or MDS after genomic characterization. This was the result of re-interpretation of the clinical and pathological features in light of supportive molecular evidence for these diagnoses. Notably, one patient retained a post-test diagnosis of UBMF despite having a DNMT3A mutation detected due to inadequate pathological and clinical features to support a diagnosis of aAA, hMDS, MDS or tMN. Five patients in this group presented with hypoplastic BMF and a monosomy 7 detectable by either conventional karyotyping, fluorescence in situ hybridization or copy number analysis by targeted sequencing. Three of these (#20, #23, #62) had SAMD9 variants that were all confirmed to be de novo through testing of parental samples. Multiple somatic reversion events were noted in these patients including acquired sequence variants as well as progressive resolution of monosomy 7 over time (see the Online Supplementary Appendix). One patient (#118) with monosomy 7 had a synonymous GATA2 variant (Thr117=) which has previously been associated with GATA2 haploinsufficiency syndrome.17 Two patients in this group had pathogenic DDX41 variants identified. One of these patients was a 69-year-old female (#82) who presented with severe neutropenia and was found to have a severely hypocellular bone marrow. Genomic characterization revealed a DDX41 missense variant (Gly173Arg) which has been described previously in 70

five patients presenting with a similar clinical phenotype.18 In addition, targeted panel sequencing in this patient identified a hotspot somatic DDX41 Arg525His variant present at approximately 3% VAF. Finally, this group included a 5-year-old male (#4) with a pre-test diagnosis of UBMF with onset just after birth. Targeted panel sequencing identified a RUNX1 nonsense variant (Arg166*) at low VAF (4.8%). This variant was quantitatively monitored over time using ddPCR and was noted to be acquired, steadily increasing and reaching 11.8% over a 12-month period. After multidisciplinary review it was decided to proceed to an allogeneic bone marrow transplant in this patient.

Contribution to diagnostic categorization from targeted panel sequencing versus WES All germline variants in genes common to the targeted panel and whole exome were detected. In eight patients, variants were detected in the whole exome in genes not covered in the targeted panel. These included variants in SAMD9, SAMD9L, DNAJC21, SRP54, KMT2D and PSTPIP1. These genes can be divided into two groups: (i) genes where the primary phenotype of pathogenic variants is an IBMFS (SAMD9, SAMD9L, DNAJC21 and SRP54) and (ii) genes where BMF is either an infrequent or not the primary clinical manifestation (KMT2D and PSTPIP1). Mutations in genes in this latter category were detected through an extended analysis of the WES data given the absence of other causative mutations found in these two patients and the remaining strong suspicion of a genetic etiology. In total, 65 acquired variants were detected across the entire cohort. All of these variants were detected by targeted panel sequencing with a bioinformatic pipeline optimized for acquired variant detection. Only 16 of 65 (24.6%) of these variants were detected by WES using a standard germline bioinformatics pipeline. Of the 49 haematologica | 2021; 106(1)

Genomic characterization of BMF syndrome patients

acquired variants not called by the WES pipeline, 24 were visible after manual inspection of aligned reads in whole exome data based on variant detection in the targeted panel data. The mean depth of targeted panel sequencing was 795x (range: 576-1,025) with 98.9% of bases sequenced to ≥100x coverage, compared to a mean depth of 198x (range: 117-260) and 77.5% ≥100x coverage for WES. The VAF of the variants that were not detected by WES (and associated pipeline) were all less than 15%.

Discussion We have described a cohort of 115 patients presenting with a diverse range of hypocellular BMF that have undergone prospective comprehensive genomic evaluation for germline and acquired causes. In our cohort we were able to identify pathogenic germline or acquired variants in 54% (62 of 115) of the entire cohort including in 52% (12 of 23) of patients with a clinical diagnosis of IBMFS, 53% (25 of 47) of patients with a clinical diagnosis of aAA/hMDS and 56% (25 of 45) of patients with clinically unclassifiable BMF. Overall, genomic characterization resulted in a change in diagnostic categorization in 30 of 115 (26.1%) patients. The observed clinical impacts in our patient cohort after comprehensive genomic characterization included the choice to perform allogeneic stem cell transplantation, disease-specific treatments (such as corticosteroid for DBA and lenalidomide for MDS with del(5q)), identification of at-risk family members, enhanced screening of patients for secondary malignancy and influence of sibling allogeneic stem cell donor choice. We have also described multiple novel genomic findings including novel pathogenic variants in TERT (Leu408Pro and Leu557Pro), RPS7 (c.75+1G>T) and SAMD9 (Lys1353Glu, Tyr1274Cys and Ile773Thr) and have confirmed recent novel findings of other groups including (i) the association of biallelic FANCM variants with chemotherapy toxicity in the absence of an overt FA phenotype,19 (ii) the association of DDX41 Gly173Arg with a hypocellular MDS phenotype18 and (iii) the presence of acquired reversion variants in SAMD9/SAMD9L.20 The diverse spectrum of inherited and acquired diseases identified in our cohort highlights the diagnostic utility of comprehensive genomic characterization for patients with a clinical presentation of BMF. Despite the range of underlying diagnoses that may manifest as hypoplastic BMF there may be minimal or no specific features on conventional (non-genomic) diagnostic testing to indicate the specific underlying diagnosis. In addition, clinical features such as abnormal digits or oral leukoplakia cannot be relied upon to suggest an underlying genetic cause as they may be absent or unrecognized due to lack of physician familiarity with these rare diseases. Moreover, as more literature and larger cohorts emerge with genetic annotation, the clinical phenotypes associated with specific genes are expanding. For example, the expansion of the phenotype of pathogenic variants in DDX41 to include a hypocellular MDS-like phenotype in the case of Gly173Arg18 rather than the initial description of presentation in later life with acute myeloid leukemia after minimal antecedent overt hematological abnormality.21 Indeed, in our cohort we observed a pathogenic DDX41 variant in a 13-year-old male (#81) with isolated severe stable haematologica | 2021; 106(1)

thrombocytopenia and no evidence of hematological malignancy. A significant proportion of patients (47.8%) with a clinical diagnosis of IBMFS did not have causative variants detected. The vast majority (10 of 11) of these were in patients with SCN. This may be explained by the significant baseline rate of “variant-negative” cases in this condition described in the literature to date22 as well as the potential for other causes of neutropenia. Given these patients had no pathogenic variants detectable despite comprehensive genomic analysis (including WES), one possible future strategy for identifying causative variants may be to evaluate the non-coding regions of existing pathogenic genes (e.g., ELANE, HAX1) for deep intronic variants resulting in splicing abnormalities that are missed by methods focused on coding regions. In the acquired BMF setting, we noted six patients who were re-categorized as either hMDS, MDS, tMN or refractory cytopenia of childhood (RCC) after genomic characterization. Whilst there are no absolutely specific molecular abnormalities that distinguish these disorders from aAA complicated by clonal hematopoiesis or age-related clonal hematopoiesis, the presence of the genomic abnormalities detected were sufficient for the treating team to re-evaluate the clinical and morphological features of these six cases and assign post-diagnoses of hMDS/MDS/RCC/tMN. This reflects the way in which this type of genomic data is increasingly integrated into diagnostic decision making in a real-world context, especially in areas where morphological interpretation is challenging and inter-observer agreement is poor.23,24 Another practical aspect regarding diagnostic workup is the limited availability of specific non-genomic tests (e.g., telomere length analysis) in the accredited setting which may have contributed to a variable rate of performance of these tests in the cohort. In our cohort 39% (45 of 115) of patients were categorized as clinically UBMF before genomic testing. The relatively large number of patients in this category may be the result of numerous factors including a referral bias for patients to this study with atypical clinicopathological features in order to access novel genomic technologies. In addition, patients could also be included at an early timepoint in their disease course and it is possible that repeated history and physical examination over time in patients (including assessing response to therapy) would have enabled more definitive categorization before genomic testing. Moreover, there is an increasing recognition of the possibility of occult IBMFS in patients diagnosed as acquired BMF (e.g., aAA), indeed in our cohort three patients harbored occult IBMFS. This recognition, in turn, may lead to a desire by treating clinical teams to more comprehensively exclude IBMFS rather than relying on a clinical diagnosis of a purely acquired BMFS - hence their inclusion in a UBMF category. Reliable clinical and laboratory features to identify those with a low likelihood of IBMFS (e.g., the presence of a PNH clone, severity/time course/persistence of neutropenia in SCN) would be of significant value in resource-limited settings in order to avoid expensive and time consuming genetic testing in this potentially expanding subgroup. The approach to genomic characterization in our cohort included both targeted BMF panel sequencing and WES for all patients. Our choice of targeted panel sequencing in addition to WES was based on the requirement for enhanced sensitivity 71

P. Blombery et al.

due to the clinical context of BMF mandating assessment for both germline and acquired genomic lesions. In addition to the detection of acquired variants in aAA/hMDS (which may be present at VAF less than 5%), clinically relevant reversion variants in inherited conditions such as SAMD9/SAMD9L can also exist which require sensitive testing methodologies. As expected (given the greater depth of sequencing able to be cost-effectively achieved), targeted panel sequencing was clearly superior at detecting low-level acquired variants when compared to WES. Using a bioinformatics pipeline optimized for acquired variant detection (GATK4/Mutect2) we were able to detect acquired variants down to a VAF of <0.5%. Despite their very low level, detection of these variants can be highly instructive for diagnostic categorization by demonstration of clonality (to distinguish from reactive conditions) as well as providing a baseline genomic profile for future monitoring of clonal evolution. Conversely, one potential benefit of WES for investigation of germline variants is in the context of phenotypic uncertainty in order to identify pathogenic variants across a broad gene list. In our cohort, two patients had pathogenic variants detected in genes which are not typically associated with BMF (KMT2D and PSTPIP1). The other genes originally not present in the targeted panel due to their recent description (SAMD9, SAMD9L, DNAJC21) have been added to a future iteration of the targeted panel. However the need for panel re-design and subsequent costly laboratory validation highlights a recognized drawback of targeted panel sequencing. Nevertheless, given the relative specificity of the clinical presentation and the requirement for sensitive detection of acquired variants, targeted panel sequencing is a methodological choice highly suited for this clinical context, compared to which, WES offers minimal additional clinical utility. Copy number analysis was possible using either of these sequencing strategies and facilitated the identification of pathogenic copy number losses (RPL35A, RPS19) as well as clinically relevant acquired monosomies. RNA sequencing was valuable for determining the effect of putative splice variants on mRNA transcripts (FANCA, DNAJC21). In conclusion, we have described the impact of genomic characterization on diagnostic categorization in a cohort of patients with hypoplastic BMF. In addition to multiple

References 1. Swerdlow SH, Campo E, Harris NL, et al. WHO classification of Tumours of Haematopoietic and Lymphoid Tissues. 4th edn ed. Lyon, France: International Agency for Research on Cancer, 2017. 2. Oostra AB, Nieuwint AW, Joenje H, de Winter JP. Diagnosis of fanconi anemia: chromosomal breakage analysis. Anemia. 2012;2012:238731. 3. Lai TP, Wright WE, Shay JW. Comparison of telomere length measurement methods. Philos Trans R Soc Lond B Biol Sci. 2018; 373(1741). 4. Ghemlas I, Li H, Zlateska B, et al. Improving diagnostic precision, care and syndrome definitions using comprehensive next-generation sequencing for the inherited bone marrow failure syndromes. J Med Genet. 2015;

novel genomic findings, this personalized genomic approach is of significant utility to rapidly providing accurate diagnostic categorization to inform clinical decisionmaking in this diverse and challenging group of diseases. The incorporation of a multidisciplinary review further strengthens the impact and clinical value of this approach. Disclosures No conflicts of interest to disclose. Contributions PB designed the research study, analyzed data, curated and interpreted genetic variation for clinical reports, recruited and managed patients, participated in multidisciplinary meetings and wrote the paper. LF analyzed data, curated and interpreted genetic variation for clinical reports, recruited and managed patients, participated in multidisciplinary meetings and wrote the paper. GR analyzed data, curated and interpreted genetic variation for clinical reports, performed experiments, participated in multidisciplinary meetings and wrote the paper. ET, JL and MMc analyzed data, curated and interpreted genetic variation for clinical reports, participated in multidisciplinary meetings and wrote the paper. DH, AG, FM, GL, JS, PBar, JR and JW recruited and managed patients, participated in multidisciplinary meetings and wrote the paper. EW, EL and MM participated in multidisciplinary meetings and wrote the paper. CG designed the research study and wrote the paper. DR designed the research study, recruited and managed patients, participated in multidisciplinary meetings and wrote the paper. Acknowledgments The authors would like to acknowledge the support of Maddie Riewoldt’s Vision, all participants of the multidisciplinary meetings and the patients and families. The Australian Regenerative Medicine Institute is supported by grants from the State Government of Victoria and the Australian Government. Current work in GL’s laboratory is supported by the National Health and Medical Research Council (NHMRC, 1086020), Australian Research Council (ARC, DP170102235) and Maddie Riewoldt’s Vision. Funding The study was funded by the State Government of Victoria (Department of Health and Human Services) and the 10 member organizations of the Melbourne Genomics Health Alliance.

52(9):575-584. 5. Kulasekararaj AG, Jiang J, Smith AE, et al. Somatic mutations identify a subgroup of aplastic anemia patients who progress to myelodysplastic syndrome. Blood. 2014; 124(17):2698-2704. 6. Yoshizato T, Dumitriu B, Hosokawa K, et al. Somatic mutations and clonal hematopoiesis in aplastic anemia. N Engl J Med. 2015;373(1):35-47. 7. Bono E, McLornan D, Travaglino E, et al. Clinical, histopathological and molecular characterization of hypoplastic myelodysplastic syndrome. Leukemia. 2019; 33(10):2495-2505. 8. Killick SB, Bown N, Cavenagh J, et al. Guidelines for the diagnosis and management of adult aplastic anaemia. Br J Haematol. 2016;172(2):187-207. 9. Ryland GL, Jones K, Chin M, et al. Novel genomic findings in multiple myeloma iden-

tified through routine diagnostic sequencing. J Clin Pathol. 2018;71(10):895-899. 10. McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010;20(9):1297-1303. 11. DePristo MA, Banks E, Poplin R, et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet. 2011;43(5):491498. 12. Van der Auwera GA, Carneiro MO, Hartl C, et al. From FastQ data to high confidence variant calls: the Genome Analysis Toolkit best practices pipeline. Curr Protoc Bioinformatics. 2013;43:11. 13. Markham JF, Yerneni S, Ryland GL, et al. CNspector: a web-based tool for visualisation and clinical diagnosis of copy number variation from next generation sequencing.

haematologica | 2021; 106(1)

Genomic characterization of BMF syndrome patients

Sci Rep. 2019;9(1):6426. 14. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405-423. 15. Shen W, Clemente MJ, Hosono N, et al. Deep sequencing reveals stepwise mutation acquisition in paroxysmal nocturnal hemoglobinuria. J Clin Invest. 2014;124(10):45294538. 16. Chen DH, Below JE, Shimamura A, et al. Ataxia-Pancytopenia Syndrome Is Caused by Missense Mutations in SAMD9L. Am J Hum Genet. 2016;98(6):1146-1158. 17. Wehr C, Grotius K, Casadei S, et al. A novel disease-causing synonymous exonic mutation in GATA2 affecting RNA splicing. Blood. 2018;132(11):1211-1215. 18. Sebert M, Passet M, Raimbault A, et al. Clinical and molecular characteristics of DDX41-mutated patients in a large cohort of sporadic MDS/AML. Blood. 2018; 132(Suppl 1):S797. 19. Catucci I, Osorio A, Arver B, et al. Individuals with FANCM biallelic mutations do not develop Fanconi anemia, but show risk for breast cancer, chemotherapy toxicity and may display chromosome fragility. Genet Med. 2018;20(4):452-457. 20. Tesi B, Davidsson J, Voss M, et al. Gain-offunction SAMD9L mutations cause a syndrome of cytopenia, immunodeficiency, MDS, and neurological symptoms. Blood. 2017;129(16):2266-2279. 21. Lewinsohn M, Brown AL, Weinel LM, et al. Novel germ line DDX41 mutations define families with a lower age of MDS/AML onset and lymphoid malignancies. Blood. 2016;127(8):1017-1023. 22. Boztug K, Klein C. Genetic etiologies of severe congenital neutropenia. Curr Opin Pediatr. 2011;23(1):21-26. 23. Sasada K, Yamamoto N, Masuda H, et al. Inter-observer variance and the need for standardization in the morphological classification of myelodysplastic syndrome. Leuk Res. 2018;69:54-59. 24. Parmentier S, Schetelig J, Lorenz K, et al. Assessment of dysplastic hematopoiesis: lessons from healthy bone marrow donors. Haematologica. 2012;97(5):723-730. 25. Willig TN, Draptchinskaia N, Dianzani I, et al. Mutations in ribosomal protein S19 gene and diamond blackfan anemia: wide variations in phenotypic expression. Blood. 1999;

haematologica | 2021; 106(1)

94(12):4294-4306. 26. Farrar JE, Vlachos A, Atsidaftos E, et al. Ribosomal protein gene deletions in Diamond-Blackfan anemia. Blood. 2011; 118(26):6943-6951. 27. Ulirsch JC, Verboon JM, Kazerounian S, et al. The genetic landscape of DiamondBlackfan anemia. Am J Hum Genet. 2018;103(6):930-947. 28. Vulliamy T, Beswick R, Kirwan MJ, Hossain U, Walne AJ, Dokal I. Telomere length measurement can distinguish pathogenic from non-pathogenic variants in the shelterin component, TIN2. Clin Genet. 2012;81(1):76-81. 29. Walne AJ, Vulliamy T, Beswick R, Kirwan M, Dokal I. TINF2 mutations result in very short telomeres: analysis of a large cohort of patients with dyskeratosis congenita and related bone marrow failure syndromes. Blood. 2008;112(9):3594-3600. 30. Vaz F, Hanenberg H, Schuster B, et al. Mutation of the RAD51C gene in a Fanconi anemiaâ&#x20AC;&#x201C;like disorder. Nat Genet. 2010;42(5):406-409. 31. Somyajit K, Saxena S, Babu S, Mishra A, Nagaraju G. Mammalian RAD51 paralogs protect nascent DNA at stalled forks and mediate replication restart. Nucleic Acids Res. 2015;43(20):9835-9855. 32. Somyajit K, Subramanya S, Nagaraju G. Distinct roles of FANCO/RAD51C protein in DNA damage signaling and repair: implications for Fanconi anemia and breast cancer susceptibility. J Biol Chem. 2012; 287(5):3366-3380. 33. Savage SA, Ballew BJ, Giri N, et al. Novel FANCI mutations in Fanconi anemia with VACTERL association. Am J Med Genet A. 2016;170A(2):386-391. 34. Mantere T, Haanpaa M, Hanenberg H, et al. Finnish Fanconi anemia mutations and hereditary predisposition to breast and prostate cancer. Clin Genet. 2015;88(1):6873. 35. Dale DC, Person RE, Bolyard AA, et al. Mutations in the gene encoding neutrophil elastase in congenital and cyclic neutropenia. Blood. 2000;96(7):2317-2322. 36. Makaryan V, Zeidler C, Bolyard AA, et al. The diversity of mutations and clinical outcomes for ELANE-associated neutropenia. Curr Opin Hematol. 2015;22(1):3-11. 37. Smith BN, Ancliff PJ, Pizzey A, Khwaja A, Linch DC, Gale RE. Homozygous HAX1 mutations in severe congenital neutropenia patients with sporadic disease: a novel mutation in two unrelated British kindreds.

Br J Haematol. 2009;144(5):762-770. 38. Carlsson G, Elinder G, Malmgren H, et al. Compound heterozygous HAX1 mutations in a Swedish patient with severe congenital neutropenia and no neurodevelopmental abnormalities. Pediatr Blood Cancer. 2009; 53(6):1143-1146. 39. Bellanne-Chantelot C, Schmaltz-Panneau B, Marty C, et al. Mutations in the SRP54 gene cause severe congenital neutropenia as well as Shwachman-Diamond-like syndrome. Blood. 2018;132(12):1318-1331. 40. Carapito R, Konantz M, Paillard C, et al. Mutations in signal recognition particle SRP54 cause syndromic neutropenia with Shwachman-Diamond-like features. J Clin Invest. 2017;127(11):4090-4103. 41. Draptchinskaia N, Gustavsson P, Andersson B, et al. The gene encoding ribosomal protein S19 is mutated in Diamond-Blackfan anaemia. Nat Genet. 1999;21(2):169-175. 42. Cronkhite JT, Xing C, Raghu G, et al. Telomere shortening in familial and sporadic pulmonary fibrosis. Am J Respir Crit Care Med. 2008;178(7):729-737. 43. Ameziane N, Errami A, Leveille F, et al. Genetic subtyping of Fanconi anemia by comprehensive mutation screening. Hum Mutat. 2008;29(1):159-166. 44. Gille JJ, Floor K, Kerkhoven L, Ameziane N, Joenje H, de Winter JP. Diagnosis of Fanconi anemia: mutation analysis by multiplex ligation-dependent probe amplification and PCR-based sanger sequencing. Anemia. 2012;2012:603253. 45. Bottega R, Nicchia E, Cappelli E, et al. Hypomorphic FANCA mutations correlate with mild mitochondrial and clinical phenotype in Fanconi anemia. Haematologica. 2018;103(3):417-426. 46. Morgan NV, Tipping AJ, Joenje H, Mathew CG. High frequency of large intragenic deletions in the Fanconi anemia group A gene. Am J Hum Genet. 1999;65(5):1330-1341. 47. Zheng Z, Geng J, Yao RE, et al. Molecular defects identified by whole exome sequencing in a child with Fanconi anemia. Gene. 2013;530(2):295-300. 48. Wehr C, Grotius K, Casadei S, et al. A novel disease-causing synonymous exonic mutation in GATA2 affecting RNA splicing. Blood. 2018;132(11):1211-1215. 49. Holzinger D, Fassl SK, de Jager W, et al. Single amino acid charge switch defines clinically distinct proline-serine-threonine phosphatase-interacting protein 1 (PSTPIP1)associated inflammatory diseases. J Allergy Clin Immunol. 2015;136(5):1337-1345.

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):74-86

Cell Therapy & Immunotherapy

Expanded circulating hematopoietic stem/ progenitor cells as novel cell source for the treatment of TCIRG1 osteopetrosis

Valentina Capo,1 Sara Penna,1,2 Ivan Merelli,3 Matteo Barcella,1 Serena Scala,1 Luca Basso-Ricci,1 Elena Draghici,1 Eleonora Palagano,4,5 Erika Zonari,1 Giacomo Desantis,1 Paolo Uva,6 Roberto Cusano,6 Lucia Sergi Sergi,1 Laura Crisafulli,4,5 Despina Moshous,7,8 Polina Stepensky,9 Katarzyna Drabko,10 Zühre Kaya,11 Ekrem Unal,12,13 Alper Gezdiric,14 Giuseppe Menna,15 Marta Serafini,2 Alessandro Aiuti,1 Silvia Laura Locatelli,16 Carmelo Carlo-Stella,16,17 Ansgar S. Schulz,18 Francesca Ficara,4,5 Cristina Sobacchi,4,5 Bernhard Gentner1 and Anna Villa1,4

San Raffaele Telethon Institute for Gene Therapy (SR-Tiget), IRCCS San Raffaele Scientific Institute, Milan, Italy; 2DIMET, University of Milano-Bicocca, Monza, Italy; 3Institute for Biomedical Technologies, National Research Council, Segrate, Italy; 4 CNR-IRGB, Milan Unit, Milan, Italy; 5Humanitas Clinical and Research Center - IRCCS, Rozzano, Italy; 6CRS4, Science and Technology Park Polaris, Pula, Italy; 7 Unité d'Immunologie, Hématologie et Rhumatologie Pédiatriques (UIHR), Assistance Publique-Hôpitaux de Paris, Hôpital Necker-Enfants Malades, Paris, France; 8INSERM UMR1163, Institut Imagine, Université Paris Descartes-Sorbonne Paris Cité, Paris, France; 9 Department of Bone Marrow Transplantation and Cancer Immunotherapy, Hadassah University Hospital, Jerusalem, Israel; 10Medical University of Lublin, Lublin, Poland; 11 Department of Pediatric Hematology, Gazi University, School of Medicine, Ankara, Turkey; 12 Erciyes University, Pediatric Hematology Oncology, Kayseri, Turkey; 13Molecular Biology and Genetic Department, Gevher Nesibe Genom and Stem Cell Institution, Genome and Stem Cell Center (GENKOK), Erciyes University, Kayseri, Turkey; 14Department of Medical Genetics, Istanbul Health Science University, Kanuni Sultan Suleyman Training and Research Hospital, Istanbul, Turkey; 15Hemato-Oncology Unit, Department of Oncology, Pausilipon Hospital, Naples, Italy; 16Department of Oncology and Hematology, Humanitas Cancer Center, Humanitas Clinical and Research Center, Rozzano, Italy; 17Department of Biomedical Sciences, Humanitas University, Rozzano, Italy and 18Department of Pediatrics and Adolescent Medicine, University Medical Center, Ulm, Germany 1

ABSTRACT

Correspondence: ANNA VILLA villa.anna@hsr.it Received: September 14, 2019. Accepted: January 9, 2020. Pre-published: January 16, 2020. https://doi.org/10.3324/haematol.2019.238261

llogeneic hematopoietic stem cell transplantation is the treatment of choice for autosomal recessive osteopetrosis caused by defects in the TCIRG1 gene. Despite recent progress in conditioning, an important number of patients are not eligible for allogeneic stem cell transplantation because of the severity of the disease and significant transplant-related morbidity. We exploited peripheral CD34+ cells, known to circulate at high frequency in the peripheral blood of TCIRG1-deficient patients, as a novel cell source for autologous transplantation of gene corrected cells. Detailed phenotypical analysis showed that circulating CD34+ cells have a cellular composition that resembles bone marrow (BM), supporting their use in gene therapy protocols. Transcriptomic profile revealed enrichment in genes expressed by hematopoietic stem and progenitor cells (HSPC). To overcome the limit of BM harvest/HSPC mobilization and serial blood drawings in TCIRG1 patients, we applied UM171-based ex vivo expansion of HSPC coupled with lentiviral gene transfer. Circulating CD34+ cells from TCIRG1defective patients were transduced with a clinically-optimized lentiviral vector expressing TCIRG1 under the control of phosphoglycerate promoter and expanded ex vivo. Expanded cells maintained long-term engraftment capacity and multi-lineage repopulating potential when transplanted in vivo both in primary and secondary NOD scid gamma common chain (NSG) recipients. Moreover, when CD34+ cells were differentiated in vitro, genetically corrected osteoclasts resorbed the bone efficiently. Overall, we provide evidence that expansion of circulating HSPC coupled to gene therapy can overcome the limit of stem cell harvest in osteopetrotic patients, thus opening the way to future gene-based treatment of skeletal diseases caused by BM fibrosis. haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

Introduction Autosomal recessive osteopetrosis (ARO) is a rare and heterogeneous genetic disease, caused by defects in the differentiation or resorption activity of osteoclasts. Patients present with dense and brittle bones, severe anemia, hepatosplenomegaly, macrocephaly, progressive deafness and blindness due to pressure on nerves, and limited bone marrow (BM) cavities.1,2 The incidence of the disease is 1 in 250,000 live births, with higher rates in specific geographic areas where consanguineous marriages are frequent.3 More than 50% of the cases are due to defects in the TCIRG1 gene, encoding for the a3 subunit of ATPase H+ transporting V0 complex, necessary for the acidification of organelles and resorption lacuna.4 The disease is usually lethal in the first decade of life, with poor quality of life. To date, the only curative treatment is hematopoietic stem cell transplantation (HSCT) from an allogeneic donor, which has to be performed as early as possible before compression of nerves and irreversible neurological damage has occurred.5,6 Children with osteopetrosis suffer from high rates of graft failure and transplant-related mortality, mostly due to severe graft-versus-host disease, liver toxicity (veno-occlusive disease), infections or lung toxicity (idiopathic pneumonia syndrome and acute respiratory distress syndrome).7,8 In particular, transplants from HLA-matched related or unrelated donors have an 80-88% 5-year disease-free survival, whereas the success rate is lower for haploidentical transplants (66%).2 Recently, improvements in the outcome and overall survival have been observed in patients treated with fludarabine-based conditioning regimens and T-cell depleted haploidentical donors.9,10 However, HLA-compatible donors are readily available only to a minority of patients. Umbilical cord blood transplantation (UCBT) has also been used as an alternative source, but overall survival at 6 years was 43%, so it is no longer recommended.11 To overcome donor-related issues, gene therapy has been proposed as an alternative strategy. In the past, a retroviral vector, in which the TCIRG1 gene was driven by the strong viral SFFV (spleen focus-forming virus) promoter, has been tested in oc/oc mice, the murine model of TCIRG1-deficient osteopetrosis,12 showing that ex vivo gene therapy for ARO is effective. More recently, TCIRG1-expressing lentiviral vectors, driven by different promoters, were designed and tested in ARO CD34+ cells in vitro and in the oc/oc mouse model.13-15 Since BM harvest cannot be performed in these patients due to severe BM fibrosis and susceptibility to bone fractures, peripheral blood (PB) CD34+ cells represent a potential source of autologous hematopoietic stem and progenitor cells (HSPC). The majority of ARO patients have high frequencies of circulating CD34+ cells, because of the limited BM cavities and the reduction of hematopoietic stem cell (HSC) niches.16,17 Of note, previous studies showed that PB of osteopetrotic patients is highly enriched in cells with myeloid and erythroid clonogenic potential.16,18 However, there is still no detailed characterization of ARO PB CD34+ cell stemness markers, a prerequisite before considering their clinical use. Finally, despite the high frequency of PB CD34+ cells, the amount of collectable HSPC for ex vivo manipulation is constrained by the severity of the disease, the young age of the patients, and the small quantity of blood that can be drawn. Data reported in literature indicate the feasibility of exchange transfusion in osteopetrotic patients as backhaematologica | 2021; 106(1)

up.16 Since gene therapy protocols usually require higher amounts of CD34+ cell/kg, we might speculate that an adequate number of autologous CD34+ cells can be obtained through the collection of both spontaneously circulating and mobilized HSPC. We hypothesized that an efficient expansion of short-term progenitors and HSC may promote the collection of an adequate cell dose, allowing timely hematopoietic recovery and durable engraftment by genetically-engineered cells, respectively. To this end, we tested an HSPC expansion protocol previously used for cord blood (CB), BM or mobilized PB CD34+ cells from healthy donors.19,20 We exploited the pyrimidoindole derivative UM171 to expand ARO-derived PB CD34+ cells with repopulating potential, after transduction with a clinically optimized TCIRG1-expressing lentiviral vector driven by the phosphoglycerate kinase (PGK1) promoter. We demonstrated that transduced and expanded cells generated functional bone-resorbing osteoclasts in vitro. Our deep phenotypic characterization revealed that ARO-derived spontaneously mobilized CD34+ cells contained bona fide primitive HSC, and that the stem cell output and BM homing capacity were maintained in NOD scid gamma common chain (NSG) mice after the expansion protocol. Overall, we have established a novel protocol that will allow transplantation of gene-corrected and expanded PB CD34+ cells in human disorders characterized by BM fibrosis.

Methods Patients and healthy donors Peripheral blood of ARO patients and healthy donors was obtained according to the Declaration of Helsinki with the approval of the local medical ethical committees. A description of patients is provided in Table 1. ARO17 and ARO18 patients have been previously described (patients 13 and 19, respectively).21 Details on healthy donors are reported in the Online Supplementary Methods.

CD34+ isolation and culture A Lymphoprep (STEMCELL Technologies) density gradient was used to isolate PB mononuclear cells (PBMC). CD34+ cells were isolated from PBMC using human CD34 MicroBead Kit and autoMACS Pro Separator (Miltenyi Biotec), according to the manufacturerâ&#x20AC;&#x2122;s instructions. CD34+ were pre-stimulated for 24 hours (h) and transduced with 1-hit of LV at MOI 100 overnight, as previously described.22,23 Hematopoietic progenitor cultures were performed plating 2,500 cells in 2.5 mL MethoCult H4434 Classic methylcellulose-based medium (STEMCELL Technologies) and cultured for 12 days. After transduction, cells were expanded using UM171 compound until day 7, as previously described.19 A fraction of expanded cells was differentiated in vitro towards the myeloid lineage for 1 week and then into osteoclasts for 2 or 3 weeks on plastic wells or bone slices (Immunodiagnostic Systems), as previously described.14

Mice Animal experimental procedures were approved by the Institutional Animal Care and Use Committee of San Raffaele Hospital and the Italian Ministry of Health. NOD scid gamma common chain (NSG) mice, obtained from Charles River Laboratories, were irradiated at 180 RAD and transplanted after 2 h, as detailed in Online Supplementary Table S1. 75

V. Capo et al.

RNAsequencing Total RNA of CD34+ cells was extracted using ReliaPrep RNA Cell Miniprep System (Promega), according to the manufacturer’s instructions. Library generation and data analysis were performed as detailed in the Online Supplementary Methods.

Osteoclast activity Osteoclasts cultured on plastic were stained using the Tartrate Resistant Acid Phosphatase (TRAP) Kit (Sigma-Aldrich), following the manufacturer's instructions. Osteoclasts differentiated

on bone slices were stained using alendronate conjugated to Alexa Fluor 488,24 TRITC-conjugated phalloidin (Sigma-Aldrich) and TO-PRO-3 (Thermo Fisher Scientific). To evaluate bone resorption, the same bone slices were used for toluidine blue staining as previously described.25 TRAP and toluidine blue images were acquired on a Zeiss AxioImager M2m microscope, while immunofluorescence staining images were acquired on a Leica TCS SP5 Laser Scanning Confocal microscope. Quantification of CTX-I was performed on culture supernatants using CrossLaps for Culture (CTX-I) ELISA

Table 1. Patients’ characteristics.

% CD34+ in WB

Patient

Sample type

Age

TCIRG1 mutation

ARO2

Peripheral blood

5.5 months

ARO6

Frozen PBMC

11 months

ARO7

Peripheral blood

8 months

ARO9

Frozen PBMC

12 months

ARO12

Frozen PBMC

2 months

ARO13

Peripheral blood

5 months

ARO16

Frozen PBMC

10 months

ARO17

Frozen PBMC

7 months

ARO18

Frozen PBMC

7 months

ARO19

Peripheral blood

12 months

c.713+1G>A, r.spl; c.1555-2A>C, r.spl c.1674-1G>A, r.spl; c.1674-1G>A, r.spl c.2321C>G, p.(Pro774Arg); c.2321C>G, p.(Pro774Arg) c.1874-1G>A, r.spl; c.1874-1G>A, r.spl c.2233+1G>A, r.spl; c.2233+1G>A, r.spl c.2233+1G>A, r.spl; c.2233+1G>A, r.spl c.2233+1G>A, r.spl; c.2233+1G>A, r.spl c.1875C>A, p.(Tyr625X); c.1875C>A, p.(Tyr625X) c.1228G>A, p.(Gly410Arg); c.2374G>C, p.(Gly792Arg) r.223_224inst; r.223_224inst

ARO20

Peripheral blood Peripheral blood

13 months 7 months

ARO21

Peripheral blood

2.5 months

ARO25

Frozen PBMC

3.5 months

ARO26

Peripheral blood Peripheral blood

5.5 months 8 months

ARO27 ARO28

Peripheral blood Peripheral blood

8 months 6 months

ARO29

Peripheral blood

6.5 months

c.2005C>T, p.(Arg669X); r.694_695insa c.1682delinsTT, p.(Gly561Valfs); c.2383_2384del, p.(Ala796Leufs) g.8280_9560del; g.8280_9560del c.2218_2219del, (p.Leu740Glufs); c.2218_2219del, (p.Leu740Glufs) c.1230del; c.1230del c.1213G>A, p.(Gly405Arg); c.1674-1G>A, r.spl c.1674-1G>A, r.spl; c.1674-1G>A, r.spl

CD34+/uL

Blood volume (mL)

Isolated CD34+

4.97

1831

2.0

3.53E+05

0.94

2.20E+04

1.59

282

3.2

2.70E+05

4.26

5.00E+05

7.52

1.30E+05

5.90

539

6.7

1.90E+05

1.93

2.00E+05

1.61

1.60E+05

8.91

1.16E+06

2.90

1466

6.0

1.89E+06

0.10

8.8 7.0

1.08E+06 4.10E+05

14.45

8.00E+05

6.70

1.02E+05

4.07

1353 899

16.5 5.6

1.11E+06 WBD only

5.20 3.23

770 341

5.6 2.7

WBD only WBD only

2.40

458

3.0

WBD only

PBMC: peripheral blood mononuclear cells; WB: whole blood; WBD: whole blood dissection analysis. na: not available. Accession number of the TCIRG1variant1 cDNA: NM_006019; the numbering used starts with nucleotide +1 for the A of the ATG-translation initiation codon. Sequence variant nomenclature conforms to the Human Genome Variation Society (HGVS) recommendations for the description of sequence variants.51

haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

Figure 1. Circulating CD34+ cells. (A) The graph shows the percentage of CD34+ in the peripheral blood (PB) of ARO patients, divided in the indicated age groups. (B) Pie charts show distribution of 25 subsets within the CD45+ gate in ARO patient PB (n=5), healthy donor PB (n=4) and bone marrow (BM) (n=7). Black outer ticks indicate 100 cells/mL. Percentages indicate the frequency of hematopoietic stem and progenitor cell (HSPC) population (CD34+ LIN–) on total CD45+ cells. Stacked bar graphs indicate the absolute cell count/mL of the HSPC compartment, composed of ten primitive subsets: CMP: common myeloid progenitors; DC: dendritic cells; Ep: erythroid progenitors; ETP: early T progenitors; GMP: granulocyte-monocyte progenitors; HSC: hematopoietic stem cells; iPMN: immature polymorphonucleated cells; MEP: megakaryo-erythroid progenitors; MKp: megakaryocyte progenitors; MLP: multi-lymphoid progenitors; MPP: multipotent progenitors; NKT: natural killer-T cells; NK: natural killer cells; PMN: mature polymorphonucleated cells. (C-E) Graphs show cell count/mL of HSC (C), MPP (D), and MLP (E) in ARO patient PB, healthy donor PB and BM. Data show mean±standard deviation (A and B) or mean±standard error of mean (C-E). Statistical significance was determined by non-parametric Mann Whitney test (A) or Kruskal-Wallis test with Dunn’s multiple comparison post test (C-E).

haematologica | 2021; 106(1)

V. Capo et al.

Figure 2. RNA sequencing. (A) Principal component analysis of ARO, cord blood (CB) and mobilized peripheral blood (mPB) samples. (B) Heatmaps show unsupervised hierarchical clustering on differentially expressed genes (DEG) of ARO patients versus CB (left) or mPB (right). (C) Heatmap shows normalized enrichment score (NES) from gene set enrichment analysis (GSEA) in the indicated categories.

haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

Figure 3. Transduction and expansion. (A) Schematic representation of the two vector constructs driving TCIRG1 expression: the PGK.TCIRG1 vector and PGK.TCIRG1/dNGFR bidirectional vector. LTR: long terminal repeat; PGK: phosphoglycerate kinase promoter; TCIRG1: T-cell immune regulator 1; dNGFR: deleted nerve growth factor receptor; CMV: cytomegalovirus promoter. (B) Time-line of transduction, expansion and in vivo and in vitro studies. (C) Absolute counts of hematopoietic progenitor colonies obtained in MethoCult cultures. BFU-E: burst-forming unit of erythroid cells; CFU-GM: colony-forming unit of granulocyte/macrophage. Data show mean±standard deviation.

(Immunodiagnostic Systems), according to the manufacturer’s instructions.

Statistical analysis Results are shown as mean±standard deviation (SD) or standard error of the mean (SEM). Statistical significance was determined by non-parametric Mann Whitney test, Kruskal-Wallis test, oneway ANOVA, two-way ANOVA, Dunn’s post-test, Bonferroni post-test or Spearman correlation, as detailed in figure legends. *P<0.05, ** P<0.01, ***P<0.001, ****P<0.0001. Additional methods are presented in the Online Supplementary Methods.

Results High number of circulating stem/progenitor cells in TCIRG1-defective autosomal recessive patients Human HSPC are known to circulate at low frequencies in steady-state PB of healthy adult donors (0.11%±0.01).26 Interestingly, patients affected by TCIRG1-defective ARO show high CD34+ cell counts in the PB,16 due to the reduced BM cavities. To confirm this finding in our cohort of patients, we analyzed the frequencies of circulating CD34+ cells in 17 TCIRG1-mutated patients. The vast majority had elevated CD34+ frequencies, with age-dependent differences (Figure 1A). In particular, ARO patients under 6 months of age had a mean percentage of 4.57%±2.3, while children of 7-12 months showed 2.38%±1.7 CD34+ cells. We further characterized the composition of ARO PB haematologica | 2021; 106(1)

CD34+ cells by deep multi-parametric phenotyping, since CD34+ cells are composed by subpopulations with distinct differentiation and survival capability. Blood samples from five ARO patients (age range 5.5-8 months, mean 6.8 months) were analyzed using whole blood dissection (WBD) technology, a novel flow cytometry protocol, which evaluates and quantifies all major hematopoietic lineages including ten distinct HSPC subpopulations circulating in the periphery.27 As reference, we also analyzed PB samples from four age-matched pediatric healthy donors (age range: 4-7 months, mean: 5.3 months) and BM samples of seven pediatric healthy donors (age range: 3-17 years, mean: 10.5 years) (Figure 1B). The hematopoietic composition of PB from TCIRG1-deficient patients was more similar to pediatric BM than to PB from age-matched healthy children, as evidenced by the presence of classical BM-restricted progenitors such as erythroblasts, myeloblasts and B-cell progenitors (Online Supplementary Figure S1). Quantitative analysis of the absolute number/mL of various HSPC subpopulations further confirmed similarities with BM. Importantly, we found that ARO patients display a content of primitive phenotypic HSC (CD90+ CD45RA–) comparable to BM of pediatric HD (Figure 1C). Additionally, ARO patients showed increased number of primitive multipotent progenitors (MPP) and multi-lymphoid progenitors (MLP) with respect to BM of healthy children and, most importantly, to the PB of healthy age-matched controls (Figure 1D and E). Next, we performed whole trascriptome profiling through RNA sequencing (RNA-seq) to determine the 79

V. Capo et al.

Figure 4. Hematopoietic stem and progenitor cells (HSPC) expansion. (A) Fold expansion, calculated as a ratio between total CD34+ cells at the start of the culture and at the end of expansion phase in untransduced (UT) or transduced (PGK.TCIRG1 and PGK.TCIRG1/dNGFR) patient cells. (B) Absolute counts of the CD34high CD90+ endothelial protein C receptor-positive (EPCR+) population, which is enriched for the long-term hematopoietic stem cells (HSC), at day 0 (freshly isolated CD34+ cells) and day 7 (end of expansion phase). Each color represents a different untransduced patient, cultured with 35 or 100 nM UM171, as indicated in figure. (C) Frequency of the CD90+ EPCR+ population on the CD34high cells at day 0 and at day 7. As in (A), each color represents a different untransduced patient, cultured with 35 or 100 nM UM171, as indicated in figure. Statistical significance was determined by non-parametric one-way ANOVA with Dunns post test (A) or non-parametric Mann Whitney test (B and C). *P<0.05, **P<0.01. Data show meanÂąstandard deviation.

transcriptional signature of ARO CD34+ cells, in comparison to CD34+ cells circulating in CB and mobilized PB (mPB) healthy donor cells, the two main sources for HSCT. Explorative data analysis performed by principal component analysis (PCA) revealed three clusters (ARO, CB and mPB) corresponding to the cell source (Figure 2A). ARO samples are closer to CB cell cluster rather than to mPB. Using unsupervised hierarchical clustering on differentially expressed genes (DEG), ARO patients clustered together when compared to both CB and mPB (Figure 2B and Online Supplementary Figure S2), confirming the unique nature of the osteopetrotic circulating CD34+ cells. To study the HSPC composition at the RNA level, we used different gene lists reported in literature.28-36 Gene set enrichment analysis (GSEA) revealed that ARO CD34+ cells were enriched for committed progenitors (MLP, CLP 80

and GMP) in comparison to mPB, similarly to CB. No definitive conclusions could be drawn on the most primitive HSC population, since heterogeneous results were obtained depending on the gene set used (Figure 2C). Overall, these analyses confirmed that CD34+ cells are present at high numbers in the PB of TCIRG1-defective patients and that these cells are enriched for primitive subsets both phenotypically and transcriptionally, providing the first indication that unmobilized PB could be a suitable stem cell source for gene therapy.

Transduction and expansion of TCIRG1-defective CD34+ cells We designed a gene therapy approach based on the transduction with TCIRG1-expressing LV coupled with a HSPC expansion protocol, to overcome the limitation haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

Figure 5. Osteoclast bone resorption. (A) Immunofluorescence staining of osteoclasts differentiated from CD34+ cells and cultured on bone slices. Images show the merge result of alendronate-AF-488 (resorption pits in green), phalloidin-TRITC (actin rings in red) and TO-PRO-3 (nuclei in blue). (Top) Representative images of osteoclasts obtained from mobilized peripheral blood (mPB) CD34+ cells from two healthy donors. (Bottom) Representative images of untransduced (UT) or transduced (PGK.TCIRG1) osteoclasts from seven ARO patients. Vector copy numbers (VCN) of the transduced cells are indicated for each patient. Images were acquired with Leica TCS SP5 laser scanning confocal, equipped with HC PL FLUOTAR 10X (NA 0.3) dry and Leica Application Suite Advanced Fluorescence (LASAF) software. (B) Toluidine blue staining of bone slices that were previously cultured with osteoclasts from healthy donor mPB CD34+ cells and ARO patient UT or transduced (PGK.TCIRG1) cells. VCN of the transduced cells are shown for each patient. Images were acquired with Nikon ECLIPSE E600 microscope equipped with Nikon DSRi2 camera, using Plan Fluor 4x/0.13 objective and NIS-Elements F 4.30.01 software. (C) C-terminal telopeptide fragment of type I collagen (CTX-I) quantification in the osteoclast culture supernatant at day 0, day 8, and day 21. Healthy donor mPB (mPB HD) CD34+ cells in white, ARO patient untransduced (ARO UT) cells in gray and ARO patient transduced (ARO PGK.TCIRG1) in black. Statistical significance was determined by two-way ANOVA with Bonferroni post-test (C). ****P<0.0001. All data show meanÂąstandard.

haematologica | 2021; 106(1)

V. Capo et al. A

Figure 6. Transplants in primary NOD scid gamma common chain (NSG) recipients. (A) Percentage of human hematopoietic (huCD45+) cells in the peripheral blood (PB) of primary NSG recipients over time. Each color represents a different patient, as indicated in figure. UM171 dosage used for CD34+ cell expansion is also indicated. Continuous line indicates untransduced (UT) cells, dotted line indicates cells transduced with PGK.TCIRG1 vector and dashed line indicates cells transduced with PGK.TCIRG1/dNGFR vector. (B) Percentage of human hematopoietic (huCD45+) cells on total CD45+ cells in spleen, bone marrow (BM) and thymus at sacrifice. (C) Percentage of CD19+ (B lymphocytes, in blue), CD3+ (T lymphocytes, in green), CD13+ (myeloid cells, in pink), and CD34+ (hematopoietic stem and progenitor cells, in yellow) subsets on the total human CD45+ cells. (D) Vector copy number (VCN) of in vitro transplanted cells and in tissues (spleen, BM and thymus). Data show mean (A) or meanÂąstandard deviation (B and C).

imposed by the restricted amount of TCIRG1-defective CD34+ cells. In order to reduce the total culture period, we adapted the gene-correction protocol currently in use for gene therapy clinical trials,22,23 using only 1 hit of LV transduction. We designed two different LV expressing TCIRG1 cDNA under the control of the phosphoglycerate kinase (PGK1) promoter (Figure 3A). The first vector (PGK.TCIRG1) was designed as a clinically-applicable vector, using the same backbone currently used in clinical trials. The second LV (PGK.TCIRG1/dNGFR) has a bidirectional design containing a marker gene, allowing easy detection/selection of transduced cells for research purposes.37 Circulating CD34+ cells, isolated from 2-16.5 mL of peripheral blood from 12 TCIRG1-mutated ARO patients (Table 1), were pre-stimulated with cytokines and then transduced either with PGK.TCIRG1 or PGK.TCIRG1/dNGFR (Figure 3B). Similar frequencies of 82

burst-forming unit of erythroid cells (BFU-E) and colonyforming unit of granulocyte/macrophage (CFU-GM) colonies in transduced and non-transduced cells (Figure 3C) demonstrated that transduction did not impact on the clonogenic potential of HSPC. To compensate for HSPC shortage and the difficulties in collecting them from the PB of very young patients, we exploited a recently described ex vivo expansion protocol based on UM171 small molecule.19,20,38 After transduction, cells were expanded for 5 additional days in the presence of early-acting cytokines and the pyrimidoindole derivative UM171, reaching a total culture time of 7 days (Figure 3B), as practised in the UM171-based CB expansion trials.39 In line with published data,38 UM171 exposure induced the expansion of the total CD34+ cells. We observed a 10.4-mean-fold expansion, with differences probably due to patient-to-patient variability, regardless of UM171 dose haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

Figure 7. Transplants in secondary NOD scid gamma common chain (NSG) recipients. (A) Percentage of human hematopoietic (huCD45+) cells in the peripheral blood (PB) of secondary NSG recipients over time. Each color represents a different patient, as indicated in figure. Continuous line indicates untransduced (UT) cells, dotted line indicates cells transduced with PGK.TCIRG1 vector and dashed line indicates cells transduced with PGK.TCIRG1/dNGFR vector. (B) Percentage of human hematopoietic (huCD45+) cells on total CD45+ cells in spleen, bone marrow (BM) and thymus at sacrifice. (C) Percentage of CD19+ (B lymphocytes, in blue), CD3+ (T lymphocytes, in green), CD13+ (myeloid cells, in pink), and CD34+ (hematopoietic stem and progenitor cells, in yellow) subsets on the total human CD45+ cells. (D) Vector copy number (VCN) of in vitro transplanted cells and in tissues (spleen, BM and thymus). Data show mean (D) or meanÂąstandard deviation (B and C).

(Figure 4A). Cells exposed to the TCIRG1-expressing LV showed a similar fold expansion as compared to their untransduced counterparts, confirming that UM171 expansion protocol can also be used in the setting of gene therapy.19 Importantly, the HSC-containing CD34+ CD90+ endothelial protein C receptor-positive (EPCR+) cell subpopulation expanded both in terms of absolute counts (Figure 4B) and relative frequency within the CD34+ population (Figure 4C and Online Supplementary Figure S3A and B). The triple positive CD34+ CD90+ EPCR+ population was present at comparable frequency in both transduced and untransduced cells (Online Supplementary Figure S3C).

In vitro correction of osteoclast defect Transduced and expanded CD34+ cells were differentiated in vitro into myeloid progenitors and then into osteoclasts, in order to assess the correction of osteoclast resorptive function. As expected, osteoclasts derived from TCIRG1-mutated haematologica | 2021; 106(1)

patients were able to normally fuse and differentiate into mature, multinucleated, tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts, although not functional (Online Supplementary Figure S4). We verified that osteoclasts from ARO patients did not express TCIRG1 protein. Transduction with the LV PGK.TCIRG1 restored the expression of the TCIRG1 protein, as assessed by western blot (Online Supplementary Figure S5). To assess the functionality of gene-corrected cells, we cultured osteoclasts derived from untransduced and transduced patient cells on bone slices and analyzed the bone resorption by immunofluorescence and confocal microscopy after 2-3 weeks. The assembled nuclei (stained with TO-PRO-3) and the presence of actin rings (stained with phalloidin) confirmed osteoclast differentiation on the bone slices (Figure 5A). Untransduced patient cells were not able to resorb bone slices, as shown by the absence of alendronate-positive resorption pits. Conversely, gene-corrected patient-derived CD34+ were 83

V. Capo et al.

able to form fully functional osteoclasts, even at low vector copy number/genome (VCN). These findings were confirmed by toluidine blue staining, that was performed on bone slices after cell removal. Resorption pits were visible only on bone discs in the presence of patient corrected cells (Figure 5B). To quantify the resorption levels, we measured the C-terminal telopeptide fragment of type I collagen (CTX-I), a degradation marker of collagen released during bone and cartilage resorption. We quantified the CTX-I concentration in the culture supernatant at day 0 (start of the culture), at day 8 (intermediate time point), and at day 21 (end of the culture) by ELISA. Transduced patient cells resorbed bone slices at levels comparable to healthy donor cells at both days 8 and 21. Conversely, untransduced patient cells were not able to resorb bone (Figure 5C). The CTX-I levels did not show a statistically significant correlation with the VCN, probably due to the variability of the osteoclast differentiation assay (Online Supplementary Figure S6). However, high levels of resorption were found also in cells with low VCN, supporting the hypothesis that osteoclasts may be functional even in case of a minor fraction of corrected nuclei.

In vivo repopulating capacity of expanded CD34+ cells To evaluate the capability of gene-corrected AROCD34+ cells to engraft and to fully differentiate in all hematopoietic lineages, we transplanted a median of 0.75x106 expanded CD34+ cells (range: 0.5-1x106 cells) in NSG mice, which corresponded to 0.11-0.47 day0 equivalents per mouse (Figure 3B and Online Supplementary Table S1). Mice were bled at 6 and 10 weeks, and euthanized at 13 weeks to study human cell engraftment. We observed the presence of stable human hematopoietic (CD45+) multilineage cell engraftment in PB, starting from 6 weeks after transplant (Figure 6A). The engraftment showed donor-dependent variability, not correlating with the transplanted day 0 equivalents, the transduction status of the cells, or the UM171 dosage. At 13 weeks post transplant, we sacrificed the mice and analyzed the hematopoietic organs. In accordance to the results obtained in the PB, we observed similar level of human CD45+ cell engraftment in the spleen and in the BM. In the thymus, nearly 100% of the cells were of human origin as expected (Figure 6B). The majority of human cells were B cells (CD19+) as a consequence of the human cell differentiation bias in NSG mice. Nonetheless, we could observe the presence of myeloid (CD13+) and T (CD3+) cells in the hematopoietic organs and also of CD34+ HSPC in the BM (Figure 6C). Importantly, transduced cells were retrieved in the analyzed tissues, in line with the VCN of the in vitro cultures (Figure 6D), indicating long-term engraftment ability by TCIRG1-expressing HSPC. After sacrifice of NSG mice, human CD34+ cells were re-isolated from the BM and transplanted in secondary NSG recipients. We isolated 0.7-1x106 CD34+ cells from each mouse and transplanted them into a different recipient. We observed low but stable human engraftment overtime in secondary recipients up to 13 weeks after transplantation, indicating that our transduction and expansion strategy allowed the maintenance of long-term HSC (Figure 7A). Similarly to the results obtained in primary recipients, variable CD45+ frequencies were observed in the spleen and in the BM, whereas the thymus showed nearly 100% of human CD45+ cells (Figure 7B). In the sec84

ondary grafts, multi-lineage cell repopulation was observed, with the expected prevalence of CD19+ cells (B lymphocytes) and the presence of CD34+ cells in the BM (Figure 7C). VCN in the hematopoietic organs was consistent with the transduction levels of the primary recipients (Figure 7D), indicating the long-term maintenance of genecorrected primitive HSPC.

Discussion Up to now, hematopoietic stem cell transplantation has been the treatment of choice in patients with severe forms of autosomal recessive osteopetrosis, a rare genetic disease characterized by defective osteoclast function and BM fibrosis. Infusion of autologous gene corrected HSPC may represent an attractive therapy to avoid the risk of severe graft-versus-host-disease reactions and limit complications caused by intensive myeloablative conditioning. Gene therapy would also allow treatment of patients without compatible donors or whose severe clinical conditions and/or age preclude conventional therapy.7,40 In addition, mild forms of TCIRG1 osteopetrosis have been recently identified in adult patients, raising concerns about the risk of life-threatening complications during conventional therapy.41 However, contrary to other inherited diseases treated with gene therapy,42 the development of novel strategies is required in osteopetrosis to overcome clinical limitations that may hamper gene therapy applicability.43 For this reason, we developed a tailored approach for the treatment of TCIRG1-mutated osteopetrosis exploiting autologous HSPC gene correction and expansion. Although previous studies have demonstrated the feasibility of gene therapy,12-15 its clinical applicability remains constrained by the scarcity of a BM niche and the limited amount of blood that can be drawn in these patients to collect circulating hematopoietic precursors. Remarkably, CD34+ cells isolated from PB of ARO patients have been reinfused successfully in two transplanted osteopetrotic patients with no engraftment or other complications, providing clinical evidence that circulating CD34+ cells can engraft and speed count recovery.16 In spite of these observations, no studies on the phenotypic composition and transcriptome of these spontaneously mobilized HSPC have been published. To assess relative frequencies and absolute counts/mL of blood cellular components, we took advantage of a multiparametric flow cytometry analysis that allows the dissection of 23 different blood cell types, including HSPC, myeloid and lymphoid progenitors.27 Notably, HSC counts/mL were comparable to those observed in the BM of healthy children, supporting the feasibility of exploiting non-mobilized PB as an easily accessible source for hematopoietic stem cells in these patients. These results were corroborated by the transcriptomic profile of circulating CD34+ cells from ARO patients, showing a positive enrichment for committed progenitor signatures, in particular for granulocyte-monocyte progenitor (GMP) cells. As recently described,44 osteoclast defects can be rescued by monocytic cell transfusion in osteopetrotic mice. The simultaneous presence of HSC and myeloid lineage cells makes the circulating CD34+ cells a favorable cell source for gene therapy. The heterogeneity of the population, containing both committed haematologica | 2021; 106(1)

Circulating HSPC expansion to treat osteopetrosis

and primitive progenitors, would allow for an initial reconstitution after transplant sustained by differentiated progenitors, followed by a later output of transduced cells from long-term engrafting cells.45 The long-term survival capability of PB ARO-CD34+ cells is supported by our in vivo transplantation analyses, indicating that expanded gene-modified PB HSPC from ARO patients are able to maintain long-term multi-lineage engineered hematopoiesis both in primary and in secondary transplanted mice. To further enhance the number of HSPC that can be collected by exchange transfusion, here we proposed an innovative strategy coupling transduction protocol with efficient ex vivo expansion of genetically modified HSC. We exploited the small molecule pyrimidoindole derivate UM171, that has been demonstrated to stimulate the in vitro expansion of human HSC and to enhance lentiviral transduction efficiency of CB derived CD34+ cells, maintaining their short- and long-term repopulating potential.19,20,38,46 The availability of UM171-expanded gene corrected CD34+ cells may allow the cryopreservation of corrected HSPC as backup for potential repeated administration. We can speculate that patients with low chimerism may benefit from repeated infusions in the absence of conditioning, as modeled in osteopetrotic mice.44,47 Remarkably, we demonstrated multi-lineage engraftment of PB CD34+ cells in immunodeficient mice upon ex vivo manipulation, including transduction and/or expansion. Of note, for the first time, we assessed the maintenance of long-term repopulating capacity in NSG secondary recipients of UM171-expanded CD34+ cells, indicating the presence of bona fide LT-HSC in the expanded and corrected cell population. In order to pursue the clinical applicability of our approach, we designed a TCIRG1-expressing LV based on the vector approved and currently in use in the clinical trial for mucopolysaccharidosis type I (clinical trial TigetT10_MPSIH; clinicaltrials.gov identifier: NCT03488394).48,49 In our in vitro studies TCIRG1-corrected CD34+ with low VCN are able to resorb bone substrate, indicating that a limited number of corrected nuclei could provide for adequate protein levels in the mature multinucleated osteoclast. These results are in line with previous studies showing in vivo bone resorption in oc/oc mice with 1-3% of BM engraftment, likely due to the

References 1. Teti A, Teitelbaum SL. Congenital disorders of bone and blood. Bone. 2019;11:971-81. 2. Sobacchi C, Schulz A, Coxon FP, Villa A, Helfrich MH. Osteopetrosis: genetics, treatment and new insights into osteoclast function. Nat Rev Endocrinol. 2013;9(9):522-536. 3. Palagano E, Menale C, Sobacchi C, Villa A. Genetics of osteopetrosis. Curr Osteoporos Rep. 2018;16(1):13-25. 4. Frattini A, Orchard PJ, Sobacchi C, et al. Defects in TCIRG1 subunit of the vacuolar proton pump are responsible for a subset of human autosomal recessive osteopetrosis. Nat Genet. 2000;25(3):343-346. 5. Wu CC, Econs MJ, DiMeglio LA, et al. Diagnosis and management of osteopetrosis: consensus guidelines from the osteopet-

haematologica | 2021; 106(1)

fusion of corrected hematopoietic progenitors to uncorrected cells.14,47,50 Importantly, the unregulated expression of TCIRG1 protein driven by the cellular ubiquitous PGK promoter does not impact on the clonogenic potential of HSPC, as shown by methylcellulose cultures. In conclusion, here we provide a clinically applicable multi-step approach for autosomal recessive osteopetrosis based on the use of an optimized vector, an accessible HSPC source and reliable HSC expansion culture conditions. Our strategy may overcome all the major limitations related to the low number of CD34+ cells retrievable from ARO patients. Namely, PB CD34+ cells with primitive phenotype were easily collected and transduced with the corrective PGK.TCIRG1 vector. Then, the UM171based HSC expansion culture overcame the limited access to high blood volumes in severely affected children. These major improvements would allow the implementation of a clinical trial for autosomal recessive osteopetrosis in the next future and the exploitation of circulating CD34+ cells in other clinical conditions characterized by BM fibrosis. Disclosures No Conflicts of interest to disclose. Contributions VC, SP, SS, LBR, ED, LSS and LC performed experiments and analyzed the data; EP, EZ, GD and MS provided intellectual input, reagents, and protocols; IM, MB, PU and RC performed and analyzed RNAseq experiments; DM, PS, KD, ZK, EU, AG, GM, AA, SLL, CCS and AS provided patient samples and data; VC, SP, FF, CS and AV contributed to write the manuscript; BG and AV designed and coordinated the research. All authors critically revised the manuscript. Acknowledgments The authors would like to thank Miep Helfrich for helping with immunofluorescence and for valuable discussion. We would also like to thank ALEMBIC, an advanced microscopy laboratory established by IRCCS Ospedale San Raffaele and UniversitĂ Vita-Salute San Raffaele. Funding This research was supported by a grant from the Telethon Foundation (TGT16C05) to AV and partially by a fellowship from the European Calcified Tissue Society (ECTS) to VC.

rosis working group. J Clin Endocrinol Metab. 2017;102(9):3111-3123. 6. Stepensky P, Grisariu S, Avni B, et al. Stem cell transplantation for osteopetrosis in patients beyond the age of 5 years. Blood Adv. 2019;3(6):862-868. 7. Teti A, Econs MJ. Osteopetroses, emphasizing potential approaches to treatment. Bone. 2017;102:50-59. 8. Steward CG. Hematopoietic stem cell transplantation for osteopetrosis. Pediatr Clin North Am. 2010;57(1):171-180. 9. Orchard PJ, Fasth AL, Le Rademacher JL, et al. Hematopoietic stem cell transplantation for infantile osteopetrosis. Blood. 2015; 126(2):270-276. 10. Schulz AS, Moshous D, Steward CG, Villa A, Sobacchi C. Osteopetrosis consensus guidelines for diagnosis, therapy and followup. ESID EBMT. 2015;1-35.

11. Chiesa R, Ruggeri A, Paviglianiti A, et al. Outcomes after unrelated umbilical cord blood transplantation for children with osteopetrosis. Biol Blood Marrow Transplant. 2016;22(11):1997-2002. 12. Johansson MK, De Vries TJ, Schoenmaker T, et al. Hematopoietic stem cell-targeted neonatal gene therapy reverses lethally progressive osteopetrosis in oc/oc mice. Blood. 2007;109(12):5178-5185. 13. Moscatelli I, LĂśfvall H, Schneider Thudium C, et al. Targeting NSG mice engrafting cells with a clinically applicable lentiviral vector corrects osteoclasts in infantile malignant osteopetrosis. Hum Gene Ther. 2018; 29(8):938-949. 14. Moscatelli I, Thudium CS, Flores C, et al. Lentiviral gene transfer of TCIRG1 into peripheral blood CD34+ cells restores osteoclast function in infantile malignant

V. Capo et al. osteopetrosis. Bone. 2013;57(1):1-9. 15. Löfvall H, Rothe M, Schambach A, Henriksen K, Richter J, Moscatelli I. Hematopoietic stem cell-targeted neonatal gene therapy with a clinically applicable lentiviral vector corrects osteopetrosis in oc/oc mice. Hum Gene Ther. 2019; 30(11):1395-1440. 16. Steward CG, Blair A, Moppett J, et al. High peripheral blood progenitor cell counts enable autologous backup before stem cell transplantation for malignant infantile osteopetrosis. Biol Blood Marrow Transplant. 2005;11(2):115-121. 17. Blin-Wakkach C, Rouleau M, Wakkach A. Roles of osteoclasts in the control of medullary hematopoietic niches. Arch Biochem Biophys. 2014;561:29-37. 18. Marcus JR, Fibach E, Aker M. Circulating myeloid and erythroid progenitor cells in malignant osteopetrosis. Acta Haematol. 1982;67(3):185-189. 19. Zonari E, Desantis G, Petrillo C, et al. Efficient ex vivo engineering and expansion of highly purified human hematopoietic stem and progenitor cell populations for gene therapy. Stem Cell Reports. 2017; 8(4):977-990. 20. Fares I, Chagraoui J, Gareau Y, et al. Pyrimidoindole derivatives are agonists of human hematopoietic stem cell self-renewal. Science. 2014;345(6203):1509-1512. 21. Mazzolari E, Forino C, Razza A, Porta F, Villa A, Notarangelo LD. A single-center experience in 20 patients with infantile malignant osteopetrosis. Am J Hematol. 2009;84(8):473-479. 22. Aiuti A, Biasco L, Scaramuzza S, et al. Lentiviral hematopoietic stem cell gene therapy in patients with Wiskott-Aldrich syndrome. Science. 2013;341(6148):12331511233151. 23. Biffi A, Montini E, Lorioli L, et al. Lentiviral hematopoietic stem cell gene therapy benefits metachromatic leukodystrophy. Science. 2013;341(6148):1233158-1233158. 24. Coxon FP. Fluorescence imaging of osteoclasts using confocal microscopy. Methods Mol Biol. 2012;816:401-424. 25. Menale C, Campodoni E, Palagano E, et al. Mesenchymal stromal cell-seeded biomimetic scaffolds as a factory of soluble RANKL in Rankl-deficient osteopetrosis. Stem Cells Transl Med. 2019;8(1):22-34. 26. Cimato TR, Furlage RL, Conway A, Wallace PK. Simultaneous measurement of human hematopoietic stem and progenitor cells in blood using multicolor flow cytometry.

Cytometry B Clin Cytom. 2016;90(5):415423. 27. Basso-Ricci L, Scala S, Milani R, et al. Multiparametric whole blood dissection: a one-shot comprehensive picture of the human hematopoietic system. Cytometry A. 2017;91(10):952-965. 28. Ramalho-Santos M, Yoon S, Matsuzaki Y, Mulligan RC, Melton DA. “Stemness”: transcriptional profiling of embryonic and adult stem cells. Science. 2002;298(5593):597-600. 29. Jaatinen T, Hemmoranta H, Hautaniemi S, et al. Global gene expression profile of human cord blood-derived CD133+ cells. Stem Cells. 2005;24(3):631-641. 30. Eppert K, Takenaka K, Lechman ER, et al. Stem cell gene expression programs influence clinical outcome in human leukemia. Nat Med. 2011;17(9):1086-1094. 31. Georgantas RW, Tanadve V, Malehorn M, et al. Microarray and serial analysis of gene expression analyses identify known and novel transcripts overexpressed in hematopoietic stem cells. Cancer Res. 2004; 64(13):4434-4441. 32. Doulatov S, Vo LT, Chou SS, et al. Induction of multipotential hematopoietic progenitors from human pluripotent stem cells via respecification of lineage-restricted precursors. Cell Stem Cell. 2013;13(4):459-470. 33. Laurenti E, Doulatov S, Zandi S, et al. The transcriptional architecture of early human hematopoiesis identifies multilevel control of lymphoid commitment. Nat Immunol. 2013;14(7):756-763. 34. Park IK, He Y, Lin F, et al. Differential gene expression profiling of adult murine hematopoietic stem cells. Blood. 2002; 99(2):488-498. 35. Zheng S, Papalexi E, Butler A, Stephenson W, Satija R. Molecular transitions in early progenitors during human cord blood hematopoiesis. Mol Syst Biol. 2018; 14(3):e8041. 36. Clien L, Kostadraia M, Martens JHA, et al. Transcriptional diversity during lineage commitment of human blood progenitors. Science. 2014;345(6204):1251033. 37. Amendola M, Venneri MA, Biffi A, Vigna E, Naldini L. Coordinate dual-gene transgenesis by lentiviral vectors carrying synthetic bidirectional promoters. Nat Biotechnol. 2005;23(1):108-116. 38. Fares I, Chagraoui J, Lehnertz B, et al. EPCR expression marks UM171-expanded CD34+ cord blood stem cells. Blood. 2017; 129(25):3344-3351. 39. Cohen S, Roy J, Lachance S, et al. Single

UM171 expanded cord blood permits transplantation of better HLA matched cords with excellent GvHD relapse free survival. Blood. 2018;132(Suppl 1):4658. 40. Askmyr M, Flores C, Fasth A, Richter J. Prospects for gene therapy of osteopetrosis. Curr Gene Ther. 2009;9(3):150-159. 41. Palagano E, Blair HC, Pangrazio A, et al. Buried in the middle but guilty: intronic mutations in the TCIRG1 gene cause human autosomal recessive osteopetrosis. J Bone Miner Res. 2015;30(10):1814-1821. 42. Gawthorpe P. Gene therapy. Nurs Stand. 2003;17(33):29. 43. Penna S, Capo V, Palagano E, Sobacchi C, Villa A. One disease, many genes: implications for the treatment of osteopetroses. Front Endocrinol (Lausanne). 2019;10:85. 44. Jacome-Galarza CE, Percin GI, Muller JT, et al. Developmental origin, functional maintenance and genetic rescue of osteoclasts. Nature. 2019;568(7753):541-545. 45. Scala S, Basso-Ricci L, Dionisio F, et al. Dynamics of genetically engineered hematopoietic stem and progenitor cells after autologous transplantation in humans. Nat Med. 2018;24(11):1683-1690. 46. Ngom M, Imren S, Maetzig T, et al. UM171 enhances lentiviral gene transfer and recovery of primitive human hematopoietic cells. Mol Ther Methods Clin Dev. 2018;10:156164. 47. Flores C, De Vries TJ, Moscatelli I, et al. Nonablative neonatal bone marrow transplantation rapidly reverses severe murine osteopetrosis despite low-level engraftment and lack of selective expansion of the osteoclastic lineage. J Bone Miner Res. 2010; 25(9):2069-2077. 48. Gentner B, Bernardo ME, Zonari E, et al. Exvivo gene therapy for Hurler disease: initial results from a phase I/II clinical study. Mol Ther. 2019;27(4S1):abstr 2. 49. Squeri G, Passerini L, Ferro F, et al. Targeting a pre-existing anti-transgene T cell response for effective gene therapy of MPS-I in the mouse model of the disease. Mol Ther. 2019;27(7):1215-1227. 50. Thudium CS, Moscatelli I, Löfvall H, et al. Regulation and function of lentiviral vectormediated TCIRG1 expression in osteoclasts from patients with infantile malignant osteopetrosis: implications for gene therapy. Calcif Tissue Int. 2016;99(6):638-648. 51. Den Dunnen JT, Dalgleish R, Maglott DR, et al. HGVS recommendations for the description of sequence variants: 2016 update. Hum Mutat. 2016;37:564-569.

haematologica | 2021; 106(1)

ARTICLE

Chronic Lymphocytic Leukemia

Genomic arrays identify high-risk chronic lymphocytic leukemia with genomic complexity: a multicenter study

Alexander C. Leeksma,1,2,3 Panagiotis Baliakas,4 Theodoros Moysiadis,5,6 Anna Puiggros,7,8 Karla Plevova,9,10 Anne-Marie van der Kevie-Kersemaekers,11 Hidde Posthuma,11 Ana E. Rodriguez-Vicente,12 Anh Nhi Tran,6 Gisela Barbany,6 Larry Mansouri,6 Rebeqa Gunnarsson,13 Helen Parker,14 Eva van den Berg,15 Mar Bellido,16 Zadie Davis,17 Meaghan Wall,18 Ilaria Scarpelli,19 Anders Österborg,20,21 Lotta Hansson,20,21 Marie Jarosova,9,10 Paolo Ghia,22 Pino Poddighe,23 Blanca Espinet,7,8 Sarka Pospisilova,9,10 Constantine Tam,24 Loïc Ysebaert,25 Florence Nguyen-Khac,26 David Oscier,17 Claudia Haferlach,27 Jacqueline Schoumans,19 Marian Stevens-Kroef,28 Eric Eldering,2,3 Kostas Stamatopoulos,5,6 Richard Rosenquist,6 Jonathan C. Strefford,14 Clemens Mellink11 and Arnon P. Kater1,3 on behalf of ERIC, the European Research Initiative on CLL

Department of Hematology and Lymphoma and Myeloma Center Amsterdam (LYMMCARE), Amsterdam University Medical Centers, University of Amsterdam, Amsterdam, the Netherlands; 2Department of Experimental Immunology, Amsterdam University Medical Centers, University of Amsterdam, Amsterdam, the Netherlands; 3Cancer Center Amsterdam and Amsterdam Infection and Immunity Institute, Amsterdam, the Netherlands; 4Department of Immunology, Genetics and Pathology, Science for Life Laboratory, Uppsala University, Uppsala, Sweden; 5Institute of Applied Biosciences, Center for Research and Technology Hellas, Thessaloniki, Greece; 6Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm, Sweden; 7Laboratori de Citogenètica Molecular, Servei de Patologia, Hospital del Mar, Barcelona, Spain; 8Grup de Recerca Translacional en Neoplàsies Hematològiques, Programa de Recerca en Càncer, Institut Hospital del Mar d’Investigacions Mèdiques (IMIM), Barcelona, Spain; 9 Department of Internal Medicine Hematology and Oncology, University Hospital Brno and Faculty of Medicine, Masaryk University, Brno, Czech Republic; 10Central European Institute of Technology (CEITEC), Masaryk University, Brno, Czech Republic; 11Department of Clinical Genetics, Amsterdam University Medical Centers, University of Amsterdam, Amsterdam, the Netherlands; 12Department of Hematology, IBSAL, IBMCC, University of Salamanca, CSIC, Cancer Research Center, University Hospital of Salamanca, Salamanca, Spain; 13Division of Clinical Genetics, Department of Laboratory Medicine, Lund University, Lund, Sweden; 14Cancer Genomics, Academic Unit of Cancer Sciences, Faculty of Medicine, University of Southampton, Southampton, UK; 15Department of Genetics, University Medical Center Groningen, University of Groningen, Groningen, the Netherlands; 16Department of Hematology, University Medical Center Groningen, University of Groningen, Groningen, the Netherlands; 17Department of Molecular Pathology, Royal Bournemouth Hospital, Bournemouth, UK; 18Cytogenetics Department, St Vincent Hospital, Melbourne, Victoria, Australia; 19Oncogenomic Laboratory, Department of Hematology, Lausanne University Hospital (CHUV), Lausanne, Switzerland; 20 Department of Oncology-Pathology, Karolinska Institutet, Stockholm, Sweden; 21 Department of Hematology, Karolinska University Hospital, Stockholm, Sweden; 22 Division of Experimental Oncology, IRCCS Ospedale San Raffaele e Università Vita-Salute San Raffaele, Milan, Italy; 23Department of Clinical Genetics, Amsterdam University Medical Centers, Vrije Universiteit, Amsterdam, the Netherlands; 24Department of Haematology, St Vincent Hospital Melbourne and Peter MacCallum Cancer Center; University of Melbourne, Melbourne, Victoria, Australia; 25Institut Universitaire du Cancer de Toulouse-Oncopôle, Toulouse, France; 26INSERM U1138; Université Pierre et Marie Curie-Paris; Service d'Hématologie Biologique, Hôpital Pitié-Salpêtrière, APHP, Paris, France; 27MLL Munich Leukemia Laboratory, Munich, Germany and 28Radboud University Medical Center, Department of Human Genetics, Nijmegen, the Netherlands

Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):87-97

ABSTRACT

omplex karyotype identified by chromosome-banding analysis has been shown to have prognostic value in chronic lymphocytic leukemia (CLL). Genomic arrays offer high-resolution genomewide detection of copy-number alterations (CNA) and could therefore be well equipped to detect the presence of a complex karyotype. Current knowledge on genomic arrays in CLL is based on outcomes of single-cenhaematologica | 2021; 106(1)

Correspondence: ARNON P. KATER a.p.kater@amc.nl Received: October 9, 2019. Accepted: January 22, 2020. Pre-published: January 23, 2020. https://doi.org/10.3324/haematol.2019.239947

A.C. Leeksma et al.

ter studies, in which different cutoffs for CNA calling were used. To further determine the clinical utility of genomic arrays for CNA assessment in CLL diagnostics, we retrospectively analyzed 2,293 arrays from 13 diagnostic laboratories according to established standards. CNA were found outside regions captured by CLL fluorescence in situ hybridization probes in 34% of patients, and several of them, including gains of 8q, deletions of 9p and 18p (P<0.01), were linked to poor outcome after correction for multiple testing. Patients (n=972) could be divided into three distinct prognostic subgroups based on the number of CNA. In multivariable analysis only high genomic complexity, defined as ≥5 CNA, emerged as an independent adverse prognosticator for time to first treatment (hazard ratio: 2.15; 95% confidence interval: 1.36-3.41; P=0.001) and overall survival (hazard ratio: 2.54, 95% confidence interval: 1.54-4.17; P<0.001; n=528). Lowering the size cutoff to 1 Mb in 647 patients did not significantly improve risk assessment. Genomic arrays detected more chromosomal abnormalities and, in terms of risk stratification, performed at least as well as simultaneous chromosome banding analysis as carried out in 122 patients. Our findings indicate that genomic array is an accurate tool for CLL risk stratification.

Introduction Chronic lymphocytic leukemia (CLL) is both clinically and genetically highly heterogeneous. The advent of efficacious, albeit expensive, targeted treatment regimens for CLL has highlighted the need for more accurate risk stratification. Diverse CLL-specific biomarkers are used to predict clinical course and survival in CLL including Rai1 and Binet2 staging, immunoglobulin heavy variable (IGHV) gene somatic hypermutation status,3,4 TP53 mutation status and fluorescence in situ hybridization (FISH)-detected specific chromosomal alterations.5,6 Chromosome banding analysis (CBA) of in vitro mitogen-stimulated peripheral blood samples provides a lowresolution whole-genome view of the cytogenetic landscape of CLL, thereby overcoming the targeted nature of FISH which precludes a comprehensive assessment of the genomic landscape of CLL.7-12 Relatively small studies employing CBA in CLL identified the presence of a complex karyotype (CK; defined as ≥3 cytogenetically-visible structural and/or numerical aberrations) as a prognostic marker for refractoriness not only to chemoimmunotherapy,9,12,13 but also to novel targeted agents such as ibrutinib14 and venetoclax.15 Within the European Research Initiative on CLL (ERIC), we recently performed a retrospective study of 5,290 patients with available CBA data and found that cytogenetic complexity with ≥5 but not ≥3 chromosomal aberrations is an independent prognostic factor with adverse outcome in CLL.12 These observations demonstrate that the genome-wide analysis of genetic aberrations provides valuable additional prognostic information compared to FISH analysis alone, hence refining risk-adapted stratification of patients. Genomic array platforms such as array-based comparative genomic hybridization and single nucleotide polymorphism arrays allow the entire genome to be screened for copy-number alterations (CNA) in a single experiment. Current array platforms enable the identification of CNA of 10-100 kb, and have been implemented in the diagnostic work-up of hematologic malignancies, including CLL.16-22 In contrast to CBA analyses, input of array platforms only includes DNA and obviates the need for fresh isolated tumor cells and the need for in vitro mitogen stimulation. In order to focus on tumor-associated abnormalities and to avoid the detection of benign constitutional variants (present in the population), a conservative size cutoff of ≥5 megabases (Mb) has been recommended 88

clinically for CNA other than those occurring in specific loci recurrently affected in CLL, i.e., del(13)(q14) (RB1/DLEU/MIR15A/MIR16-1) [del(13q)], del(11)(q22.3) (ATM) [del(11q)], and del(17)(p13.1) (TP53) [del(17p)].23,24 Consequently, it remains unclear whether lowering the cutoff might identify clinically relevant CNA, with potential to improve risk stratification, or reduce assay sensitivity through the inclusion of non-pathogenic somatic or normal germline variations. Array-based studies have been performed with lower size cutoffs than 5 or even 1 Mb, zooming in on (new) potentially relevant chromosome regions containing cancer-specific genes.19,21,23 These studies have been inconclusive with respect to the value of such higher resolution approaches for risk assessment. A number of genomic array studies have observed an association between genomic complexity and adverse outcome in CLL.12,19,25-27 These studies were performed in single centers, used different cutoffs for CNA calling, and had relatively short follow-ups, precluding definitive conclusions regarding the applicability of genomic arrays for risk assessment in CLL. The emergence of ≥5 chromosomal abnormalities determined by CBA as an independent prognosticator and the discovery of heterogeneity within CK patients, urged for a reappraisal of the performance of genomic arrays. Realizing this, ERIC conducted the present multicenter analysis of genomic arrays in 2,293 CLL patients from 13 European CLL laboratories.

Methods Patients and established biomarkers Overall, 2,293 CLL patients, diagnosed according to International Workshop on CLL criteria,28 were included in the present multicenter, retrospective study. Thirteen different CLL diagnostic laboratories connected to ERIC were involved (Online Supplementary Figure S1). In total, 572 of the 2,293 patients have been included in previous publications.19,20,29 Cases included for survival analyses were mostly recruited within 2 years (69.1%) or 1 year (59.1%) after diagnosis. An overview of the patients’ characteristics (demographics and biological features) is provided in Online Supplementary Table S1. TP53 mutation analysis (Online Supplementary Table S2) and determination of the somatic hypermutation status of the clonotypic rearranged IGHV genes were performed and interpreted according to ERIC guidelines (Online Supplementary Methods).30,31 In a subgroup of patients (260/2,293) interphase FISH analysis haematologica | 2021; 106(1)

Genomic complexity in chronic lymphocytic leukemia

using probes for the regions 13q14 (RB1 and/or DLEU), 11q22.3 (ATM), 17p13.1 (TP53) and trisomy 12 was performed concurrently with genomic array analysis (i.e., same date of sampling). Simultaneous CBA (i.e., same date of sampling) was conducted in 122/2,293 patients. Metaphases for CBA were induced with stimulation protocols based on phorbol-12-myristate-13-acetate or immunostimulatory cytosine guanine dinucleotide-oligonucleotide DSP30 plus interleukin as previously described.12 The study was performed according to national and international ethical and legal recommendations with approval from the local Ethics Review Committees of the participating institutes

Genomic array analysis Genomic arrays were performed using several commercially available high-density (array-based comparative genomic hybridization/single nucleotide polymorphism array) platforms applied for diagnostics in the participating centers (Online Supplementary Table S3 and Online Supplementary Methods). CNA routinely assessed in CLL by targeted FISH revealed by genomic arrays in this study were included in all analyses irrespective of size, while for other CNA a size cutoff of â&#x2030;Ľ5 Mb was used according to Schoumans et al.24 Copy-number neutral loss of heterozygosity events, detected by single nucleotide polymorphism arrays, were not included in the counting of CNA. CNA with a distance â&#x2030;¤5 Mb between the two CNA, and putative chromothripsis events were counted as one event.

Statistical analysis Statistical analysis was performed using SPSS version 24 and R. A Pearson c2 test was used to evaluate the independence between categorical factors, and a Kruskal-Wallis test was applied to compare the distributions of continuous variables between groups of patients. Time-dependent receiver operating characteristic (ROC) curve analysis was performed evaluating different time-points from the date of array analysis (Online Supplementary Methods). Overall survival (OS) and time to first treatment (TTFT) were calculated from the date of sampling for array analysis to the last date of follow-up. Only patients untreated at the date of sampling (n=972) were included in the survival analyses unless otherwise stated. The median follow-up was 44 months with a total of 353 survival events. Kaplan-Meier survival curves were compared by log-rank testing. Univariable Cox regression analysis was used to evaluate the effect of CNA, for which there were at least ten events and established clinico-biological features, on OS. Factors found to be statistically significant in univariable analysis were included in multivariable analysis (for TTFT and OS). A P-value <0.05 signified a statistically significant result. Correction for multiple tests was applied when appropriate using the Bonferroni approach.

Data sharing statement A list of curated array profiles is provided in the Online Supplementary Excel File. Please contact either c.h.mellink@amc.nl or a.p.kater@amc.nl for the original data.

Results Genomic arrays detect aberrations not captured by fluorescence in situ hybridization in over one-third of patients with chronic lymphocytic leukemia CNA including aneuploidy were detected in 78.9% of patients enrolled in this study. Reflecting the heterogeneous genetic landscape of CLL, the total number of CNA haematologica | 2021; 106(1)

per patient varied considerably (median 1; range, 0-17). Increased numbers of CNA were identified in samples from previously treated patients (2.23 vs. 1.66; P<0.0001) (Online Supplementary Figure S2A) and in patients with aggressive disease (Binet B/C; 2.16 vs. 1.57, unmutated IGHV (U-CLL); 2.15 vs. 1.38, and TP53 mutation and/or del(17p) (TP53abn); 2.93 vs. 1.45; P<0.0001) (Online Supplementary Figure 2B-D). In 34.1% of patients, CNA were outside the established CLL regions captured by FISH (13q14, 11q22.3 [ATM], 17p13.1 [TP53] and chromosome 12) and included chromosomal deletions (n=1,064), chromosomal gains (n=597), monosomies (n=69), and trisomies (n=68) (Figure 1A-E). Complex patterns (n=32), indicative of chromothripsis, were also detected. Cases with putative chromothripsis were all del(11q) or TP53abn-positive, and chromosomes 2 (n=3), 3 (n=6), 6 (n=3), 8 (n=4) and 17 (n=3) were mainly targeted (Figure 1F). Putative chromothripsis was associated with adverse outcome (median OS: 3.7 years, 95% confidence interval [95% CI]: 0.6-6.8 years; P<0.001) (Online Supplementary Figure S3A) and a high total number of CNA (median 6). Surivival of cases with chromothripsis was also worse than that of cases with TP53abn or del(11q) without chromothripsis (P=0.03) (Online Supplementary Figure S3B). We focused on the ten most frequent CNA other than del(13q), del(11q), trisomy 12 and del(17p) in our cohort (indicated by asterisks in Online Supplementary Figure S4) and investigated whether these CNA correlated with IGHV gene somatic hypermutation status and CNA captured by FISH probes applied in CLL. This resulted in several significant associations, such as (i) U-CLL with gain of 2p and 8q and loss of 6q (Online Supplementary Figure S5); (ii) del(11q) with gain of 2p and 8q and loss of 4p; (iii) trisomy 12 with loss of 14q; (iv) del(13q) with loss of 18p and 14q; and, (v) del(17p) with gain of 3q and 8q, and loss of 4p, 8p, 15q and 18p (Online Supplementary Figure S6AD). A summary of all significant correlations of CNA (corrected P<0.01) detected in the entire cohort with predefined CLL prognostic subgroups is presented in Table 1 and for all correlations in Online Supplementary Table S4. In a subgroup of 260 patients with FISH and genomic array results available from the same sample, we observed 92.2% concordance (906/983 of regions analyzed by FISH) between the two approaches. Compared to genomic arrays, FISH detected statistically more cases of del(13q) (58.6% vs. 49.8%, P<0.001) and del(11q) (31.3% vs. 25.3%, P=0.003); while the detection rates of trisomy 12 (12.9% vs. 10.4%, P=0.07) and del(17p) (9.3 vs. 8.9%, P=1.00) were not statistically different (Online Supplementary Table S5). FISH-identified CNA not captured by genomic arrays (60/983) were mostly subclonal (median FISH clone size of 19%). Vice versa, 17/983 CNA were detected by genomic arrays only, including six cases of del(17p) (median size 19.11 Mb) and four of del(11q) (median size 17.67 Mb). To investigate the discriminatory power of FISH versus genomic arrays, a concordance index was calculated for genomic arrays versus FISH for del(11q) and del(17p) for both TTFT and OS. The higher number of del(11q) detected by FISH did result in a higher concordance index for TTFT (57% vs. 53%) while the concordance index for OS was 54% for both techniques. The concordance index for del(17p) was similar for genomic arrays and FISH for both TTFT (52% vs. 51%) and OS (57% vs. 55%). These data confirm that genomic 89

A.C. Leeksma et al. Figure 1. Overview of copy-number alterations detected by genomic arrays. (A) Percentages of patients with del(13)(q14), del(11)(q22.3) (ATM), trisomy 12 (+12) or del(17)(p13.1) (TP53) detected by genomic arrays irrespective of size. (B-E) Percentages of patients with different chromosomal losses (B), gains (C), monosomies (D) and trisomies (E). (F) Percentages of patients with putative chromothripsis events containing TP53abn in black and cases without TP53abn but del(11q)-positive in white. In four del(11q) cases, TP53 mutation status was not determined.

haematologica | 2021; 106(1)

Genomic complexity in chronic lymphocytic leukemia Table 1. Correlations of copy-number alterations with predefined prognostic subgroups of chronic lymphocytic leukemia.

CLL subgroup U-CLL

del(11)(q22.3)

trisomy 12

del(13)(q14)

del(17)(p13.1)

losses: 13q14*, 11q22.3, 17p13.1, 14q, 6q, 8p, 13q.other, 4p gains: 2p, 8q § trisomies: none † chromothripsis: none losses: 8p, 18p, 4p, 3q, 11p, 4q, 12p, 6q gains: 2p, 8q, 22q, 21q, 7p, 6p trisomies: 12*, 22 chromothripsis: none losses: 13q14*, 11q22.3*, 14q, 8p* gains: 2p* trisomies: 19, 18 chromothripsis: none losses: 14q*, 13q.other, 18p gains: 17q*, 13q trisomies: 12* chromothripsis: none losses: 8p, 18p, 4p, 15q, 9p, 3p, 4q, 13q.other, 6p, 2q, 20p, 10q, 9q, 18q, 2p, 19p, 10p, 5p, 5q, 17q, Xp, 9, 21q, 13, Yq, 7q, 1p, 11p, 14q gains: 8q, 3q, 17q, 15q, 11q, 3p, 5q, 1p, 13q, 11p trisomies: none chromothripsis: 8, 5, 6, 17

Significantly correlated copy-number abnormalities (corrected for multiple testing; P<0.01). Negative correlations are indicated by an asterisk (*). CLL: chronic lymphocytic leukemia; U-CL: CLL with unmutated immunoglobulin heavy chain variable gene (IGHV).

arrays show high concordance with FISH with the advantage that (other) potentially clinically relevant abnormalities can be elucidated on a genome-wide scale.

Recurrent copy-number alterations detected by genomic arrays are associated with adverse outcome The effect of CNA detected by genomic arrays on OS was tested in untreated patients at the date of sampling (n=972), mostly within 2 years from diagnosis (69.1%; median time after diagnosis, 7 months). A significant impact of well-established risk factors for CLL, including Binet B/C, U-CLL, TP53abn and del(11q) (Figure 2 and Online Supplementary Figure S7) on OS was confirmed. Next, we analzed the effects of CNA outside the established CLL regions which were detected in at least ten cases (Figure 2). Minimal common regions of deletion or amplification are presented in Online Supplementary Table S6. Gains of 8q (P=0.01; encompassing MYC in >95% of cases), loss of 9p (P<0.001; involving SMARCA2 [46%] and CDKN2A [85%]) and 18p (P<0.001; including USP14 [100%]) were significantly associated with a shorter OS in univariable Cox regression analysis after correction for multiple testing.

Patients with chronic lymphocytic leukemia can be grouped into three prognostic subgroups based on copy-number alteration profile ROC curve analysis revealed that based on the total number of CNA above the 5 Mb cutoff, including aneuploidy events (see the Online Supplementary Methods for details of this analysis), the cohort could be subdivided into three prognostic subgroups: low-genomic complexity (GC) (0-2 CNA; n=793), intermediate-GC (3-4 CNA; n=122) and high-GC (≥5 CNA; n=57). The results of the ROC curve analysis were replicated by a different statistical approach, using maximally selected rank statistics haematologica | 2021; 106(1)

(see the Online Supplementary Methods for details of this analysis).32 Results were supported with bootstrapping based on 100 random bootstrap samples from the original sample (with replacement) and showed relatively high percentages for the Youden index and the maximally selected estimated cut-point (Online Supplementary Methods). Low-GC cases were associated with indolent disease: Binet A, IGHV mutated-CLL, and low incidences of TP53abn and del(11q) (P<0.01). Intermediate-GC and high-GC cases were associated with a more advanced clinical stage compared to low-GC cases (P=0.003). High-GC cases were enriched for TP53abn and U-CLL (P<0.001) compared to intermediate-GC and low-GC cases (Table 2). Calculation of associations of established biomarkers and CNA with prognostic subgroups identified differences between the prognostic GC subgroups (Figure 3). Interestingly, loss of 9p and loss of 15q were strongly associated with high-GC (P<0.001), not significantly associated with intermediateGC and negatively associated with low-GC (P<0.001). A positive correlation with high-GC was observed for all CNA except for trisomies 12,18 and 19. Low-GC was associated with a more favorable clinical outcome (median OS: 10.17 years, 95% CI: 8.92-11.42 years) compared to the outcomes of cases with intermediate-GC (median OS: 7.02 years, 95% CI: 4.79-9.25 years, P=0.001) and high-GC, who had the shortest OS (median OS: 3.05 years, 95% CI: 1.14-4.96 years, P<0.001) (Figure 4A). Similar results were observed for TTFT (Figure 4B) between the three different prognostic subgroups (P=0.001).

High genomic complexity is an independent prognostic risk factor in chronic lymphocytic leukemia Univariable Cox regression analysis revealed that the GC subgroups intermediate-GC and high-GC (P<0.001) as well as established CLL prognostic factors – male sex, age 91

A.C. Leeksma et al.

Figure 2. Effect of copy-number alterations on clinical outcome in chronic lymphocytic leukemia. Forest plot representing the results of univariable Cox regression analysis of copy-number alterations detected by array (with at least 10 events) and the effect of chronic lymphocytic leukemia (CLL) prognostic factors. A hazard ratio of less than 1.00 indicates a lower risk of overall survival. The size of each square is proportional to the amount of data available. U-CLL denotes patients with an unmutated immunoglobulin heavy chain variable gene and TP53abn denotes patients with del(17p) and/or a TP53 mutation. Loss13q.other are patients with a del(13q) not containing the established 13q14 minimally deleted region detectable by the diagnostic fluorescence in situ hybridization probe, and TRIS denotes patients with a trisomy. Only patients untreated at the date of DNA sampling were included for survival analysis. 95% CI: 95% confidence interval.

>70, Binet stage B/C disease, U-CLL, TP53abn and del(11q) – were associated with adverse outcome (P<0.05). (Online Supplementary Tables S7 and S8). HighGC cases had significantly shorter TTFT and OS in the CLL subsets U-CLL (P<0.001), IGHV gene-mutated CLL (P=0.001) and TP53abn/del(11q) (P=0.01) compared to low-GC cases (Online Supplementary Figures S8-S10). Furthermore, including the established factors – male gender, age >70, Binet stage B/C, U-CLL, TP53abn and del(11q) – in the multivariable Cox regression model, high-GC retained statistical significance for TTFT (hazard ratio [HR]: 2.15; 95% CI: 1.36-3.41; P=0.001) (Table 3) and OS (HR: 2.54; 95% CI: 1.54-4.17; P<0.001) (Table 4). Similarly, high-GC retained statistical significance as a binary predictor (with the categories ≥5 CNA and <5 92

CNA) including the above-mentioned factors in the multivariable Cox regression model for TTFT and OS (P=0.002) (Online Supplementary Tables S9 and S10).

Lower copy-number alteration size cutoff does not significantly improve risk stratification In routine clinical practice, a 5 Mb threshold for reporting CNA other than del(13q), del(11q) and del(17p) is used, as described in the publication of Schoumans et al.24 In order to identify whether the recommended 5 Mb threshold or a 1 Mb threshold may be a better discriminator for prognostication, a subgroup of genomic arrays (n=647; with both 1 Mb and 5 Mb cutoff results available) was analyzed applying both cutoffs. Lowering the size cutoff to 1 Mb resulted in the detection of 290 additional haematologica | 2021; 106(1)

Genomic complexity in chronic lymphocytic leukemia Table 2. Correlations of the three genomic complexity subgroups (based on the complexity of the array profile) with other chronic lymphocytic leukemia risk factors

Low-GC

Intermediate-GC

High-GC

P-values

500/761, 66% 179/793, 23% 250/752, 33% 212/515, 41% 29/602, 5% 84/793, 11%

81/115, 70% 31/122, 25% 55/118, 47% 50/81, 62% 26/102, 25% 52/122, 43%

35/53, 66% 20/57, 35% 26/54, 48% 25/33, 76% 27/54, 50% 26/57, 46%

0.606 0.089 0.003 <0.001 <0.001 <0.001

N=972 Male >70 years Binet B/C U-CLL TP53abn† del(11)(q22.3)

TP53abn= del(17)(p13.1) and/or TP53 mutation. GC: genomic complexity: U-CLL: chronic lymphocytic leukemia with unmutated IGHV genes, GC categories: low-GC=0-2 copynumber alterations (CNA); intermediate-GC=3-4 CNA; high-GC=≥5 CNA detected by array. A Pearson c2 test was used to calculate the P-values with a combination of subgroups.

†

aberrations in these 647 patients and revealed an increased percentage of high-GC (≥5 CNA) cases (10.4 vs. 7.3; P<0.001). Next, the discriminatory power was determined by calculating a concordance index to study the effects of lowering the array cutoff to 1 Mb (Online Supplementary Methods) on TTFT and OS. The concordance index was identical for TTFT (77%) and slightly decreased for OS (74% vs. 75%), indicating no further improvement when lowering the cutoff to 1Mb.

Genomic arrays do not underperform compared with chromosome-banding analysis

CLL patients with ≥5 chromosomal abnormalities detected by CBA were recently shown to have a dismal prognosis and CBA is considered the gold standard for defining a CK.12 Direct comparison of 122 patients with simultaneous CBA and genomic array results available, showed chromosomal abnormalities in 86.1% of patients by genomic arrays and in 73.8% by CBA (P=0.015) (Online Supplementary Figure S11). Analysis of a subset of patients (n=96) for whom prognostic factors gender, age, Binet stage, del(11q), del(17p) and IGHV mutation status were known, revealed a higher concordance index for OS (77% vs. 72%) and similar concordance index for TTFT (68% vs. 67%) for genomic arrays versus CBA.

Discussion In our study CLL patients with ≥5 CNA emerged as a separate subgroup with aggressive disease and an adverse outcome, which was independent of several well-established factors in CLL, such as IGHV and TP53 status. A disadvantage of a collection of retrospective multicenter data is the risk of variation introduced by combining results from different genomic array platforms. Nevertheless, marked variation in this study is less likely since the data were methodically reviewed and uniformly interpreted using commonly accepted guidelines which provide tools for filtering and calling CNA and single nucleotide polymorphism data.17,23,24,33 Another clear disadvantage of such real-world retrospective data is that clinical correlations, although indicative, must be considered with great caution as treatments were not uniform and we were limited to the date of sample collection (instead of diagnosis) for the TTFT analysis. In current International Working Group CLL guidelines,34 targeted FISH is recommended for the detection of genomic aberrations. In our multicenter study overall concordance between arraybased comparative genomic hybridization data and (targeted) FISH results (subgroup, n=260) was high (>90%). haematologica | 2021; 106(1)

Figure 3. Heatmap of correlations for different categories of genomic complexity based on total number of copy-number alterations. Genomic complexity (GC) was subdivided into three prognostic subgroups: low-GC (0-2 copy-number alterations [CNA]), intermediate-GC (3-4 CNA) and high-GC (≥5 CNA). Colors are based on the results of Kendall’s tau_b which assumes values between -1 and 1. Correlation levels are depicted according to the color scale with increased correlation ranging from green to red. Only statistically significant correlations with a P<0.01 are colored. Non-significant correlations are shown as uncolored crossed squares. U-CLL: chronic lymphocytic leukemia with an unmutated immunoglobulin heavy chain variable gene; TP53abn: cases with del(17p) and/or a TP53 mutation; Loss13q.other: cases with a del(13q) not containing the established 13q14 minimally deleted region detectable by the diagnostic fluorescence in situ hybridization probe; TRIS trisomy.

Nevertheless, genomic arrays resulted in the detection of fewer CNA for regions captured by FISH, which could be explained by the superior detection limit of FISH.17,25,28,35-38 However, the increased detection of CNA by FISH was not directly associated with a better risk assessment, in agreement with the findings in other genomic array stud93

A.C. Leeksma et al.

ies,17,18,22 providing a rationale that genomic arrays can be used, besides FISH, in CLL diagnostics.17 Direct comparison of genomic arrays with FISH in prospective clinical trials is necessary before genomic arrays can be accepted to replace FISH in routine clinical practice for CLL prognostics. Since our data were collected retrospectively from patients during follow-up after various chemoimmunotherapy regimens, we could not correlate the effects

of CNA on OS with specific treatments. CNA such as del(4p), del(9p) and del(18p) co-occurred with other CNA and were associated with other CLL biomarkers. For this reason, the prognostic impact of these individual CNA must be considered with caution. Furthermore, the applicability of our results needs to be established for novel agents, as the prognostic value following chemoimmunotherapy might be different. This is exemplified by a recent update of the RESONATE-2 study in which

Figure 4. Effect of genomic complexity on clinical outcome in chronic lymphocytic leukemia. (A, B) Kaplan-Meier plots presenting the overall survival (A) and time to first treatment (B) of patients with chronic lymphocytic leukemia divided into three categories of genomic complexity (GC) based on the total number of copy-number alterations (CNA): low-GC (0-2 CNA), intermediate-GC (3-4 CNA) and high-GC (â&#x2030;Ľ5 CNA) and defined by receiver operating characteristic curve analysis. 95% CI: 95% confidence interval.

haematologica | 2021; 106(1)

Genomic complexity in chronic lymphocytic leukemia

Table 3. Multivariable analysis for time to first treatment.

Table 4. Multivariable analysis for overall survival.

N=528

95% CI

P-values

Male >70 years Binet B/C U-CLL TP53abn† del(11)(q22.3) GC (3 categories) Intermediate-GC vs. low-GC High-GC vs. low-GC

1.05 1.10 3.87 3.25 1.12 1.22

0.84-1.32 0.83-1.46 3.04-4.94 2.51-4.21 0.76-1.65 0.92-1.63

0.674 0.490 <0.001 <0.001 0.567 0.166

1.00 2.15

0.73-1.38 1.36-3.41

0.984 0.001

† TP53abn= del(17)(p13.1) and/or TP53 mutation. HR: hazard ratio; 95% CI: 95% confidence interval of the hazard ratio; U-CLL: chronic lymphocytic leukemia with unmutated IGHV genes; GC: genomic complexity, GC categories: low-GC=0-2 copy-number alterations (CNA); intermediateGC=3-4 CNA; high-GC=≥5 CNA detected by array.

del(11q) was prognostically favorable in ibrutinib-treated patients.39 Additionally, del(8p) was recently correlated with resistance to ibrutinib.40 The inferior outcome of patients with loss of 9p is in line with the findings of a recent study on the effect of loss of 9p24.3 containing SMARCA2.41 Loss of 9p24.3 and/or mutations in the SWISNF chromatin remodeling complex were identified in all nonresponders to ibrutinib and venetoclax. Deletion of SMARCA2 results in upregulation of the prosurvival protein BCL-XL which could protect cells from apoptosis.41 The differences in OS and TTFT in the distinct GC subcategories based on ROC analysis in this study are in agreement with recent findings regarding CBA-defined CK and the independent effect of high-CK (with ≥5 numerical and/or structural abnormalities),12 which were characterized by a distinct distribution of aberrations affecting a broad spectrum of chromosomes.12 Similarly, high-GC detected by genomic arrays was negatively associated with trisomy 12 and, in contrast to intermediateGC, not positively associated with trisomy 18 and 19 (Figure 3). High-CK cases detected by CBA and lacking del(17p) or a Sanger-detected (‘clonal’) TP53 mutation were all negative for subclonal TP53 mutations when sequenced by next-generation sequencing.12 It is conceivable that recently identified potential drivers for high-GC in CLL, such as SAMHD1 and SETD2, attributed to highGC in a subset of patients.9,42-44 The resolutions of the genomic array platforms used in this study are higher than the resolution that can be achieved by CBA. Differences in the detection of chromosomal aberrations between genomic arrays and CBA in our study could be related to the lower resolution of CBA and/or the use of phorbol-12myristate-13-acetate in most CBA cases instead of the more sensitive method involving cytosine guanine dinucleotides plus interleukin 2.12 In contrast to genomic arrays, CBA selects for cells with proliferative potential as a prognostic marker and is able to detect balanced translocations. Although balanced rearrangements (such as balanced translocations) cannot be detected by genomic arrays, chromosome regions involved in unbalanced translocations can be recognized more precisely by arrays, possibly revealing new recurrent abnormalities. Our finding that, in terms of risk stratification, genomic arrays do not underperform compared with CBA implies that the inability of genomic arrays to detect balanced rearrangements does not impede adequate risk assessment. This haematologica | 2021; 106(1)

N=528

95% CI

P-values

Male >70 years Binet B/C U-CLL∗ TP53abn† del(11)(q22.3) GC (3 categories) Intermediate-GC vs. low-GC High-GC vs. low-GC

1.26 2.48 1.47 3.84 1.56 0.92

0.96-1.64 1.86-3.31 1.13-1.91 2.85-5.17 1.03-2.36 0.67-1.26

0.094 <0.001 0.004 <0.001 0.037 0.605

1.19 2.54

0.81-1.73 1.54-4.17

0.373 <0.001

†TP53abn= del(17)(p13.1) and/or TP53 mutation. HR: hazard ratio; 95% CI: 95% confidence interval of the hazard ratio; U-CLL: chronic lymphocytic leukemia with unmutated IGHV genes; GC: genomic complexity, GC categories: low-GC=0-2 copy-number alterations (CNA); intermediate-GC=3-4 CNA; high-GC=≥5 CNA detected by array.

could be explained by the fact that, in contrast to other mature B-cell tumors, CLL is defined by the presence of CNA with a paucity of balanced rearrangements.11,45 In CLL diagnostics genomic arrays are cost-effective as a stand-alone method when compared to FISH plus CBA. Only recently, the integration of genomic arrays has been recommended by the Cancer Genomics Consortium working group as having evidence-based clinical utility for a broad spectrum of hematologic malignancies.33,46-49 In conclusion, our study highlights the strength of genome-wide approaches for risk stratification in CLL and we identified CLL patients with high-GC, defined as ≥5 CNA detected by genomic arrays, as an independent subgroup with dismal clinical outcome. We, therefore, recommend implementation of comprehensive genomic profiling in future clinical trials. Disclosures This study was partly funded by an unrestricted contribution from Janssen Pharmaceuticals and from GILEAD Sciences SA. The funding sources had no role in identifying statements, abstracting data, synthesizing results, grading evidence or preparing the manuscript, or in the decision to submit the manuscript for publication. A.C.L. is supported by the Peters van der Laan foundation. J.C.S. was funded by Bloodwise (11052, 12036), the Kay Kendall Leukaemia Fund (873), Cancer Research UK (C34999/A18087, ECMC C24563/A15581), Wessex Medical Research and the Bournemouth Leukaemia Fund. K.P., M.J., and S.P. are supported by the project MHCR DRO no. 65269705, the research infrastructures NCMG LM2015091, and EATRIS-CZ LM2015064, and the project CEITEC2020 LQ1601, funded by MEYS CR. R.R. is supported by Swedish Cancer Society, the Swedish Research Council, the Knut and Alice Wallenberg Foundation, Karolinska Institutet, Karolinska University Hospital, and Radiumhemmets Forskningsfonder, Stockholm. Contributions ACL, PB, TM, CM, DO, RR, JCS, KS, and APK were responsible for conception and design of the study. ACL, PB, TM, A-MK, CM, and APK were involved in analysis and interpretation of data. PB, CM, AP, PG, KP, EB, RG, GB, LM, EE, AÖ, LH, HP, BE, SP, FN-K, PP, AR-V, ANT, MB, I.S., Z.D., MW, CH, MJ, LY, CT, DO, MS-K. were involved in acquisition of clinical data of patient samples. ACL, PB, and TM, performed statistical analyses. ACL, PB, and TM, made the figures. 95

A.C. Leeksma et al.

A.C.L., PB, KS, DO, JCS, RR, and APK wrote the first draft; and all authors were involved in revising the manuscript and approved the final version. Acknowledgments The authors gratefully acknowledge all patients who contributed to this study. This study was partly funded by an unrestricted contribution from Janssen Pharmaceuticals and from GILEAD Sciences SA. The funding sources had no role in identifying statements, abstracting data, synthesizing results, grading evidence or preparing the manuscript, or in the decision to submit the manuscript for publication. ACL is supported by the Peters

References 1. Rai KR, Sawitsky A, Cronkite EP, Chanana AD, Levy RN, Pasternack BS. Clinical staging of chronic lymphocytic leukemia. Blood. 1975;46(2):219-234. 2. Binet JL, Auquier A, Dighiero G, et al. A new prognostic classification of chronic lymphocytic leukemia derived from a multivariate survival analysis. Cancer. 1981;48(1):198-206. 3. Hamblin TJ, Davis Z, Gardiner A, Oscier DG, Stevenson FK. Unmutated Ig V(H) genes are associated with a more aggressive form of chronic lymphocytic leukemia. Blood. 1999;94(6):1848-1854. 4. Fais F, Ghiotto F, Hashimoto S, et al. Chronic lymphocytic leukemia B cells express restricted sets of mutated and unmutated antigen receptors. J Clin Invest. 1998;102(8):1515-1525. 5. Dohner H, Stilgenbauer S, Benner A, et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med. 2000;343(26):1910-1916. 6. Krober A, Bloehdorn J, Hafner S, et al. Additional genetic high-risk features such as 11q deletion, 17p deletion, and V3-21 usage characterize discordance of ZAP-70 and VH mutation status in chronic lymphocytic leukemia. J Clin Oncol. 2006;24(6): 969-975. 7. Blanco G, Puiggros A, Baliakas P, et al. Karyotypic complexity rather than chromosome 8 abnormalities aggravates the outcome of chronic lymphocytic leukemia patients with TP53 aberrations. Oncotarget. 2016;7(49):80916-80924. 8. Haferlach C, Dicker F, Schnittger S, Kern W, Haferlach T. Comprehensive genetic characterization of CLL: a study on 506 cases analysed with chromosome banding analysis, interphase FISH, IgV(H) status and immunophenotyping. Leukemia. 2007;21 (12):2442-2451. 9. Herling CD, Klaumunzer M, Rocha CK, et al. Complex karyotypes and KRAS and POT1 mutations impact outcome in CLL after chlorambucil-based chemotherapy or chemoimmunotherapy. Blood. 2016;128(3): 395-404. 10. Rigolin GM, Saccenti E, Guardalben E, et al. In chronic lymphocytic leukaemia with complex karyotype, major structural abnormalities identify a subset of patients with inferior outcome and distinct biological characteristics. Br J Haematol. 2018;181(2): 229-233. 11. Baliakas P, Iskas M, Gardiner A, et al. Chromosomal translocations and kary-

van der Laan foundation. JCS was funded by Bloodwise (11052, 12036), the Kay Kendall Leukaemia Fund (873), Cancer Research UK (C34999/A18087, ECMC C24563/A15581), Wessex Medical Research and the Bournemouth Leukaemia Fund. KP, MJ, and SP are supported by the project MHCR DRO n. 65269705, the research infrastructures NCMG LM2015091, and EATRIS-CZ LM2015064, and the project CEITEC2020 LQ1601, funded by MEYS CR. RR is supported by the Swedish Cancer Society, the Swedish Research Council, the Knut and Alice Wallenberg Foundation, Karolinska Institutet, Karolinska University Hospital, and Radiumhemmets Forskningsfonder, Stockholm.

otype complexity in chronic lymphocytic leukemia: a systematic reappraisal of classic cytogenetic data. Am J Hematol. 2014;89 (3):249-255. 12. Baliakas P, Jeromin S, Iskas M, et al. Cytogenetic complexity in chronic lymphocytic leukemia: definitions, associations and clinical impact. Blood. 2019;133(11): 1205-1216 13. Le Bris Y, Struski S, Guieze R, et al. Major prognostic value of complex karyotype in addition to TP53 and IGHV mutational status in first-line chronic lymphocytic leukemia. Hematol Oncol. 2017;35(4):664670. 14. Thompson PA, O'Brien SM, Wierda WG, et al. Complex karyotype is a stronger predictor than del(17p) for an inferior outcome in relapsed or refractory chronic lymphocytic leukemia patients treated with ibrutinibbased regimens. Cancer. 2015;121(20): 3612-3621. 15. Anderson MA, Tam C, Lew TE, et al. Clinicopathological features and outcomes of progression of CLL on the BCL2 inhibitor venetoclax. Blood. 2017;129(25): 3362-3370. 16. Sargent R, Jones D, Abruzzo LV, et al. Customized oligonucleotide array-based comparative genomic hybridization as a clinical assay for genomic profiling of chronic lymphocytic leukemia. J Mol Diagn. 2009;11(1):25-34. 17. Gunn SR, Mohammed MS, Gorre ME, et al. Whole-genome scanning by array comparative genomic hybridization as a clinical tool for risk assessment in chronic lymphocytic leukemia. J Mol Diagn. 2008;10(5): 442-451. 18. O'Malley DP, Giudice C, Chang AS, et al. Comparison of array comparative genomic hybridization (aCGH) to FISH and cytogenetics in prognostic evaluation of chronic lymphocytic leukemia. Int J Lab Hematol. 2011;33(3):238-244. 19. Gunnarsson R, Mansouri L, Isaksson A, et al. Array-based genomic screening at diagnosis and during follow-up in chronic lymphocytic leukemia. Haematologica. 2011;96(8):1161-1169. 20. Puiggros A, Puigdecanet E, Salido M, et al. Genomic arrays in chronic lymphocytic leukemia routine clinical practice: are we ready to substitute conventional cytogenetics and fluorescence in situ hybridization techniques? Leuk Lymphoma. 2013;54(5): 986-995. 21. Edelmann J, Holzmann K, Miller F, et al. High-resolution genomic profiling of chronic lymphocytic leukemia reveals new recurrent genomic alterations. Blood.

2012;120(24):4783-4794. 22. Stevens-Kroef MJ, van den Berg E, Olde Weghuis D, et al. Identification of prognostic relevant chromosomal abnormalities in chronic lymphocytic leukemia using microarray-based genomic profiling. Mol Cytogenet. 2014;7(1):3. 23. Simons A, Sikkema-Raddatz B, de Leeuw N, Konrad NC, Hastings RJ, Schoumans J. Genome-wide arrays in routine diagnostics of hematological malignancies. Hum Mutat. 2012;33(6):941-948. 24. Schoumans J, Suela J, Hastings R, et al. Guidelines for genomic array analysis in acquired haematological neoplastic disorders. Genes Chromosomes Cancer. 2016;55(5):480-491. 25. Ouillette P, Collins R, Shakhan S, et al. Acquired genomic copy number aberrations and survival in chronic lymphocytic leukemia. Blood. 2011;118(11):3051-3061. 26. Yu L, Kim HT, Kasar S, et al. Survival of Del17p CLL depends on genomic complexity and somatic mutation. Clin Cancer Res. 2017;23(3):735-745. 27. Knight SJ, Yau C, Clifford R, et al. Quantification of subclonal distributions of recurrent genomic aberrations in paired pre-treatment and relapse samples from patients with B-cell chronic lymphocytic leukemia. Leukemia. 2012;26(7):15641575. 28. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760. 29. Parker H, Rose-Zerilli MJ, Parker A, et al. 13q deletion anatomy and disease progression in patients with chronic lymphocytic leukemia. Leukemia. 2011;25(3):489-497. 30. Malcikova J, Tausch E, Rossi D, et al. ERIC recommendations for TP53 mutation analysis in chronic lymphocytic leukemiaupdate on methodological approaches and results interpretation. Leukemia. 2018;32 (5):1070-1080. 31. Rosenquist R, Ghia P, Hadzidimitriou A, et al. Immunoglobulin gene sequence analysis in chronic lymphocytic leukemia: updated ERIC recommendations. Leukemia. 2017; 31(7):1477-1481. 32. Hothorn T, Zeileis A. Generalized maximally selected statistics. Biometrics. 2008; 64(4):1263-1269. 33. Chun K, Wenger GD, Chaubey A, et al. Assessing copy number aberrations and copy-neutral loss-of-heterozygosity across the genome as best practice: an evidencebased review from the Cancer Genomics Consortium (CGC) working group for

haematologica | 2021; 106(1)

Genomic complexity in chronic lymphocytic leukemia

chronic lymphocytic leukemia. Cancer Genet. 2018;228-229:236-250. 34. Hallek M, Cheson BD, Catovsky D, et al. Guidelines for diagnosis, indications for treatment, response assessment and supportive management of chronic lymphocytic leukemia. Blood. 2018;131(25):27452760. 35. Hagenkord JM, Monzon FA, Kash SF, Lilleberg S, Xie Q, Kant JA. Array-based karyotyping for prognostic assessment in chronic lymphocytic leukemia: performance comparison of Affymetrix 10K2.0, 250K Nsp, and SNP6.0 arrays. J Mol Diagn. 2010;12(2):184-196. 36. Lehmann S, Ogawa S, Raynaud SD, et al. Molecular allelokaryotyping of early-stage, untreated chronic lymphocytic leukemia. Cancer. 2008;112(6):1296-1305. 37. Tyybakinoja A, Vilpo J, Knuutila S. Highresolution oligonucleotide array-CGH pinpoints genes involved in cryptic losses in chronic lymphocytic leukemia. Cytogenet Genome Res. 2007;118(1):8-12. 38. Kolquist KA, Schultz RA, Slovak ML, et al. Evaluation of chronic lymphocytic leukemia by oligonucleotide-based microarray analysis uncovers novel aberrations not detected by FISH or cytogenetic analysis. Mol Cytogenet. 2011;16;4:25. 39. Barr PM, Robak T, Owen C, et al. Sustained efficacy and detailed clinical follow-up of first-line ibrutinib treatment in older patients with chronic lymphocytic

haematologica | 2021; 106(1)

leukemia: extended phase 3 results from RESONATE-2. Haematologica. 2018;103 (9):1502-1510. 40. Burger JA, Landau DA, Taylor-Weiner A, et al. Clonal evolution in patients with chronic lymphocytic leukaemia developing resistance to BTK inhibition. Nat Commun. 2016;20;7:11589. 41. Agarwal R, Chan YC, Tam CS, et al. Dynamic molecular monitoring reveals that SWI-SNF mutations mediate resistance to ibrutinib plus venetoclax in mantle cell lymphoma. Nat Med. 2019;25(1):119-129. 42. Clifford R, Louis T, Robbe P, et al. SAMHD1 is mutated recurrently in chronic lymphocytic leukemia and is involved in response to DNA damage. Blood. 2014;123(7):1021-1031. 43. Parker H, Rose-Zerilli MJ, Larrayoz M, et al. Genomic disruption of the histone methyltransferase SETD2 in chronic lymphocytic leukaemia. Leukemia. 2016;30 (11):2179-2186. 44. Mar BG, Chu SH, Kahn JD, et al. SETD2 alterations impair DNA damage recognition and lead to resistance to chemotherapy in leukemia. Blood. 2017;130(24):26312641. 45. Leeksma AC, Taylor J, Wu B, et al. Clonal diversity predicts adverse outcome in chronic lymphocytic leukemia. Leukemia. 2018;33(2):390-402. 46. Kanagal-Shamanna R, Hodge JC, Tucker T, et al. Assessing copy number aberrations

and copy neutral loss of heterozygosity across the genome as best practice: an evidence based review of clinical utility from the Cancer Genomics Consortium (CGC) working group for myelodysplastic syndrome, myelodysplastic/myeloproliferative and myeloproliferative neoplasms. Cancer Genet. 2018;228-229:197-217. 47. Pugh TJ, Fink JM, Lu X, et al. Assessing genome-wide copy number aberrations and copy-neutral loss-of-heterozygosity as best practice: an evidence-based review from the Cancer Genomics Consortium working group for plasma cell disorders. Cancer Genet. 2018;228-229:184-196. 48. Xu X, Bryke C, Sukhanova M, et al. Assessing copy number abnormalities and copy-neutral loss-of-heterozygosity across the genome as best practice in diagnostic evaluation of acute myeloid leukemia: an evidence-based review from the Cancer Genomics Consortium (CGC) myeloid neoplasms working group. Cancer Genet. 2018;228-229:218-235. 49. Mikhail FM, Biegel JA, Cooley LD, et al. Technical laboratory standards for interpretation and reporting of acquired copy-number abnormalities and copy-neutral loss of heterozygosity in neoplastic disorders: a joint consensus recommendation from the American College of Medical Genetics and Genomics (ACMG) and the Cancer Genomics Consortium (CGC). Genet Med. 2019;21(9):1903-1915.

ARTICLE Ferrata Storti Foundation

Chronic Lymphocytic Leukemia

Analysis of retrotransposon subfamily DNA methylation reveals novel early epigenetic changes in chronic lymphocytic leukemia Timothy M. Barrow,1 Nicole Wong Doo, 2,3 Roger L. Milne, 2,4,5 Graham G. Giles, 2,4,5 Elaine Willmore,6 Gordon Strathdee7 and Hyang-Min Byun7

Faculty of Health Sciences and Wellbeing, University of Sunderland, Sunderland, UK; Cancer Epidemiology Division, Cancer Council Victoria, Melbourne, Australia; 3 Concord Hospital, University of Sydney, Sydney, Australia; 4Center for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, Australia; 5Precision Medicine, School of Clinical Sciences at Monash Health, Monash University, Clayton, Australia; 6CR UK Drug Discovery Unit, Translational and Clinical Research Institute, Newcastle University, Newcastle upon Tyne, UK and 7Newcastle University Center for Cancer, Biosciences Institute, Newcastle University, Newcastle upon Tyne, UK 1

Haematologica 2021 Volume 106(1):98-110

ABSTRACT

etrotransposons such as LINE-1 and Alu comprise >25% of the human genome. While global hypomethylation of these elements has been widely reported in solid tumours, their epigenetic dysregulation is yet to be characterised in chronic lymphocytic leukemia (CLL), and there has been scant consideration of their evolutionary history that mediates sensitivity to hypomethylation. Here, we developed an approach for locus- and evolutionary subfamily-specific analysis of retrotransposons using the Illumina Infinium Human Methylation 450K microarray platform, which we applied to publicly-available datasets from CLL and other haematological malignancies. We identified 9,797 microarray probes mapping to 117 LINE-1 subfamilies and 13,130 mapping to 37 Alu subfamilies. Of these, 10,782 were differentially methylated (P <0.05) in CLL patients (n=139) compared with healthy individuals (n=14), with enrichment at enhancers (P=0.002). Differential methylation was associated with evolutionary age of LINE-1 (r2=0.31, P=0.003) and Alu (r2=0.74, P=0.002) elements, with greater hypomethylation of older subfamilies (L1M, AluJ). Locus-specific hypomethylation was associated with differential expression of proximal genes, including DCLK2, HK1, ILRUN, TANK, TBCD, TNFRSF1B and TXNRD2, with higher expression of DCLK2 and TNFRSF1B associated with reduced patient survival. Hypomethylation at nine loci was highly frequent in CLL (>90% patients) but not observed in healthy individuals or other leukaemias, and was detectable in blood samples taken prior to CLL diagnosis in 9 of 82 individuals from the Melbourne Collaborative Cohort Study. Our results demonstrate differential methylation of retrotransposons in CLL by their evolutionary heritage that modulates expression of proximal genes. FDR

Correspondence: TIMOTHY BARROW timothy.barrow@sunderland.ac.uk Received: June 11, 2019. Accepted: January 7, 2020. Pre-published: January 9, 2020. https://doi.org/10.3324/haematol.2019.228478

Introduction Retrotransposons, including long interspersed elements (LINE) and short interspersed elements (SINE), comprise more than 25% of the human genome. The most widely studied LINE and SINE are LINE-1 (L1) and Alu respectively, which can be further categorised into subfamilies according to sequence variants that inform upon their evolutionary heritage: L1 can be broadly categorised into L1M (mammalian-specific, oldest), L1P (primate-specific, intermediate) and L1H (human-specific, youngest) subfamilies; and Alu into AluJ (oldest), AluS (intermediate) and AluY (youngest).1,2 Retrotransposons are considered to be mobile within the human genome, but the ability to retrotranspose has been lost in the oldest L1 and Alu subfamilies due to sequence changes over time, while, among others, haematologica | 2021; 106(1)

Retrotransposon methylation in CLL

members of the L1PA, L1HS, AluS and AluY families have retained this ‘jumping’ ability.3 Of these, the AluY subfamily is the most active,4 while there are approximately 68 L1 sequences in the human genome that remain active in retrotransposition.5 De novo retrotransposition of an Alu element has been estimated to occur in 1 in every 21 births, and an L1 element in 1 in 212 births.6 Somatic retrotransposition is largely supressed through epigenetic mechanisms that silence expression of these elements, but hypomethylation of L1 and Alu has been reported in response to a range of environmental exposures, including benzene7 and tobacco,8 with differential sensitivity to environmental pollutants according to the evolutionary age of the subfamily.9 Deregulation of retrotransposon methylation is common in many cancers10 and is associated with patient prognosis.11 It can serve to activate the expression of oncogenes, such as through locus-specific hypomethylation events and transcription from alternative transcriptional start sites located within retrotransposon elements.12 At the genome-wide level, it can also lead to chromosomal instability and the subsequent acquisition of genetic defects,13,14 with important consequences for patient prognosis. Hypomethylation of these elements is associated with their re-activation and can lead to retrotransposition events, which are frequent in colorectal, lung, prostate and breast cancers (47–93% of tumors)15 and are biased towards hypomethylated regions of the genome.16 Retrotransposition can initiate carcinogenesis in epithelial tissues17 and has been identified in adenomas, metaplasia and histologically-normal tissue surrounding tumors.18,19 The epigenome in chronic lymphocytic leukemia (CLL) is characterised by widespread hypomethylation at the point of diagnosis20 and continues to be reshaped during disease progression, with epigenetic changes associated with genetic evolution.21 However, to date there has been scant study of the role of retrotransposons in the development and progression of leukamia, with only pyrosequencing-based analysis of single L1 and Alu subfamilies performed.22,23 In this study, we used the Illumina Infinium HumanMethylation 450 BeadChip microarray platform (HM450K) to analyse locus- and subfamily-specific changes in the methylation of retrotransposons in CLL. We identified the HM450K probes mapping to retrotransposons and employed these to analyse the methylation of subfamilies by their evolutionary age. Sites identified as differentially methylated in CLL were analysed in other B-cell malignancies to assess their specificity to the disease, and in pre-diagnostic baseline samples from cohort study participants who went on to develop CLL to determine whether they are detectable prior to diagnosis.

Methods Publicly-available HM450K and gene expression microarray datasets Differentially methylated retrotransposon loci were identified in a discovery cohort of 139 CLL patients and CD19+ B-cells from 14 healthy individuals from the study of Kulis et al.,20 obtained through the International Cancer Genome Consortium (European Genome-phenome Archive accession number: EGAS00001000272; Online Supplementary Table S1). Validation was performed in a cohort of 24 CLL patients attending clinic in North-East England (Online Supplementary Table S2). Ethical haematologica | 2021; 106(1)

approval for the validation study was granted by the North East Newcastle & North Tyneside 1 Research Ethics Committee (REC reference number 17/NE/0361). All donors provided written informed consent. Leading hits were analyzed in publicly-available datasets available through Gene Expression Omnibus (GEO). Details of studies and participants are summarized in the Online Supplementary Table S1. Data from healthy individuals was obtained from the studies of Hannum et al. (n=656; GSE40279),24 the European Prospective Investigation into Cancer and Nutrition (EPIC, n=329; GSE51057)25 and the Young Finns Study (n=184; GSE69270).26 DNA was extracted from whole blood (Hannum et al.)24 or buffy coat (EPIC and Young Finns Study) and therefore represent unfractionated leukocyte preparations. Examination in other haematological malignancies was performed using data from patients with acute lymphoblastic leukemia (ALL) (n=797; GSE47051),27 chronic myeloid leukemia (CML) (n=12; GSE106600),28 acute myeloid leukaemia (AML) (n=68; GSE62298),29 and lymphoma (n=31; GSE42372).30 Analysis of methylation by hematological cell type was achieved using data from the discovery cohort and the studies of Reinius et al. (n=6; GSE35069)31 and Lee et al. (n=4-6; GSE45461).32 Correlations between retrotransposon DNA methylation and the expression of proximal genes were examined within the discovery cohort.20 Comparison of gene expression in normal CD19+ B cells (n=32) and CLL (n=188) was performed using gene expression microarray data from GSE50006.

Pre-diagnostic samples from the Melbourne Collaborative Cohort Study Leading hits unique to CLL patients were examined in prospective samples taken up to 18 years prior to diagnosis in order to reveal potential early epigenetic changes in disease development. HM450K data from 82 prospective CLL cases and 82 age-matched controls within the Melbourne Collaborative Cohort Study (MCCS) were used to examine DNA methylation changes prior to the diagnosis of CLL. The study protocol and performance of HM450K microarrays have been previously described.33

Identification of microarray probes mapping to retrotransposon subfamilies HM450K probes mapping to CpG sites within retrotransposons were identified using the Data Integrator function of the UCSC Genome Browser34 with annotation from RepeatMasker35 for the human genome build GRCh37/hg19 (Online Supplementary Methods).

Statistical analyses Analysis of methylation change by subfamily evolutionary age was performed by linear regression using estimates of time since sequence amplification (millions of years ago, MYA) from the studies of Kapitonov et al.36 and Khan et al.37 Differentially methylated loci in CLL were identified by t-test, with correction for false discovery rate (FDR) by the Benjamini-Hochberg method. All statistical analyses were performed in R (version 3.4.0) using the ggplot2, heatmap2, qqman and corrplot packages, and GraphPad Prism (version 7.0b).

Results Distribution of HM450K probes mapping to retrotransposons We identified 22,927 probes mapping to retrotransposons, 9,797 of which mapped to 117 L1 subfamilies and 99

T.M. Barrow et al. A

Figure 1. Distribution of HM450K probes mapping to L1 and Alu elements. (A-B) The number of probes mapping to each subfamily are displayed. A total of 9,797 probes map to 117 L1 subfamilies, and 13,130 probes to 37 Alu subfamilies. Their categorisation into old (L1M, AluJ), intermediate (L1P, L1PB, AluS), young (L1HS, L1PA, AluY) and related (HAL, FAM) subfamilies are indicated. (C-D) Distribution of probes mapping to retrotransposons (‘Retrotransposons’; n=22,927) by genomic feature (C) and in relation to CpG islands (D), in comparison to all probes on the HM450K array (‘HM450K’; n=485,512). Annotation for genomic features and CpG islands was extracted from the Illumina annotation file. TSS1500 ≤ 1500 bp upstream of the transcription start site; TSS200 ≤ 200 bp upstream of the transcription start site: UTR: untranslated region; Body: gene body; N: north, upstream of a CpG island; S: south, downstream of a CpG island.

13,130 to 37 Alu subfamilies. L1 elements were categorised into oldest (L1M, mammalian-wide), intermediate (L1P, primate-specific) and youngest (L1HS, human-specific and L1PA, primate-amplified) subfamilies. Alu elements were categorised into oldest (AluJ), intermediate (AluS) and youngest (AluY) subfamilies.1 The distribution of subfamilies revealed a high proportion mapping to older L1 (L1M) and intermediate Alu (AluS) subfamilies (Figure 1A–B). A total of 7,581 probes map to L1M elements, 364 to L1P, 1,435 to L1PA, 62 to human-specific L1HS, and 355 to HAL1. For Alu, 3,563 probes map to AluJ elements, 7,626 to AluS, 1,468 to AluY, 73 to FAM, 310 to FLAM, and 90 to FRAM. A comprehensive list of probes by subfamily is provided in the Online Supplementary Table S3. The relative proportions of all Alu probes mapping to the AluJ (28%), AluS (60%) and AluY (12%) subfamilies did not significantly differ from their natural abundance in the human genome (26%, 62% and 13% respectively; P>0.70, Fisher’s exact test) as identified by RepeatMasker.35 Similarly, the proportion of L1 probes mapping to the L1M (80%), L1P (4%) and L1HS/L1PA (16%) subfamilies did not significantly differ from their frequency in the human genome (81%, 6% and 13% respectively; P>0.70). Therefore, the HM450K platform offers a representative means to interrogate retrotransposon methylation in the human genome. In comparison to all >450,000 probes on the HM450K microarray, retrotransposon probes show relative enrichment for mapping to intergenic regions, and depletion of probes mapping to within 200 bases of transcriptional 100

start sites (TSS200) and first exons (Figure 1C). Additionally, there is substantial depletion of probes mapping to CpG islands, with relative enrichment at north and south shelves (Figure 1D).

Retrotransposon elements are highly methylated in normal B cells Retrotransposon elements predominantly showed high levels of methylation in normal CD19+ B cells, but displayed significant methylation variability by genomic feature location and by evolutionary age of the subfamily (Figure 2). Those CpG sites mapping to Alu sequences showed a highly significant trend of increasing methylation from the older to younger elements (P<2.2x10-16, ANOVA), with mean methylation levels (b ± standard deviation [SD]) of 0.72±0.21 for AluJ, 0.74±0.18 for AluS, and 0.78±0.17 for AluY. L1 elements did not display such a linear trend by evolutionary age, with the mean methylation of L1P elements (b: 0.73±0.19) lower than those of L1M (0.76±0.20) and L1H/PA (0.76±0.14) (P<0.008, ANOVA). Analysis by genomic region revealed that those retrotransposons mapping to <200 bp upstream of TSS showed the greatest variability in methylation levels, while those mapping to 5’UTR and first exon regions displayed the least (Figure 2). Furthermore, at these regions L1M elements displayed greater variability than younger L1H/L1PA elements, while AluY similarly displayed less variability of methylation patterns in comparison with the older AluJ sequences. haematologica | 2021; 106(1)

Retrotransposon methylation in CLL

Table 1. Leading differentially methylated retrotransposon loci in chronic lymphocytic leukemia.

Probe ID

Element

chr

cg20820557 cg04258086 cg05922591 cg17505852 cg11665613 cg00981250 cg16564946 cg10795552 cg22894805 cg17342709 cg27020349 cg08641155 cg18406010 cg02575319 cg10474404 cg25995870 cg04312999 cg09149541 cg18286080 cg12115081 cg14266770 cg19679081 cg22813097 cg16386046 cg17084653 cg23985408 cg07134930 cg02948444 cg25947773 cg16273734 cg16348358 cg10664272 cg17713912 cg07293188 cg04211501 cg13988440 cg10180165 cg23840797 cg14985591 cg11848483 cg20959920 cg27447753 cg26924822 cg21518709 cg10163122 cg23177739 cg12079885 cg15129876 cg23876355 cg09153458

AluSx L1ME3B AluY L1M4b FRAM HAL1b L1PA16 L1MB7 L1MB8 AluSg AluSx1 AluSp AluSx1 L1M5 L1MB7 L1MC5 L1MEd AluSx1 L1MC4a L1PB3 L1MB5 AluSq2 AluY L1MD2 AluSx L1ME3 L1MEe L1MB7 L1M4b AluSq2 FLAM_C AluJo HAL1 L1MB7 L1ME3B L1MC2 L1MB3 HAL1b AluSz L1MB5 L1MC4 L1MB3 AluJb L1MEd AluJo L1ME3C L1ME5 AluSz AluSz L1MEg

6:53453216 16:85966544 19:55174624 22:30003602 1:12268883 6:144330345 6:32304275 5:138605898 15:41983772 2:235319934 6:144643174 16:70469236 14:51137625 16:81602666 3:129148046 19:38571162 21:45399399 22:19898335 6:30291349 4:151038390 9:136721182 12:14514672 19:18629910 16:89394863 22:26822031 5:171410890 2:240176050 17:80741558 2:223771010 1:181087919 1:32731477 2:85638053 7:103449892 3:38209834 16:85966435 11:69240805 17:40810558 6:144330232 10:126274649 8:144543485 17:75238948 2:162047157 4:26332762 18:3063385 6:34633132 10:71087363 12:125328475 11:82865068 17:47723651 4:979307

Gene

Genomic feature Island status Enhancer TRUE

LILRB4 NF2 TNFRSF1B PLAGL1 C6orf10

Body Body 3'UTR TSS1500 Body

S_Shelf TRUE S_Shore N_Shelf

MGA; MIR626

Body; TSS200

UTRN ST3GAL2

Body 5'UTR

CMIP C3orf25 SIPA1L3 AGPAT3 TXNRD2 HCG18 DCLK2 VAV2

Body TSS1500 5'UTR Body Body Body Body Body

ELL ANKRD11

Body 5'UTR

FBXW11 HDAC4 TBCD ACSL3

Body Body Body 5'UTR

LCK CAPG RELN OXSR1

5'UTR TSS1500 Body Body

TUBG2 PLAGL1 LHPP ZC3H3

TSS1500 TSS1500 Body Body

TANK RBPJ

Body Body

TRUE TRUE N_Shelf S_Shelf S_Shore N_Shore N_Shelf S_Shelf N_Shelf

TRUE TRUE

TRUE TRUE N_Shelf N_Shelf N_Shelf TRUE TRUE S_Shore TRUE TRUE TRUE N_Shelf TRUE S_Shelf

N_Shore S_Shore N_Shelf N_Shelf TRUE N_Shelf

C6orf106 HK1 SCARB1

Body Body Body

SPOP 5'UTR SLC26A1; IDUA Body; TSS1500

TRUE TRUE S_Shelf N_Shelf TRUE N_Shore

Healthy

CLL

Δb

PFDR

0.86 0.84 0.81 0.87 0.86 0.88 0.84 0.86 0.79 0.84 0.93 0.82 0.85 0.88 0.80 0.84 0.87 0.88 0.87 0.85 0.85 0.84 0.83 0.81 0.82 0.86 0.80 0.88 0.83 0.84 0.81 0.90 0.81 0.88 0.90 0.82 0.80 0.81 0.86 0.89 0.86 0.82 0.75 0.84 0.83 0.85 0.90 0.83 0.85 0.83

0.16 0.26 0.22 0.18 0.28 0.32 0.29 0.26 0.25 0.18 0.25 0.34 0.39 0.30 0.32 0.38 0.33 0.39 0.34 0.28 0.34 0.34 0.40 0.34 0.39 0.35 0.26 0.40 0.34 0.44 0.38 0.49 0.31 0.43 0.11 0.21 0.51 0.30 0.44 0.37 0.46 0.49 0.36 0.43 0.38 0.47 0.62 0.47 0.46 0.33

0.70 0.58 0.59 0.69 0.58 0.56 0.55 0.60 0.54 0.65 0.68 0.48 0.46 0.58 0.49 0.46 0.54 0.49 0.52 0.56 0.51 0.50 0.42 0.47 0.43 0.51 0.54 0.48 0.48 0.40 0.43 0.42 0.50 0.45 0.79 0.61 0.29 0.51 0.41 0.52 0.41 0.34 0.39 0.41 0.44 0.38 0.28 0.36 0.38 0.50

1.88 x 10-104 3.83 x 10-87 1.44 x 10-78 3.81 x 10-78 6.35 x 10-73 2.09 x 10-72 2.97 x 10-70 1.89 x 10-69 3.28 x 10-68 2.48 x 10-66 6.10 x 10-66 2.48 x 10-64 4.56 x 10-63 9.92 x 10-63 2.85 x 10-62 1.07 x 10-61 2.75 x 10-61 3.12 x 10-61 1.68 x 10-59 4.25 x 10-59 4.90 x 10-58 2.08 x 10-57 5.92 x 10-57 2.17 x 10-56 1.14 x 10-55 1.69 x 10-55 3.00 x 10-55 3.33 x 10-54 3.45 x 10-53 3.01 x 10-52 7.64 x 10-52 3.92 x 10-51 7.45 x 10-51 3.20 x 10-50 3.99 x 10-50 9.69 x 10-50 1.13 x 10-49 1.14 x 10-48 1.29 x 10-48 1.63 x 10-48 3.75 x 10-47 1.08 x 10-46 2.62 x 10-46 2.75 x 10-46 3.80 x 10-46 6.27 x 10-46 1.51 x 10-45 2.09 x 10-45 8.21 x 10-45 4.56 x 10-44

The 50 most significantly differentially methylated loci between chronic lymphocytic leukemia (CLL) patients (‘CLL’, n=139) and normal CD19+ B cells from healthy individuals (‘Healthy’, n=14). Mean methylation levels (b) among the healthy individuals and CLL patients are provided, with the mean change in methylation (Δb) and FDR-adjusted P-values. The genomic location (‘chr’), genomic feature, relation to CpG islands and enhancer regions from the Illumina annotation file are provided for each locus.

haematologica | 2021; 106(1)

101

T.M. Barrow et al. AluJ

L1M

AluS

L1P

AluY

L1H

Figure 2. Methylation of L1 and Alu elements by genomic feature in normal CD19+ B cells. Methylation (b) of retrotransposon probes by Alu (A) and L1 (B) subfamilies and genomic feature, as reported by the Illumina annotation file. Retrotransposon probes were categorised as those mapping to within 1,500 or 200 bases of the transcriptional start site (TSS1500, TSS200), the 5’ untranslated region (5’UTR), first exon (1st exon), gene body (Body) and 3’ untranslated region (3’UTR). Lines indicate median values, boxes the interquartile range (IQR), whiskers the highest and lowest values within 1.5*IQR, and outliers are displayed as individual points.

Differential methylation of retrotransposon subfamilies by evolutionary age To identify changes in retrotransposon methylation in CLL, we analysed a publicly-available dataset of 139 CLL patients and normal CD19+ B-cells from 14 healthy individuals,20 serving as the discovery cohort. Our analysis revealed that L1 and Alu elements are hypomethylated in CLL in a subfamily-specific manner. Greater changes in methylation were observed in the older AluJ subfamilies (mean Δb:-0.04) than in the intermediate age AluS (Δb:-0.03) and youngest AluY subfamilies (Δb:-0.02) (Figure 3A; P <0.0001, ANOVA). L1 elements displayed a similar trend for greater hypomethylation among the oldest L1M subfamilies (mean Δb: -0.06) in comparison with the intermediate L1P (Δb: -0.05) and young L1HS & L1PA (Δb:-0.02) subfamilies (P <0.0001, ANOVA), but displayed considerably greater variability by individual subfamily (Figure 3B). Analysis of subfamilies by age of insertion into the human genome revealed a clear trend of greater hypomethylation of Alu elements in those integrated earliest (Figure 3C; r2=0.74, P=0.002). A similar but more subtle trend was also observed for hypomethylation of L1 subfamilies by time since integration (Figure 3D; r2=0.31, P=0.018). Trend

Trend

Locus-specific hypomethylation of retrotransposons in CLL Epigenome-wide analysis of locus-specific retrotransposon methylation in the same 139 CLL patients and 14 healthy individuals revealed 10,782 loci distributed widely across the genome to be differentially methylated 102

in CLL (P <0.05; Figure 4A), with 5,896 mapping to Alu elements and 4,886 to L1 elements. Of these, 9,670 (89.7%) were hypomethylated and 1,112 (10.3%) were hypermethylated (Figure 4B-C), with differential methylation primarily occurring at sites that are highly methylated in normal B cells ( >0.7 for 7,295 loci). The differentially methylated loci showed enrichment at enhancer regions (P=0.002, Fisher’s exact test), but no significant enrichment at CpG islands, shelves or shores (P=0.10-0.73). Unsupervised clustering identified distinct patterns of retrotransposon methylation in CLL in comparison with normal B cells, and an association with IGHV mutational status among CLL patients (Figure 4D). The magnitude of hypomethylation was frequently large, with 43 of the 50 most significant loci displaying changes in methylation (Δb) of >0.40 (Table 1). At these 50 leading loci, methylation was uniformly high in normal CD19+ B cells from healthy individuals (range b: 0.71– 0.95), while hypomethylation was highly frequent in CLL and more common amongst patients with unmutated IGHV (Figure 4E). We noted that single base changes (SNP) have been identified at the C or G position of target CpG sites for 43 of these 50 most significantly differentially methylated loci, but with extremely low minor allele frequencies (<0.01 for 41 of the 43 loci) in European populations (Online Supplementary Table S4). FDR

Analysis in a validation cohort To confirm our findings in the discovery cohort, we performed epigenome-wide analysis of retrotransposon haematologica | 2021; 106(1)

Retrotransposon methylation in CLL

Figure 3. Differential methylation of L1 and Alu subfamilies in chronic lymphocytic leukemia by evolutionary age. (A-B) Mean change in methylation (Δb) for probes mapping to each of the 37 Alu (A) and 117 L1 (B) subfamilies by the FDR-adjusted P-value. (C-D) Mean methylation change (Δb) of Alu (C) and L1 (D) subfamilies by time since integration into the genome (millions year ago, MYA). Mean and 95% Confidence Intervals (95% CI) are displayed.

methylation in a validation cohort of 24 CLL patients (Online Supplementary Table S2), in comparison to the same 14 healthy individuals from the discovery cohort. A total of 9,488 retrotransposon loci were differentially methylated in the validation CLL cohort (P <0.05) with a very high level of correlation in the magnitude of methylation changes observed in the discovery and validation cohorts relative to normal CD19+ B cells (rho=0.76, P<2.2x10-16, Online Supplementary Figure S1). Of the 50 leading loci in the discovery cohort (Table 1), 41 were differentially methylated in the validation cohort (P <0.0000001) with changes in methylation (Δb) of >0.40; the probes for the other nine loci were missing from the validation cohort dataset and could not be examined (Online Supplementary Table S5). FDR

class-switched memory B cells (n=3) available from the study of Kulis et al.20 Of the 10,782 loci previously identified as differentially methylated in CLL, 8,898 (82.5%) did not display substantial changes in methylation (Δb>0.2) during B-cell maturation. Importantly only one of the leading 50 loci (Table 1), cg13988440, displayed such a change. Therefore, in contrast to many other studies of the CLL epigenome, our analysis has predominantly revealed changes in DNA methylation that are specific to the disease.

FDR

DNA methylation of retrotransposon loci during B cell differentiation Most epigenetic changes observed in CLL mirror those that occur during the process of B-cell development.21 In order to account for differential methylation that may be the product of B-cell maturation or proliferation, as opposed to being unique to CLL, we utilised publiclyavailable DNA methylation microarray data from CD19+ CD27– IgD+ naïve B cells (n=3) and CD19+ CD27+ IgA/G+ haematologica | 2021; 106(1)

Effect of locus-specific hypomethylation upon proximal gene expression In order to explore the functional consequences of locus-specific retrotransposon hypomethylation events, we examined the effects of differential methylation at the leading 50 loci (Table 1) upon the expression of proximal genes using expression microarray data from the same 139 CLL patients within the discovery cohort. Thirty-seven of the leading 50 hits from the discovery cohort map to genes, with the other 13 mapping to intergenic regions that are >2 kb from the nearest transcriptional start site (Table 1). Of the 37 hits mapping to genes, differential methylation at 29 of the loci was confirmed in the validation cohort (Figure 5A-B; Online Supplementary Figure 1). 103

T.M. Barrow et al. A

Figure 4. Locus-specific differential methylation of retrotransposon elements in chronic lymphocytic leukemia. (A) Manhattan plot of retrotransposon loci showing differential methylation between chronic lymphocytic leukemia (CLL) patients (n=139) and normal CD19+ B cells from healthy individuals (n=14). The threshold (red line) indicates P<5x10-8. (B-C) Volcano plots for probes mapping to Alu (B) and L1 (C) elements. The change in methylation (Δb) and P-value (-log10(p)) are displayed, with significantly different loci (P <0.05) highlighted in blue. (D) Heatmap displaying unsupervised clustering of the 200 most significantly differentially methylated loci. The colour scale indicates methylation level, from low (yellow) to high (blue). Healthy individuals (turquoise, n=14) and CLL patients with mutated (orange, n=75), unmutated (dark red, n=57) or unknown (grey, n=6) IGHV status are indicated. (E) Methylation of the six most significantly differentially methylated loci in healthy individuals (turquoise, n=14) and CLL patients with IGHV mutated (orange, n=75) and unmutated (dark red, n=57) disease. Lines indicate median values, boxes the interquartile range (IQR), whiskers the highest and lowest values within 1.5*IQR, and outliers are displayed as individual points. FDR

Methylation at 7 of these 29 loci was correlated with expression of the proximal gene (Figure 5C–D; P <0.05, Spearman rank correlation), illustrating the capacity for retrotransposon hypomethylation to modulate gene transcription. These seven genes were C6orf106 (ILRUN), DCLK2, HK1, TANK, TBCD, TNFRSF1B (TNFR2) and TXNRD2 (Online Supplementary Table S6). All showed negative correlations between methylation and expression (rho=-0.25 to -0.59), with the exception of TBCD which demonstrated a positive correlation (rho=0.25). Each of these seven genes displayed uniformly low levels of methylation at their promoter regions, thereby potentially facilitating retrotransposon-mediated regulation of gene expression (Online Supplementary Figure S2). In an independent cohort of 188 CLL patients and CD19+ B cells isolated from 32 healthy individuals, all seven genes were determined to be differentially expressed in CLL (Figure 5E–F; P <0.05, Mann Whitney U test). In order to examine the clinical relevance of these genes, we examined their expression in the discovery cohort in relation to patient outcomes. High expression of DCLK2 and TNFRSF1B were associated with reduced patient surFDR

FDR

104

vival (Figure 5G–H; P <0.05, log-rank test), while high expression of HK1, ILRUN, TANK, and TBCD were associated with improved patient survival. In order to examine whether the associations with impaired patient prognosis were driven by IGHV status, analysis was then performed separately for IGHV mutated (n=75) and unmutated (n=57) patients, where status was known. Among IGHV mutated patients, significant associations with survival were retained for ILRUN and TANK expression (P <0.05), while DCLK2 P =0.0532) approached significance. HK1, TBCD and TNFRSF1B no longer achieved statistical significance in this restricted patient group (all P <0.13). Among IGHV unmutated patients, significant associations were present for expression of TANK (P =0.0210) and TNFRSF1B (P =0.0005). FDR

FDR

Stability of retrotransposon methylation across hematopoietic cell types It is recognised that studies of DNA methylation in blood samples must take into account changes in the relative proportions of the cell types.38 We therefore examined our 10 leading loci (Table 1) in publicly-available HM450K haematologica | 2021; 106(1)

Retrotransposon methylation in CLL

Figure 5. Impact of retrotransposon methylation upon the expression of proximal genes. (A-B) Methylation at cg12115081 (A) and cg11665613 (B) in normal CD19+ B cells (‘Bcell’; n=14) and chronic lymphocytic leukemia (CLL) patients within the discovery cohort (‘Disc’; n=139) and validation cohort (‘Val’; n=24). (C-D) Correlation between retrotransposon methylation and proximal gene expression for cg12115081 and DCLK2 expression (C) and for cg11665613 with TNFRSF1B expression (D). (E-F) Expression of DCKL2 (E) and TNFRSF1B (F) in normal CD19+ B-cells (n=32) and CLL patients (n=188) in an independent cohort. (G-H) Kaplan-Meier plots for patient overall survival stratified by expression level (high/low) of DCLK2 (G) and TNFRSF1B (H) in the Discovery cohort.

datasets from isolated hematopoietic cell types to reveal differences in retrotransposon methylation by leukocyte cell type. We employed data obtained from samples of whole blood, peripheral blood mononuclear cells, CD16+ neutrophils, eosinophils, CD14+ monocytes, multipotent progenitors, pre-B cells, immature B cells, CD19+ B cells, CD4+ T-cells, CD8+ T cells, and CD56+ natural killer cells (each n=6). Nine of the loci (cg20820557, cg04258086, cg05922591, cg17505852, cg11665613, cg00981250, cg16564946, cg10795552, and cg17342709) were highly methylated across all cell types and displayed little variation between them (Figure 6A-B). In contrast, cg22894805 showed high methylation in lymphoid cells but low levels in myeloid cells (Figure 6C).

Specificity of differential retrotransposon methylation to CLL In order to determine whether the observed changes are specific to CLL or may also be present in other lymphoid and myeloid malignancies, we analysed the 10 most significantly altered loci (Table 1) in publicly-available HM450K datasets from healthy individuals and patients with other hematological malignancies. At each of these 10 loci we had observed mean methylation levels (b) of <0.32 in CLL patients in comparison to >0.79 in normal CD19+ B cells. First, to assess the variability in methylation of these loci in the general population, we leveraged three largescale studies of blood samples from a total of 1,169 healthy individuals: the Hannum et al. study of human ageing (n=656);24 the European Prospective Investigation into Cancer and Nutrition (EPIC, n=329);25 and the Young Finns study (n=184).26 Our analysis confirmed the loci to be almost exclusively highly methylated in healthy indihaematologica | 2021; 106(1)

viduals (Figure 6D–E). Eight of the loci (cg20820557, cg04258086, cg05922591, cg17505852, cg11665613, cg00981250, cg10795552, and cg17342709) displayed mean methylation levels (b) of 0.86–0.96 in each of the three cohorts of healthy individuals, with low variability across the populations (Online Supplementary Table S7). cg16564946 was also highly methylated but displayed more variation within the cohorts (mean b: 0.74±0.11 – 0.79±0.07). Methylation at cg22894805 measured in these unfractionated leukocyte samples was lower than previously observed in isolated CD19+ B-cells (Table 1 and Online Supplementary Table S7) due to variation in methylation by leukocyte cell type (Figure 6C), but was still different in each study compared to CLL patients (all P<0.0001; Mann Whitney U test). In order to determine whether the identified loci may also be differentially methylated in other hematological malignancies, we examined datasets from patients with ALL (n=797),27 CML (n=12),28 AML (n=68),29 and lymphoma (diffuse large B-cell lymphoma and Burkitt’s lymphoma, n=31).30 Hypomethylation was predominantly confined to CLL at nine of the loci (cg20820557, cg04258086, cg05922591, cg17505852, cg11665613, cg00981250, cg16564946, cg10795552, and cg17342709), with mean methylation (b) levels of >0.74 in ALL, CML and AML (Figure 6D-E and Online Supplementary Table S7). Mean methylation levels were uniformly lower in lymphoma (b: 0.61–0.73), with up to 23% of patients having methylation levels of <0.40 at the loci. In contrast to the other nine loci, and in concordance with our analysis by hematopoietic cell type, cg22894805 had low methylation levels in myeloid malignancies (AML, CML) and higher levels in lymphoid malignancies (ALL, lymphoma) (Figure 6F). 105

T.M. Barrow et al. A

Figure 6. Differentially methylated loci in other haematological malignancies and leukocyte cell types. (A-C) Methylation (b) of three loci by blood cell type. Publiclyavailable datasets from whole blood (WB, n=6), peripheral blood mononuclear cells (PBMC, n=6), CD16+ neutrophils (Neutro, n=6), eosinophils (Eosin, n=6), CD14+ monocytes (Mono, n=6), multipotent progenitor cells (MPP, n=6), Pre-B I and Pre-B II cells (both n=6), immature B cells (B_imm, n=4), CD19+ B cells (n=14), CD4+ and CD8+ T cells (both n=6), and CD56+ natural killer cells (NK, n=6). Lines indicate median values, boxes the interquartile range (IQR), whiskers the highest and lowest values within 1.5*IQR, and outliers are displayed as individual points. (D-F) Methylation (β) of the same three loci in publicly-available datasets from three studies of healthy individuals (Hannum et al.24, n=656; EPIC, n=329; the Young Finns Study (YFS), n=184), CLL patients within the discovery cohort (n=139), and patients with ALL (n=797), AML (n=68), CML (n=12), and diffuse large B-cell and Burtkitt’s lymphomas (Lymph, n=31).

Together, our results indicate that the largest observed differences in retrotransposon methylation are highly specific to CLL and are not the product of changes in cell type proportions, including hypomethylation at cg11665613 that is associated with reactivation of TNFRSF1B expression.

CLL cases, with the remainder displaying normal, high levels of methylation at each locus (Figure 7A). Methylation across the 10 loci was very highly correlated (Figure 7B), with the same nine individuals displaying hypomethylation at each locus. There was weak evidence of a positive association between methylation and time to diagnosis (P =0.06–0.12) at all loci except for cg10795552 (P =0.51). Notably, hypomethylation was more commonly present among cases diagnosed <7 years after blood draw (7 of 24, 29%) than those diagnosed more than 7 years after (2 of 58, 3%) (Figure 7C). ROC curve analysis identified that a methylation cut-off of b=0.90 for cg2080557 provides 45% sensitivity and 91% specificity (area under curve [AUC]=0.66, P=0.0003) in predicting diagnosis with CLL (Figure 7D). In order to determine whether hypomethylation is highly localised or more regional, we examined methylation at other CpG sites captured by the HM450K array that are within 200 bp of the target loci. Only two of these loci had regional CpG sites in close proximity that are captured by the array: two CpG sites within 119 bp of cg04258086 (cg04211501 and cg06951664); and two CpG sites up to 184 bp downstream of cg00981250 (cg23840797 and FDR

Locus-specific hypomethylation of retrotransposons prior to diagnosis We sought to establish whether these epigenetic changes might constitute early events that are detectable prior to patient diagnosis. We analysed methylation of the 10 leading loci in blood samples taken from 82 individuals up to 18 years prior to their being diagnosed with CLL and 82 individually matched controls who remained free of disease at the same age at follow-up. For nine of the loci (cg20820557, cg04258086, cg05922591, cg17505852, cg11665613, cg00981250, cg16564946, cg10795552, and cg17342709) lower methylation levels were prospectively associated with the diagnosis of CLL (all P<0.003, Mann Whitney U test). Additionally, there was weak evidence of an association for cg22894805 (P=0.0501). Reduced methylation levels were, however, confined to 9 of the 82 106

FDR

haematologica | 2021; 106(1)

Retrotransposon methylation in CLL

Figure 7. Hypomethylation of retrotransposon loci prior to chronic lymphocytic leukemia diagnosis. Analysis of 10 retrotransposon loci observed to be hypomethylated in chronic lymphocytic leukemia (CLL) patients, analysed by HM450K in prospective samples taken from future CLL patients within the Melbourne Collaborative Cohort Study. (A) Methylation (Î˛) of five loci in 82 future CLL patients (Case) and 82 age-matched individuals who remained free of the disease (Control) (***P<0.005, ****P<0.0001). (B) Correlation in methylation between the ten loci among the 82 future CLL cases. The colour scale indicates negative (red) and positive (blue) correlations. (C) Correlation between methylation (Î˛) of cg20820557 and time to diagnosis (years) in the 82 future CLL cases. (D) ROC curve analysis for cg20820557 in the prediction of CLL diagnosis among 82 future CLL cases and 82 age-matched healthy controls who remained free of the disease (area under curve [AUC]=0.66, P=0.0003). (E) Regional analysis of methylation (Î˛) at HM450K probes mapping to within 200 bases of cg04258086 and cg00981250 (indicated by asterisks, *). Methylation at loci within the same patient are indicated by connecting lines.

cg18593039). For both regions, the neighbouring CpG sites mapped to within the same retrotransposon element as the target (L1ME3B and HAL1 respectively). Methylation at the neighbouring CpG sites was observed to be highly similar to that of the target loci (Figure 7E), demonstrating that the observed epigenetic changes are not confined to a single CpG site.

Discussion To the best of our knowledge, our study is the first to utilise the Illumina Infinium microarray platform to analyse locus- and subfamily-specific methylation of retrotransposons in human disease. The potential for this approach has previously been demonstrated through use of the HM27K and HM450K microarray platforms to analyse L1 and Alu elements categorised as young (L1H, AluY), intermediate (L1P, AluS) and old (L1M, AluJ) in healthy tissues39 and a lymphoblastoid cell line.40 We have built on this to identify, for the first time, differential methylation of LINE-1 and Alu elements by their evoluhaematologica | 2021; 106(1)

tionary heritage in leukemia. Importantly, our analysis has identified novel epigenetic changes specific to CLL that are highly frequent among patients, that modulate the expression of proximal genes, and that can be detected prior to diagnosis in some individuals. Together, our results suggest that epigenetic dysregulation of retrotransposons may be implicated in the development and progression of CLL. There remains significant ignorance regarding the genetic and epigenetic origins and evolution of CLL. It has recently been demonstrated that the genome and epigenome show co-evolution during disease progression.21,41 This co-evolution predominantly consists of hypomethylation away from promoters and transcriptional start sites at regions that are predominantly highly methylated in normal B cells21 which, hence, may describe general loss of epigenome maintenance as opposed to sitespecific selection for gene silencing or activation. Our study has revealed hypomethylation of retrotransposon elements in a subfamily-specific manner which, in accordance with the findings of Oakes et al.,21 primarily occurs away from CpG islands in regions that are highly methy107

T.M. Barrow et al.

lated in normal B cells. Our findings may thus throw light on the underlying cause of this epigenetic evolution in CLL. Interestingly, we observed that this hypomethylation is significantly enriched at enhancer regions, which may implicate the dysregulation of retrotransposons in the transcriptional re-wiring of CLL cells. To date there have only been two studies of retrotransposon methylation in CLL, both of which reported associations with genetic instability. Fabris et al.22 reported significant hypomethylation of L1 and Alu elements that was associated with 17p deletions, while Hoxha et al.23 reported associations between L1 and Alu methylation with telomere length. This is corroborated by recent findings that telomere length is associated with del(17p), del(11q) and mutations in ATM and SF3B1,42 together linking methylation of retrotransposons, telomere length and the acquisition of genetic abnormalities in CLL. This is further supported by the extensive evidence linking loss of retrotransposon methylation to their activation and genetic instability in solid tumours.13,14 However, the studies by Fabris et al.22 and Hoxha et al.23 used a pyrosequencingbased approach that was limited to the L1HS and AluSx subfamilies, on account of their relative abundance in the genome and subsequent suitability as surrogate markers of global DNA methylation patterns. Such an approach does not facilitate locus-specific analysis. While providing novel information about the disease, these two studies were unable to address the full complexity of retrotransposon DNA methylation in CLL. Indeed, we have demonstrated the utility and necessity of performing subfamilyand locus-specific analysis to fully appreciate their implication in malignancies. Comprehensive analysis of retrotransposons by high-throughput sequencing approaches is highly challenging due to the difficulty in mapping sequencing reads containing repetitive sequences to singular genomic regions, requiring tailored bioinformatics approaches such as RetroSeq.43 Hence, the approach that we have developed, to analyse retrotransposons using the widely-utilised HM450K platform, enables a comparatively simple and cost-effective means to perform this demanding work. We have utilised this locus-specific analytical approach to reveal that hypomethylation of retrotransposons can modulate the expression of proximal genes in CLL. In particular, we identified hypomethylation of an L1PB3 sequence within the second intron of DCLK2 and a FRAM sequence within the 3’UTR of TNFRSF1B that are inversely associated with expression of the genes, both of which are overexpressed in CLL and for which higher expression levels are associated with reduced patient survival. TNFRSF1B encodes a member of the TNF-receptor superfamily that interacts with TRAF2 to promote cell proliferation and survival via stimulation of the non-canonical NF-kB pathway, and it is known to be overexpressed in CLL.44,45 Myeloid cell lines treated with the DNMT inhibitor 5-aza-2’-deoxycytidine show increased expression of TNFRSF1B,46 and our study has demonstrated that hypomethylation of the FRAM element within the 3’UTR is associated with increased expression of the gene. While many studies have focussed primarily on epigenetic changes at promoter and enhancer regions, DNA methylation at 3’UTR has been shown to be associated with differential gene expression and patient outcomes in solid tumours,47 although the underlying mechanisms have not been elucidated. Increased expression of the TNFRSF1B 108

promotes cell survival and resistance to apoptosis,48 which may explain our observation of worse patient prognosis with high TNFRSF1B expression. In contrast, the function of DCLK2 in CLL biology is less clear. This gene encodes a regulator of microtubule polymerisation, and while family members have been identified as potential prognostic markers in gastric and colorectal cancers,49,50 DCLK2 has not previously been reported as differentially methylated or expressed in CLL. Further work is required to establish the functional role of this gene in the disease. We also report that hypomethylation of an intronic AluSx1 element within TXNRD2 was associated with increased expression of the gene in CLL. While we observed no association with patient survival, a germline variant within 300 bp of this Alu element has been reported to be associated with shorter time to first treatment.51 The gene encodes a thioredoxin reductase that promotes malignant cell survival by providing protection against reactive oxygen species,52 and targeting these enzymes by use of chemotherapeutic agents such as auranofin may be effective in the treatment of CLL.53 In addition to the impact on the expression of proximal genes, epigenetic dysregulation of L1 and Alu elements may further be implicated in CLL biology through retrotransposition events. Recent work elsewhere has identified enrichment of transposable elements in the transcriptome of CLL patients, but without being able to identify the underlying cause.54 Our study has revealed widespread locus-specific epigenetic dysregulation of retrotransposon elements in the genome, which may serve to explain their increased expression in CLL, but the full implications of these observations remains to be elucidated. Hypomethylation of retrotransposon elements leads to their re-activation55 and potential retrotransposition into the genome. Somatic retrotransposition has been widely studied in solid tumours and is a highly frequent event in cancers of the lung, colon, breast and prostate.15 However, to date there have been no studies of this phenomenon in leukemias. Our findings, together with those observations elsewhere of reactivation of transposable elements54 and association of retrotransposon methylation with the acquisition of 17p deletions,22 indicate that the study of this phenomenon may help to identify novel genetic drivers of the disease. The most frequently mutated driver genes (NOTCH1, SF3B1, TP53, MYD88 and BIRC3) are mutated in only 2-12% of CLL patients,56 and large-scale studies have been unable to identify driver mutations in 25-40% of patients,57,58 underlining the need for further research in this area. While aberrant methylation was more frequently observed in the older and (presumed) inactive subfamilies such as L1M, hypomethylation of active AluY and L1HS elements was also identified, with two AluY sequences on chromosome 19 amongst the leading hits from our analysis. We noted that rare genetic variants have been reported at the target CpG sites for many of the leading hits from this study. Genome-wide association studies have identified more than 20 genetic susceptibility loci for CLL,59 yet together these loci only account for 25% of heritable risk60 and therefore genetic predisposition to CLL remains to be fully explained. SNP are enriched within Alu elements,61 and such variants at CpG sites can lead to their elimination and thus loss of methylation and reactivation. However, the extremely low minor allele frequencies for the variants mapping to the 50 leading loci and our obserhaematologica | 2021; 106(1)

Retrotransposon methylation in CLL

vation of normal methylation in most individuals prior to their CLL diagnosis suggest that these loci do not correspond to inherited predisposition to the disease. It nonetheless remains to be established whether the observed hypomethylation could originate through the acquisition of genetic changes during disease initiation, as the high frequency of methylated CpG sites within L1 and Alu sequences results in a high point mutation rate through the deamination of methylated-cytosine. As our study has primarily involved analysis of publicly-available datasets, we have been unable to sequence regions of interest to examine regional changes in DNA methylation and detect genetic variants that may be implicated in the observed changes. Furthermore, we have not been able to examine whether these epigenetic changes lead to re-expression of the elements and somatic retrotransposition. However, our study has significant strengths. We have developed an approach that enables the locus-specific interrogation of L1 and Alu element methylation across the genome using the widely-utilized HM450K platform, thereby facilitating its ready application to studies elsewhere. We have also comprehensively examined the leading hits from our analysis, including examination in 1,169 healthy individuals to determine the frequency of hypomethylation in the general population, fractionated leukocytes to identify cell type-specific differences, and other hematological malignancies to determine the specificity of the observed changes to CLL. In summary, we have identified locus- and subfamilyspecific hypomethylation of L1 and Alu elements that is highly frequent and specific to CLL, modulates the expression proximal genes, and was observable in a proportion of future CLL cases prior to their diagnosis. Further work is required to elucidate how these epigenetic changes may be implicated in leukaemogenesis and to

References 1. Mighell AJ, Markham AF, Robinson PA. Alu sequences. FEBS Lett. 1997;417(1):1-5. 2. Price AL, Eskin E, Pevzner PA. Wholegenome analysis of Alu repeat elements reveals complex evolutionary history. Genome Res. 2004;14(11):2245-2252. 3. Mills RE, Bennett EA, Iskow RC, Devine SE. Which transposable elements are active in the human genome? Trends Genet. 2007; 23(4):183-191. 4. Konkel MK, Walker JA, Hotard AB, et al. Sequence analysis and characterization of active human Alu subfamilies based on the 1000 Genomes Pilot Project. Genome Biol Evol. 2015;7(9):2608-2622. 5. Beck CR, Collier P, Macfarlane C, et al. LINE-1 retrotransposition activity in human genomes. Cell. 2010;141(7):1159-1170. 6. Xing J, Zhang Y, Han K, et al. Mobile elements create structural variation: analysis of a complete human genome. Genome Res. 2009;19(9):1516-1526. 7. Bollati V, Baccarelli A, Hou L, et al. Changes in DNA methylation patterns in subjects exposed to low-dose benzene. Cancer Res. 2007;67(3):876-880. 8. Breton CV, Byun HM, Wenten M, Pan F, Yang A, Gilliland FD. Prenatal tobacco smoke exposure affects global and gene-spe-

haematologica | 2021; 106(1)

investigate somatic retrotransposition as a potential driver of CLL.

Data availability The datasets used in this study are publicly-available through the European Genome-phenome Archive (https://www.ebi.ac.uk/ega/home) and the Gene Expression Omnibus repository (https://www.ncbi.nlm.nih.gov/geo/). Accession numbers for all datasets used in this study can be found within Online Supplementary Table S1. Disclosures No Conflicts of interest to disclose. Contributions TB and HB conceived the study; TB identified HM450K loci mapping to retrotransposons and performed all data analysis; NWD, RM and GG provided the HM450K data from the MCCS and EW and GS provided HM450K data from the validation cohort; TB and HB performed the data interpretation, with input from NWD, RM, GG, EW and GS; TB and HB drafted the manuscript, which was then revised by all authors. All authors had final approval of the submitted manuscript. Acknowledgments We would like to thank all patients involved in the study. We acknowledge the Edinburgh Clinical Research Facility for the performance of the DNA methylation microarrays from the validation cohort, and the Newcastle University Bioinformatics Support Unit for their assistance with the analysis of DNA methylation microarray data. Funding This work was supported by funding from Bright Red (awarded to Timothy Barrow) and the JGW Patterson Foundation (grant number 30015.088.045/PA/IXS, awarded to Gordon Strathdee).

cific DNA methylation. Am J Respir Crit Care Med. 2009;180(5):462-467. 9. Byun HM, Motta V, Panni T, et al. Evolutionary age of repetitive element subfamilies and sensitivity of DNA methylation to airborne pollutants. Part Fibre Toxicol. 2013;10:28. 10. Xiao-Jie L, Hui-Ying X, Qi X, Jiang X, Shi-Jie M. LINE-1 in cancer: multifaceted functions and potential clinical implications. Genet Med. 2016;18(5):431-439. 11. Benard A, van de Velde CJ, Lessard L, et al. Epigenetic status of LINE-1 predicts clinical outcome in early-stage rectal cancer. Br J Cancer. 2013;109(12):3073-3083. 12. Wolff EM, Byun HM, Han HF, et al. Hypomethylation of a LINE-1 promoter activates an alternate transcript of the MET oncogene in bladders with cancer. PLoS Genet. 2010;6(4):e1000917. 13. Daskalos A, Nikolaidis G, Xinarianos G, et al. Hypomethylation of retrotransposable elements correlates with genomic instability in non-small cell lung cancer. Int J Cancer. 2009;124(1):81-87. 14. Symer DE, Connelly C, Szak ST, et al. Human L1 retrotransposition is associated with genetic instability in vivo. Cell. 2002; 110(3):327-338. 15. Tubio JMC, Li Y, Ju YS, et al. Mobile DNA in cancer. Extensive transduction of nonrepetitive DNA mediated by L1 retrotrans-

position in cancer genomes. Science. 2014; 345(6196):1251343. 16. Lee E, Iskow R, Yang L, et al. Landscape of somatic retrotransposition in human cancers. Science. 2012;337(6097):967-971. 17. Scott EC, Gardner EJ, Masood A, Chuang NT, Vertino PM, Devine SE. A hot L1 retrotransposon evades somatic repression and initiates human colorectal cancer. Genome Res. 2016;26(6):745-755. 18. Ewing AD, Gacita A, Wood LD, et al. Widespread somatic L1 retrotransposition occurs early during gastrointestinal cancer evolution. Genome Res. 2015;25(10):15361545. 19. Doucet-O’Hare TT, Rodić N, Sharma R, et al. LINE-1 expression and retrotransposition in Barrett’s esophagus and esophageal carcinoma. Proc Natl Acad Sci U S A. 2015; 112(35):E4894-4900. 20. Kulis M, Heath S, Bibikova M, et al. Epigenomic analysis detects widespread gene-body DNA hypomethylation in chronic lymphocytic leukemia. Nat Genet. 2012;44(11):1236-1242. 21. Oakes CC, Claus R, Gu L, et al. Evolution of DNA methylation is linked to genetic aberrations in chronic lymphocytic leukemia. Cancer Discov. 2014;4(3):348-361. 22. Fabris S, Bollati V, Agnelli L, et al. Biological and clinical relevance of quantitative global methylation of repetitive DNA sequences in

109

T.M. Barrow et al. chronic lymphocytic leukemia. Epigenetics. 2011;6(2):188-194. 23. Hoxha M, Fabris S, Agnelli L, et al. Relevance of telomere/telomerase system impairment in early stage chronic lymphocytic leukemia. Genes Chromosomes Cancer. 2014;53(7):612-621. 24. Hannum G, Guinney J, Zhao L, et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell. 2013;49(2):359-367. 25. Demetriou CA, Chen J, Polidoro S, et al. Methylome analysis and epigenetic changes associated with menarcheal age. PLoS One. 2013;8(11):e79391. 26. Kananen L, Marttila S, Nevalainen T, et al. Aging-associated DNA methylation changes in middle-aged individuals: the Young Finns study. BMC Genomics. 2016;17:103. 27. Nordlund J, Bäcklin CL, Wahlberg P, et al. Genome-wide signatures of differential DNA methylation in pediatric acute lymphoblastic leukemia. Genome Biol. 2013; 14(9):r105. 28. Maupetit-Mehouas S, Court F, Bourgne C, et al. DNA methylation profiling reveals a pathological signature that contributes to transcriptional defects of CD34. Mol Oncol. 2018;12(6):814-829. 29. Ferreira HJ, Heyn H, Vizoso M, et al. DNMT3A mutations mediate the epigenetic reactivation of the leukemogenic factor MEIS1 in acute myeloid leukemia. Oncogene. 2016;35(23):3079-3082. 30. Matsunaga A, Hishima T, Tanaka N, et al. DNA methylation profiling can classify HIVassociated lymphomas. AIDS. 2014; 28(4):503-510. 31. Reinius LE, Acevedo N, Joerink M, et al. Differential DNA methylation in purified human blood cells: implications for cell lineage and studies on disease susceptibility. PLoS One. 2012;7e41361. 32. Lee ST, Xiao Y, Muench MO, et al. A global DNA methylation and gene expression analysis of early human B-cell development reveals a demethylation signature and transcription factor network. Nucleic Acids Res. 2012;40(22):11339-11351. 33. Wong Doo N, Makalic E, Joo JE, et al. Global measures of peripheral blood-derived DNA methylation as a risk factor in the development of mature B-cell neoplasms. Epigenomics. 2016;8(1):55-66. 34. Kent WJ, Sugnet CW, Furey TS, et al. The human genome browser at UCSC. Genome Res. 2002;12(6):996-1006. 35. Smit AFA, Hubley R, Green P. RepeatMasker Open-4.0. 2013 36. Kapitonov V, Jurka J. The age of Alu subfam-

110

ilies. J Mol Evol. 1996;42(1):59-65. 37. Khan H, Smit A, Boissinot S. Molecular evolution and tempo of amplification of human LINE-1 retrotransposons since the origin of primates. Genome Res. 2006;16(1):78-87. 38. Houseman EA, Accomando WP, Koestler DC, et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics. 2012;13:86. 39. Price EM, Cotton AM, Peñaherrera MS, McFadden DE, Kobor MS, Robinson W. Different measures of “genome-wide” DNA methylation exhibit unique properties in placental and somatic tissues. Epigenetics. 2012;7(6):652-663. 40. Zheng Y, Joyce BT, Liu L, et al. Prediction of genome-wide DNA methylation in repetitive elements. Nucleic Acids Res. 2017; 45(15):8697-8711. 41. Landau DA, Clement K, Ziller MJ, et al. Locally disordered methylation forms the basis of intratumor methylome variation in chronic lymphocytic leukemia. Cancer Cell. 2014;26(6):813-825. 42. Strefford JC, Kadalayil L, Forster J, et al. Telomere length predicts progression and overall survival in chronic lymphocytic leukemia: data from the UK LRF CLL4 trial. Leukemia. 2015;29(12):2411-2414. 43. Keane TM, Wong K, Adams DJ. RetroSeq: transposable element discovery from nextgeneration sequencing data. Bioinformatics. 2013;29(3):389-390. 44. Hui D, Satkunam N, Al Kaptan M, Reiman T, Lai R. Pathway-specific apoptotic gene expression profiling in chronic lymphocytic leukemia and follicular lymphoma. Mod Pathol. 2006;19(9):1192-1202. 45. Wang J, Coombes KR, Highsmith WE, Keating MJ, Abruzzo LV. Differences in gene expression between B-cell chronic lymphocytic leukemia and normal B cells: a metaanalysis of three microarray studies. Bioinformatics. 2004;20(17):3166-3178. 46. Laurenzana A, Petruccelli LA, Pettersson F, et al. Inhibition of DNA methyltransferase activates tumor necrosis factor alphainduced monocytic differentiation in acute myeloid leukemia cells. Cancer Res. 2009;69 (1):55-64. 47. McGuire MH, Herbrich SM, Dasari SK, et al. Pan-cancer genomic analysis links 3’UTR DNA methylation with increased gene expression in T cells. EBioMedicine. 2019; 43:127-137. 48. Baud V, Karin M. Signal transduction by tumor necrosis factor and its relatives. Trends Cell Biol. 2001;11(9):372-377. 49. Dai J, Li ZX, Zhang Y, et al. Whole genome

messenger RNA profiling identifies a novel signature to predict gastric cancer survival. Clin Transl Gastroenterol. 2019; 10(1): e00004. 50. Gao T, Wang M, Xu L, Wen T, Liu J, An G. DCLK1 is up-regulated and associated with metastasis and prognosis in colorectal cancer. J Cancer Res Clin Oncol. 2016; 142(10):2131-2140. 51. Mosquera Orgueira A, Antelo Rodríguez B, Alonso Vence N, et al. The association of germline variants with chronic lymphocytic leukemia outcome suggests the implication of novel genes and pathways in clinical evolution. BMC Cancer. 2019;19(1):515. 52. Lincoln DT, Ali Emadi EM, Tonissen KF, Clarke FM. The thioredoxin-thioredoxin reductase system: over-expression in human cancer. Anticancer Res. 2003;23(3B):24252433. 53. Fiskus W, Saba N, Shen M, et al. Auranofin induces lethal oxidative and endoplasmic reticulum stress and exerts potent preclinical activity against chronic lymphocytic leukemia. Cancer Res. 2014;74(9):25202532. 54. Ferreira PG, Jares P, Rico D, et al. Transcriptome characterization by RNA sequencing identifies a major molecular and clinical subdivision in chronic lymphocytic leukemia. Genome Res. 2014;24(2):212-226. 55. Schulz WA, Steinhoff C, Florl AR. Methylation of endogenous human retroelements in health and disease. Curr Top Microbiol Immunol. 2006;310:211-250. 56. Baliakas P, Hadzidimitriou A, Sutton LA, et al. Recurrent mutations refine prognosis in chronic lymphocytic leukemia. Leukemia. 2015;29(2):329-336. 57. Landau DA, Tausch E, Taylor-Weiner AN, et al. Mutations driving CLL and their evolution in progression and relapse. Nature. 2015;526(7574):525-530. 58. Puente XS, Beà S, Valdés-Mas R, et al. Noncoding recurrent mutations in chronic lymphocytic leukaemia. Nature. 2015; 526 (7574):519-524. 59. Berndt SI, Skibola CF, Joseph V, et al. Genome-wide association study identifies multiple risk loci for chronic lymphocytic leukemia. Nat Genet. 2013;45(8):868-876. 60. Law PJ, Berndt SI, Speedy HE, et al. Genome-wide association analysis implicates dysregulation of immunity genes in chronic lymphocytic leukaemia. Nat Commun. 2017;8:14175. 61. Ng SK, Xue H. Alu-associated enhancement of single nucleotide polymorphisms in the human genome. Gene. 2006;368:110-116.

haematologica | 2021; 106(1)

ARTICLE

Chronic Myeloid Leukemia

The quiescent fraction of chronic myeloid leukemic stem cells depends on BMPR1B, Stat3 and BMP4-niche signals to persist in patients in remission Sandrine Jeanpierre,1,2,3,4,5* Kawtar Arizkane,1,2,3,4* Supat Thongjuea,6 Elodie Grockowiak,1,2,3,4 Kevin Geistlich,1,2,3,4 Lea Barral,1,2,3,4 Thibault Voeltzel,1,2,3,4 Anissa Guillemin,7 Sandrine Gonin-Giraud,7 Olivier Gandrillon,7 Franck-Emmanuel Nicolini,1,2,3,4,5 Adam J. Mead,8 Véronique Maguer-Satta1,2,3,4# and Sylvain Lefort1,2,3,4#

CNRS UMR5286, Centre de Recherche en Cancérologie de Lyon, 69000 Lyon, France; Inserm U1052, Centre de Recherche en Cancérologie de Lyon, 69000 Lyon, France; 3 Université de Lyon, 69000, Lyon, France; 4Department of Signaling of Tumor Escape, Lyon, France; 5Centre Léon Bérard, 69000 Lyon, France; 6MRC WIMM Centre for Computational Biology, Weatherall Institute of Molecular Medicine, NIHR Oxford Biomedical Research Centre, John Radcliffe Hospital, Oxford, UK; 7Laboratoire de Biologie et Modélisation de la Cellule, LBMC - Ecole Normale Supérieure - Lyon, Université Claude Bernard Lyon - Centre National de la Recherche Scientifique: UMR5239 - Institut National de la Santé et de la Recherche Médicale: U1210 - Ecole Normale Supérieure de Lyon, 69007 Lyon, France and 8Haemopoietic Stem Cell Biology Laboratory, Weatherall Institute of Molecular Medicine, University of Oxford, Oxford, UK 1

Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):111-122

*SJ and KA contributed equally as co-first author.

VM-S and SL contibuted equaly as co-senior authors.

ABSTRACT

hronic myeloid leukemia arises from the transformation of hematopoietic stem cells by the BCR-ABL oncogene. Though transformed cells are predominantly BCR-ABL-dependent and sensitive to tyrosine kinase inhibitor treatment, some BMPR1B+ leukemic stem cells are treatment-insensitive and rely, among others, on the bone morphogenetic protein (BMP) pathway for their survival via a BMP4 autocrine loop. Here, we further studied the involvement of BMP signaling in favoring residual leukemic stem cell persistence in the BM of patients having achieved remission under treatment. We demonstrate by single-cell RNASequencing analysis that a sub-fraction of surviving BMPR1B+ leukemic stem cells are co-enriched in BMP signaling, quiescence and stem cell signatures, without modulation of the canonical BMP target genes, but enrichment in actors of the Jak2/Stat3 signaling pathway. Indeed, based on a new model of persisting CD34+CD38– leukemic stem cells, we show that BMPR1B+ cells display co-activated Smad1/5/8 and Stat3 pathways. Interestingly, we reveal that only the BMPR1B+ cells adhering to stromal cells display a quiescent status. Surprisingly, this quiescence is induced by treatment, while non-adherent BMPR1B+ cells treated with tyrosine kinase inhibitors continued to proliferate. The subsequent targeting of BMPR1B and Jak2 pathways decreased quiescent leukemic stem cells by promoting their cell cycle re-entry and differentiation. Moreover, while Jak2-inhibitors alone increased BMP4 production by mesenchymal cells, the addition of the newly described BMPR1B inhibitor (E6201) impaired BMP4-mediated production by stromal cells. Altogether, our data demonstrate that targeting both BMPR1B and Jak2/Stat3 efficiently impacts persisting and dormant leukemic stem cells hidden in their BM microenvironment.

Introduction Chronic myeloid leukemia (CML) represents a unique model of neoplasia driven via stem cell (SC) transformation, which is triggered by the BCR-ABL oncogene. Without treatment, this disease progresses to an inexorable fatal blast crisis. Apart haematologica | 2021; 106(1)

Correspondence: VÉRONIQUE MAGUER-SATTA veronique.maguer-satta@lyon.unicancer.fr SYLVAIN LEFORT sylvain.lefort@lyon.unicancer.fr Received: July 22, 2019 Accepted: January 27, 2020. Pre-published: January 30, 2020. https://doi.org/10.3324/haematol.2019.232793

111

S. Jeanpierre et al.

from allogeneic stem cell transplantation, until 2001, interferon-α (IFN-α) was extensively used for the treatment of chronic phase (CP) CML and induced stable remission in some patients.1 This was correlated with long-term survival,2 but was also responsible for numerous side effects leading to drug dose reduction and lack of response. Tyrosine kinase inhibitors (TKI), such as imatinib (IM), are targeted therapies that specifically inhibit BCR-ABL tyrosine kinase activity. TKI revolutionized the management of CML patients and currently represent the first-line treatment. However, TKI fail to eliminate leukemic stem cells (LSC), which persist in patients displaying complete cytogenetic remission (CCyR),3,4 as evidenced by the indefinite detection of residual BCR-ABL+ cells in their blood. These cells are most likely responsible for the high rate of relapse after treatment discontinuation, even in patients with a durable undetectable status.5 Indeed, LSC are highly heterogeneous, and single-cell transcriptomic analyses linked specific subpopulations to a primitive quiescent phenotype that persists throughout therapy, conferring cells upon treatment cessation with the ability to proliferate extensively, regenerate or acquire resistance.6,7 As LSC are insensitive to TKI, this strongly suggests that their survival is independent of BCR-ABL activity, relying on other factors, including interactions with the bone marrow (BM) microenvironment.8-12 Among the cytokines produced by the BM stroma, bone morphogenetic proteins (BMP) are a group of growth factors involved in many processes, including stem cell regulation. During embryogenesis, BMP4 is crucial for mesodermal cell commitment to the hematopoietic lineage.13 BMP4-deficient mice display a reduced number of hematopoietic stem cells (HSC), while the engraftment of wild-type HSC into BMP4-deficient mice impairs the repopulating activity of these cells.14 Additionally, BMP4 regulates the maintenance of human primitive cord blood progenitors.15 In adults, BMP regulate distinct characteristics of HSC and parameters of their niche.16 For example, BMP4 induces CD34+ progenitor differentiation into megakaryocytes,17 whereas BMP2 favors erythropoietic commitment.18 The BMP pathway that co-regulates the fate and proliferation of HSC and interactions with their niche has been reported in several types of leukemia,19 such as acute myloid leukemia (AML),20,21 and is dysregulated in CML.22-24 Indeed, CP-diagnosed CML patients show higher concentrations of soluble BMP2 and BMP4 in their BM compared to normal BM donors. Furthermore, the BMP receptor type-1B (BMPR1B) is over-expressed in CP-diagnosed CD34+ CML cells. Soluble BMP, secreted into the tumor niche maintain a fraction of BMPR1B+ LSC and amplify leukemic progenitors, contributing to disease progression.22 These alterations sustain a permanent pool of LSC and progenitors expressing high levels of BMPR1B receptor, which evolve upon treatment25 and progressively implement a BMP4 autocrine loop, leading to TKI-resistance.26,27 Having unraveled a majority of the interactions/mechanisms underlying the BMP-associated survival of BMPR1B+ LSC in the BM niche, we herein further investigated the spatio-temporal setting of persistence mechanisms in CML TKI-responsive patients after remission. It remains a major therapeutic challenge to achieve complete recovery through eradication of BMPR1B+ LSC. We identify BMP4/BMPR1B as a major bi-directional signal driving 112

microenvironment-dependent LSC persistence alongside the Jak2-Stat3 pathway. Our data further show that targeting both BMP and Jak2/Stat3 signaling efficiently promotes LSC cell cycle re-entry and differentiation, enabling their efficient eradication.

Methods Cells Bone marrow samples were obtained either from healthy BM donors for allogeneic transplant or from peripheral blood (PB), or from the BM of CML patients diagnosed to be in CP prior to or post TKI treatment (from patients achieving CCyR). All donors provided written informed consent in accordance with the Declaration of Helsinki. Studies were approved by local ethics committee bylaws. Definitions of response respected the last European LeukemiaNet Recommendations.28 Mesenchymal stem cells (MSC) were isolated from BM samples and cultured as described.26,29 TF1 cell lines were obtained and authenticated as described.22,30 TF1-BA, TF1-BAP cells were cultured in RPMI1640 containing 10% fetal calf serum (FCS), with or without 2 mM IM. HS5 and HS27A cell lines were obtained from ATCC and cultured in DMEM and RPMI1640 containing 10% FCS.

RNA analysis The quantitative reverse transcription (RT) polymerase chain reaction (PCR) protocol, reagents, and primers are described by Laperrousaz et al.22 Single-cell RNA sequencing and gene set enrichment analyses were performed as previously described.7 The BMP signature was designed as presented in the Online Supplementary Table S1. Briefly, we took 245 cells from groups A and B derived from 16 patients, of whom 11 and 5 were good and poor responders, respectively.7 All cells analyzed in single-cell RNA sequencing were BCR-ABL+ cells as verified by qualitative (q)PCR analysis. High-throughput microfluidic-based RT-qPCR was performed following the Fluidigm protocols (PN 68000088 K1) as described.31

Flow cytometry Extracellular staining of cells was performed using PECy5–conjugated anti-CD34, APC-conjugated anti-CD38, PE-conjugated anti-CD71 (BD Biosciences) or PE-conjugated anti-GPA (Invitrogen). For intracellular staining, cells were fixed with 4% paraformaldehyde, and washed and permeabilized with 80% ethanol before adding PE–conjugated anti-P-Smad1/5/8 (Cell Signaling) or AF647-conjugated anti-P-Stat3 (Biolegend), or PE-conjugated anti-Ki67 (BD Biosciences).

Bone morphogenetic protein quantification Bone marrow plasma or HS5/HS27A supernatants were precleared of cellular debris by rapid centrifugation and processed for BMP4 ELISA quantification (R&D Systems) as previously described.22

Western blotting Western blot was performed using TF1-BA and TF1-BAP cells treated with IM (2 mM), AG490 (25 mM) or E6201 (100 nM) during 24 hours (h), proteins were then extracted using RIPA buffer and 40 mg of proteins were loaded. Membranes were incubated with monoclonal-antibodies against P-ERK (#4370), ERK (#4695), P-STAT3 (#9145), STAT3 (#9139), Caspase 3 (#9662). P-ABL (#2865), P-CRKL (#3181), P-Smad1/5/8 (#13820), GAPDH (#8884) (Cell Signalling Technology), BCR-ABL (ThermoFisher, #MA1153) and Smad (SantaCruz Technology, #sc6031).21 haematologica | 2021; 106(1)

Targeting BMPR1B and Jak2/Stat3 awakes dormant LSC in CML

Functional assays Colony forming cell (CFC) and long-term culture-initiating cell (LTC-IC) assays were performed as reported.22

Chemical inhibitors Cells were treated with AG490 (Sigma-Aldrich) and/or E6201 (Spirita Oncology, LLC), a BMPR1B inhibitor, also shown to inhibit MEK1/2 and FLT3 without affecting BMPR1A or JAK232-34 (Online Supplementary Table S2).

Statistical analysis Mean comparisons were performed using the bilateral MannWhitney test (paired or unpaired as required) and correlations were determined with the Pearson bilateral test, using the GaphPadPrism software (San Diego, CA, USA). *P≤0.05, **P≤0.01, ***P≤0.001, were considered statistically significant.

Results At remission, persistent chronic myeloid leukemia cells are enriched in bone morphogenetic protein and quiescence signaling pathways Recent single-cell analyses revealed the heterogeneity of the SC compartment in the BM of CML patients with distinct sub-fractions of LSC that displayed different molecular signatures.7 This study demonstrated the selective persistence of a BCR-ABL SC subset (group A) already present at diagnosis and from which leukemic cells emerge at relapse upon disease progression. This LSC sub-fraction progressively accumulates at the expense of another LSC sub-fraction (group B) in patients treated with TKI. Interestingly, in contrast to group B, BCR-ABL+ cells from group A are highly quiescent. At the beginning of the study reported here, we hypothesized that BMP signaling could be involved in the maintenance of group A LSC in CML patients that achieved remission, due to the involvement of this pathway in TKI resistance. To address this and gain further insight into mechanisms underlying LSC persistence, we analyzed RNAsequencing (RNASeq) data obtained from Giustacchini et al. for BMP pathway actors. Compared to the highly proliferative group B that disappeared upon TKI treatment, cells from group A displayed an enhanced BMP signaling pathway (Figure 1A and Online Supplementary Table S1). Both persistent LSC from group A and B expressed BMPR1B transcripts at variable levels (Figure 1B, left panel). However, unlike our previous observations in resistant LSC,26 persistent LSC were devoid of autocrine BMP4 production, suggesting their dependence upon BMP4-producing cells of their niche (Figure 1B, right panel). Interestingly, neither group exhibited a significant difference in the comparative level of classical BMP-target genes, as illustrated in Online Supplementary Figure S1A for ID1, ID2, RUNX1 and RUNX2 gene expression. In addition, single cell analyses of BMPR1B+ cells from each group revealed that LSC of group A scarcely expressed Id1 or Runx1 (8% of group A cells) (Online Supplementary Figure S1B, left panel) unlike cells of group B that frequently (50% of group B cells) coexpressed these genes with BMPR1B (Online Supplementary Figure S1B, right panel). Taken together, our data indicate that the selective persistence of a quiescent BCR-ABL+ LSC subset (group A) upon treatment is correlated in these cells with a signifihaematologica | 2021; 106(1)

cant enrichment in the BMP signaling molecular signature, unrelated to the expression of BMP target genes.

Persisting quiescent leukemic stem cells are characterized by bone morphogenetic protein and Stat3 co-activation We then investigated whether BMPR1B+ cells could activate alternative non-canonical pathways in LSC of groups A and B to enable them to persist in the BM microenvironment despite TK treatment. Based on these single cell RNAseq data, we identified a specific correlation between BMPR1B and STAT3 pathways exclusively in group A (quiescent cells), though this was not correlated with the expression of IL6ST (gp130) (Figure 1C, left panel). Conversely, BMPR1B expression was associated with kinases like Tyk2 and MAP3K8 in the proliferative group B (Figure 1C, right panel). No correlation between BMPR1B and Stat5 signaling was identified in either group (Online Supplementary Figure S1C). We next evaluated the functional link between these two pathways. However, it was not technically possible to access viable BMPR1B+ quiescent LSC subpopulation by cell sorting BM cells from patients in remission due to their extremely low frequency and the fact that it was impossible to distinguish BCR-ABL cells from dominant non-BCR-ABL stem cells. Therefore, owing to the scarcity of the sorted BMPR1B+-LSC sub-fraction from the BM of CML patients in remission, we developed a novel CML model of TKI-persistent immature cells. We used an established TF1-BCR-ABL (TF1-BA) cell line22 selected to persist upon chronic imatinib-IM (2 mM of TKI IM) exposure, designated as TF1-BAP (TF1-BCR-ABL persistent) (Figure 2A). Persistence of TF1-BAP cells was accompanied by a slightly higher level of BMPR1B as observed by immunfluorescence, western blot and qPCR (Figure 2B and C and Online Supplementary Figure S2A). Similarly to RNAseq data from primary BM cells of patients in remission (Figure 1B), TF1-BAP cells do not implement an autocrine BMP4 production, but inversely seemed to repress its production (Online Supplementary Figure S2B), corroborating our previous report in CD34+ CML cells from patients that achieved CCyR.26 The TF1 human cell line displayed immature properties similar to human hematopoietic stem and progenitor cells, including the immature CD34+CD38– (stem) and CD34+CD38+ (progenitor) phenotypes, and their ability to generate colony-forming progenitor cells (CFC) in a semi-solid medium (see Figure 2A and Online Supplementary Figure S3A) from both erythroid and myeloid lineages.35 Chronic treatment of TF1-BA cells with TKI led to a severe reduction in the CD34+CD38+ sub-fraction in ensuing TF1-BAP cells, which were thus characterized by a CD34+CD38– phenotype (Figure 2D). We then evaluated the CFC output (Figure 2E and Online Supplementary Figure S3A) or LTC-IC potential (Figure 2F and Online Supplementary Figure S3B) following TKI persistence. We observed a drastic decrease in the ability of TF1-BAP cells to form CFC compared to TF1-BA (Figure 2E), suggesting a loss of progenitor activity. Moreover, TF1-BAP cells display increased LTC-IC activity (Figure 2F) according to a higher CD34+CD38– phenotype (Figure 2D). However, the W5-CFC colonies, representative of BCR-ABL TF1 immature cells derived LTC-IC, were difficult to observe, as we had frequently experienced with primary CML cells. We also detected a higher expression 113

S. Jeanpierre et al.

level of stemness-associated transcription factors (NANOG, MYC and FoxO1 and FoxO3a)36-38 as well as an increase in genes involved in LSC survival (CD123, TPOR, HIF2a, ALOX5, TWIST-1, BCL2 and BCL-XL)8,36,3941 in TF1-BAP compared to TF1-BA cells (Online Supplementary Figure S2A). Taken together, these data confirmed the leukemic stem-like status of TF1-BAP cells. We also show that, despite increased BCR-ABL protein levels, TF1-BAP display similar activation of P-CRKL (Figure 2G), and TKI equally impairs P-BCR-ABL in TF1-BA and TF1-BAP cells (while having no effect on P-Smad1/5/8) (Online Supplementary Figure S4A), indicating that there is no major involvement of the BCR-ABL kinase activity in the persistent phenotype. Therefore, this set of data further sustains that previously described in primary CML-LSC.42 Lastly, we used intracellular flow

cytometry to measure the level of activation of BMP and Stat3 signaling, and revealed an increase in P-Smad1/5/8 and double positive P-Smad1/5/8-P-Stat3 levels in TF1BAP cells compared to IM-sensitive TF1-BA cells (Figure 2H). However, no difference was observed in total PStat3 levels between the TF1-BA and TF1-BAP, as evidenced from western blot and flow cytometry analyses (Figure 2H and Online Supplementary Figure S4B), suggesting that in TKI persistent cells a specific co-activation of Smad/Stat3 occurs. Altogether these data indicate that the TF1-BAP cell line is a representative model of the primary CML persistent LSC group A characterized by a BMPR1B+ expression, stem-like status, the absence of autonomous BMP4 production, and the co-activation of P-Smad1/5/8-P-Stat3 signals.

Figure 1. BMPR1B-positive chronic myeloid leukemia (CML) persistent cells signal through the Stat3 pathway. (A) Gene set enrichment analysis (GSEA) of group A versus group B BCR-ABL+ stem cells (SC) at remission (n=122 and n=123, respectively). Gene set shown is bone morphogenetic proteins (BMP) signaling pathway. (B) Beeswarm plot of BMPR1B (left) or BMP4 (right) expression between group A (light blue; n=122) and group B BCR-ABL+ SC from patients at remission (purple; n=123). Number of cells analyzed and displaying amplification for the selected genes are shown below the plot. Average gene expression level is indicated by red squares; boxes represent the median and quartiles of gene expression levels. (C) Heat map shows hierarchical clustering of IL6_JAK_STAT3 signaling pathway genes in BMPR1B+ cells (n=8) from group A BCR-ABL+ SC (left) or group B BCR-ABL+ SC (right). FDR: false discovery rate.

114

haematologica | 2021; 106(1)

Targeting BMPR1B and Jak2/Stat3 awakes dormant LSC in CML

BMP4/BMPR1B contribute to maintaining leukemic cells in a non-differentiated state by regulating Stat3 signaling Next, we evaluated the importance of BMP and Stat3 signaling in protecting BCR-ABL+ cells from TKI-induced cell death. We used the synthetic small molecule inhibitor E6201, which has a strong affinity for BMPR1B (Online Supplementary Table S2). Having shown the relevance of the TF1-BAP persistent CML cell model, with increased BMPR1B expression and P-Smad1/5/8-P-Stat3 activity, we also evaluated the impact of a Jak2 inhibitor (AG490) used

alone or in combination with E6201. We confirmed in TF1-BAP that E6201 inhibited P-Smad1/5/8 (Figure 3A), while, as expected, AG490 decreased P-Stat3 (Online Supplementary Figure S5A). Neither E6201 nor AG490 affected the kinase activity of BCR-ABL as revealed by the absence of a significant effect on the main BCR-ABL and ERK targets, P-Ckrl and P-ERK, respectively (Online Supplementary Figure S5B). In addition, no significant change in apoptosis was observed by monitoring caspase3 cleavage (Online Supplementary Figure S5C). Interestingly, E6201 alone also significantly decreased P-Stat3 levels

Figure 2. The TF1-BAP model displays similar features as chronic myeloid leukemia (CML) persistent cells. (A) Strategy for generating TF1-BA and TF1-BAP cell lines. (B) Representative images of BMPR1B staining of TF1-BA and TF1-BAP cells. (C) Western blot analysis showing BMPR1B levels in TF1-BA and TF1-BAP cells; Scatter plot showing BMPR1B (GAPDH used as an internal control) n=5. (D) Representative FACS plots of TF1-BA and TF1-BAP cells analyzed for their CD34 and CD38 content. (E) Dot plot showing colony forming cell (CFC) activity from TF1-BA and TF1-BAP cells. (F) Dot plot showing long-term culture-initiating cell (LTC-IC) activity from TF1-BA and TF1-BAP cells. (G) Western blot analysis showing BCR-Abl and P-CRKL levels in TF1-BA and TF1-BAP cells. Bar graph showing BCR-Abl and P-CRKL (GAPDH used as an internal control); n=4. (H) TF1-BA and TF1-BAP cells were analyzed for their content in P-Smad1/5/8 and P-Stat3. Dot plots represent percentage of PSmad1/5/8 (left panel), P-Stat3 (middle panel) or double positive P-Smad1/5/8-P-Stat3 (right panel). (E-H) Unpaired t-test significant: *Pâ&#x2030;¤0.05, **Pâ&#x2030;¤0.01. ns: not significant.

haematologica | 2021; 106(1)

115

S. Jeanpierre et al. A

Figure 3. Dual targeting of BMPR1B and Stat3 impairs proliferation and promotes differentiation of chronic myeloid leukemia (CML) persistent cells. (A and B) TF1-BAP cells treated with E6201 (100 nM) for 24 hours (h) and analyzed for their content of P-Smad1/5/8 (A) or P-Stat3 (B). (C) TF1-BAP cells treated or not (NT) with AG490 (25 mM) or E6201 (100 nM) for 24 h and analyzed for their content of double positive P-Smad1/5/8-P-Stat3. (D-F) TF1-BAP cells were treated for 3 days with imatinib (IM) (2 mM), and/or AG490 (25 mM), and/or E6201 (100 nM), and then (D) cell viability was determined by counting viable cells (Trypan blue exclusion), (E) colony forming cell (CFC) activity was assessed, or (F) cells were analyzed for their content of CD38 (left panel), CD71 (middle panel) or GPA (right panel). Unpaired (C-E) or paired (A and B, F) t-test: *P≤0.05, **P≤0.01, ***P≤0.001. NT: not treated.

(Figure 3B). Moreover, unlike AG490 (that had no impact on the level of P-Smad1/5/8-P-Stat3 signal detected by flow cytometry), E6201 alone or in combination with AG490 decreased the cell subset presenting co-activation of P-Smad1/5/8 and P-Stat3 (Figure 3C and Online Supplementary Figure S5D). At the functional level, we observed that either AG490 or E6201 or their combination were sufficient to decrease the proliferation of persistent TF1-BAP cells (Figure 3D). The effect of both inhibitors was not modulated by the presence or absence of TKI, which, as expected, by itself had no effect on TF1-BAP cell proliferation (Figure 3D). Treatment of TF1-BAP with AG490 or E6201, alone or in combination, significantly increased the progenitor outcome as evidenced by CFC assays (Figure 3E). We then used flow cytometry to measure the appearance of cell surface markers indicative of early progenitor expansion (CD38 and CD71) and differentiation (GPA) into a nonmyeloid lineage to evaluate whether this increase in CFC and proliferation arrest could be due to the engagement of the pluripotent immature leukemic cell towards a more differentiated stage. Results indicate that inhibition of the JAK2/Stat3 and BMP signaling induced TF1-BAP cells to differentiate into a more committed progenitor status, as indicated by the increased number of CD38+ (3.26-fold) or CD71+ cells (1.72-fold) and a trend to increase GPA cell membrane expression (1.42-fold) (Figure 3F). Collectively, these results suggest that simultaneous 116

inhibition of BMP and JAK2 signaling could induce a decrease in persistent LSC through the induction of their differentiation towards progenitors.

Targeting of BMP4/BMPR1B impairs chronic myeloid leukemia persistent cells mediated by their niche As we have previously shown that BMP2 and BMP4 secreted by the niche are drivers of LSC survival and expansion at diagnosis and following the development of resistance,22,26 we analyzed whether patients having achieved CCyR displayed normal levels of these molecules following treatment. The level of both BMP2 and BMP4 was significantly reduced in CCyR patients compared to values obtained at diagnosis (Online Supplementary Figure S6A and B). However, a higher (2.7fold) concentration of BMP2 and BMP4 persists in the BM plasma of patients in remission compared to healthy donors (Figure 4A), indicating that, despite eradication of a majority of leukemic cells, soluble BMP remain abnormally elevated in the BM niche. Our former findings also revealed that BMP4, which contributes to LSC survival under TKI treatment, is provided by the leukemic niche.22,26 Consequently, we evaluated whether combinations of BMP and/or Jak2 inhibitors (E6201 and AG490, respectively) impacted its production by BM mesenchymal stem cells (MSC). Surprisingly, AG490 alone increased BMP4 expression in both CML (Figure 4B) and healthy MSC (Online Supplementary Figure S6C), while TKI (IM) haematologica | 2021; 106(1)

Targeting BMPR1B and Jak2/Stat3 awakes dormant LSC in CML

Figure 4. Chronic myeloid leukemia (CML) niche promotes tyrosine kinase inhibitor (TKI) persistence of leukemic cells. (A) ELISA quantification of BMP2 and BMP4 in bone marrow (BM) plasma obtained from healthy donors and complete cytogenetic remission (CcyR) patients. Results from individual samples are expressed in picograms per milliliter, and horizontal lines represent mean values±standard error of mean (SEM) of the indicated number of analyzed samples. (B) Mesenchymal stem cells from CML patients were treated with imatinib (IM), and/or AG490, and/or E6201 for 3 days. BMP4-transcript expression from three samples is expressed in arbitrary units, and lines represent mean values±SEM, compared to untreated (NT). (C) TF1-BA and TF1-BAP cells treated or not with BMP4 (20 ng/mL for 24 hours) were analyzed for their content in P-Smad1/5/8 (left panel), or double positive P-Smad1/5/8-P-STAT3 (right panel). (D) TF1-BA cells were cultured in suspension or seeded on top of a HS27A layer and then treated for 6 days with IM (1.5 μM) before viable cell number was analyzed. Scatter plot showing fold to input (day 6/day 0), with dashed line representing no proliferation compared to input; n=7. (E) Experimental protocol for analysis of TKI persistence of leukemic cells promoted by stroma. TF1-BA cells were seeded on top of a HS27A layer for 3 or 6 days with imatinib (IM) (1.5 mM), before adherent TF1-BA were harvested and further divided into suspended cells without IM or seeded on top of a HS27A layer with IM. (F) After 3 days, cell numbers were analyzed (left and middle) or colony forming cell (CFC) assays were performed (right). Scatter plots showing fold to input: Day 6 / Day 3 (left), Day 9 / Day 6 (middle), or CFC activity (right). Mann-Whitney test: *P≤0.05, **P≤0.01.

haematologica | 2021; 106(1)

117

S. Jeanpierre et al.

Figure 5. Targeting of chronic myeloid leukemia (CML) niche impacts the quiescence of persisting cells. (A-C) TF1-BAP cells were seeded on top of a HS27A layer and then treated for 3 days with imatinib (IM) (2 mM) (A), with or without AG490 (25 mM) + E6201 (100 nM) (B and C), and addition of anti-BMP4 (C). Then TF1-BAP (GFP tagged) cells adherent and non-adherent fractions were harvested and analyzed separately for their content of Ki67-PE. Unpaired t-test: *P≤0.05, **P≤0.01, ***P≤0.001. NT: not treated.

118

haematologica | 2021; 106(1)

Targeting BMPR1B and Jak2/Stat3 awakes dormant LSC in CML

Figure 6. Proposed model to illustrate the regulation of quiescence stage of persisting chronic myeloid leukemia (CML) leukemic stem cells by their microenvironment. (Left) Leukemic stem cell dormancy of cells that interact with stromal cells of the niche is partly driven by BMP4 binding to its BMPR1B receptor over-expressed in those cells. This leads to phosphorylation of the Smad signaling (P-Smad 1/5/8) canonical pathway. It co-operates with the activation of Jak2/Stat3 signal through P-Stat3 induced directly by BMPR1B and/or by other cytokines binding to their receptors repressing proliferation while inducting quiescence genes expressions. (Right) Using a chemical inhibitor toward BMPR1B (E6201) in combination with a Jak2 inhibitor (AG490), targeting LSC could induce their proliferation and repress their quiescence. In addition, E6021 is preventing BMP4-induced production by stromal cells upon AG490 exposure, allowing a simultaneous dual targeting of the LSC and their leukemic protecting microenvironment.

and E6021 had no effect individually. BMP4 production induced by AG490 treatment was prevented by adding E6201 (Figure 4B). In order to evaluate whether soluble BMP contribute to LSC persistence, we measured the impact of BMP4 treatment on Smad and Stat3 signaling. TF1-BA and TF1-BAP cells treated with exogenous BMP4 further increased their P-Smad1/5/8 and double positive P-Smad1/5/8-P-Stat3 levels (Figure 4C), while this treatment had no effect on cells displaying only P-Stat3 activation (Online Supplementary Figure S6D). To evaluate the functional importance of BMP4 production by MSC, we then performed co-culture experiments of TF1-BA or TF1-BAP and HS27A cells (a model of BMP4producing MSC).22 We first observed that, despite the fact that TF1-BA cells in suspension were sensitive to IM, the addition of HS27A cells increased their persistence following 6 days of TKI treatment (Figure 4D). Drug withdrawal led TF1-BA cells to re-establishing a high level of proliferation after 3 (Figure 4E, left panel) or 6 (Figure 4E, right panel) days of co-culture with HS27A in the presence of TKI. In addition, release from the niche leads to clonogenic expansion in favor of TKI-treatment arrest (Figure 4F). In order to evaluate the effect of BMP4 provided by the stroma on long-term persistent LSC, we separately analyzed adherent and non-adherent TF1-BAP cell fractions following 3 days of HS27A co-culture (Figure 5A). Flow cytometry analysis provided Ki67 expression (proliferation marker) of hematopoietic cells exclusively by gating GFP-expressing TF1-BAP cells. Figure 5A clearly shows a haematologica | 2021; 106(1)

decrease in Ki67+ cells in adherent TKI-treated TF1-BAP cells, while the non-adherent fraction expressed similar Ki67 levels as untreated cells (normalized to 1). This effect on proliferation appeared to be dependent on the activation of BMP and Jak2 signaling, as it was reversed in the presence of both AG490 and E6201 (Figure 5B). The importance of BMP4 to mediate TKI-induced quiescence was confirmed by the absence of any significant effect when TF1-BAP cells were co-cultured with HS5 MSC (isolated from the BM of the same patient as HS27A) that express lower levels of BMP4 (Online Supplementary Figure S7A), on which TKI did not induce a decrease in Ki67+ cells (Online Supplementary Figure S7B). Lastly, the addition of anti-BMP4 blocking antibody further demonstrates that the decrease in Ki67 following stroma co-culture under TKI treatment is impaired by inhibiting the BMP4 signal (Figure 5C), and can potentiate the effect of AG490 and E6201. Overall, these data suggest that BMP4 controls Stat3 activation through Smad1/5/8 in persistent LSC to favor the maintenance of low cycling cells close to BMP4expressing stromal cells.

Discussion In CML, the oncogene BCR-ABL alters numerous signaling pathways that control HSC, such as the BMP pathway that sustains the survival of BMPR1B+ LSC and pro119

S. Jeanpierre et al. genitors.22 Under continuous pressure of the BCR-ABLspecific TKI-treatment, LSC evolved and established a BMP4 autocrine loop, which contributes to treatment escape.26 Furthermore, a sub-fraction of LSC survives in both TKI-sensitive or -resistant CML patients.7 Here, we revealed that, in the BM of CML patients in remission, residual LSC that displayed a quiescent molecular signature are also enriched in BMP signaling. Moreover, we show that a sub-fraction of quiescent cells presents a correlation between BMPR1B expression and the Stat3 pathway, but not with Stat5. Both BMPR1B26 and Stat343 have been implicated in CML LSC resistance/persistence to TKI treatment independently of the BCR-ABL oncogene.26,44 Targeting Jak2 and/or Stat3 alone or in combination with other pathways has been proposed to constitute an efficient way to target LSC and counteract their TKI resistance.45-47 Here, using primary cells and a novel model of CD34+CD38â&#x20AC;&#x201C; TKI persistent LSC, we demonstrate that BMPR1B expression is linked to P-Smad1/5/8 and P-Stat3 activation, and that concomitant targeting of these pathways impairs cell proliferation in suspension. This raised the hypothesis that the BMP pathway is responsible for the Stat3 activation in a small subset of persisting LSC. Indeed, we showed that BMP4 controls the concomitant activation of Smad1/5/8 and Stat3 in TKI persistent CD34+ cells, identifying the soluble factor produced by stromal cells and involved in the JAK/STAT3 activation as suggested by Kuepper et al.43 This is in accordance with results obtained in embryonic SC, showing that BMP controls Stat3 expression to sustain self-renewal and prevents cell differentiation through the induction of Id gene expression.48 At a molecular level, it is interesting to note that transcription factor binding site analysis revealed that Smad1 and Stat3 often co-occupy similar loci49 and that Jak2 signaling regulates CML LSC.8,45 Moreover, a direct link between the BMP and Jak/Stat pathway has been established in the neural system through the identification of a molecular complex between Smad1 and Stat3.50 We have previously demonstrated that BMP4 regulates human normal megakaryopoiesis, which is prevented in the presence of the Jak2 inhibitor, AG490, but not of MEK1/2 inhibitors.17 In addition, we have shown that TPO-induced megakaryopoiesis was partly mediated by the induction of a BMP4/BMPR1 autocrine loop.17 In the context of stem cell dormancy, TPO has been identified in an in vivo mouse model as a master cytokine controlling HSC quiescence in the BM niche through the Jak2 pathway.51 Moreover, BMP4 was shown to induce SC quiescence in different tissues such as the eyes,52 neural system,53-55 and skin.56 In keratinocytes, BMP4 induces SC dormancy by controlling DNA hypomethylation.56 In the neural system, BMP4 decreases cell proliferation through P27Kip1 regulation53 and prevents neural SC differentiation.55 Altogether, these observations suggest that BMP4 could directly control CML LSC quiescence through its modulation of the JAK/Stat3 pathway in a non-canonical signal, dependent upon BMPR1B kinase activity. We now provide evidence that residual BM BMPR1B+ LSC include (at least) two distinct subsets of leukemic persisting cells. In vitro data obtained with the TF1-BAP model of persisting immature BCR-ABL+ cells indicate that cell adhesion to stromal BMP4-producing MSC results in a drastically decreased proliferation in the presence of TKI, mediated by BMPR1B and Jak2 signaling. Conversely, non-adherent 120

BMPR1B+ TF1-BAP cells do not display a decrease in Ki67 upon TKI treatment. These data suggest that LSC could be maintained into a quiescent stage through a synergistic mechanism between the BMP signaling and adhesion to BM stromal cells. Indeed, in the drosophila model, it has been demonstrated that integrin signaling enhances BMP pathway activation through interactions with the BMP receptors independently of the FAK-signaling.57 In addition, the authors showed that integrin themselves are a target of BMP signaling. Moreover, extracellular matrix molecules such as integrin promote Stat3-dependent mitochondrial functions and cell survival through integrin-FAK signaling.58 Murine BMP4 regulates the expression of integrin alpha4 in HSC through a Smad-independent non-canonical signal.16 BMP4 induces HSC adhesion to stromal cells resulting in an increase in stem cell maintenance and expansion.17 Taken together, these data suggest that integrin could contribute to maintaining and amplifying the BMP4 signal to promote adherent LSC persistence. Since LSC of group A appeared to constitute a reservoir from which leukemic resistant or progressing cells could re-emerge,7 we hypothesized that these characteristics may enable the targeting of those persistent and quiescent LSC in their niche. Indeed, we then demonstrated that the combination of AG490-E6201 is optimal to prevent BMP4-induced MSC protection, reinforcing the importance of combining BMPR1B inhibition with Jak2 targeting. Taken together, our results raise the hypothesis that close proximity of a subset of LSC to BMP4-producing MSC of the leukemic niche, increases the binding rate of BMP4 to its receptor BMPR1B to promote cell cycle exit in co-operation with integrin signaling and amplified by TKI exposure (Figure 6). This subpopulation of dormant LSC then re-enter the cell cycle upon TKI treatment cessation to slowly expand, progress,7 and mediate patient relapse after several months, as observed in 60% of the cases reported in the multicenter Stop Imatinib (STIM) clinical trial of treatment arrest.59 In conclusion, we identified BMP4/BMPR1B as a major bi-directional signal driving microenvironment-dependent LSC persistence alongside the Jak2-Stat3 pathway. Our data may prompt new therapeutic strategies to efficiently target LSC and their microenvironment, and prevent disease persistence during the treatment of CML and other leukemic malignancies. Disclosures No conflicts of interest to disclose. Contributions SJ, KA, ST, EG, KG and SL performed experiments and data analysis; TV developed the LSC persistent model; F-EN provided CML patients and donor samples and related clinical follow-up; AJM and ST designed protocols and data analysis for single cell RNA-Seq experiments; AG, SG-G and OG designed protocols and data analysis for Fluidigm RNA quantification; SL and VM-S designed the study, analyzed data and wrote the manuscript. All authors approved the final manuscript. Acknowledgments We thank P. Battiston (Cytometry Facility, CRCL, Lyon) for her technical assistance, and B. Manship for English editing. We thank Dr T. Myers from Spirita Oncology for providing us with E6201. haematologica | 2021; 106(1)

Targeting BMPR1B and Jak2/Stat3 awakes dormant LSC in CML

Funding Funding is acknowledged from LMC France patients’ association and its president Mrs Mina Daban, “Fondation de France” 2014-0047501 and 2017-00076282/Fondation Ramona Ehrman Amador, “Association Laurette Fugain”

References 1. Kantarjian HM, O'Brien S, Cortes JE, et al. Complete cytogenetic and molecular responses to interferon-alpha-based therapy for chronic myeloid leukemia are associated with excellent long-term prognosis. Cancer. 2003;97(4):1033-1041. 2. Simonsson B, Gedde-Dahl T, Markevarn B, et al. Combination of pegylated IFNalpha2b with imatinib increases molecular response rates in patients with low- or intermediate-risk chronic myeloid leukemia. Blood. 2011;118(12):3228-3235. 3. Chu S, McDonald T, Lin A, et al. Persistence of leukemia stem cells in chronic myeloid leukemia patients in prolonged remission with imatinib treatment. Blood. 2011; 118(20):5565-5572. 4. Chomel J-C, Bonnet M-L, Sorel N, et al. Leukemic stem cell persistence in chronic myeloid leukemia patients with sustained undetectable molecular residual disease. Blood. 2011;118(13):3657-3660. 5. Rousselot P, Charbonnier A, Cony-Makhoul P, et al. Loss of major molecular response as a trigger for restarting tyrosine kinase inhibitor therapy in patients with chronicphase chronic myeloid leukemia who have stopped imatinib after durable undetectable disease. J Clin Oncol. 2014;32(5):424-430. 6. Warfvinge R, Geironson L, Sommarin MNE, et al. Single-cell molecular analysis defines therapy response and immunophenotype of stem cell subpopulations in CML. Blood. 2017;129(17):2384-2394. 7. Giustacchini A, Thongjuea S, Barkas N, et al. Single-cell transcriptomics uncovers distinct molecular signatures of stem cells in chronic myeloid leukemia. Nat Med. 2017; 23(6):692-702. 8. Zhang B, Li L, Ho Y, et al. Heterogeneity of leukemia-initiating capacity of chronic myeloid leukemia stem cells. J Clin Invest. 2016;126(3):975-991. 9. Graham SM, Jorgensen HG, Allan E, et al. Primitive, quiescent, Philadelphia-positive stem cells from patients with chronic myeloid leukemia are insensitive to STI571 in vitro. Blood. 2002;99(1):319-325. 10. Hamilton A, Helgason GV, Schemionek M, et al. Chronic myeloid leukemia stem cells are not dependent on Bcr-Abl kinase activity for their survival. Blood. 2012;119(6):15011510. 11. Houshmand M, Simonetti G, Circosta P, et al. Chronic myeloid leukemia stem cells. Leukemia. 2019;33(7):1543-1556. 12. Agarwal P, Isringhausen S, Li H, et al. Mesenchymal niche-specific expression of Cxcl12 controls quiescence of treatmentresistant leukemia stem cells. Cell Stem Cell. 2019;24(5):769-784.e6. 13. Hogan BL. Bone morphogenetic proteins: multifunctional regulators of vertebrate development. Genes Dev. 1996;10(13):15801594. 14. Goldman DC, Bailey AS, Pfaffle DL, Masri AA, Christian JL, Fleming WH. BMP4 regulates the hematopoietic stem cell niche.

haematologica | 2021; 106(1)

ALF2014-03, Ligue contre le Cancer (Haute Savoie, Loire, Puy de Dôme and Rhone), “Association ALTE-SMP”, PLASCAN grants to VM-S, SL, and/or F-EN. PhD fellowship was obtained from “Ligue contre le cancer” (KA).

Blood. 2009;114(20):4393-4401. 15. Hutton JF, Rozenkov V, Khor FSL, D'Andrea RJ, Lewis ID. Bone morphogenetic protein 4 contributes to the maintenance of primitive cord blood hematopoietic progenitors in an ex vivo stroma-noncontact co-culture system. Stem Cells Dev. 2006;15(6):805-813. 16. Khurana S, Buckley S, Schouteden S, et al. A novel role of BMP4 in adult hematopoietic stem and progenitor cell homing via Smad independent regulation of integrin-alpha4 expression. Blood. 2013;121(5):781-790. 17. Jeanpierre S, Nicolini FE, Kaniewski B, et al. BMP4 regulation of human megakaryocytic differentiation is involved in thrombopoietin signaling. Blood. 2008;112(8):31543163. 18. Maguer-Satta V, Bartholin L, Jeanpierre S, et al. Regulation of human erythropoiesis by activin A, BMP2, and BMP4, members of the TGFbeta family. Exp Cell Res. 2003;282(2): 110-120. 19. Zylbersztejn F, Flores-Violante M, Voeltzel T, Nicolini FE, Lefort S, Maguer-Satta V. The BMP pathway: a unique tool to decode the origin and progression of leukemia. Exp Hematol. 2018;61:36-44. 20. Raymond A, Liu B, Liang H, et al. A role for BMP-induced homeobox gene MIXL1 in acute myeloid leukemia and identification of type I BMP receptor as a potential target for therapy. Oncotarget. 2014;5(24):1267512693. 21. Voeltzel T, Flores-Violante M, Zylbersztejn F, et al. A new signaling cascade linking BMP4, BMPR1A, DeltaNp73 and NANOG impacts on stem-like human cell properties and patient outcome. Cell Death Dis. 2018;9(10):1011. 22. Laperrousaz B, Jeanpierre S, Sagorny K, et al. Primitive CML cell expansion relies on abnormal levels of BMPs provided by the niche and on BMPRIb overexpression. Blood. 2013;122(23):3767-3777. 23. Toofan P, Irvine D, Hopcroft L, Copland M, Wheadon H. The role of the bone morphogenetic proteins in leukaemic stem cell persistence. Biochem Soc Trans. 2014; 42(4):809-815. 24. Gerber JM, Gucwa JL, Esopi D, et al. Genome-wide comparison of the transcriptomes of highly enriched normal and chronic myeloid leukemia stem and progenitor cell populations. Oncotarget. 2013;4(5):715728. 25. Toofan P, Busch C, Morrison H, et al. Chronic myeloid leukaemia cells require the bone morphogenic protein pathway for cell cycle progression and self-renewal. Cell Death Dis. 2018;9(9):927. 26. Grockowiak E, Laperrousaz B, Jeanpierre S, et al. Immature CML cells implement a BMP autocrine loop to escape TKI treatment. Blood. 2017;130(26):2860-2871. 27. Cosset E, Hamdan G, Jeanpierre S, et al. Deregulation of TWIST-1 in the CD34+ compartment represents a novel prognostic factor in chronic myeloid leukemia. Blood. 2011;117(5):1673-1676. 28. Baccarani M, Deininger MW, Rosti G, et al. European LeukemiaNet recommendations

for the management of chronic myeloid leukemia: 2013. Blood. 2013;122(6):872-884. 29. Schallmoser K, Rohde E, Reinisch A, et al. Rapid large-scale expansion of functional mesenchymal stem cells from unmanipulated bone marrow without animal serum. Tissue Eng Part C Methods. 2008;14(3):185196. 30. Sheng Y, Ji Z, Zhao H, et al. Downregulation of the histone methyltransferase SETD2 promotes imatinib resistance in chronic myeloid leukaemia cells. Cell Prolif. 2019; 52(4):e12611. 31. Richard A, Boullu L, Herbach U, et al. Singlecell-based analysis highlights a surge in cellto-cell molecular variability preceding irreversible commitment in a differentiation process. PLoS Biol. 2016;14(12):e1002585. 32. Byron SA, Loch DC, Wellens CL, et al. Sensitivity to the MEK inhibitor E6201 in melanoma cells is associated with mutant BRAF and wildtype PTEN status. Mol Cancer. 2012;11:75. 33. Zhang W, Borthakur G, Gao C, et al. The dual MEK/FLT3 inhibitor E6201 exerts cytotoxic activity against acute myeloid leukemia cells harboring resistance-conferring FLT3 mutations. Cancer Res. 2016;76(6):1528-1537. 34. Babiker HM, Byron SA, Hendricks WPD, et al. E6201, an intravenous MEK1 inhibitor, achieves an exceptional response in BRAF V600E-mutated metastatic malignant melanoma with brain metastases. Invest New Drugs. 2019;37(4):636-645. 35. Kitamura T, Tange T, Terasawa T, et al. Establishment and characterization of a unique human cell line that proliferates dependently on GM-CSF, IL-3, or erythropoietin. J Cell Physiol. 1989;140(2):323-334. 36. Abraham SA, Hopcroft LE, Carrick E, et al. Dual targeting of p53 and c-MYC selectively eliminates leukaemic stem cells. Nature. 2016;534(7607):341-346. 37. Das B, Pal B, Bhuyan R, et al. MYC regulates the HIF2alpha stemness pathway via Nanog and Sox2 to maintain self-renewal in cancer stem cells versus non-stem cancer cells. Cancer Res. 2019;79(16):4015-4025. 38. Pellicano F, Scott MT, Helgason GV, et al. The antiproliferative activity of kinase inhibitors in chronic myeloid leukemia cells is mediated by FOXO transcription factors. Stem Cells. 2014;32(9):2324-2337. 39. Chen Y, Hu Y, Zhang H, Peng C, Li S. Loss of the Alox5 gene impairs leukemia stem cells and prevents chronic myeloid leukemia. Nat Genet. 2009;41(7):783-792. 40. Nievergall E, Ramshaw HS, Yong AS, et al. Monoclonal antibody targeting of IL-3 receptor alpha with CSL362 effectively depletes CML progenitor and stem cells. Blood. 2014;123(8):1218-1228. 41. Carter BZ, Mak PY, Mu H, et al. Combined targeting of BCL-2 and BCR-ABL tyrosine kinase eradicates chronic myeloid leukemia stem cells. Sci Transl Med. 2016; 8(355):355ra117. 42. Hamilton A, Helgason GV, Schemionek M, et al. Chronic myeloid leukemia stem cells are not dependent on Bcr-Abl kinase activity

121

S. Jeanpierre et al. for their survival. Blood. 2012;119(6):15011510. 43. Kuepper MK, Butow M, Herrmann O, et al. Stem cell persistence in CML is mediated by extrinsically activated JAK1-STAT3 signaling. Leukemia. 2019;33(8):1964-1977. 44. Eiring AM, Kraft IL, Page BD, O'Hare T, Gunning PT, Deininger MW. STAT3 as a mediator of BCR-ABL1-independent resistance in chronic myeloid leukemia. Leuk Suppl. 2014;3(Suppl 1):S5-S6. 45. Gallipoli P, Cook A, Rhodes S, et al. JAK2/STAT5 inhibition by nilotinib with ruxolitinib contributes to the elimination of CML CD34+ cells in vitro and in vivo. Blood. 2014;124(9):1492-1501. 46. Gleixner KV, Schneeweiss M, Eisenwort G, et al. Combined targeting of STAT3 and STAT5: a novel approach to overcome drug resistance in chronic myeloid leukemia. Haematologica. 2017;102(9):1519-1529. 47. Chakraborty SN, Leng X, Perazzona B, Sun X, Lin YH, Arlinghaus RB. Combination of JAK2 and HSP90 inhibitors: an effective therapeutic option in drug-resistant chronic myeloid leukemia. Genes Cancer. 2016;7(56):201-208. 48. Ying QL, Nichols J, Chambers I, Smith A. BMP induction of Id proteins suppresses dif-

122

ferentiation and sustains embryonic stem cell self-renewal in collaboration with STAT3. Cell. 2003;115(3):281-292. 49. Chen X, Xu H, Yuan P, et al. Integration of external signaling pathways with the core transcriptional network in embryonic stem cells. Cell. 2008;133(6):1106-1117. 50. Nakashima K, Yanagisawa M, Arakawa H, et al. Synergistic signaling in fetal brain by STAT3-Smad1 complex bridged by p300. Science. 1999;284(5413):479-482. 51. Bersenev A, Wu C, Balcerek J, Tong W. Lnk controls mouse hematopoietic stem cell selfrenewal and quiescence through direct interactions with JAK2. J Clin Invest. 2008; 118(8):2832-2844. 52. Balenci L, Wonders C, Coles BL, Clarke L, van der Kooy D. Bone morphogenetic proteins and secreted frizzled related protein 2 maintain the quiescence of adult mammalian retinal stem cells. Stem Cells. 2013;31(10):2218-2230. 53. Andreu Z, Khan MA, Gonzalez-Gomez P, et al. The cyclin-dependent kinase inhibitor p27 kip1 regulates radial stem cell quiescence and neurogenesis in the adult hippocampus. Stem Cells. 2015;33(1):219-229. 54. Han X, Yu L, Chen Q, et al. BMP4/LIF or RA/Forskolin suppresses the proliferation of

neural stem cells derived from adult monkey brain. Stem Cells Int. 2017;2017:7012405. 55. Bond AM, Peng CY, Meyers EA, McGuire T, Ewaleifoh O, Kessler JA. BMP signaling regulates the tempo of adult hippocampal progenitor maturation at multiple stages of the lineage. Stem Cells. 2014;32(8):2201-2214. 56. Lee J, Kang S, Lilja KC, et al. Signalling couples hair follicle stem cell quiescence with reduced histone H3 K4/K9/K27me3 for proper tissue homeostasis. Nat Commun. 2016;7:11278. 57. Sawala A, Scarcia M, Sutcliffe C, Wilcockson SG, Ashe HL. Peak BMP responses in the drosophila embryo are dependent on the activation of integrin signaling. Cell Rep. 2015;12(10):1584-1593. 58. Visavadiya NP, Keasey MP, Razskazovskiy V, et al. Integrin-FAK signaling rapidly and potently promotes mitochondrial function through STAT3. Cell Commun Signal. 2016; 14(1):32. 59. Mahon FX, Rea D, Guilhot J, et al. Discontinuation of imatinib in patients with chronic myeloid leukaemia who have maintained complete molecular remission for at least 2 years: the prospective, multicentre Stop Imatinib (STIM) trial. Lancet Oncol. 2010;11(11):1029-1035.

haematologica | 2021; 106(1)

ARTICLE

Coagulation & its Disorders

Inhibitor incidence in an unselected cohort of previously untreated patients with severe hemophilia B: a PedNet study Christoph Male,1 Nadine G Andersson,2,3 Anne Rafowicz,4 Ri Liesner,5 Karin Kurnik,6 Kathelijn Fischer,7 Helen Platokouki,8 Elena Santagostino,9 Hervé Chambost,10 Beatrice Nolan,11 Christoph Königs,12 Gili Kenet,13 Rolf Ljung2 and H. Marijke van den Berg14 on behalf of the PedNet study group*

Department of Paediatrics, Medical University of Vienna, Vienna, Austria; 2Department of Clinical Sciences, Division of Pediatrics, Lund University, Lund, Sweden; 3Center for Thrombosis and Hemostasis, Skåne University Hospital, Malmö, Sweden; 4Centre de Référence pour le Traitement des Maladies Hémorragiques (CRTH), Hôpital Bicêtre, Kremlin Bicêtre, Paris, France; 5Hemophilia Center, Department of Hematology, Great Ormond Street Hospital for Children, London, UK; 6Dr. V. Haunersches Kinderspital, University of Munich, Munich, Germany; 7Van Creveld Kliniek, University Medical Center Utrecht, Utrecht, the Netherlands; 8Hemophilia-Hemostasis Unit, St. Sophia Children’s Hospital, Athens, Greece; 9Fondazione, IRCCS Ca’ Granda, Ospedale Maggiore Policlinico, Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, Milan, Italy; 10 APHM, La Timone Children’s Hospital, Center for Bleeding Disorders, and Aix Marseille University, INSERM, INRA, C2VN, Marseille, France; 11Department of Paediatric Hematology, Children’s Health Ireland at Crumlin, Dublin, Ireland; 12J .W. Goethe University Hospital, Department of Pediatrics, Frankfurt, Germany; 13National Hemophilia Center, Ministry of Health, Sheba Medical Center, Tel Hashomer, Israel and 14 PedNet Hemophilia Research Foundation, Baarn, the Netherlands 1

Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):123-129

*Members and centers of the PedNet study group are listed in the Appendix

ABSTRACT

he incidence of factor IX (FIX) inhibitors in severe hemophilia B (SHB) is not well defined. Frequencies of 3-5% have been reported but most studies to date have been small, including patients with different severities, and without prospective follow up for inhibitor incidence. The study objective was to investigate the inhibitor incidence in patients with SHB followed up for to 500 exposure days (ED), the frequency of allergic reactions, and the relationship with genotypes. Consecutive previously untreated patients (PUP) with SHB enrolled into the PedNet cohort were included. Detailed data was collected for the first 50 ED, followed by the annual collection of the inhibitor status and allergic re-actions. The presence of inhibitors was defined by at least two consecutive positive samples. Additionally, data on FIX gene mutation was collected. One hundred and fifty-four PUP with SHB were included; 75% were followed up until 75 ED, and 43% until 500 ED. Inhibitors developed in 14 patients (seven high-titer). The median number of ED at inhibitor manifestation was 11 (interquartile range [IQR]: 6.5-36.5). The cumulative inhibitor incidence was 9.3% (95% Confidence Interval [CI]: 4.4-14.1) at 75 ED, and 10.2% (95% CI: 5.1-15.3) at 500 ED. Allergic reactions occurred in four (28.6%) inhibitor patients. Missense mutations were most frequent (46.8%) overall but not associated with inhibitors. Nonsense mutations and deletions with large structural changes comprised all mutations among inhibitor patients and were associated with an inhibitor risk of 26.9% and 33.3%, respectively. In an unselected, well-defined cohort of PUP with SHB, the cumulative inhibitor incidence was 10.2% at 500 ED. Nonsense mutations and large deletions were strongly associated with the risk of inhibitor development. The ‘PedNet Registry’ is registered at clinicaltrials.gov; identifier: NCT02979119. haematologica | 2021; 106(1)

Correspondence: CHRISTOPH MALE christoph.male@meduniwien.ac.at Received: October 11, 2019. Accepted: January 9, 2020. Pre-published: January 9, 2020. https://doi.org/10.3324/haematol.2019.239160

123

C. Male et al.

Introduction Hemophilia A (HA) and B (HB) are X-linked inherited bleeding disorders, characterised by deficiency of factor VIII (FVIII) and factor IX (FIX), respectively. HB occurs at a frequency of about 1 in every 25-30.000 newborn males, comprising about 15% of all patients with hemophilia.1 Patients with severe HB (SHB) have a FIX level <0.01 IU/mL and account for approximately 30-40% of persons with HB.2,3 Both hemophilia subtypes suffer from recurrent joint bleeds, soft-tissue bleeds, and muscle bleeds. Several studies in adult patients reported a more severe phenotype for HA.4,5 The treatment history has a large effect on the joint outcome of hemophilia, even though the reason for these different outcomes in adults is still unclear. A study in a large cohort of children did not detect differences in the age of the first joint bleed.6 Longer follow-up of patients with SHA and SHB on primary prophylaxis will provide the answers whether clear phenotypic differences exist. A clear difference between hemophilia subtypes has been observed in inhibitor frequency, with studies reporting incidences of 1-5% in HB2,7-11 compared to 2535% in HA.12,13 Moreover, in HB, FIX inhibitors occur almost exclusively in the severe form,11 while FVIII inhibitors occur in both SHA and non-severe HA.14 The inhibitor incidence in HB is not well defined. Due to small numbers of patients with HB, sufficiently large cohorts of previously untreated patients (PUP) are difficult to obtain. Thus, no representative prospective studies reporting the inhibitor incidence are available to date.15 Many studies reporting on the inhibitor frequency in HB included all severities and even the definition of SHB varied between <0.01 and <0.02 IU/mL. The period of highest risk for inhibitor development is believed to be during the first 50-75 exposure days (ED) which can take many years during on demand therapy. Studies on HB published before widespread use of prophylaxis might have missed late inhibitors. Gene defects such as large deletions and nonsense mutations have been reported to increase the risk of inhibitors in HB.7,9,16-19 The high proportion of missense mutations in HB resulting in circulating amounts of FIX antigen is considered as one of the factors responsible for the overall lower risk of inhibitors.7,9,17,19 The lower proportion of the severe phenotype of HB compared to HA may be another reason.11 Moreover, the similarity between FIX and the other vitamin K-dependent factors (FII, FVII, FX) has been hypothesized to render therapeutic FIX less immunogenic.20 Although the development of an inhibitor in HB is a rare event, it is associated with significant morbidity, related not only to a higher risk of hemorrhagic complications but also to the frequent occurrence of allergic reactions.16,21-23 How often allergic reactions occur in association with inhibitor manifestation is unclear. Nephrotic syndrome often complicates immune tolerance induction treatment in HB and might be one reason for the low success rates of 30-35%.24 International registries on HB reported on the risk factors for inhibitors such as genotypes but the selection of patients in such registries makes it difficult to draw definitive conclusions on their impact.9,16-18 The objective of our study was to investigate inhibitor development in SHB patients followed up for up to 500 124

ED, the frequency of allergic reactions at inhibitor manifestation, and the effect of FIX genotype on the inhibitor risk.

Methods Patients The PedNet study is a birth cohort enrolling all PUP with HA and HB (FVIII or FIX less than 25%) born after January 2000 treated at participating centers. For the current study, we included consecutive PUP with SHB (FIX activity <0.01 IU/mL) followed until January 1, 2018, if informed consent was available, data quality was sufficient, and they had been exposed to FIX concentrate. A list of the participating centers that contributed to this study can be found in the Appendix. Patients who were referred to the participating centers because of the presence of an inhibitor were excluded. Patients were enrolled and followed according to the PedNet study protocol, NCT02979119. Study approval was obtained from the institutional review boards of every center. Written informed consent was obtained from the parents or guardians of all participants.

Measurements Baseline information on diagnosis of HB, including age at diagnosis, basal FIX plasma level, FIX gene mutation, ethnicity, family history of hemophilia, family history of inhibitors were collected in uniform web-based case report forms. During the first 50 ED, detailed information on each ED, dose, product type and reason for treatment was collected. After 50 ED, treatment data was collected in annual follow-up forms per individual patient. This included all changes in the treatment regimens, bleeds and surgeries, and inhibitor development, and other adverse events. An ED was defined as a calendar day on which one or more infusions of FIX were given.

Genetic analysis Characterization of mutations in the FIX gene was performed locally at the participating centers and the reports on genotype were centrally evaluated and adapted to the Human Genome Variation Society nomenclature.25 In order to assess the association of the FIX genotype with inhibitor development, mutations were categorized according to the mutation type (point mutation, deletion, duplication, insertion, polymorphism, complex) and the mutation effect (missense, nonsense, frameshift, large structural changes [>50 bp], small structural changes [<50 bp] in frame, silent, splice site mutation, promotor abnormalities).

Outcomes The primary outcome of the study was development of a clinically relevant inhibitor, defined as at least two positive inhibitor titers combined with a decreased in vivo FIX recovery. Inhibitor positivity was defined according to the cut-off level in the individual laboratory of each center, the highest cut-off level used being 0.6 Bethesda Units per mL (BU/mL). Secondary outcome was development of a high-titer inhibitor, defined as the occurrence of an inhibitor with a peak titer of more than 5 BU/mL. The number of ED at inhibitor development was defined as the last ED before the date of the first positive inhibitor test. After a single positive inhibitor test, all subsequent tests and recovery measurements were collected. For this study, additional information was collected regarding the occurrence of allergic reactions at the time of inhibitor development. Follow-up data on the clinical course of patients who developed an inhibitor has been collected and will be reported in a separate manuscript. haematologica | 2021; 106(1)

Inhibitor incidence in severe hemophilia B

Data analyses We used Kaplan-Meier curve survival analysis methods with the number of ED as the time variable. All patients were included in the analyses and censored at the ED of their last follow-up. In the analyses using “all clinically relevant inhibitors” as the outcome, censoring occurred at the last ED. In the analyses with “high titer inhibitor development” as the outcome, censoring occurred at the last exposure day in non-inhibitor patients and at the last ED before inhibitor development in patients with low-titer inhibitors. Continuous variables are summarized by median and interquartile range (IQR). Inhibitor incidences are reported as percentage with 95% Confidence Interval (95% CI). Comparisons of categorial variables between groups were done by Fisher’s exact test.

Results Of a total of 168 consecutive PUP with SHB identified in PedNet centers during the study period, 14 were excluded for lack of consent (n=2), insufficient data quality (n=5), loss-to-follow-up (n=1) or because they were not yet treated with a FIX product (n=7). Thus, 154 (91.6%) patients were included in this analysis. The demographic information of study patients is reported in Table 1. The median age at first treatment was 0.8 years, the age at the time of study evaluation was 9.6 years. About half of the patients had a positive family history for hemophilia at diagnosis. Caucasian ethnicity was present in more than 80% of the patients. Seventy-seven percent of patients were followed up until 50 ED, 75% until 75 ED, 68% until 150 ED, and 43% until 500 ED. Inhibitors were diagnosed in 14 patients, seven were classified as high-titer and 7 as low-titer. The median number of ED at inhibitor manifestation was 11 (IQR: 6.536.5), the median age was 23.2 months (IQR: 12.1-37.1). The cumulative inhibitor incidence at 75 ED was 9.3% (95 % CI: 4.4-14.1) for all inhibitors and 5% for high-titer inhibitors. Between 76 and 500 ED, only 1 low-titer inhibitor was diagnosed at 121 ED. The cumulative inhibitor incidence at 500 ED was 10.2% (95% CI: 5.115.3) (Table 2 and Figure 1). A positive family history of FIX inhibitors was significantly more frequent (21%) among inhibitor patients compared to non-inhibitor patients (2%) (Table 1). A recombinant FIX product was used initially in 75% and 71% of non-inhibitor and inhibitor patients, respectively. Peak treatment episodes of at least 5 days at initial FIX exposure occurred in 15% and 14%, respectively. The median age at the start of prophylaxis was 1.5 years for non-inhibitor patients and 1.3 years for inhibitor patients, but only five inhibitor patients had started prophylaxis before developing the inhibitor. Surgery performed before ED20 (and before inhibitor development) occurred in 34% and 29% of non-inhibitor and inhibitor patients, respectively. The FIX gene mutation genotypes were known from 124 of 154 (80.5%) patients (Table 3). Overall, the most frequent mutation type were point mutations in 95 of 124 (76.6%) patients, leading to missense mutations in 58 of 124 (46.8%) and nonsense mutations in 26 of 124 (21.0%) patients. Deletions were found in 24 of 124 (19.4%) patients, causing large structural changes in 15 of 124 (12.1%) and frameshift in 6 of 124 (4.8%) patients. Duplications were found in 2.4% and insertions in 1.6% of patients. For more details on the mutation type and effect see Table 3. haematologica | 2021; 106(1)

Table 1. Demographic information of study participants comparing patients.

Non-inhibitors Inhibitor patients patients number (%) or number (%) or median [IQR] median [IQR] Number of patients 140 Ethnicity: Caucasian 117 (84%) Positive family history of hemophilia 66 (47%) Positive family history of FIX inhibitor 3 (2%)+ Age at diagnosis (months) 3.3 [0–10.5] Age at first ED (years) 0.8 [0.4-1.2] Recombinant FIX product at 105 (75%)++ initial exposure Peak treatment at initial exposure* 21 (15%) Age at start of prophylaxis** (years) 1.5 [1.0-2.1] ED until start of prophylaxis 8.0 [5-16] Surgery before ED20 48 (34%) Age at time of study (years) 9.7 [5.6–13.6]

14 13 (93%) 8 (57%) 3 (21%)+ 3.1 [0–8.9] 0.9 [0.8-1.5] 10 (71%) 2 (14%) 1.3 [0.9-3.1]¶ 7.5 [2-18]¶ 4 (29%)¶¶ 8.8 [6.5–12.3]

+ P=0.008 (Fisher’s exact test); ++FIX product type unknown in six (4.3%) patients; *peak treatment of at least 5 consecutive exposure days (ED); **definition of prophylaxis: at least once a week, for 2 consecutive months; ¶only five inhibitor patients had started prophylaxis before inhibitor development; ¶¶surgery before ED20 and before inhibitor development. IQR: interquartile range; FIX: factor IX.

Among inhibitor patients, the most frequent mutations were nonsense mutations in 7 of 12 (58.3%), four patients with low-titer and three patients with high-titer inhibitors, and deletions with large structural changes in 5 of 12 (41.7%), three patients with low-risk and two with high-risk inhibitors. No other mutations were present among inhibitor patients. In two inhibitor patients, the genotype was not known (no genetic report was available in one; no mutation was identified in spite of repeated analysis in the other). Figure 2 displays the risk of inhibitor development by mutation. The inhibitor risk for deletions with large structural change was 33.3% (95% CI: 11.861.6) and for nonsense mutations 26.9% (95% CI: 11.647.8). For all other mutations, the inhibitor risk was zero. Allergic reactions at the time of inhibitor development were observed in 4 of 14 (28.6%, 95% CI: 8.4-58.1) patients, one patient with a high-titer inhibitor and three patients with low titer inhibitors (Table 3). Of these four patients, two patients had nonsense mutations, one had a large deletion and one had an unknown mutation.

Discussion In this unselected, prospectively followed birth cohort of 154 PUP with SHB, 14 patients developed an inhibitor. The cumulative inhibitor incidence at 75 ED was 9.3% for all inhibitors and 5% for high-titer inhibitors. At 500 ED, the cumulative inhibitor incidence for all inhibitors had increased to 10.2%. Previous large databases of HB patients in general reported inhibitor frequencies of 1.32.8% 2,8-10 while studies focusing on SHB patients reported inhibitor frequencies of 3.8-4.9%.2,7,8,11 However, these registries reported inhibitor prevalences rather than prospectively following patients for inhibitor incidence. In the European Hemophilia Safety Surveillance (EUHASS), five inhibitors were reported among 72 PUP with SHB 125

C. Male et al. Table 2. Inhibitor development in relation to number of exposure days.

Exposure days 0 1-10 11-20 21-30 31-40 41-50 51-60 61-75 76-150 151-250 251-500

Patients at risk 154 137 131 127 122 107 107 103 91 84 52

All inhibitors 0 6 10 10 11 12 12 13 14 14 14

High-titer inhibitors

Cumulative incidence (all inhibitors)

0 4 6 6 6 6 6 7 7 7 7

estimate

95% CI

0 4.1 6.9 6.9 7.6 8.4 8.4 9.3 10.2 10.2 10.2

0.9 - 7.3 2.8 - 11.0 2.8 - 11.0 3.3 – 11.9 3.8 – 13.0 3.8 – 13.0 4.4 – 14.1 5.1 - 15.3 5.1 - 15.3 5.1 - 15.3

all inhibitors

high titer inhibitors low titer inhibitors

Figure 1. Kaplan-Meier survival curve of inhibitor development until 500 exposure days

(6.9%).26 A recent systematic review of PUP studies in SHB reported a summary inhibitor incidence of 10.2%.27 However, most studies included in this analysis were small (patient numbers between 7 and 72), with inhibitor frequencies from individual studies varying between 5-14%. How can we interpret the higher incidence of inhibitors in SHB found in our study compared to most previous reports? The PedNet registry represents an unselected birth cohort. In contrast, patients with bleedings in the neonatal period will not be eligible for licensing PUP studies.28 In the current study on SHB patients, we found that 52% of patients had a negative family history, similar to prospective studies in SHA patients.12,13 This frequency is much higher than the 30% of patients reported in the literature. Patients with a negative family history are usually diagnosed after the onset of bleeding, potentially 126

increasing their risk for inhibitors.12 In our cohort of patients born after 2000, all received primary prophylaxis, which enabled us to follow them until 500 ED. This long-term follow-up might also have resulted in detecting a higher total number of inhibitors compared to other studies.27 Another reason for the higher incidence in our study may be attributed to more frequent testing for inhibitors. Inhibitors have received much more attention in the last decades and frequent testing for inhibitors has become standard of care. In SHA, increased frequencies of inhibitors have been observed since the nineties.12,13 Our group reported that significantly more low-titer inhibitors were diagnosed in SHA after the year 2000.29 In the current study of SHB, low-titer inhibitors comprised about half of all inhibitors, and the incidence of high-titer inhibitors was 5%, which is more in line with previous reports.2,7,10,11 haematologica | 2021; 106(1)

Inhibitor incidence in severe hemophilia B

Figure 2. Risk of inhibitor development by mutation

Table 3. Factor IX mutations in study participants comparing patients without and with inhibitors.

Mutation Type

Mutation effect

Mutation known 124 of 154 (80.5%) Point mutation 95 of 124 (76.6%)

Missense Nonsense Splice site Promoter Missing data on effect Deletion 24 of 124 (19.3%) Large structural change* Small structural change** Frameshift Splice site Duplication 3 of 124 (2.4%) Frameshift Small structural change** Insertion 2 of 124 (1.6%) Frameshift Mutation unknown 30 of 154 (19.5%)

Non-inhibitor patients n=140 n (%)

Inhibitor patients n=140 n (%)

Risk of inhibitor by mutation n/N, % (95% CI)

112 (80.0) 58 (51.8) 19 (17.0) 4 (3.6) 5 (4.5) 1 (0.9) 10 (8.9) 2 (1.8) 6 (5.4) 1 (0.9) 2 (1.8) 1 (0.9) 3 (2.7) 28 (25.0)

12 (85.7) 7 (58.3) 5 (41.7) 2 (14.3)

0% 7 of 26 (26.9%; 11.6-47.8) 0% 0% 0% 5 of 15 (33.3%; 11.8-61.6) 0% 0% 0% 0% 0% 0%

* Large structural change: >50bp; **small structural change: <50bp, in frame.

A positive family history for inhibitors has been recognized as an important risk factor, in our patients this was confirmed in 21% of the inhibitor patients. Higher incidences of inhibitors for SHB where reported in Sweden and Ireland that might have been influenced by the inclusion of families with more high-risk mutations or a founder effect.19,30 No inhibitor patients in our study were related which excludes this so-called â&#x20AC;&#x153;founderâ&#x20AC;? effect. The relative frequency of FIX gene mutation types/effects in our cohort was comparable to those reported from other registries.9,17 We found 46.8% missense mutations in SHB patients, in contrast to only 11.4% missense mutations among SHA patients in the RODIN study.12 A higher prevalence of missense mutations in SHB compared to SHA has previously been haematologica | 2021; 106(1)

reported.7,9,12,17,18,31 Missense mutations were only found in non-inhibitor patients, consistent with previous reports.7,9,16-19 We hypothesize that the high proportion of missense mutations in HB is a key reason for the overall lower inhibitor incidence compared to HA. Two mutation types/effects were strongly associated with inhibitor development, comprising all mutations among the inhibitor patients: nonsense mutations and deletions with large structural changes both representing null-mutations. Patients with these high-risk mutations had a 27% and 33% risk of developing an inhibitor, respectively. The findings from our representative PUP cohort confirm previous reports,7,9,16-19 suggesting a strong association between absent endogenous FIX protein due to gross and complete gene deletions and inhibitor development.32 127

C. Male et al.

For gross FIX gene defects resulting in absent endogenous FIX protein, the observed risk of inhibitor development (27-33%) is similar to HA due to gross FVIII gene mutations resulting in absent endogenous FVIII protein (inhibitor risk around 35%).33 Treatment-related factors, such as the type of FIX product, peak treatment episodes at the initial exposure to FIX concentrate, time of the start of prophylaxis, or surgery, were not associated with inhibitor development. However, we must be cautious in interpreting these findings, as the power to identify or rule out determinants of inhibitor development is limited with the small number of inhibitor patients. Allergic reactions are more frequently observed in HB compared to HA. Frequencies of allergic reactions reported in the literature for HB vary considerably between 4-60% of inhibitor patients.10,16,21-23 In our cohort, allergic reactions at the time of inhibitor manifestation were observed in four (29%) patients, one with a high-titer and three with low-titer inhibitors. Of these patients, two patients had nonsense mutations, one had a large deletion, and one had an unknown mutation. Large deletions have previously been reported to be associated with allergic reactions.16,22 Data on the course and effect of immune tolerance induction in our cohort are currently collected and will be reported in a future manuscript. Our study has some limitations, such as the lack of racial diversity which limits generalisability to some extent. The determination of hemophilia severity as well as inhibitor testing was done in local laboratories at the individual centers. A potential disadvantage is insufficient standardization, however, the advantage of this pragmatic approach is better feasibility resulting in a more representative cohort. Although we report the largest consecutive cohort of PUP with SHB to date, its size still does not allow to comprehensively evaluate whether there is an influence of treatment-related risk factors on inhibitor development. In conclusion, our study in an unselected cohort of PUP with SHB found a cumulative inhibitor incidence of 10.2% at 500 ED. Missense mutations were the most frequent mutation type but these were not associated with inhibitors, while nonsense mutations and large deletions were significantly associated with an increased risk of inhibitor development. Disclosures No Conflicts of interest to disclose. Contributions CM, AR, RL, KK, KF, HP, ES, HC, BN, CK, GK were responsible for the care for HB patients with inhibitors; CM, NA, RL, MvdB analyzed the data and drafted the manuscript, all authors contributed to writing the manuscript and have approved its final version. Acknowledgments We thank all members of the PedNet study group who are listed in the Appendix for providing the data on their HB patients. We thank Ella van Hardeveld and Marloes de Kovel for maintaining the data base and supporting the analysis. Appendix PedNet study group members and centers in Europe: A Santamaria Ortiz, Unitat Hemofilia, Hospital Vall 128

d’Hebron, Barcelona, Spain; MT Alvarèz Román, Unidad de Coagulopatías, Hopital Universitario La Paz, Madrid, Spain, M Bührlen, Gesundheit Nord, Klinikum Bremen Mitte, Prof.-HessKinderklinik, Bremen, Germany; HM van den Berg, PedNet Hemophilia Research Foundation, Baarn, the Netherlands; M Carvalho, S. Imuno-hemoterapia, Centro Hospitalar São João, Porto, Portugal; E Chalmers, Department of Hematology, Royal Hospital for Sick Children, Yorkhill, Glasgow, UK; H Chambost, Aix Marseille Univ, INSERM, INRA, C2VN, and APHM, La Timone Children Hospital, Center for Bleeding Disorders, Marseille, France; A Rosa Cid, Unidad de Hemostasia y Trombosis, Hospital Universitario y Politécnico La Fe, Valencia, Spain; S Claeyssens, Centre Regional d’Hemophilie, Centre Hospitalo Universitaire, Toulouse, France; T Stamm Mikkelsen, Department of Pediatrics, University Hospital of Aarhus at Skejby, Aarhus, Denmark; C Escuriola, HZRM Hämophilie Zentrum Rhein Main GmbH, Mörfelden-Walldorf, Germany; K Fischer, Van Creveld Kliniek, University Medical Center Utrecht, Utrecht, the Netherlands; C Van Geet, Catholic University of Leuven, Campus Gasthuisberg, Service of Pediatric Hematology, Leuven, Belgium; H Glosli, Oslo University Hospital HF, Oslo, Norway; N N Gretenkort Andersson, Department of Clinical Sciences, Division of Pediatrics, Lund University, Lund, and Center for Thrombosis and Haemostasis, Skåne University Hospital, Malmö, Sweden; R Kobelt, Hämophiliezentrum, Wabern and Children's Hospital of the University of Bern, Bern,Switzerland; C Königs, J .W. Goethe University Hospital, Department of Pediatrics, Frankfurt, Germany; K Kurnik, Dr. V. Haunersches Kinderspital, University of Munich, Munich, Germany; R Liesner, Hemophilia Center, Department of Hematology, Great Ormond Street Hospital for Children, London, UK; R Ljung, Department of Clinical Sciences, Division of Pediatrics, Lund University, Lund, Sweden; A Mäkipernaa, Children’s Hospital, University Central Hospital and University of Helsinki, Helsinki, Finland; C Male, Department of Pediatrics, Medical University of Vienna, Vienna, Austria; A Molinari, Dipartimento di Ematologia ed Oncologia, Unità Trombosi ed Emostasi, Ospedale Pediatrico Giannina Gaslini, Genova, Italy; W Muntean*, Universitäts-Klinik für Kinder- und Jugendheilkunde, Graz, Austria; B Nolan, Department of Pediatric Hematology, Children’s Health Ireland at Crumlin, Dublin, Ireland; J Oldenburg, Institut für Experimentelle Hämatologie und Transfusionsmedizin, Universitätsklinikum Bonn, Bonn, Germany; R Pérez Garrido*, Hospital General Unidad de Hemofilia, Hospitales Universitarios Virgen del Rocio, Sevilla, Spain; H Platokouki, Hemophilia-Hemostasis Unit, St. Sophia Children’s Hospital, Athens, Greece; A Rafowicz, Centre de Référence pour le Traitement des Maladies Hémorragiques (CRTH), Hôpital Bicêtre, Kremlin Bicêtre, Paris, France; S Ranta, Department of Pediatrics, Clinic of Coagulation Disorders, Karolinska Hospital, Stockholm,Sweden; E Santagostino, ME Mancuso, Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, Fondazione, IRCCS Ca’ Granda, Ospedale Maggiore Policlinico, Milan, Italy; A Thomas*, Royal Hospital for Sick Children, Edinburgh, UK and J Motwani, Department of Hematology, The Children’s Hospital, Birmingham, UK. PedNet Study group members and centers in Israel. GK, National Hemophilia Center, Ministry of Health, Sheba Medical Center, Tel Hashomer, Israel. PedNet Study group members and centers in Canada. MC, Division of Hematology/Oncology, Hospital for Sick Children, Toronto, Ontario and GR, Division of Hematology/Oncology, Hôpital St Justine, Montréal, Quebec, Canada. *No longer participating as a PedNet center. haematologica | 2021; 106(1)

Inhibitor incidence in severe hemophilia B

References 1. Peyvandi F, Garagiola I, Young G. The past and future of haemophilia: diagnosis, treatments, and its complications. Lancet. 2016; 388(10040):187-197. 2. Katz J. Prevalence of factor IX inhibitors among patients with haemophilia B: results of a large-scale North American survey. Haemophilia. 1996;2(1):28-31. 3. Goodeve AC. Hemophilia B: molecular pathogenesis and mutation analysis. J Thromb Haemost. 2015;13(7):1184-1195. 4. Schulman S, Eelde A, Holmstrom M, Stahlberg G, Odeberg J, Blomback M. Validation of a composite score for clinical severity of hemophilia. J Thromb Haemost. 2008;6(7):1113-1121. 5. Tagariello G, Iorio A, Santagostino E, et al. Comparison of the rates of joint arthroplasty in patients with severe factor VIII and IX deficiency: an index of different clinical severity of the 2 coagulation disorders. Blood. 2009;114(4):779-784. 6. Clausen N, Petrini P, Claeyssens-Donadel S, Gouw SC, Liesner R. Similar bleeding phenotype in young children with haemophilia A or B: a cohort study. Haemophilia. 2014; 20(6):747-755. 7. Belvini D, Salviato R, Radossi P, et al. Molecular genotyping of the Italian cohort of patients with hemophilia B. Haematologica. 2005;90(5):635-642. 8. Miller CH, Benson J, Ellingsen D, et al. F8 and F9 mutations in US haemophilia patients: correlation with history of inhibitor and race/ethnicity. Haemophilia. 2012;18(3):375-382. 9. Rallapalli PM, Kemball-Cook G, Tuddenham EG, Gomez K, Perkins SJ. An interactive mutation database for human coagulation factor IX provides novel insights into the phenotypes and genetics of hemophilia B. J Thromb Haemost. 2013; 11(7):1329-1340. 10. Castaman G, Bonetti E, Messina M, et al. Inhibitors in haemophilia B: the Italian experience. Haemophilia. 2013;19(5):686-690. 11. Puetz J, Soucie JM, Kempton CL, Monahan PE. Prevalent inhibitors in haemophilia B subjects enrolled in the Universal Data Collection database. Haemophilia. 2014;

haematologica | 2021; 106(1)

20(1):25-31. 12. Gouw SC, van den Berg HM, Fischer K, et al. Intensity of factor VIII treatment and inhibitor development in children with severe hemophilia A: the RODIN study. Blood. 2013;121(20):4046-4055. 13. Calvez T, Chambost H, d'Oiron R, et al. Analyses of the FranceCoag cohort support differences in immunogenicity among one plasma-derived and two recombinant factor VIII brands in boys with severe hemophilia A. Haematologica. 2018;103(1):179-189. 14. Eckhardt CL, van Velzen AS, Peters M, et al. Factor VIII gene (F8) mutation and risk of inhibitor development in nonsevere hemophilia A. Blood. 2013;122(11):1954-1962. 15. Santoro C, Quintavalle G, Castaman G, et al. Inhibitors in hemophilia B. Semin Thromb Hemost. 2018;44(6):578-589. 16. Chitlur M, Warrier I, Rajpurkar M, Lusher JM. Inhibitors in factor IX deficiency a report of the ISTH-SSC international FIX inhibitor registry (1997-2006). Haemophilia. 2009; 15(5):1027-1031. 17. Li T, Miller CH, Payne AB, Craig Hooper W. The CDC hemophilia B mutation project mutation list: a new online resource. Mol Genet Genomic Med. 2013;1(4):238-245. 18. Saini S, Hamasaki-Katagiri N, Pandey GS, et al. Genetic determinants of immunogenicity to factor IX during the treatment of haemophilia B. Haemophilia. 2015; 21(2):210-218. 19. Martensson A, Letelier A, Hallden C, Ljung R. Mutation analysis of Swedish haemophilia B families - high frequency of unique mutations. Haemophilia. 2016;22(3):440445. 20. Warrier I. Factor IX antibody and immune tolerance. Vox Sang. 1999;77(Suppl 1):S7071. 21. Warrier I, Ewenstein BM, Koerper MA, et al. Factor IX inhibitors and anaphylaxis in hemophilia B. J Pediatr Hematol Oncol. 1997;19(1):23-27. 22. Thorland EC, Drost JB, Lusher JM, et al. Anaphylactic response to factor IX replacement therapy in haemophilia B patients: complete gene deletions confer the highest risk. Haemophilia. 1999;5(2):101-105. 23. Recht M, Pollmann H, Tagliaferri A, Musso R, Janco R, Neuman WR. A retrospective study to describe the incidence of moderate

to severe allergic reactions to factor IX in subjects with haemophilia B. Haemophilia. 2011;17(3):494-499. 24. Ewenstein BM, Takemoto C, Warrier I, et al. Nephrotic syndrome as a complication of immune tolerance in hemophilia B. Blood. 1997;89(3):1115-1116. 25. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405-424. 26. Fischer K, Lassila R, Peyvandi F, et al. Inhibitor development in haemophilia according to concentrate. Four-year results from the European HAemophilia Safety Surveillance (EUHASS) project. Thromb Haemost. 2015;113(5):968-975. 27. Franchini M, Santoro C, Coppola A. Inhibitor incidence in previously untreated patients with severe haemophilia B: a systematic literature review. Thromb Haemost. 2016;116(1):201-203. 28. Fischer K, Ljung R, Platokouki H, et al. Prospective observational cohort studies for studying rare diseases: the European PedNet Haemophilia Registry. Haemophilia. 2014; 20(4):e280-286. 29. van den Berg HM, Hashemi SM, Fischer K, et al. Increased inhibitor incidence in severe haemophilia A since 1990 attributable to more low titre inhibitors. Thromb Haemost. 2016;115(4):729-737. 30. Jenkins PV, Egan H, Keenan C, et al. Mutation analysis of haemophilia B in the Irish population: increased prevalence caused by founder effect. Haemophilia. 2008;14(4):717-722. 31. Schwarz J, Astermark J, Menius ED, et al. F8 haplotype and inhibitor risk: results from the Hemophilia Inhibitor Genetics Study (HIGS) Combined Cohort. Haemophilia. 2013;19(1):113-118. 32. DiMichele D. Inhibitor development in haemophilia B: an orphan disease in need of attention. Br J Haematol. 2007;138(3):305315. 33. Schwaab R, Brackmann HH, Meyer C, et al. Haemophilia A: mutation type determines risk of inhibitor formation. Thromb Haemost. 1995;74(6):1402-1406.

129

ARTICLE Ferrata Storti Foundation

Hematopoiesis

HES1 and HES4 have non-redundant roles downstream of Notch during early human T-cell development Matthias De Decker,1 Marieke Lavaert,1 Juliette Roels,1,2 Laurentijn Tilleman,3 Bart Vandekerckhove,1,4 Georges Leclercq,1,4 Filip Van Nieuwerburgh,3 Pieter Van Vlierberghe2,4 and Tom Taghon1,4

Haematologica 2021 Volume 106(1):130-141

1 Department of Diagnostic Sciences, Ghent University; 2Department of Biomolecular Medicine, Ghent University; 3Laboratory of Pharmaceutical Biotechnology, Ghent University and 4Cancer Research Institute Ghent (CRIG), Ghent University, Ghent, Belgium

ABSTRACT

Correspondence: TOM TAGHON tom.taghon@ugent.be Received: May 3, 2019. Accepted: January 2, 2020. Pre-published: January 9, 2020.

n both mouse and human, Notch1 activation is the main initial driver to induce T-cell development in hematopoietic progenitor cells. The initiation of this developmental process coincides with Notch1-dependent repression of differentiation towards other hematopoietic lineages. Although well described in mice, the role of the individual Notch1 target genes during these hematopoietic developmental choices is still unclear in human, particularly for HES4 since no orthologous gene is present in the mouse. Here, we investigated the functional capacity of the Notch1 target genes HES1 and HES4 to modulate human Notch1-dependent hematopoietic lineage decisions and their requirement during early T-cell development. We show that both genes are upregulated in a Notch-dependent manner during early T-cell development and that HES1 acts as a repressor of differentiation by maintaining a quiescent stem cell signature in CD34+ hematopoietic progenitor cells. While HES4 can also inhibit natural killer and myeloid cell development like HES1, it acts differently on the T- versus B-cell lineage choice. Surprisingly, HES4 is incapable of repressing B-cell development, the most sensitive hematopoietic lineage with respect to Notch-mediated repression. In contrast to HES1, HES4 promotes initiation of early T-cell development, but ectopic expression of HES4, or HES1 and HES4 combined, is insufficient to induce T-lineage differentiation. Importantly, knockdown of HES1 or HES4 significantly reduces human T-cell development. Overall, we show that the Notch1 target genes HES1 and HES4 have non-redundant roles during early human T-cell development which may relate to differences in mediating Notch-dependent human hematopoietic lineage decisions.

Introduction https://doi.org/10.3324/haematol.2019.226126

130

Multipotent hematopoietic stem cells (HSC) reside in the bone marrow after birth and give rise to all the different blood cell types.1 This process is orchestrated by integrating cell-intrinsic mechanisms and external stimuli which results in the activation of molecular networks that drive differentiation towards specific hematopoietic lineages.2,3 Such events are also important during the early stages of intrathymic T-cell development. Following colonization of thymus-seeding progenitors from the bone marrow, Delta-like-ligand-4 (DLL4), which is expressed by thymic epithelial cells, triggers the Notch1 receptor on these multipotent precursors. Activation of Notch1 induces T-cell differentiation through upregulation of T-lineage genes and this also prevents differentiation towards other hematopoietic lineages.4-6 In mice, the regulatory network that drives early T-cell development has been well-studied. Activation of TCF1 contributes to the induction of T-lineage genes,7 while Gata3 upregulation is important to impair B-lineage potential.8 Subsequently, residual natural killer (NK) cell potential is repressed by BCL11B that completes T-cell commitment.9 During these steps, Notch activation remains haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development

essential to guide the cells along the T-lineage differentiation pathway and for proliferation and survival. In human, however, the regulatory network that supports these processes is less clear. Due to species-specific gene expression patterns and functions, extrapolation of murine data is often not possible. One important difference between human and mouse is the need for Notch signaling during T-cell development. In both, strong Notch1 activation is necessary to induce T-cell specification at the expense of B- and myeloid lineage potential.10 In human, however, Notch signaling then needs to be reduced for efficient differentiation towards committed thymocytes,11 while in mouse high Notch activity is maintained.12 Thereafter, Notch also differentially modulates the differentiation towards T-cell receptor (TCR)-αb and TCR-γd T-cells in mouse versus human.13-15 Activation of Notch1 results in the expression of many different target genes whose individual roles in controlling hematopoietic lineage decisions are still unclear. These include some members of the HES gene family that encode bHLH transcription factors that function as homoand heterodimers and repress their target genes through recruitment of co-repressors.16 HES1 is a well-known target and is essential for the inhibition of myeloid development during early T-lineage differentiation in mouse through repression of C/EBP-α.17 It also plays a prominent role in Notch-driven T-cell acute lymphoblastic leukemia (T-ALL).18 Importantly, the human genome also encodes HES4, a HES1 paralogue that is absent in the mouse19 (Online Supplementary Figure S1), but present in Xenopus and Zebrafish. It has been documented that, depending on the species, biological setting and tissue, HES4 can have overlapping or antagonistic roles compared to HES1.20-24 Although HES4 displays clear Notch-dependent regulation in Notch-driven cancers,25 it is unclear if HES4 can mediate non-malignant Notch-driven hematopoietic lineage decisions. Given that the individual roles of the various Notch target genes are still unclear, we studied the function of HES1 and HES4 during the early Notch-dependent hematopoietic lineage decisions in human. We show that they have both different and similar impacts on the differentiation of human hematopoietic progenitor cells (HPC) towards various blood cell types when overexpressed, and that both are essential during early human T-cell development. Methods

Isolation of CD34+ hematopoietic precursor cells Umbilical cord blood (CB) was obtained with informed consent and used with approval of the Medical Ethical Commission of Ghent University Hospital (Belgium). Mononuclear cells were isolated through Lymphoprep density-gradient (ELITech) and CD34+ HPC were purified using CD34-magnetic beads (Miltenyi Biotec). Other methods are available in the Online Supplementary Material and Methods.

Results HES4, like HES1, displays Notch-dependent expression and regulation during human T-lineage differentiation HES4 was shown to be Notch dependent in leukemic T cells,25 but whether HES4 displays Notch-dependent haematologica | 2021; 106(1)

expression during non-malignant T-cell development is unclear. Gene expression analysis showed that both HES1 and HES4 are expressed during similar Notch-dependent stages4,14 of human T-cell development (Figure 1A). This Notch-dependent HES4 expression was functionally validated through exposure of CB CD34+Lin– HPC onto OP9 stromal cells that express different Notch ligands. HES1 and HES4 expression in HPC was only induced in conditions that permit T-lineage differentiation (OP9-DLL1, OP9-DLL4 and OP9-Jagged2 (JAG2)), but not in B-lineage permissive conditions (OP9-GFP and OP9-JAG1; Figure 1B).26 Similarly, short 48 hour (h) co-culture experiments onto OP9-GFP versus OP9-DLL1 stromal cells confirmed the Notch-dependent expression of both HES genes in intrathymic CD34+ cells, cells that have already experienced in vivo Notch activation (Figure 1C). Publicly available data revealed that, in contrast to HES1, HES4 is not expressed in HSC in the bone marrow (Online Supplementary Figure S2A). However, HES4 expression is detected in monocytes while lymphocytes and dendritic cells (DC) express lower levels. Protein staining confirmed the higher HES4 levels in monocytes compared to lymphocytes (Online Supplementary Figure S2B). These findings illustrate that both HES1 and HES4 are Notch-dependently induced during human T-cell development. Given that no information is available on the role of HES4 in human hematopoiesis, we studied this using viral gene perturbation approaches in human CD34+ HPC that are differentiated towards various hematopoietic lineages in wellestablished in vitro co-culture systems (Figure 1D).

HES4 lacks the potential to repress B-cell development DLL4-dependent Notch1 activation is essential to induce T-lineage differentiation at the expense of B-cell development.6,13,27 Since both HES1 and HES4 are induced following Notch1 activation, we investigated if these genes have the potential to repress B-cell development through enforced expression in CB HPC. This resulted in 50-fold higher expression levels of each gene compared to ex vivo isolated CD34+ thymocytes that reflect physiological Notch activation (Online Supplementary Figure S3A), and protein staining suggests that slightly higher levels of HES4 were obtained compared to HES1 (Online Supplementary Figure S3B). HES1- and HES4-transduced HPC were cultured on MS-5 stromal cells in B-lineage differentiation conditions and compared to HPC transduced with the constitutive active form of Notch1 (ICN1) that mimics Notch activation and represses B-cell development.13,27 Whereas control transduced HPC efficiently differentiate into CD19+HLA-DR+ B-cells, enforced ICN1 expression inhibited B-cell development (Figure 2A-B). Remarkably, while HES1 also inhibited B-lineage differentiation, continuous HES4 expression did not impact B-cell development (Figure 2A-B), and combined overexpression of HES1 and HES4 showed a dominant repression of HES1 (Online Supplementary Figure S4A-B). Consistently, CD7+CD5+ T-cell precursors were generated in cultures where ICN1 was overexpressed, but not within control transduced conditions (Figure 2C-D). In contrast, enforced expression of either HES1 or HES4 (Figure 2C-D), or HES1 and HES4 combined (Online Supplementary Figure S4C-D), was insufficient to induce the development of CD7+CD5+ T-lineage progenitors. These findings show that HES1 and HES4 have distinct capacities to mediate the Notch1131

M. De Decker et al. A

Figure 1. HES4, like HES1, is expressed Notch dependently during human T-cell differentiation. (A) Line plots show log normalized probe intensities for HES1 (black) and HES4 (red) gene expression levels during normal T-cell development. Data shows the average expression of two independent experiments from two different donors. Error bars indicate the standard error of the mean (SEM). (B) Quantitative RT-PCR of HES1 (black bars) and HES4 (red bars) in CB CD34+Linâ&#x20AC;&#x201C; precursors cultured for 3 days on OP9-GFP, OP9-DLL1, OP9-DLL4, OP9-JAG1 and OP9-JAG2. Data shows the average expression of four independent experiments, relative to the mean of GAPDH and YHWAZ mRNA levels. Error bars indicate SEM. ***P<0.001; *P<0.05 (two-way ANOVA); ns: not significant. (C) DESeq2-normalized read counts of HES1 (black bars) and HES4 (red bars) in thymic CD34+ cells cultured for 2 days on OP9-GFP and OP9-DLL1. Graphs show the average of two independent experiments and error bars indicate SEM. (D) Schematic overview illustrating the experimental workflow and the differentiation markers used to identify the different hematopoietic lineages. 2

dependent T- versus B-cell lineage decision and that both HES proteins, individually or combined, are insufficient to impose T-cell fate.

HES4, but not HES1, promotes induction of T-lineage differentiation To further study the impact of HES1 and HES4 on early T-cell development, we performed perturbation experiments in OP9-DLL4 co-cultures that support induction of T-lineage differentiation in human HPC as illustrated through the emergence of CD7+CD5+ T-cell precursors in control transduced cells (Figure 3A). We analyzed these cultures after 2 weeks to focus on the efficiency and kinetics of induction of T-lineage specification and commitment. Ectopic expression of ICN1 promoted the generation of early CD7+CD5+ and CD5+CD1a+ T-cell precursors from HPC (Figure 3A, C) and yielded higher numbers of these cells (Figure 3B, D). Enforced expression of HES1, however, resulted in a significant reduction in the frequency and absolute number of CD7+CD5+ specified and CD5+CD1a+ committed T-cells (Figure 3A-D). In contrast, high levels of HES4 improved the differentiation of HPC towards the T-cell lineage, although less efficient compared to ICN1 overexpression (Figure 3A-D). This positive role of HES4 on early T-lineage differentiation was counteracted by HES1 when both were overexpressed (Online Supplementary Figure S5A-B). Because previous work from our lab demonstrated that JAG2, DLL1 and DLL4, but not JAG1, have the potential to initiate T-cell development in human HPC,26 we investigated if enforced HES4 expression could alter these Notch ligand requirements to induce T-lineage differentiation. Whereas overexpression of ICN1 in HPC was sufficient to induce T-cell development in OP9-GFP, OP9-JAG1 and OP9-DLL4 co-cultures, HES4132

transduced cells, like the control, could only generate CD7+CD5+ T-cell progenitors on OP9-DLL4 stromal cells (Online Supplementary Figure S6A-D). Overall, these findings illustrate that, while continuous HES1 expression inhibits T-lineage specification and commitment, HES4 can enhance induction of T-cell development in the presence of Notch signaling when present at higher levels compared to endogenous HES1 expression. However, HES4 is not sufficient to alter the Notch ligand requirements to induce T-lineage differentiation in HPC.

NK-cell development is severely hampered by HES1 and HES4 In human, Notch1 activation is permissive for NK-lineage differentiation,28 and our laboratory previously illustrated that this is mediated by the Notch1 target gene DTX1.11 Here, we studied the impact of ectopic HES1 and HES4 expression on NK-cell development through co-culture of control, ICN1-, HES1- and HES4-transduced HPC on MS-5 stromal cells with cytokines supporting NK-cell differentiation. As documented previously,28 enforced ICN1 expression enhanced NK-cell development and resulted in a significant increase in CD56+ NK-cells that express CD7, a Notch-dependent marker during early T-cell development (Figure 4A, C). In contrast, both HES1 and HES4 reduced NK-cell development, and this was the case for both CD56+CD7â&#x20AC;&#x201C; and CD56+CD7+ subsets (Figure 4A-C). To investigate if HES4 overexpression can improve the generation of CD56+CD7+ NK-cells in the presence of Notch signaling, analogous to the Notch signaling requirement for HES4 to enhance differentiation towards CD7+ T-lineage precursors, we cultured control and HES4-transduced HPC on MS5-DLL4 stromal cells in the presence of NK-lineage supporting cytokines. However, also in these haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development

Figure 2. HES4 is not sufficient to inhibit B-cell development. Flow cytometric analysis (A, C) and absolute cell numbers (B, D) of control, ICN1-, HES1- and HES4transduced CD34+Lin– CB progenitors cultured on the MS-5 feeder (Notch-OFF) for 3 weeks in the presence of the B-lineage supporting cytokines IL-7 and SCF, showing the development of CD19+HLA-DR+ B-lineage cells (A-B) and CD7+CD5+ T-lineage cells (C-D). Data shows the average of six independent experiments and error bars indicate SEM. *P<0.05 (non-parametric paired Wilcoxon test); ns: not significant; ND: not detectable. Flow cytometry plots shown are representative for six independent experiments.

conditions, continuous HES4 expression impaired the development of CD56+CD7+ NK-cells (Online Supplementary Figure S7A-B). These findings illustrate that both HES1 and HES4 can repress NK-cell development.

HES1 and HES4 inhibit the development of myeloid cells Notch1-mediated repression of myeloid lineage differentiation is critical following thymic colonization of HPC4,6 and genetic experiments in mice revealed an important role for HES1 in this process.17 To investigate if HES1 and HES4 can mediate repression of human myeloid development downstream of Notch1, we cultured control, ICN1-, HES1- and HES4-transduced HPC in conditions that support myeloid differentiation. Consistent with earlier work, enforced expression of ICN1 reduced myeloid development.13,27,29 While this was mainly reflected by a decrease in the absolute cell numbers for CD15+HLA-DR– granulocytes and CD14+CD15– monocytes, for HLADR+CD4+CD14– DC both the frequency and absolute cell numbers were decreased (Figure 5A-D). In case of continuous HES1 and HES4 expression, however, we observed a reduction in both the percentage and absolute cell number of granulocytes (Figure 5A-B) and monocytes (Figure 5A, C). In contrast, while the absolute numbers of DC was decreased by HES1 overexpression, this was not the case upon ectopic expression of HES4. Furthermore, the frequencies of DC were increased in these cultures (Figure 5A, D). Thus, both HES1 and HES4 have the potential to repress monocyte and granulocyte differentiation downstream of Notch1, whereas DC development is less impaired, consistent with the intrathymic development of these antigen-presenting cells.

HES1, but not HES4, maintains a quiescent stem cell signature in CD34+ cells Enforced expression of HES1 can maintain a stem cell signature in multiple cell types.16,30,31 To compare the effect haematologica | 2021; 106(1)

of HES1 and HES4 on HSC maintenance, we tracked the frequency and number of CD34+ HPC, following ICN1, HES1 or HES4 transduction, in OP9-DLL4 T-lineage differentiation conditions. In agreement with previous studies,13,27 ectopic expression of ICN1 induced T-cell development and resulted in a significant decrease in the percentage and absolute number of CD34+ HPC (Figure 6A-C). Although conditions with elevated HES1 levels did not yield higher numbers of CD34+ cells (Figure 6C), we observed a significant and more than 2-fold increase in their frequency (Figure 6A-B). In contrast, enforced expression of HES4 did not affect the fraction of CD34+ HPC (Figure 6A-C). In addition, since earlier work illustrated that overexpression of HES1 results in the generation of quiescent HSC,18 we evaluated cellular proliferation. While HPC that express higher levels of HES1 displayed reduced proliferation, HES4-transduced cells proliferated similarly as the control (Figure 6D). These findings suggest that HES1, but not HES4, can induce maintenance of a quiescent stem cell state by repressing differentiation and proliferation of CD34+ HPC.

RNA sequencing reveals HES1- and HES4-induced molecular pathways during early human hematopoiesis To reveal specific HES1- and HES4-induced molecular changes, we performed RNA sequencing in control, HES1and HES4-transduced CD34+Lin– HPC that were cultured in the absence or presence of Notch signaling on OP9-GFP or OP9-DLL1, respectively. Enforced expression of HES1 in the absence of Notch1 activation (OP9-GFP, Notch-OFF, Figure 7A) resulted in a significant downregulation of genes associated with the differentiation of B-cells (MYB,32 GM2A,33 DUSP233 and IGLL134), NK-cells (ERI1,35 EMP336 and SRGN37) and myeloid cells (EPX,38 CLC,39 SRGN,37 MS4A240 and MS4A341) (Figure 7A). CEBPA expression was not detected in these experiments. However, given the essential role of HES1 in repression of CEBPA-dependent myeloid differentiation in mouse,17 we evaluated the 133

M. De Decker et al. A

Figure 3. HES4, but not HES1, promotes induction of T-lineage differentiation. Flow cytometry analysis (A, C) and absolute cell numbers (B, D) of control, ICN1-, HES1- and HES4-transduced CD34+ Lin- CB precursors cultured on the OP9-DLL4 feeder (Notch-ON) for 2 weeks in the presence of the T-lineage specific cytokines IL-7, SCF and FLT3L, showing the development of CD7+ CD5+ (A-B) and CD5+ CD1a+ (C-D) T-lineage cells. Data shows the average of six independent experiments and error bars indicate the standard error of the mean (SEM). *P<0.05 (non-parametric paired Wilcoxon test). Flow cytometry plots shown are representative for six independent experiments.

impact of HES1 and HES4 on CEBPA expression in the HL-60 leukemic cell line and, consistently, HES1 was very potent in repressing CEBPA expression, while HES4 was less efficient (Online Supplementary Figure S8A). In OP9DLL1-cultured CD34+ HPC (Notch-ON, Figure 7B), we observed significant repression of Notch signaling and Tlineage associated genes by HES1, including NOTCH1, RBPJ, MYB42 and DAD143 (Figure 7B). Surprisingly, on OP9-GFP stromal cells where no or little Notch1 triggering occurs, we also observed downregulation of these genes upon HES1 overexpression (Figure 7C). Also IRF8 expression, which is induced by Notch signaling for DC deve-lopment,44 was repressed by HES1 (Figure 7B). On both OP9 stromal co-cultures, expression of IKZF1, which is essential for differentiation of lymphoid-primed multipotent progenitors into common lymphoid progenitors (CLP),45 was downregulated by high levels of HES1 (Figure 7A-B). In contrast, HES1 overexpression showed significant upregulation of genes associated with stem cell maintenance, such as CD34, CD44, HOXB246, HOXB546, ESAM47 and ID148, and genes involved in cell cycle arrest, including IGFBP4, CDKN2D and CDKN1C, which was independent of Notch signaling (Figure 7A-B). These findings were confirmed by gene set enrichment analysis (GSEA) that revealed significant enrichment of HSC markers and genes involved in CD34+ quiescence in the HES1-transduced condition (Figure 7D-E). Genes associated with the maintenance of multipotent HSC at the expense of both CLP and granulocyte-monocyte progenitor (GMP) differentiation were also enriched upon HES1 overexpression (Online Supplementary Figure S8BC). In contrast, HES4 overexpression did not result in large gene expression changes. Among the significantly downregulated genes, we found CLC and SRGN (Figure 7A, G), which were previously associated with respectively myeloid and NK-cell development.37,39 Moreover, upon HES4 overexpression, we observed a slight, but 134

non-significant upregulation of NOTCH1, RBPJ and IKZF1 in the absence of Notch signaling (Figure 7C), while this was not the case when the cells were cultured on the OP9-DLL1 feeder (Figure 7B). Furthermore, high levels of HES1 or HES4 resulted in a significant downregulation of HES1 itself (Figure 7F-G). Comparison of the promotor sequences of the putative HES1 and HES4 target genes did not reveal a difference in the relative abundance of E-box, N-box and C-site motifs between HES1 and HES4 targets (Online Supplementary Figure S9A). However, using HOMER motif analysis, we observed a difference in the top enriched transcription factor binding sites within HES1 and HES4 target gene promoters (Online Supplementary Figure S9B-C). Thus, using RNA sequencing, we identified potential downstream targets of HES1 and HES4 that may modulate Notch-driven lineage decisions during human hematopoiesis and this may involve co-regulation by specific transcription factors.

HES1 and HES4 are both essential for inducing human T-cell development in HPC To investigate if HES1 and HES4 are required downstream of Notch during early human T-lineage differentiation, we performed HES1 and HES4 short hairpin RNA (shRNA)-mediated knockdown experiments in HPC that were cultured on OP9-DLL4 in T-lineage supporting conditions. For both genes, we used two independent shRNA which induced 40-60% knockdown (Figure 8A). Compared to the control, both the frequency and absolute numbers of CD7+CD5+ T-lineage specified and CD5+CD1a+ committed T-cell progenitors were significantly reduced upon knockdown of either HES1 or HES4 (Figure 8B-E). However, we did not observe a diversion to the B- or myeloid lineage in these cultures (data not shown), probably as a result of insufficient knockdown. Nevertheless, these findings illustrate that both HES1 and haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development

Figure 4. HES1 and HES4 repress NK cell development. (A) Flow cytometric analysis of control, ICN1-, HES1- and HES4-transduced CB CD34+Linâ&#x20AC;&#x201C; precursors cultured on the MS-5 feeder (Notch-OFF) for 3 weeks in the presence of the NK-lineage specific cytokines IL-7, SCF, FLT3L and IL-15, showing the development of CD56+CD7and CD56+CD7+ NK cells. Plots shown are representative for six independent experiments. (B-C) Absolute numbers of CD56+CD7â&#x20AC;&#x201C; (B) and CD56+CD7+ (C) NK cells generated in corresponding cultures shown in (A). Data shows the average of six independent experiments. Error bars indicate the standard error of the mean (SEM). *P<0.05 (non-parametric paired Wilcoxon test); ns: not significant.

HES4 have essential and non-redundant roles during early human T-cell development.

Discussion It is well established that Notch1 activation in thymusseeding precursors allows progenitor cells to differentiate towards the T-cell lineage, while development towards other hematopoietic cell types is repressed.4-6,13 However, the individual roles of various Notch1 target genes during these developmental choices are still unknown. Therefore, we investigated the potential of HES1 and HES4 to mediate human Notch1-dependent hematopoietic lineage decisions. We illustrate that both genes are essential downstream of Notch1 to induce human T-cell development and that they have both similar and different potentials to repress the differentiation towards alternative, non T-cell lineages. In murine hematopoiesis, HES1 is a repressor of intrathymic myeloid differentiation during the Notchdependent induction of T-cell development.17 HES1 is also required for sustaining leukemic growth in Notch-driven T-ALL, both in mouse and human.18,49 Previous overexpression studies in human have shown that HES1 inhibits B- and NK-lineage differentiation, but failed to reveal a block in monocyte development.27 In agreement with previous work in mice,17 we show here that HES1 can repress myeloid development in human and that this is mediated through repression of several genes with known roles in monocyte, granulocyte and dendritic cell development, thereby providing novel molecular insights in this process. The discrepancy in monocyte differentiation compared to previous work27 may relate to differences in expression levels following transduction since we used a codon-optihaematologica | 2021; 106(1)

mized construct to facilitate comparison with HES4. This could also explain the more efficient block in B-cell development in our results. Our novel RNA sequencing data reveal mechanisms through which HES1 specifically represses T-, B- and NKassociated genes, thereby validating the developmental impacts in our co-cultures. In light of the requirement for HES1 downstream of Notch1 during T-cell development, the repression of T-lineage differentiation by HES1 is striking. While this may result from the high expression levels in our perturbation settings, this may also reflect stage-specific requirements for HES1 in this process that are recurrent and may involve HES1 oscillation as observed during embryonic development.16 Although it is unclear if the latter occurs in developing thymocytes, Notch activation is required at different stages of early T-cell development and recurrent Notch activation events therein seem essential to balance between proliferation and differentiation.4,50 While HES1 is required to repress myeloid development in early T-cell precursors, it could later be essential to maintain thymocytes in a quiescent state, which is necessary during TCR rearrangements.4 Although less evident in human, this fits the high expression level of Hes1 in DN3a thymocytes that are active TCR-rearranging cells.51 A role for HES1 as a regulator of cell quiescence also concurs with the observation that continuous expression in HPC maintains cells in a CD34+ precursor state, consistent with previous studies.18,27 Here, we show that this is correlated with the upregulation of genes involved in stem cell maintenance, quiescence and cell cycle arrest. However, while HES1 is dispensable in mice for HSC maintenance,18 our data suggests that HES1 represses differentiation in human by maintaining a quiescent stem cell signature. 135

M. De Decker et al. A

Figure 5. Myeloid differentiation is hampered by HES1 and HES4. (A) Flow cytometry analysis of control, ICN1-, HES1- and HES4-transduced CB CD34+Lin- progenitors cultured on the MS-5 feeder (Notch-OFF) for 1 week in the presence of the myeloid-lineage specific cytokines SCF, FLT3L, TPO, GM-CSF and G-CSF, showing the development of CD15+HLA-DR– granulocytes (blue), CD14+CD15– monocytes (red) and HLA-DR+CD4+CD14– dendritic cells (green). t-SNE plots shown are representative of six independent experiments. (B-D) Absolute numbers of CD15+HLA-DR- granulocytes (B), CD14+CD15– monocytes (C) and HLA-DR+CD4+CD14– dendritic cells (D) generated in corresponding cultures shown in (A). Data shows the average of six independent experiments and error bars indicate the standard error of the mean (SEM). *P<0.05 (non-parametric paired Wilcoxon test); ns: not significant.

For the first time, we study the role of HES4 as a downstream modulator of Notch during non-malignant human hematopoiesis. In agreement with previous studies that compared the function of HES1 and HES4 in brain and bone development,20,21 our data suggests that both HES proteins have similar, and also different roles during human Notch-dependent hematopoietic lineage decisions. Compared to HES1, which is a strong repressor with a clear impact on HPC differentiation, enforced HES4 expression has a similar impact on myeloid and NK-cell development, but not on B- and T-cell differentiation. We believe that this lower repressive capacity of HES4 compared to HES1 is not due to lower expression levels following transduction since we obtained a similar mRNA increase for both HES genes, and protein staining even suggested higher amounts of HES4 protein compared to HES1. These different repressive capacities may, however, reflect a more stringent role for HES genes in the repression of myeloid fate compared to B-lineage differentiation, which is in agreement with a genetic study in mice that showed that intrathymic Notch signaling is more important for the repression of myeloid development than for preventing B-lineage diversion.52 One exception to the robust repression of myeloid development by HES genes concerns their impact on DC development. Although HES1 overexpression did result in a significant reduction in the absolute number of DC, their frequency was increased compared to control transduced cells. Strikingly, 136

HES4 was not sufficient to robustly repress DC development and this higher tolerance towards HES genes is in agreement with the intrathymic and Notch-dependent development of DC.53 Interestingly, in contrast to HES1, HES4 did promote T-cell development in the presence of Notch1 signaling. This places HES4 in line with other Notch targets, such as IL7R and TCF7, that have clear stimulating roles on T-cell development following Notch1 activation.11,54 However, HES4 by itself, or in combination with HES1, is insufficient to induce T-cell development,26 indicating that other downstream targets, or combinations thereof, are important for driving this process. Nevertheless, the clear distinct roles of HES1 and HES4 on B- and T-cell development may be important to control the initial Notch-dependent B- versus T-cell lineage decision in which HES1 represses the B-cell lineage while HES4 permits further T-cell development. Following this initial bifurcation, both HES1 and HES4 may then be required to repress NK-lineage differentiation while initially still permitting DC development. As such, HES1 and HES4 may have unique roles during these early stages of human T-cell differentiation, which are not-redundant as illustrated through our loss-of-function studies. Unfortunately, these knockdown studies did not result in an increased development of other, non T-cell lineages, but we believe this may result from the incomplete loss of HES1 and HES4 expression in these experiments. In contrast to HES1, HES4 overexpression did not result haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development

Figure 6. HES1, but not HES4, induces stem cell maintenance and quiescence in CD34+ cells. (A) Flow cytometric analysis of control, ICN1-, HES1- and HES4-transduced CB CD34+Linâ&#x20AC;&#x201C; precursors cultured on the OP9-DLL4 feeder for 2 weeks in the presence of the T-lineage specific cytokines IL-7, SCF and FLT3L, showing the maintenance of CD34+ HPC. Plots shown are representative of six independent experiments. (B-C) Frequency (B) and absolute numbers (C) of CD34+ cells maintained in corresponding cultures shown in (A). Data shows the average of six independent experiments. Error bars indicate the standard error of the mean (SEM). *P<0.05 (non-parametric paired Wilcoxon test); ns: not significant. (D) Flow cytometry analysis of control (grey), HES1 (black)- and HES4 (red)-transduced CB CD34+Linâ&#x20AC;&#x201C; HPC cultured on the OP9-DLL4 feeder for 2 weeks in the presence of the T-lineage supporting cytokines IL-7, SCF and FLT3L, showing proliferation as measured by the dilution of the CellTrace Violet dye at indicated time points.

in large gene expression changes. This might reflect the immature state of the transduced HPC since gene expression profiling was performed 48 h after transduction in an effort to discover direct regulatory events. However, the lack of HES4 expression in HSC suggests a minimal role for this protein at this immature stage, and the precise role for HES4 in gene regulation may therefore only be revealed at later differentiation stages, such as following Notch1 activation in immature thymocytes. Other genome-wide approaches, such as ChIP-seq in HES4-expressing thymocytes, will be required to identify other HES4 target genes. Remarkably, while we did not observe clear differences in the relative abundance of specific HES binding motifs (E-box, N-box and C-site) in the promotors of the predicted HES1 and HES4 target genes, both HES proteins may differentially regulate downstream target gene expression through interactions with other haematologica | 2021; 106(1)

transcription factors that are distinct for HES1 and HES4. Such mechanism could explain their different impact on hematopoietic lineage decisions. Intriguingly, we identified HES1 as a downstream target of both HES1 and HES4, thereby confirming the autoregulation of HES116 and suggesting a novel feedback loop between HES4 and HES1 that may also contribute to a possible transient repression of HES1, thereby allowing early thymocytes to escape from the repressive effects of HES1 to support their further differentiation. Such a direct repressive role of HES4 on HES1 expression seems essential given that protein interactions through the formation of HES1-HES4 heterodimers seemingly fail to counteract the strong repressive activity of HES1 homodimers. We anticipate that single cell (sc)RNA sequencing will provide more detailed insights into the actual sequence of HES1 and HES4 expression during early human T-cell development. 137

M. De Decker et al.

138

haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development Figure 7 (previous page). Gene expression analysis of HES1 and HES4 during human hematopoiesis. (A) Heat map representation of differentially expressed genes in control versus HES1-, or HES4-transduced CB CD34+Lin– cells cultured for 1 day on OP9-GFP (Notch-OFF), highlighting genes involved in the differentiation towards the B-, natural killer (NK)- and myeloid lineage (top) and HSC maintenance and quiescence (bottom). (B) Heat map representation of differentially expressed genes in control versus HES1-, or HES4-transduced CB CD34+Lin– progenitors cultured for 1 day on OP9-DLL1 (Notch-ON), highlighting genes involved in the differentiation towards the T-lineage (top) and HSC maintenance and quiescence (bottom). (C) Heat map representation of differentially expressed genes in control versus HES1-, or HES4-transduced CB CD34+Lin– cells cultured for 1 day on OP9-GFP (Notch-OFF), highlighting genes associated with Notch signaling and T-cell development. (D-E) GSEA shows a significant enrichment of HSC markers (D) and genes involved in quiescence (E) in the HES1-transduced condition compared to control. (F-G) Volcano plot representation of control versus HES1 (F)-, or HES4 (G)-transduced CB CD34+Lin– precursors cultured for 1 day on OP9-GFP, highlighting the expression of the top 10 differentially expressed genes. Red dots indicate significance (padj<0.05; n=3).

Figure 8. Non-redundant roles for HES1 and HES4 during early human T-cell development. (A) Quantitative RT-PCR of HES1 (black bars) and HES4 (red bars) in CB CD34+Lin– cells transduced with a control or two different HES1 or two different HES4 single hairpin RNA (shRNA), respectively, and cultured for 3 days on OP9-DLL4 to induce Notch activation. Data shows the average expression of four independent experiments, relative to the mean of ACTB and GAPDH mRNA levels, and normalized to expression in the control shRNA condition. Error bars indicate the standard error of the mean (SEM). **P<0.01; *P<0.05 (paired Student’s t-test). (B-C) Flow cytometry analysis of control, HES1 and HES4 shRNA-transduced CD34+Lin– hematopoietic progenitor cells (HPC) cultured on the OP9-DLL4 stromal cells for 2 weeks in the presence of the T-lineage supporting cytokines IL-7, SCF and FLT3L, showing the development of CD7+CD5+ (B) and CD5+CD1a+ (C) T-cell precursors. Contour plots shown are representative for six independent experiments. (D-E) Absolute numbers of CD7+CD5+ T-lineage specified (D) and CD5+CD1a+ T-lineage committed (E) progenitors generated in corresponding cultures shown in (B) and (C), respectively. Data shows the average of six independent experiments and error bars indicate the standard error of the mean (SEM). *P<0.05 (non-parametric paired Wilcoxon test).

haematologica | 2021; 106(1)

139

M. De Decker et al.

Given that HES4 is not present in the mouse genome, this work thus improves our understanding of the mechanisms through which Notch activation controls human blood cell development. This is an important issue since we have previously documented that Notch activation has different impacts on T-cell development in human versus mouse.4,14,55 Our work shows that both HES1 and HES4 are important to support the Notch1-dependent induction of human Tcell development and that this may be regulated through both different and similar repressive mechanism on other hematopoietic lineages. These findings thus underscore the distinct regulatory functions of each member of the complex network of Notch1 downstream target genes. Disclosures No conflicts of interest to disclose. Contributions MDD performed experiments; MDD, ML, JR and LT ana-

References 1. Doulatov S, Notta F, Laurenti E, Dick JE. Hematopoiesis: a human perspective. Cell Stem Cell. 2012;10(2):120-136. 2. Zhu J, Emerson SG. Hematopoietic cytokines, transcription factors and lineage commitment. Oncogene. 2002;21(21):32953313. 3. Novershtern N, Subramanian A, Lawton LN, et al. Densely interconnected transcriptional circuits control cell states in human hematopoiesis. Cell. 2011;144(2):296-309. 4. Taghon T, Waegemans E, Van de Walle I. Notch signaling during human T cell development. Curr Top Microbiol Immunol. 2012;360:75-97. 5. Rothenberg EV, Moore JE, Yui MA. Launching the T-cell-lineage developmental programme. Nat Rev Immunol. 2008;8(1):921. 6. De Smedt M, Hoebeke I, Reynvoet K, Leclercq G, Plum J. Different thresholds of Notch signaling bias human precursor cells toward B-, NK-, monocytic/dendritic-, or Tcell lineage in thymus microenvironment. Blood. 2005;106(10):3498-3506. 7. Weber BN, Chi AW, Chavez A, et al. A critical role for TCF-1 in T-lineage specification and differentiation. Nature. 2011; 476(7358):63-68. 8. Garcia-Ojeda ME, Klein Wolterink RG, Lemaitre F, et al. GATA-3 promotes T-cell specification by repressing B-cell potential in pro-T cells in mice. Blood. 2013; 121(10):1749-1759. 9. Hosokawa H, Romero-Wolf M, Yui MA, et al. Bcl11b sets pro-T cell fate by site-specific cofactor recruitment and by repressing Id2 and Zbtb16. Nat Immunol. 2018;19(12): 1427-1440. 10. Waegemans E, Van de Walle I, De Medts J, et al. Notch3 activation is sufficient but not required for inducing human T-lineage specification. J Immunol. 2014;193(12):59976004. 11. Van de Walle I, Dolens AC, Durinck K, et al. GATA3 induces human T-cell commitment by restraining Notch activity and repressing NK-cell fate. Nat Commun. 2016;7:11171. 12. Koch U, Fiorini E, Benedito R, et al. Delta-

140

lyzed data; BV, GL and PVV provided critical reagents; FVN performed RNA sequencing experiments; MDD and TT designed research and wrote the paper. Acknowledgments The authors thank Juan-Carlos Zuniga-Pflucker (University of Toronto) for OP9 stromal cells, Dr. Francois and Dr. Van Nooten (Department of Cardiac Surgery; University Hospital Ghent) for thymus tissue, Dr. Conny Matthys (Navelstrengbloedbank; University Hospital Ghent) for cord blood samples and Karin Weening (University of Ghent) for assistance with vector cloning. Funding This work was supported by grants of the Fund for Scientific Research Flanders (FWO), the Concerted Research Action of Ghent University (GOA), the Foundation against Cancer (STK) and the Interuniversity Attraction Poles Program (IUAP) from the Belgian Science Policy.

like 4 is the essential, nonredundant ligand for Notch1 during thymic T cell lineage commitment. J Exp Med. 2008;205(11): 2515-2523. 13. De Smedt M, Reynvoet K, Kerre T, et al. Active form of Notch imposes T cell fate in human progenitor cells. J Immunol. 2002; 169(6):3021-3029. 14. Van de Walle I, De Smet G, De Smedt M, et al. An early decrease in Notch activation is required for human TCR-alphabeta lineage differentiation at the expense of TCR-gammadelta T cells. Blood. 2009;113(13):29882998. 15. Van de Walle I, Waegemans E, De Medts J, et al. Specific Notch receptor-ligand interactions control human TCR-alphabeta/gammadelta development by inducing differential Notch signal strength. J Exp Med. 2013; 210(4):683-697. 16. Kageyama R, Ohtsuka T, Kobayashi T. The Hes gene family: repressors and oscillators that orchestrate embryogenesis. Development. 2007;134(7):1243-1251. 17. De Obaldia ME, Bell JJ, Wang X, et al. T cell development requires constraint of the myeloid regulator C/EBP-alpha by the Notch target and transcriptional repressor Hes1. Nat Immunol. 2013;14(12):1277-1284. 18. Wendorff AA, Koch U, Wunderlich FT, et al. Hes1 is a critical but context-dependent mediator of canonical Notch signaling in lymphocyte development and transformation. Immunity. 2010;33(5):671-684. 19. Zhou M, Yan J, Ma Z, et al. Comparative and evolutionary analysis of the HES/HEY gene family reveal exon/intron loss and teleost specific duplication events. PLoS One. 2012;7(7):e40649. 20. Bai G, Cheung I, Shulha HP, et al. Epigenetic dysregulation of hairy and enhancer of split 4 (HES4) is associated with striatal degeneration in postmortem Huntington brains. Hum Mol Genet. 2015;24(5):1441-1456. 21. McManus M, Kleinerman E, Yang Y, et al. Hes4: A potential prognostic biomarker for newly diagnosed patients with high-grade osteosarcoma. Pediat Blood Cancer. 2017; 64(5). 22. El Yakoubi W, Borday C, Hamdache J, et al. Hes4 controls proliferative properties of neural stem cells during retinal ontogenesis.

Stem Cells. 2012;30(12):2784-2795. 23. Murato Y, Hashimoto C. Xhairy2 functions in Xenopus lens development by regulating p27(xic1) expression. Dev Dyn. 2009; 238(9):2179-2192. 24. Cakouros D, Isenmann S, Hemming SE, et al. Novel basic helix-loop-helix transcription factor hes4 antagonizes the function of twist-1 to regulate lineage commitment of bone marrow stromal/stem cells. Stem Cells Dev 2015;24(11):1297-1308. 25. Stoeck A, Lejnine S, Truong A, et al. Discovery of biomarkers predictive of GSI response in triple-negative breast cancer and adenoid cystic carcinoma. Cancer Discov. 2014;4(10):1154-1167. 26. Van de Walle I, De Smet G, Gartner M, et al. Jagged2 acts as a Delta-like Notch ligand during early hematopoietic cell fate decisions. Blood. 2011;117(17):4449-4459. 27. Hoebeke I, De Smedt M, Van de Walle I, et al. Overexpression of HES-1 is not sufficient to impose T-cell differentiation on human hematopoietic stem cells. Blood. 2006;107(7):2879-2881. 28. De Smedt M, Taghon T, Van de Walle I, De Smet G, Leclercq G, Plum J. Notch signaling induces cytoplasmic CD3 epsilon expression in human differentiating NK cells. Blood. 2007;110(7):2696-2703. 29. Klinakis A, Lobry C, Abdel-Wahab O, et al. A novel tumour-suppressor function for the Notch pathway in myeloid leukaemia. Nature. 2011;473(7346):230-233. 30. Lahmann I, Brohl D, Zyrianova T, et al. Oscillations of MyoD and Hes1 proteins regulate the maintenance of activated muscle stem cells. Genes Dev. 2019;33(910):524-535. 31. Sueda R, Imayoshi I, Harima Y, Kageyama R. High Hes1 expression and resultant Ascl1 suppression regulate quiescent vs. active neural stem cells in the adult mouse brain. Genes Dev. 2019;33(9-10): 511â&#x20AC;&#x201C;523. 32. Thomas MD, Kremer CS, Ravichandran KS, Rajewsky K, Bender TP. c-Myb is critical for B cell development and maintenance of follicular B cells. Immunity. 2005;23(3):275286. 33. Young K, Borikar S, Bell R, Kuffler L, Philip V, Trowbridge JJ. Progressive alterations in multipotent hematopoietic progenitors

haematologica | 2021; 106(1)

HES1 and HES4 during T-cell development

underlie lymphoid cell loss in aging. J Exp Med. 2016;213(11):2259-2267. 34. Mansson R, Zandi S, Anderson K, et al. B-lineage commitment prior to surface expression of B220 and CD19 on hematopoietic progenitor cells. Blood. 2008;112(4):1048-1055. 35. Thomas MF, Abdul-Wajid S, Panduro M, et al. Eri1 regulates microRNA homeostasis and mouse lymphocyte development and antiviral function. Blood. 2012;120(1):130142. 36. Crinier A, Milpied P, Escaliere B, et al. Highdimensional single-cell analysis identifies organ-specific signatures and conserved NK cell subsets in humans and mice. Immunity. 2018;49(5):971-986. 37. Kolset SO, Pejler G. Serglycin: a structural and functional chameleon with wide impact on immune cells. J Immunol. 2011;187(10): 4927-4933. 38. Gorgens A, Radtke S, Mollmann M, et al. Revision of the human hematopoietic tree: granulocyte subtypes derive from distinct hematopoietic lineages. Cell Rep. 2013;3(5): 1539-1552. 39. Grootens J, Ungerstedt JS, Nilsson G, Dahlin JS. Deciphering the differentiation trajectory from hematopoietic stem cells to mast cells. Blood Adv. 2018;2(17):2273-2281. 40. Zheng S, Papalexi E, Butler A, Stephenson W, Satija R. Molecular transitions in early progenitors during human cord blood hematopoiesis. Mol Syst Biol. 2018; 14(3):e8041. 41. Ishibashi T, Yokota T, Satoh Y, et al. Identification of MS4A3 as a reliable marker

haematologica | 2021; 106(1)

for early myeloid differentiation in human hematopoiesis. Biochem Biophys Res Commun. 2018;495(3):2338-2343. 42. Allen RD 3rd, Bender TP, Siu G. c-Myb is essential for early T cell development. Genes Dev. 1999;13(9):1073-1078. 43. Hong NA, Kabra NH, Hsieh SN, Cado D, Winoto A. In vivo overexpression of Dad1, the defender against apoptotic death-1, enhances T cell proliferation but does not protect against apoptosis. J Immunol. 1999; 163(4):1888-1893. 44. Lee J, Zhou YJ, Ma W, et al. Lineage specification of human dendritic cells is marked by IRF8 expression in hematopoietic stem cells and multipotent progenitors. Nat Immunol. 2017;18(8):877-888. 45. Dege C, Hagman J. Mi-2/NuRD chromatin remodeling complexes regulate B and Tlymphocyte development and function. Immunol Rev. 2014;261(1):126-140. 46. Ali MAE, Fuse K, Tadokoro Y, et al. Functional dissection of hematopoietic stem cell populations with a stemness-monitoring system based on NS-GFP transgene expression. Sci Rep. 2017;7(1):11442. 47. Yokota T, Oritani K, Butz S, et al. The endothelial antigen ESAM marks primitive hematopoietic progenitors throughout life in mice. Blood. 2009;113(13):2914-2923. 48. Perry SS, Zhao Y, Nie L, Cochrane SW, Huang Z, Sun XH. Id1, but not Id3, directs long-term repopulating hematopoietic stemcell maintenance. Blood. 2007;110(7):23512360. 49. Espinosa L, Cathelin S, D'Altri T, et al. The

Notch/Hes1 pathway sustains NF-kappaB activation through CYLD repression in T cell leukemia. Cancer Cell. 2010;18(3):268281. 50. Taghon T, Van de Walle I, De Smet G, et al. Notch signaling is required for proliferation but not for differentiation at a well-defined beta-selection checkpoint during human Tcell development. Blood. 2009;113(14):32543263. 51. Wong GW, Knowles GC, Mak TW, Ferrando AA, Zuniga-Pflucker JC. HES1 opposes a PTEN-dependent check on survival, differentiation, and proliferation of TCRbeta-selected mouse thymocytes. Blood. 2012;120(7):1439-1448. 52. Feyerabend TB, Terszowski G, Tietz A, et al. Deletion of Notch1 converts pro-T cells to dendritic cells and promotes thymic B cells by cell-extrinsic and cell-intrinsic mechanisms. Immunity. 2009;30(1):67-79. 53. Martin-Gayo E, Gonzalez-Garcia S, GarciaLeon MJ, et al. Spatially restricted JAG1Notch signaling in human thymus provides suitable DC developmental niches. J Exp Med. 2017;214(11):3361-3379. 54. Gonzalez-Garcia S, Garcia-Peydro M, Alcain J, Toribio ML. Notch1 and IL-7 receptor signalling in early T-cell development and leukaemia. Curr Top Microbiol Immunol. 2012;360:47-73. 55. Taghon T, Rothenberg EV. Molecular mechanisms that control mouse and human TCRalphabeta and TCR-gammadelta T cell development. Semin Immunopathol. 2008; 30(4):383-398.

141

ARTICLE Ferrata Storti Foundation

Hematopoiesis

Cyba-deficient mice display an increase in hematopoietic stem cells and an overproduction of immunoglobulins Rodrigo Prieto-Bermejo,1,2 Marta Romo-González,1,2* Alejandro PérezFernández,1,2* Ignacio García-Tuñón,3 Manuel Sánchez-Martín2,4,5 and Ángel Hernández-Hernández1,2

Departamento de Bioquímica y Biología Molecular, Universidad de Salamanca; 2Instituto de Investigación Biomédica de Salamanca (IBSAL); 3Unidad de Diagnóstico Molecular y Celular del Cáncer, Centro de Investigación del Cáncer-IBMCC (USAL-CSIC); 4Servicio de Transgénesis, Nucleus, Universidad de Salamanca and 5Departamento de Medicina, Universidad de Salamanca, Salamanca, Spain 1

Haematologica 2021 Volume 106(1):142-153

*MR-G and AP-F contributed equally to this work.

ABSTRACT

he regulation of protein function by reversible oxidation is increasingly recognized as a key mechanism for the control of cellular signaling, modulating crucial biological processes such as cell differentiation. In this scenario, NADPH oxidases must occupy a prominent position. Our results show that hematopoietic stem and progenitor cells express three p22phox -dependent NADPH oxidase members (NOX1, NOX2 and NOX4). By deleting the p22phox coding gene (Cyba), here we have analyzed the importance of this family of enzymes during in vivo hematopoiesis. Cyba-/- mice show a myeloid bias, and an enrichment of hematopoietic stem cell populations. By means of hematopoietic transplant experiments we have also tried to dissect the specific role of the NADPH oxidases. While the absence of NOX1 or NOX2 provides a higher level of reconstitution, a lack of NOX4 rendered the opposite result, suggesting a functional specificity among the different NADPH oxidases. Cyba-/- cells showed a hampered activation of AKT1 and a sharp decrease in STAT5 protein. This is in line with the diminished response to IL-7 shown by our results, which could explain the overproduction of immunoglobulins observed in Cyba-/- mice.

Correspondence: ÁNGEL HERNÁNDEZ-HERNÁNDEZ angelhh@usal.es Received: July 19, 2019. Accepted: January 2, 2020. Pre-published: January 9, 2020. https://doi.org/10.3324/haematol.2019.233064

142

Introduction From being considered harmful metabolic by-products, reactive oxygen species (ROS) have turned out to be important regulators of cellular biology, by acting as bona fide secondary messengers.1–3 NADPH oxidases are the only cellular system specialized in the production of ROS.4 The founding member of this family is phagocyte oxidase, a multiprotein complex that produces huge amounts of superoxide during respiratory burst, which is required for the elimination of pathogens.4 The complex consists of two integral membrane proteins (p22phox, and the catalytic subunit, named gp91phox or NOX2), three cytosolic subunits (p40phox, p47phox, p67phox), and the Rac GTPase.4,5 Inactivating mutations affecting the complex produce chronic granulomatous disease (CGD), a hematological disorder characterized by the occurrence of repetitive infections.6 For quite some time the NADPH oxidase from phagocytes seemed to be the only oxidase. However, in 1999 cloning and characterization of NOX1 was reported,7 which was followed by the discovery of other homologues oxidases. Nowadays the family comprises seven members, which can be classified into two groups, those dependent on p22phox (NOX1, NOX2, NOX3, NOX4), and those p22phox -independent that can be activated by calcium (NOX5, DUOX1 and DUOX2).4,5 These enzymes are present in all eukaryotic cells, including unicellular organisms,8 and several members of the family are commonly expressed simultaneously.4 This broad distribution and its regulation by extracellular signaling make NADPH oxidases a key element in redox signaling. ROS and NADPH oxidases are involved in the control of cell fate.2,9–11 haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

Hematopoiesis is a paradigmatic example of cell differentiation, because hematopoietic stem cells (HSC) must produce all mature blood lineages. There is increasing evidence suggesting the importance of redox signaling for hematopoietic differentiation, as well as for the contribution of an elevated level of ROS in the development of leukemia.2,3 Our previous data has shown the requirement of NADPH oxidase-produced ROS for in vitro megakaryocytic differentiation.12 However, even though NADPH oxidases were discovered in the hematopoietic system, knowledge of the importance of these enzymes for hematopoiesis in vivo is scarce. This analysis has surely been hampered by the fact that hematopoietic cells express several NADPH oxidase isoforms (our unpublished data and 13). Here we have analyzed the relevance of NADPH oxidases during in vivo hematopoiesis. Mouse hematopoietic progenitor cells express NOX1, NOX2 and NOX4, all of them p22phox-dependent. In order to assess the importance of NADPH oxidase activity for in vivo hematopoiesis we have generated p22phox-deficient mice (Cyba-/-). The lack of p22phox induces a myeloid bias. Moreover, there is an increased proportion of HSC in the bone marrow (BM). In competitive transplant experiments, cells deficient in NOX1 (Nox1-/-), NOX2 (Cybb-/-) and p22phox (Cyba-/-) showed an increased reconstitution ability with respect to control cells. However, NOX4 (Nox4-/-) deficient cells did not show this effect. The response of Cyba-/- cells to IL-7 is severely impaired, which can be explained by a hampered activation of AKT1 and by the reduction in STAT5 protein levels. These redox signaling alterations could be the cause of the increased production of immunoglobulins observed in Cyba-/- mice.

University of Salamanca) as previously described.15 Recipient mice were injected intravenously through the lateral tail vein with 3x106 BM cells from ES Cyba-/- or from wild-type C57BL/6 mice. Competitive transplant experiments were performed by injecting 1.5x106 BM cells from NADPH oxidase deficient mice (Nox1-/-, Cybb-/-, Nox4-/- or ES Cyba-/-) or from C57BL/6 control mice, together with 1.5x106 BM competitor cells from wild-type C57BL/6 mice. Cell origin was assessed by the differential expression of the CD45.1 and CD45.2 isotypes. Secondary transplants were carried out 18 weeks post-first transplant, by injecting 3x106 BM cells from the primary transplanted mice into lethally irradiated C57BL/6 mice.

Methods

Statistical analyses and data report

Animals C57BL/6 mice were from the University of Salamanca Animal Facility Unit. Albino C57BL/6 mice (B6(Cg)-Tyrc-2J/J), Nox1 (B6.129X1-Nox1tm1Kkr/J), Nox2 (B6.129S-Cybbtm1Din/J) and Nox4 (B6.129-Nox4tm1Kkr/J) deficient mice were from Jackson Laboratory (Bar Harbor, ME, USA). Cyba-/- models were generated as detailed in the Online Suplementary Materials and Methods. All procedures were approved by the Bioethics Committee at the University of Salamanca.

In vivo bromodeoxyuridine incorporation assay Mice were intraperitoneally treated with 100 mg/kg 5-bromo 2' -deoxyuridine (BrdU) and sacrificed 14 hours postinjection. The percentage of BrdU+ proliferating cells was analyzed by flow cytometry as previously described.15

RNA sequencing analysis cDNA libraries were compiled using the Illumina TruSeq RNA Library Preparation Kit v2, with 2 mg of total RNA from immunopurified Lin– cells.15 Single-end 150 nt length sequencing was performed on an Illumina NextSeq 500 System using a Mid Output kit v2.5 from Illumina. Reads obtained were compared against the GRCm38 mouse genome using the bioinformatic facility tool of the University of Salamanca (https://ranaseq.eu/home). GEO accession: GSE131725.

Immunoblotting Immunoblotting and quantification of bands were performed as previously described,16 and using fluorescently labelled secondary antibodies with an Odyssey Infrared Imaging System (Li-Cor). VINCULIN was used as a loading control.

Data are expressed as mean values ± standard deviation. In dot graphs, each dot represents an individual value and the horizontal line denotes the mean value. Data were analyzed with SPSS 23 software. Two-tailed unpaired Student’s t test was used, and differences were considered statistically significant when P<0.05 (*), P<0.01 (**) and P<0.001 (***).

Results p22phox knockout mice (Cyba-/-) are viable and present a bias towards the myeloid lineage

ROS detection Peripheral blood (PB) was lysed to eliminate red blood cells. ROS levels were detected by cell staining with 10 mM DCFDA (2 ,7 -dichlorodihydrofluorescein diacetate)12 or with 100 mM luminol14 after stimulation with 2 mM PMA (phorbol-12-myristate-13-acetate).

Colony forming unit (CFU) assays Cells were grown in methylcellulose semisolid medium supplemented with a cocktail of cytokines (50 ng/mL SCF, 20 ng/mL IL3, 20 ng/mL GM-CSF and 3 U/mL EPO), or with individual cytokines (IL-3, GM-CSF, and IL-7) at 20 ng/mL. 200,000 spleen and 10,000 BM cells (200,000 with IL-7) were used. Cells were grown at 37ºC and 5% CO for 12 days. 2

BM transplantation C57BL/6 recipient mice were lethally irradiated with two doses of 5 Gy from a Cs source (Gammacell 1,000 Elite, Nucleus Facility, haematologica | 2021; 106(1)

In agreement with our former analysis on human hematopoietic cells,13 mice progenitor Lin– cells express several NOX isoforms (NOX1, NOX2 and NOX4) (Figure 1A and Online Supplementary Figure S1). Since all of them are p22phox-dependent,4,5 we reasoned that a p22phox-deficient mouse model would allow the analysis of the relevance of NADPH oxidase activity for in vivo hematopoiesis. p22phox knockout mice were generated using either embryonic stem (ES) cells, or CRISPR/Cas9 technology. These mice will be named hereafter as ES Cyba-/- and CR Cyba-/- respectively. Mice were genotyped by PCR and Southern blotting (Online Supplementary Figure S2). The absence of the p22phox protein was confirmed by Western blotting (Figure 1B). Cyba-/- mice were viable and displayed equilibrium defects denoted by a clear inclination of the head. This head-tilt phenotype has been reported before for a p22phox hypomorphic mutant,17 and it 143

R. Prieto-Bermejo et al.

Figure 1. The lack of Cyba gene induces a myeloid bias. (A) NADPH oxidase expression in Lin- progenitor cells analyzed by RT-PCR. The tissues indicated in the figure were used as positive controls (n=2). (B) Representative blots of p22phox protein expression in wild-type, ES Cyba-/- and CR Cyba-/- mice (n=4). (C) Peripheral blood cells were stimulated with 2 mM phorbol 12-myristate 13-acetate (PMA) for 30 min. The level of reactive oxygen species (ROS) was analyzed in the granulocyte population. A representative flow cytometry experiment and the mean ± standard deviation (SD) of the ROS fold-increase of seven different experiments are shown. (D) 200,000 peripheral blood (PB) cells were incubated with 100 mM luminol, and chemiluminescence was monitored for 15 min at 37°C after addition of 2 mM PMA. A representative experiment showing the RLU (relative luminescence units) and the mean ± SD of the chemiluminescence fold-increase at 10 min of three different experiments are shown. (E) Analysis of hematopoietic lineages in ES Cyba-/- and control mice. (F) Analysis of hematopoietic lineages in CR Cyba-/- and control mice.

144

haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

Figure 2. Bone marrow cells from Cyba-/- mice are enriched in hematopoietic stem cells. (A) Representative flow cytometry analysis of the hematopoietic stem cells (HSC) populations in CR Cyba-/- and control mice. (B) Percentages of the different stem and progenitor populations in CR Cyba-/- and control mice. (C) Percentages of the different stem and progenitor populations in ES Cyba-/- and control mice. (D) Analysis of stem and progenitor cell populations at 20 weeks post-transplantation in lethally irradiated mice transplanted with wild-type or ES Cyba-/- cells. LK: lineage-committed progenitors; LSK: Lin-Sca-1+ c-Kit+ hematopoietic stem cells; LMPP: lymphoid-primed multipotent progenitors; LT-HSC: long-term HSC; ST-HSC: short-term HSC; GMP: granulocyte-monocyte progenitors; MEP: megakaryocyte-erythrocyte progenitors; CMP: common myeloid progenitors, CLP: common lymphoid progenitors.

haematologica | 2021; 106(1)

145

R. Prieto-Bermejo et al.

is suggested to be due to the requirement of NOX3 activity for otoconia synthesis.18 The lack of NADPH oxidase activity in Cyba-/- mice was checked by measuring the levels of extracellular and intracellular ROS after stimulating NADPH oxidase with PMA (Figure 1C-D and Online Supplementary Figure S3). ES Cyba-/- mice showed a significant increase in CD11b+ myeloid cells in PB and spleen. A general increase of CD11b+ myeloid cells is also observed in the BM, though not statistically significant (Figure 1E). These results support the existence of a myeloid bias in ES Cyba-/- mice, a notion that is further supported by analysis of CR Cyba-/mice, which shows a significant increase in myeloid cells in BM (CD11b+Gr1– cells) and in spleen (CD11b+Gr1+ cells) (Figure 1F).

Cyba-/- mice show an enrichment of hematopoietic stem progenitor cells in the BM In the BM, no differences in cellularity or the percentage of Lin– progenitor cells were observed between knockout (ES Cyba-/- or CR Cyba-/-) and wild-type mice (data not shown). Within the Lin– cell population we observed an enrichment in HSC (multipotent long-term HSC [LT-HSC]

and short-term HSC [ST-HSC]) in both CR Cyba-/- (Figure 2A-B) and ES Cyba-/- mice (Figure 2C). This increase was around two-fold for both LT- and ST-HSC. No significant differences were found in more-committed myeloid progenitor cells (megakaryocyte-erythrocyte progenitor [MEP], granulocyte-monocyte progenitor [GMP] and common myeloid progenitor [CMP] cells) (Figure 2A-C). In order to test whether the enrichment of HSC in Cyba-/- mice is cell-autonomous or determined by the niche, BM cells from ES Cyba-/- and wild-type mice were transplanted into lethally irradiated wild-type mice. There was an enrichment of hematopoietic stem progenitor cells (HSPC) populations in animals transplanted with ES Cyba-/- BM cells (LT-HSC, ST-HSC, lymphoid-primed multipotent progenitor [LMPP], MEP and common lymphoid progenitor [CLP] cells) (Figure 2D). These results strongly suggest that the higher proportion of HSC in the Cyba-/- mice does not depend on the influence of the niche, instead it is a cell-autonomous phenomenon.

Cyba-/- HSC show an increased proliferation capacity in vivo In vivo BrdU incorporation experiments were performed

Figure 3. Bone marrow progenitor cells from Cyba-/- mice proliferate more than control cells. CR Cyba-/- and control mice were injected intraperitoneally (IP) with 100 mg/kg 5-bromo-2'-deoxyuridine (BrdU), and sacrificed 14 hours later. BrdU incorporation in the hematopoietic stem cells (HSC) bone marrow (BM) populations was analyzed by flow cytometry. (A) A representative experiment of the BrdU incorporation in hematopoietic stem cells (LSK cells) cells is shown. (B) Percentage of BrdU+ cells in the different HSC populations. LMPP: lymphoid-primed multipotent progenitors; LT-HSC: long-term HSC; ST-HSC: short-term HSC.

Figure 4. Colony forming assays with bone marrow and spleen cells from Cyba-/- mice. The colony forming unit (CFU) ability of bone marrow (BM) and spleen cells in response to complete medium or individual cytokines was tested. (A) CFU ability of BM and spleen cells in response to complete medium (n=6 and n=9, respectively). (B) BM CFU in the presence of IL-3, GM-CSF or IL-7 (n=6). The mean ± standard deviation (SD) of the number of colonies normalized with respect to the control is shown.

146

haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

to analyze the proliferation capacity of HSPC from CR Cyba-/- and control mice. Our results showed a significant increase in BrdU incorporation in LinÂŻSca-1+c-Kit+ hematopoietic stem cells (LSK cells), mainly in ST-HSC and LMPP, but also a general increase in LT-HSC (Figure 3). This higher proliferation potential would explain the enrichment of HSC in Cyba-/- mice.

Cyba-/- cells show a lower clonogenic capacity in response to IL-7 We next analyzed the self-renewal potential of BM and spleen cells in response to different cytokines by CFU assays. In the presence of complete media (SCF, IL-3, GMCSF and EPO) BM CR Cyba-/- cells produced a slightly higher number of colonies, though not statistically significant, while spleen CR Cyba-/- cells produced a significantly

higher number (Figure 4A). BM cells from CR Cyba-/- mice showed a general trend towards a higher clonogenic capacity in the presence of cytokines that induce myeloid differentiation (IL-3 and GM-CSF), while there was a significant decrease in the number of colonies in the presence of IL-7, a cytokine that drives B-cell differentiation in mice19 (Figure 4B). These results are consistent with the myeloid bias observed in vivo.

Nox1, Cybb and Cyba deficient cells show an increased hematopoietic reconstitution capacity The results so far support the relevance of NADPH oxidase activity for hematopoiesis. Therefore, an interesting angle would be to delineate the importance of the different NADPH oxidase enzymes. Thus, through competitive transplantation experiments we analyzed the

Figure 5. The deficiency of NOX1 and NOX2 increases hematopoietic reconstitution capacity in competitive transplant experiments. The percentage of chimerism throughout time in periperal blood (PB) for the different primary and secondary transplants is shown: (A) Primary and (B) secondary transplants for Nox1-/- and control mice (n=11 and n=6, respectively). (C) Primary and (D) secondary transplants for Cybb-/- and control mice (n=11 and n=6, respectively). (E) Primary and (F) secondary transplants for Nox4-/- and control mice (n=10 and n=6, respectively). Graphics show mean Âą standard deviation (SD).

haematologica | 2021; 106(1)

147

R. Prieto-Bermejo et al.

hematopoietic reconstitution ability of cells lacking NOX1 (Nox1-/-), NOX2 (Cybb-/-) and NOX4 (Nox4-/-). BM cells from these mice strains were challenged with the same amount of BM cells from wild-type mice, and were transplanted into lethally irradiated recipient wild-type mice (Figure 5). The lack of NOX1 (Nox1-/-) and NOX2 (Cybb-/-) conferred a mild but significant increase in reconstitution compared to wild-type cells in primary recipients (Figure 5A, C). Interestingly, Nox4-/- cells showed the opposite, with a slightly lower reconstitution capacity in primary transplants (Figure 5E). In order to test this issue further, secondary transplants were performed in the three settings. While Nox1-/- and Cybb-/- once again showed a higher reconstitution capacity than wild-type cells (Figure 5B, D), Nox4-/- cells displayed the same percentage of chimerism (50%) as in the primary transplant (Figure 5F). These results validate those obtained in the primary transplants and suggest a differential role for each Nox isoform during in vivo hematopoiesis. In agreement with the results described for Nox1-/- and Cybb-/-, competitive transplant experiments performed

with ES Cyba-/- mice and wild-type littermates offered a very subtle enhancement of reconstitution of Cyba-/- cells (Figure 6A-B), which could also be observed in secondary transplants (Online Supplementary Figures 4A-B). When we analyzed the lineages within the ES Cyba-/transplanted cells, there was a significant increase in CD11b+ Gr1â&#x20AC;&#x201C; and CD3+ cells, and a decrease in CD19+ and B220+ cells (Figure 6C). These features were also evidenced in secondary transplants (Online Supplementary Figure S4C). Moreover, there was also an enrichment of LT-HSC and lineage-committed progenitors (LK) and a general increase in LMPP and LSK in the ES Cyba-/- transplanted cells (Figure 6D). These features are much like those found in the ES Cyba-/- mice (Figure 1E and Figure 2C), strongly suggesting once again that the hematopoietic alteration we have found in Cyba-/- mice is the result of cell-endogenous causes and not due to the influence of the niche. Unlike Cyba-/- cells, in the transplant experiments performed with Nox1-/- and Cybb-/- deficient cells, no clear differences were found in regard to the different hematopoietic lineages (Online Supplementary Figure S5A). No signifi-

148

Figure 6. The deficiency of p22phox induces a slight increase in hematopoietic reconstitution capacity in competitive transplant experiments. Lethally irradiated wild-type recipient mice were subjected to competitive hematopoietic transplant experiments with ES Cyba-/-. Wild-type littermates were used as control. (A) Percentage of chimerism throughout time in peripheral blood (PB) in primary transplant. Graphic show mean Âą SD (n=6). (B) Percentage of chimerism in PB, bone marrow (BM) and spleen 20 weeks posttransplantation. (C) Hematopoietic mature lineages in PB, BM and spleen of the transplan-ted animals coming from Cyba-/- and control cells. (D) Hematopoietic progenitor stem cells (HSPC) in BM of the transplanted animals coming from Cyba-/- and control cells. LMPP: lymphoid-primed multipotent progenitors; LT-HSC: long-term HSC; STHSC: short-term HSC.

haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

Figure 7. Transcriptome analysis of Linâ&#x20AC;&#x201C; cells from Cyba-/- and control mice. Lin- cells were immunopurified from the bone marrow (BM) of two CR Cyba-/- and two control mice. The gene expression profile was analyzed by RNA sequencing. (A) Heat-map showing Z-scores of the expression values (normalized as transcript per million [TPM] values) of genes with significantly altered expression in Cyba-/- cells. (B) Network of functional analysis results for processes. Each node represents a functional category. The percentage of significant differentially-expressed genes annotated by the different categories on the GO database is represented by colors, as shown by the scale in the figure. Connecting lines show the relationships between different functional categories; the width of the lines is proportional to the genes in common. (C) Network of functional analysis results for pathways.

haematologica | 2021; 106(1)

149

R. Prieto-Bermejo et al.

1.00 0.59 1.00 0.79

1.00 0.02 1.00 0.13

1.00

1.06 1.00 1.26

1.00 5.78 1.00 5.58 Figure 8. p22phox deficiency increases myeloid gene expression, immunoglobulin serum levels and induces the downregulation of STAT5 protein. (A) The expression of several myeloid genes that were upregulated according to RNA sequencing data was analyzed by quantitave RT-PCR (qRT-PCR) in Lin- cells. Results show the relative expression of these genes in CR Cyba-/- mice with respect to control mice. The discontinuous line represents expression in control mice. (B) Relative expression of Cybb in Linâ&#x20AC;&#x201C; cells from CR Cyba-/- mice with respect to control mice analyzed by qRT-PCR. (C) Immunoglobulin protein levels in the bone marrow (BM) and spleen of CR Cyba-/- and control mice (n=4). (D) Immunoglobulin serum levels in CR Cyba-/- and control mice. Levels of IgA, IgG and IgM in serum were analyzed by ELISA. Results are shown as mg/dL. The steady state level in BM and spleen of the following proteins was analyzed: (E) STAT5 and CRKL (n=4), (F) STAT3 (n=3). (G) Stat5a and Stat5b expression was analyzed by qRT-PCR in Lin- cells. Results show the relative expression of these genes in CR Cyba-/- mice with respect to control mice. (H) MYC protein levels in BM and spleen from CR Cyba-/- and control mice (n=3).

150

haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

cant differences were found regarding HSPC, though Cybb-/- cells showed a general increase in the LT-HSC and ST-HSC populations (Online Supplementary Figure S5B). Interestingly, lineage analysis of Nox4-/- cells offered some similarities to the differences described in the transplants performed with ES Cyba-/- mice, such as, for example, a significant increase in CD11b+ Gr1– myeloid cells in BM, and the same changes in CD3+ and CD19+ cells (Online Supplementary Figure S5C).

Cyba deficiency increases the expression of immunoglobulin genes In order to gain some insight regarding the previous results we performed transcriptome profile analyses in Lin– cells from control and CR Cyba-/- mice by RNA sequencing . We found 47 upregulated and five downregulated genes in Cyba-/- cells (Figure 7A). In line with the myeloid bias described above, among the upregulated genes there were several genes encoding proteins highly expressed in myeloid cells, such as Itgam (CD11b), Pirb (CD85A), Lilrb4 (CD85K), Ccr1(CD191) and Lrg1. Intriguingly, there was also an upregulation of Cybb mRNA expression in the Cyba-/- cells. These results were corroborated by quantitative RT-PCR (qRT-PCR) (Figure 8A-B). However, the most striking result was the upregulation of 18 immunoglobulin genes of both heavy variable (Ighv) and k light variable (Igkv) chains. In line with this, we detected an increased expression of immunoglobulins in the BM and spleen of CR Cyba-/- mice (Figure 8C), suggesting an increased production of immunoglobulins by CR Cyba-/- mice. Confirming this hypothesis, CR Cyba-/- mice showed a significant increase in IgA, in IgG, in serum, and a general increase in IgM that did not reach statistical significance (Figure 8D). All in all, these results are in agreement with an exacerbated production of immunoglobulins by CR Cyba-/- mice. Functional analyses of these results (Figure 7B) suggested that the innate immune response and immunoglobulin production processes are severely affected by the lack of Cyba, with a high proportion of genes involved in these processes among the miss-regulated genes in Cyba-/- mice. Accordingly, the response to bacteria and defense against viruses were among the altered processes. Moreover, as expected, processes related to superoxide anion generation are also hampered. Of note, other fundamental processes such cell migration and cell adhesion also seem to be affected by the lack of Cyba (Figure 7B). In the same line, pathway analyses also suggest the alteration of the innate immune response, and pathways related with cell adhesion and migration (Figure 7C). Signaling pathways such as VEGF, Rho GTPases and PI3K-AKT-mTOR appear among the potentially altered pathways (Figure 7C). In support of this, we observed that activation of AKT1 in response to several hematopoietic cytokines was somewhat impaired in Cyba-/- cells (Online Supplementary Figure S6A).

Cyba deficiency leads to the downregulation of STAT5 Given the possible implication of NADPH oxidases in the regulation of the signaling pathways governing hematopoiesis2,3 we analyzed the activation of several signaling pathways by immunoblotting. No differences were found in the activation of ERK or in levels of b-CATENIN (Online Supplementary Figures 6B-C). However, our analyhaematologica | 2021; 106(1)

ses revealed a sharp decrease in STAT5 protein in BM and spleen Cyba-/- cells (Figure 8E and Online Supplementary Figure 6D). In contrast, no differences were found regarding the expression of STAT3 protein (Figure 8F), which highlights the specificity of the decrease in STAT5 in Cyba-/- mice. No differences in Stat5a and Stat5b mRNA between Cyba-/- and control mice were detected in the RNA-seq analyses (Figure 7), which was corroborated by qRT-PCR (Figure 8G). Therefore, this strongly suggests that the decrease in STAT5 in Cyba-/- mice occurs at the protein level. Moreover, the levels of CRKL, a protein that can form functional heterodimers with STAT5,20 were also downregulated in Cyba-/- mice (Figure 8E). Finally, the levels of MYC, another transcription factor important for hematopoiesis,21 were also upregulated in Cyba-/- mice (Figure 8H).

Discussion There is accumulating evidence supporting the importance of ROS and redox signaling in hematopoietic differentiation.3 NADPH oxidases would be ideal candidates as an adjustable source of ROS during hematopoiesis, given that they can be activated by extracellular signals, including hematopoietic cytokines.22–25 Our previous results support this working hypothesis, since we have shown the relevance of ROS production by NADPH oxidases for regulating in vitro megakaryocytic differentiation.12 Therefore, we questioned whether NADPH oxidase activity would be required for in vivo hematopoiesis. The difficulty of this study lies in the fact that many cells express several NADPH oxidases simultaneously. Mouse hematopoietic progenitors express three different p22phoxdependent NOX (NOX1, NOX2 and NOX4), therefore, we started by generating Cyba-/- mice. Flow cytometry analyses supported the existence of a myeloid bias in ES Cyba-/- mice, which could also be observed in the CR model, though at a weaker level. Nevertheless, we found the upregulation of myeloid genes in Lin– cells from CR Cyba-/- mice, also supporting the existence of a myeloid bias in the CR Cyba-/- mice. Although the genetic background of the ES and CR models (C57BL/6N and C57BL/6J respectively) are very closely related,26 we cannot rule out that this might account for the differences regarding the myeloid bias. Moreover, the weaker phenotype in the CR Cyba-/- mice could be in line with phenomena associated with the CRISPR/Cas9 tool, such as the activation of genetic compensation mechanisms,27 and the residual expression of the targeted protein.28 A remarkable increase in HSC populations is observed in Cyba-/- mice, which correlates with a higher cellular proliferation capacity, as shown by the in vivo BrdU incorporation assays. It has been suggested before that MYC upregulation can lead to the expansion of HSPC,21 so the increased expression of MYC in BM Cyba-/- mice could explain this effect in our cells. Thrombopoietin (TPO) has recently been identified as a crucial cytokine for the control of HSC quiescence.29 TPO signaling depends on the activation of STAT5.30 In turn, the importance of STAT5 for maintaining HSC quiescence has also been reported.31 Considering the dramatic decrease in STAT5 protein levels in Cyba-/- mice, hampered TPO signaling as a consequence of diminished STAT5 activation could shift the balance 151

R. Prieto-Bermejo et al.

from quiescence towards an increase in cell proliferation. In line with this hypothesis NOX2 and NOX4 ROS production seems to be required for maintaining the stemness of induced pluripotent stem cells.32 We have addressed the relevance of NOX isoforms for hematopoiesis through competitive transplantation experiments. Cells lacking NOX1 (Nox1-/-) and NOX2 (Cybb-/-) had a higher level of hematopoietic reconstitution with respect to wild-type cells. The hematopoietic reconstitution ability of Cybb-/-cells has been reported through competitive transplant experiments, showing an enhanced hematopoietic reconstitution of adult BM Cybb-/- cells at the beginning of the experiment, which is lost at later time points.33 Our results would be similar to those reported by Weisser et al.33 since at short time points we also found a slight reconstitution advantage in Cybb-/- cells, though in our case said advantage is kept throughout time in the primary transplant. Moreover, we performed secondary transplants that confirmed the greater reconstitution ability of not only Cybb-/- cells, but also of Nox1-/- and Cyba-/cells. These results support those obtained by us in primary transplants. While Weisser et al.33 used a limited number of purified HSC that were modified by lentivirus infection before the transplant, we used 3x106 whole BM cells for the transplant. We wonder whether the different progenitor cell numbers or the presence of more mature progenitors in our assays could be responsible for the sustained advantage of reconstitution observed. Another interesting area may be the upregulation of IL-1b signaling caused by the hyperinflammation linked to the CGD. According to the data presented by Weisser et al.33 IL-1b signaling can lead to the impaired reconstitution capacity of the HSC. The group comments in their methods section that animals with overt infections were not included in their study, which suggests they may have used animals with some kind of inflammatory pathology. In our case, none of the animals used showed any sign of inflammation or disease, so it is likely that a different level of IL-1 signaling could explain the differences in our results. While preparing our manuscript, another article was published reporting a lower reconstitution ability for Cybb-/cells in primary transplants.34 Adane et al.34 also used whole BM for their assays, but again a more limited cell number was used, and unlike us, their findings were not confirmed by secondary transplants. Together these results show a certain variability in the outcome of these assays, which is also supported by the fact that Weisser et al.33 did not find any difference in the transplants when using Cybb-/- fetal liver cells.33 Despite these differences, Weisser et al.33 report that Cybb-/- mice show an increased number of HPC, with a higher proliferation rate. They also show an increase in the formation of myeloid CFU by Cybb-/- BM cells. All these results are in line with our findings for Cyba-/- mice, and combined would support the importance of NOX2 and p22phox for the control of HSPC function. Nox4-/- transplanted cells showed certain lineage alterations, similar to those observed in the Cyba-/- mice. All this considered, we hypothesize that NOX1 and NOX2 could be involved in the regulation of HSC homeostasis, while NOX4 might be important for lineage decisions given its implication in cell differentiation.2 The lack of phagocyte oxidase activity in Cyba-/- mice would lead to a defective immune response, as suggested 152

by analysis of our RNA-seq results. Moreover, alterations of the NOD-like receptor signaling pathway, involved in the recognition of pathogen derived patterns,35 or the alteration of antigen processing and presentation pathways36 suggests that adaptive immunity could also be compromised in Cyba-/- mice. Additionally, we detected an exacerbated production of immunoglobulins in Cyba-/- mice, in agreement with the high level of IgG in CGD human subjects.37 An interesting question is the molecular mechanism leading to this scenario. IL-7 signaling drives B cell lymphopoiesis in mice.38 STAT5 and the PI3K/AKT pathway are the downstream effectors of IL-7 during B cell development.39 The lack of Stat5a and Stat5b blocks B cell development.19 STAT5 is not only required for maintaining cell survival during B-cell differentiation,40 but also for regulating Ig rearrangement,41 so much so that IL-7 signaling prevents premature Igk rearrangement through STAT5 activation.19 Moreover, there are studies suggesting that STAT5b deficiency in humans leads to an increased production of immunoglobulin E42 and G43. Therefore, we hypothesize that STAT5 downregulation and the reduced response to IL-7 must be key factors for the upregulation of immunoglobulins observed in Cyba-/- mice. Analyses of STAT5 protein stability in the literature are rather scarce; STAT5 can be degraded by calpain44 and caspase-3.45 Moreover, it has also been shown that the redox sensitive phosphatase DUSP4 can trigger STAT5 protein degradation through proteasome and lysosome pathways.46 In line with that, it could be hypothesized that a lower level of ROS in Cyba-/- mice may allow a higher level of DUSP4 activity, thus leading to STAT5 protein degradation. In summary, by downregulating NADPH oxidase activity in hematopoietic cells through the deletion of the Cyba gene, we show the relevance of such activity for in vivo hematopoiesis. The lack of Cyba induces a myeloid bias and promotes the enrichment of HSC populations, together with an increased proliferation potential. Moreover, the lack of p22phox hinders the activation of the PI3K-AKT pathway, and induces STAT5 downregulation. This would jeopardize the activation of IL-7 signaling, which would explain the increased immunoglobulin production in Cyba-/- mice. Disclosures No conflicts of interest to disclose Contributions RPB performed experiments, analyzed data, assembled figures, revised the manuscript. MRG and APF performed experiments and revised the manuscript. IGT and MSM generated the Cyba-/- mice. AHH conceived and designed experiments, performed experiments, analyzed data and wrote the manuscript. Acknowledgments We thank L. Stockdale for reviewing the English version of this manuscript. Funding This work was funded by the Spanish Ministry of Economy and Competitiveness (MINECO) (BFU2014-56490-R) and Ramรณn Areces Foundation (CIV17A2822). RPB, MRG and APF, were recipients of pre-doctoral fellowships from the Regional Government of Castilla and Leon, Spain and ERDF funds. haematologica | 2021; 106(1)

NADPH oxidases in hematopoiesis

References 1. Prieto-Bermejo R, Hernández-Hernández Á. The importance of NADPH oxidases and redox signaling in angiogenesis. Antioxidants (Basel). 2017;6(2):32. 2. Sardina JL, Lopez-Ruano G, SanchezSanchez B, Llanillo M, HernándezHernández Á. Reactive oxygen species: are they important for haematopoiesis? Crit Rev Oncol Hematol. 2012;81(3):257-274. 3. Prieto-Bermejo R, Romo-González M, PérezFernández A, Ijurko C, HernándezHernández Á. Reactive oxygen species in haematopoiesis: leukaemic cells take a walk on the wild side. J Exp Clin Cancer Res. 2018;37(1):1-18. 4. Bedard K, Krause K-H. The NOX family of ROS-generating NADPH oxidases: physiology and pathophysiology. Physiol Rev. 2007;87(1):245-313. 5. Nauseef WM. Biological roles for the NOX family NADPH oxidases. J Biol Chem. 2008;283(25):16961-16965. 6. Battersby AC, Cale AM, Goldblatt D, Gennery AR. Clinical manifestations of disease in X-linked carriers of chronic granulomatous disease. J Clin Immunol. 2013; 33(1573-2592):1276-1284. 7. Suh Y-A, Arnold RS, Lassegue B, et al. Cell transformation by the superoxide-generating oxidase Mox1. Nature. 1999;401 (6748):7982. 8. Rinnerthaler M, Buttner S, Laun P, et al. Yno1p/Aim14p, a NADPH-oxidase ortholog, controls extramitochondrial reactive oxygen species generation, apoptosis, and actin cable formation in yeast. Proc Natl Acad Sci. 2012;109(22):8658-8663. 9. Kubli DA, Sussman MA. Eat, breathe, ROS: controlling stem cell fate through metabolism. Expert Rev Cardiovasc Ther. 2017; 15(1477-9072):345-356. 10. Maryanovich M, Gross A. A ROS rheostat for cell fate regulation. Trends Cell Biol. 2013;23(3):129-134. 11. Jung H, Choi I. Thioredoxin-interacting protein, hematopoietic stem cells, and hematopoiesis. Curr Opin Hematol. 2014; 21(4):265-270. 12. Sardina JL, López-Ruano G, Sánchez-Abarca LI, et al. p22phox-dependent NADPH oxidase activity is required for megakaryocytic differentiation. Cell Death Differ. 2010;17(12):1842-1854. 13. Sanchez-Sanchez B, Gutierrez-Herrero S, Lopez-Ruano G, et al. NADPH oxidases as therapeutic targets in chronic myeloid leukemia. Clin Cancer Res. 2014; 20(15):4014-4025. 14. Yamazaki T, Kawai C, Yamauchi A, Kuribayashi F. A highly sensitive chemiluminescence assay for superoxide detection and chronic granulomatous disease diagnosis. Trop Med Health. 2011;39(2):41-45. 15. López-Ruano G, Prieto-Bermejo R, Ramos TL, et al. PTPN13 and b-catenin regulate the quiescence of hematopoietic stem cells and their interaction with the bone marrow niche. Stem Cell Rep. 2015;5(4):516-531. 16. Sardina JL, López-Ruano G, Prieto-Bermejo

haematologica | 2021; 106(1)

R, et al. PTPN13 regulates cellular signalling and b-catenin function during megakaryocytic differentiation. Biochim Biophys Acta Mol Cell Res. 2014;1843(12):2886-2899. 17. Nakano Y, Longo-Guess CM, Bergstrom DE, Nauseef WM, Jones SM, Bánfi B. Mutation of the Cyba gene encoding p22phox causes vestibular and immune defects in mice. J Clin Invest. 2008;118(3):1176-1185. 18. Paffenholz R, Bergstrom RA, Passutto F, et al. Vestibular defects in head-tilt mice result from mutations in Nox3, encoding an NADPH oxidase. Genes Dev. 2004; 18(5):486-491. 19. Clark MR, Mandal M, Ochiai K, Singh H. Orchestrating B cell lymphopoiesis through interplay of IL-7 receptor and pre-B cell receptor signalling. Nat Rev Immunol. 2014; 14(2):69-80. 20. Brierley MM, Fish EN. Stats: multifaceted regulators of transcription. J Interf Cytokine Res. 2006;25(12):733-744. 21. Wilson A, Murphy MJ, Oskarsson T, et al. cMyc controls the balance between hematopoietic stem cell self-renewal and differentiation. Genes Dev. 2004; 18(22): 27472763. 22. Huang H, Kim HJ, Chang EJ, et al. IL-17 stimulates the proliferation and differentiation of human mesenchymal stem cells: implications for bone remodeling. Cell Death Differ. 2009;16(1476-5403):1332-1343. 23. Sharma P, Chakraborty R, Wang L, et al. Redox regulation of interleukin-4 signaling. Immunity. 2008;29(4):551-564. 24. Hurtado-Nedelec M, Csillag-Grange MJ, Boussetta T, et al. Increased reactive oxygen species production and p47phox phosphorylation in neutrophils from myeloproliferative disorders patients with JAK2 (V617F) mutation. Haematologica. 2013;98(15928721):1517-1524. 25. Zhu QS, Xia L, Mills GB, Lowell CA, Touw IP, Corey SJ. G-CSF induced reactive oxygen species involves Lyn-PI3-kinase-Akt and contributes to myeloid cell growth. Blood. 2006;107(0006-4971):1847-1856. 26. Pettitt SJ, Liang Q, Rairdan XY, et al. Agouti C57BL/6N embryonic stem cells for mouse genetic resources. Nat Methods. 2009; 6(7):493-495. 27. Ma Z, Zhu P, Shi H, et al. PTC-bearing mRNA elicits a genetic compensation response via Upf3a and COMPASS components. Nature. 2019;568(7751):259-263. 28. Smits AH, Ziebell F, Joberty G, et al. Biological plasticity rescues target activity in CRISPR knockouts. bioRxiv. 2019;16(11): 1087-1093. 29. de Graaf CA, Metcalf D. Thrombopoietin and hematopoietic stem cells. Cell Cycle. 2011;10(10):1582-1589. 30. Drayer AL, Boer A-K, Los EL, Esselink MT, Vellenga E. Stem cell factor synergistically enhances thrombopoietin-induced STAT5 signaling in megakaryocyte progenitors through JAK2 and Src Kinase. Stem Cells. 2005;23(2):240-251. 31. Schepers H, Wierenga ATJ, Vellenga E, Schuringa JJ. STAT5-mediated self-renewal of normal hematopoietic and leukemic stem cells. Jak-Stat 2012;1(1):13-25.

32. Kang X, Wei X, Jiang L, et al. Nox2 and Nox4 regulate self-renewal of murine inducedpluripotent stem cells. IUBMB Life. 2016; 68(12):963-970. 33. Weisser M, Demel UM, Stein S, et al. Hyperinflammation in patients with chronic granulomatous disease leads to impairment of hematopoietic stem cell functions. J Allergy Clin Immunol. 2016;138(1):219228.e9. 34. Adane B, Ye H, Khan N, et al. The hematopoietic oxidase NOX2 regulates selfrenewal of leukemic stem cells. Cell Rep. 2019;27(1):238-254.e6. 35. Kapetanovic R, Cavaillon JM. Early events in innate immunity in the recognition of microbial pathogens. Expert Opin Biol Ther. 2007;7(6):907-918. 36. Murat P, Tellam J. Effects of messenger RNA structure and other translational control mechanisms on major histocompatibility complex-I mediated antigen presentation. Wiley Interdiscip Rev RNA. 2015;6(2):157171. 37. Wu J, Wang WF, Zhang YD, Chen TX. Clinical features and genetic analysis of 48 patients with chronic granulomatous disease in a single center study from Shanghai, China (2005-2015): new Studies and a literature review. J Immunol Res. 2017;2017: 8745254. 38. Petkau G, Turner M. Signalling circuits that direct early B-cell development. Biochem J 2019;476(5):769-778. 39. Hamel KM, Mandal M, Karki S, Clark MR. Balancing proliferation with Igκ recombination during B-lymphopoiesis. Front Immunol. 2014;5:139. 40. Malin S, McManus S, Cobaleda C, et al. Role of STAT5 in controlling cell survival and immunoglobulin gene recombination during pro-B cell development. Nat Immunol. 2010;11(2):171-179. 41. Heltemes-Harris LM, Farrar MA. The role of STAT5 in lymphocyte development and transformation. Curr Opin Immunol. 2012; 24(2):146-152. 42. Klammt J, Neumann D, Gevers EF, et al. Dominant-negative STAT5B mutations cause growth hormone insensitivity with short stature and mild immune dysregulation. Nat Commun. 2018;9(1):2105. 43. Nadeau K, Hwa V, Rosenfeld RG. STAT5b deficiency: an unsuspected cause of growth failure, immunodeficiency, and severe pulmonary disease. J Pediatr. 2011;158(5):701708. 44. Oda A, Wakao H, Fujita H. Calpain is a signal transducer and activator of transcription (STAT) 3 and STAT5 protease. Blood. 2002; 99(5):1850-1852. 45. Pietschmann K, Bolck HA, Buchwald M, et al. Breakdown of the FLT3-ITD/STAT5 axis and synergistic apoptosis induction by the histone deacetylase inhibitor panobinostat and FLT3-specific inhibitors. Mol Cancer Ther. 2012;11(11):2373-2383. 46. Hsiao W-Y, Lin Y-C, Liao F-H, Chan Y-C, Huang C-Y. Dual-specificity phosphatase 4 regulates STAT5 protein stability and helper T cell polarization. PLoS One. 2015;10(12): e0145880.

153

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):154-162

Hodgkin Lymphoma

Targeted genotyping of circulating tumor DNA for classical Hodgkin lymphoma monitoring: a prospective study Vincent Camus,1,2 Mathieu Viennot,2 Justine Lequesne,3 Pierre-Julien Viailly,2 Elodie Bohers,2 Lucile Bessi,2 Bénédicte Marcq,1,2 Pascaline Etancelin,2,4 Sydney Dubois,1,2 Jean Michel Picquenot,2,5 Elena-Liana Veresezan,2,5 Marie Cornic,3 Lucie Burel,3 Justine Loret,3 Stéphanie Becker,6 Pierre Decazes,6 Pascal Lenain,1 Stéphane Lepretre,1,2 Emilie Lemasle,1 Hélène Lanic,1 Anne-Lise Ménard,1 Nathalie Contentin,1 Hervé Tilly,1,2 Aspasia Stamatoullas1,2 and Fabrice Jardin1,2

1 Department of Hematology, Center Henri Becquerel; 2University of Rouen, INSERM U1245, Center Henri Becquerel; 3Clinical Research Unit, Centre Henri Becquerel; 4 Department of Genetic Oncology, Center Henri Becquerel; 5Department of Pathology, Center Henri Becquerel and 6Department of Nuclear Medicine and Radiology, Center Henri Becquerel and QuantIF (Litis EA4108 – FR CNRS 3638), Rouen, France

ABSTRACT

Correspondence: VINCENT CAMUS vincent.camus@chb.unicancer.fr Received: October 16, 2019. Accepted: February 12, 2020. Pre-published: February 20, 2020.

he relevance of circulating tumor DNA (ctDNA) analysis as a liquid biopsy and minimal residual disease tool in the management of classical Hodgkin lymphoma (cHL) patients was demonstrated in retrospective settings and remains to be confirmed in a prospective setting. We developed a targeted Next-Generation sequencing (NGS) panel for fast analysis (AmpliSeq® technology) of nine commonly mutated genes in biopies and ctDNA of cHL patients. We then conducted a prospective trial to assess ctDNA follow-up at diagnosis and after two cycles (C2) of chemotherapy. Sixty cHL patients treated by first line conventional chemotherapy (BEACOPPescalated [21.3%], ABVD/ABVD-like [73.5%] and other regimens [5.2%, for elderly patients]) were assessed in this noninterventional study. The median age of the patients was 33.5 years (range: 20-86). Variants were identified in 42 (70%) patients. Mutations of NFKBIE, TNFAIP3, STAT6, PTPN1, B2M, XPO1, ITPKB, GNA13 and SOCS1 were found in 13.3%, 31.7%, 23.3%, 5%, 33.3%, 10%, 23.3%, 13.3% and 50% of patients, respectively. ctDNA concentration and genotype were correlated with clinical characteristics and presentation. Regarding early therapeutic response, 45 patients (83%, not available [NA] =6) had a negative positron emission tomography (PET) after C2 (Deauville Score 1-3). The mean of DeltaSUV after C2 was -78.8%. ctDNA after C2 was analysed in 54 patients (90%). ctDNA became rapidly undetectable in all cases after C2. Variant detection in ctDNA is suitable to depict the genetic features of cHL at diagnosis and may help to assess early treatment response, in association with PET. Clinical Trial reference: NCT02815137. max

https://doi.org/10.3324/haematol.2019.237719

154

Introduction Classical Hodgkin lymphoma (cHL) patients have greatly benefited from multiagent chemotherapy and improved radiation techniques and 65-90% of patients can achieve disease-free survival after 5 years, depending on stage and clinical risk factors.1 Those with a rapid response to initial treatment have the best outcomes and may benefit from truncated, less-toxic treatment regimens.2 Nevertheless, approximately 2025% of patients will ultimately experience either primary refractoriness to chemotherapy (within 3 months of doxorubicin-based chemotherapy), early disease relapse (within 12 months after the end of first-line treatment) or late disease relapse,3 underlying the need to understand the mechanisms involved and to identify predictive biomarkers. Until recent years, the genetic overview of cHL was poorly described because of the scarcity of Hodgkin and Reed-Sternberg (HRS) tumor cells, which usually account for haematologica | 2021; 106(1)

Liquid biopsy for classical Hodgkin Lymphoma

only about 0.1-3% of cells in the tissue,4 complicating biological analyzes that lacked sensitivity. Due to the scarcity of HRS cells in cHL patient biopsies, finding recurrent mutations could be easier in the plasma circulating tumor DNA (ctDNA) of these patients, with potentially fewer heterogeneity issues than tumor tissue testing. The concept of "liquid biopsy" has recently been highlighted in a series of cHL patients,6 for whom high-throughput sequencing (HTS) of a panel of target genes was performed, with successful detection of soma-tic variants both in the tumor and in the plasma. In a previous retrospective study, we investigated the prevalence and clinical relevance of XPO1 E571K in cHL cases7 and demonstrated that this recurrent mutation was detectable in both tumor and plasma, indicating that XPO1 mutations represent a new genetic biomarker, useful both at the time of diagnosis and as a minimal residual disease (MRD) marker. In order to confirm the results of this previous work, we have developed a prospective observational study (NCT02815137) with a focused and limited array of genes to implement liquid biopsy at different times throughout the standard initial cHL management and to identify in this larger number of patients at least one molecular abnormality with a role in oncogenesis or of prognostic or therapeutic impact that could be used as a biomarker of follow-up during treatment. Our Next-Generation sequencing (NGS) and digital droplet PCR (ddPCR) panel was designed after analysis of literature data on recurrent somatic mutations described in cHL and the validity of this panel was established on a series of 24 patients with cHL in a previous publication.8 A comparison with previous results recently published by Spina et al.6 showed very similar rates of mutations detected in ctDNA targeting STAT6, B2M, XPO1, NFKBIE and TNFAIP3. The goal of the present work was to establish the feasibility and validity of prospective ctDNA monitoring, in order to propose new strategies for both tailored diagnosis and treatment based on the simple detection and quantification of acquired mutations in the plasma of cHL patients.

Methods Patients For exploratory and feasibility purposes, we prospectively included in this biological study adult patients treated by adriamycin (doxorubicin), bleomycin, vinblastine, dacarbazine (ABVD) or BEACOPPescalated for newly diagnosed stage I-IV cHL at the Henri Becquerel Center (Rouen, France) between 2016 and 2018. According to these inclusion criteria, 64 patients were initially included but four were then excluded (two histological diagnostic errors, one underage patient, one loss to follow-up just after inclusion) leading to a final number of 60 patients analyzed. The exclusion criteria were: patients who had i) already started any treatment (including steroids) before signing the informed consent, ii) contraindication to positron emission tomography (PET), iii) positive HIV, hepatitis B or C serology, and iv) pregnant women. As this was an observational biological study, physicians were blinded to the results of the molecular analysis and these did not alter the routine care of the patients. All experiments were performed in accordance with the Helsinki Declaration and the study was approved by an institutional review board and ethics committee (N° ID-RCB: 2016A00202-49). All patients gave informed consent for specimen collection, clinical data collection and biomarker analysis. Patients were treated according to routine local recommendations for the haematologica | 2021; 106(1)

management of patients with a diagnosis of de novo cHL.

Blood specimens Blood samples were obtained by blood draws at diagnosis and after two cycles (C2) of chemotherapy, and another blood sample was obtained in case of disease relapse. Eighteen mL of blood samples were collected in EDTA tubes at the scheduled timepoints during patient management. Samples were centrifuged within 2 hours for 10 minutes at 2,600 g (3,500 rpm) at 4°C to isolate the plasma, which was then stored at -80°C.

NGS and statistical analysis NGS of targeted genes was performed on an Ion Torrent Personal Genome Machine™ (PGM, Thermo Fisher Scientific). Our panel targets the four basepair hotspot deletion in NFKBIE (exon 1) and the coding regions of the following genes: ITPKB (exons 2 to 8), PTPN1 (exons 1 to 10), TNFAIP3 (exons 2 to 9), SOCS1 (exon 2), STAT6 (exons 12 and 14), B2M (exons 1 to 3), XPO1 (exons 15 to 18) and GNA13 (exons 1 to 4). Based on a theoretical sensitivity of 1%, and on a minimum number of mutated reads equal to 50, the minimum depth of desired sequencing was set at 5,000X. The circulating ctDNA concentrations were expressed in haploid genome equivalents (hGE) per mL of plasma (hGE/mL) and calculated by multiplying the mean variant allelic frequency (VAF) for all mutations used for detection calling by the concentration of cell-free DNA (cfDNA) (pg/mL of plasma) and dividing by 3.3, using the assumption that each haploid genomic equivalent weighs 3.3 pg, as previously described by Scherer et al.9 Taking into account previous experiences in our laboratory,8,10 we retained a threshold of 0.5% of VAF as the lower detection limit for the present study. Statistics were performed with R software v3.3.2 (see the Online Supplementary Materials and Methods).

Results Somatic mutations in cHL at the time of initial diagnosis The main clinical features, disease characteristics, and first line treatments of the 60 patients are summarized in Table 1. Briefly, the median age was 33.5 years old, 91.4% of the patients had an excellent performance status (ECOG 0-1) and the majority of patients (55%) displayed B symptoms. The predominance of the scleronodular subtype was confirmed (70%) and we observed a predominance of localized stage I-II disease (51.7%). Regarding disease extension, the bone marrow biopsy, when performed, was normal in the majority of cases (30 of 37 patients, 81.1%). The presence of a bulky mass greater than 10 centimeters (cm) concerned 15% of the patients. Table 2 presents the somatic variants detected by HTS sequencing of biopsy and plasma ctDNA. Patients were considered mutated if they had a mutation in their biopsy-extracted genomic DNA and/or plasma ctDNA. Variants were found in 42 (70%) patients: 21 of 30 (71%) and 41 of 60 (68.8%) of available biopsy and ctDNA samples respectively. In all, 145 different variants were identified: (i) 127 variants in plasma samples, of which were 66 not found in the corresponding biopsies; (ii) 79 in the biopsy samples, of which 18 variants were not found in the corresponding plasma samples. The median VAF were superior in ctDNA than in biopsy (1.99% vs. 1.6%, P=0.024). Concordance between genetic profiles of biopsy and ctDNA (in terms of double-positivity or double-negativity of samples) was acceptable for 25 of 30 patients (83.3%, Cohen’s k =0.56 [range: 0.23-0.89]). This indicates that 20 155

V. Camus et al.

patients with a positive genetic biopsy profile also presented the same genetic ctDNA profile, and five patients with a negative genetic biopsy profile also presented no mutations in ctDNA, leading to 25 patients with concordance. More precisely, similarity at the level of the variants in paired biopsy and plasma samples was found for 61 variants (variant identified both in the plasma and in the biopsy of the same patient). Significant higher plasma ctDNA concentration levels at diagnosis was associated with the following unfavorable clinical characteristics: sedimentation rate ≥50 mm, albumin ≤40 g/L, advanced stage disease, lymphopenia <0.6 G/L, presence of B symptoms; Hasenclever International Prognostic Score ≥3, ≥ 4 involved nodal areas, presence of anemia (hemoglobin level <10.5 g/dL) and elevated lactate dehydrogenase (LDH) (>480 UI/L) (Table 3). Mutations of NFKBIE, TNFAIP3, STAT6, PTPN1, B2M, XPO1, ITPKB, GNA13 and SOCS1 were respectively found in the following proportions of patients (mean number of variants by patient [range]): 13.3% (1.1 [1-2]), 31.7% (1.1 [1-2]), 23.3% (1.4 [1-2]), 5% (1.3 [1-2]), 33.3% (1.3 [1-3]), 10% (1 [1-1]), 23.3% (1.5 [1-4]), 13.3% (1.1 [1-2]) and 50% (1.9 [1-5]). As a median value, four somatic variants per mutated patient (range: 1-12) were identified. The entire coding sequence of SOCS1 was sequenced and was found to be the most frequently mutated gene in this study with 55 somatic variants identified in gDNA and/or ctDNA of 30 patients (50%), some thus presenting two (or more) somatic variants but there is no recurrent variant identified for this gene. Figure 1A shows the number and types of somatic variants identified by gene and patient both in biopsy and ctDNA. In exons 1 to 3 of the B2M gene, 23 somatic variants were found in 20 patients (33.3% of patients). STAT6 is frequently mutated with 10 somatic variants identified in 14 patients (23.3% of patients). These variants are not mutually exclusive and can present on the same allele when the patient is double-mutated (five double-mutated patients). Among these mutations, the N417Y mutation is the most frequent, present in six patients (10%). The coding sequence of the TNFAIP3 gene was sequenced in its entirety, which allowed the identification of 21 somatic variants in 19 patients (31.7% of patients). Two different variants were identified on exons 15 to 18 of the XPO1 gene, in six patients (10%). Unsupervised hierarchical clustering was performed among the nine genes to represent the association of alterations (Figure 1B). For 8 of the 20 patients presenting similar mutations in both biopsy and plasma samples, the variants identified were strictly identical. For the 12 others, there were discrepancies (somatic variants common between the two tissues, with additional variants in one or both of the tissues). For these 20 patients, the median VAF in the biopsy and the matched plasma were equal to 1.61% and 2.3% respectively (P=0.04). Mean baseline ctDNA concentration was 323.3 hGE/mL (range: 0-2684.9) at diagnosis. Figure 3A shows the distribution of ctDNA median VAF for the 41 mutated patients with “positive plasma” at diagnosis.

PET measurements, therapeutic response and survival analysis There was a moderate positive Spearman’s correlation between metabolic tumor volume (MTV) and ctDNA concentration (r=0.47, P<0.001) and between MTV and ctDNA median VAF (r=0.57, P<0.001) for the 41 “positive” plasma 156

samples at diagnosis. Fifty-four interim PET exams were performed after C2: 45 patients displayed a complete metabolic response and nine patients displayed a partial metabolic response (Deauville Scale [DS]).4 Among these nine patients, eight were in complete remission at end of treatTable 1. Patient characteristics.

All (N=60) Age > 60 years 8 (13.3%) Median age, years (range) 33.5 (20-86) Male 32 (53.3%) BMI <18.5 4 (6.7%) 18.5-25 31 (51.7%) ≥25 25 (41.6%) Stage III/IV (vs. I/II) 29 (48.3%) LDH > ULN 11 (18.6%, NA=1) B symptoms 33 (55%) ECOG PS ≥2 5 (8.6%, NA=2) Mediastinal mass ratio >0.33 13 (52%, NA=35) Smokers 36 (63.2%, NA = 3) Diagnostic biopsy mediastinal 10 (23.3%) versus extra-mediastinal Sclero-nodular cHL (vs. other subtypes) 42 (70%) Bone marrow involvment 5 (13.5%, NA = 23) EORTC favorable stage I-II versus unfavorable 4 (6.7%) IPS (Hasenclever) 0-2 (vs. 3-7) 39 (65%) EBV positive disease 17 (31.5%, NA=6) Bulky disease (≥10 cm) 9 (15%) Splenomegaly 6 (10.3%, NA=2) Median number of involved nodal areas (range) 3 (0-5) Treatment 2 ABVD/ABVD-like 1 (1.7%) 2 ABVD/ABVD-like + IFRT 6 (10%) 2 ABVD/ABVD-like + 2 BEACOPPescalated + IFRT 1 (1.7%) 3 ABVD/ABVD-like 1 (1.7%) 3 ABVD/ABVD-like + IFRT 1 (1.7%) 4 ABVD/ABVD-like + IFRT 19 (31.7%) 6-8 ABVD/ABVD-like 3 (5%) AHL 2011 based strategy 23 (38.3%) Other Regimen 4 (6.7%) Other Regimen + IFRT 1 (1.7%) IFRT (n=28) 20 Gy 6 (21.4%) 30 Gy 21 (75%) 36 Gy 1 (3.6%) DS1-3 after C2 versus DS 4-5 46 (85.2%, NA=6) Complete metabolic response versus partial response after C245 (83.3%, NA=6). BMI: body mass index; IPI: international prognostic index; NA: not available; aaIPI: age adjusted IPI; cHL: classical Hodgkin lymphoma; IFRT: involved field radiotherapy; Gy: Grays; DS: Deauville score; C2: two cycles of chemotherapy; IPS: international prognostic score. LDH: lactate dehydrogenase; ULN: upper limit of normal; EBV: Eppstein-Barr virus; ECOG: Eastern Cooperative Oncology Group; EORT: European Organisation for Research and Treatment of Cancer; ABVD: adriamycin (doxorubicin), bleomycin, vinblastine, dacarbazine. AHL 2011 based strategy BEACOPPescalated C2 and then PET performed after C2 was used to guide subsequent treatment: the patient received four additional cycles of ABVD for patients with negative PET2, and two additional cycles of BEACOPPescalated for patients with positive PET2 and then another PET was performed after C4. If patients had a negative PET4, they received two additional cycles of BEACOPPescalated; if patients had a positive PET, they received salvage treatment.

haematologica | 2021; 106(1)

Liquid biopsy for classical Hodgkin Lymphoma

Figure 1. Number and types of somatic variants identified by gene and patient both in biopsy and circulating tumor DNA. (A) Heatmap representing somatic variants detected in circulating tumor DNA (ctDNA) in the mutated patients (the "Burden" track represents the mutational load [â&#x20AC;&#x153;burdenâ&#x20AC;?] of a patient based on the maximum/minimum number of mutations found in the complete cohort). (B) Unsupervised hierarchical clustering performed among the nine genes to represent the association of alterations (numbers within the clustering represent the number of observed associations between two genes).

ment and one elderly patient ( universal patient identification number [UPN] 39), who had no detectable mutations either at diagnosis or after C2, relapsed and died from disease progression 1 year after diagnosis. Patient UPN43 had no blood collection after C2 (organizational issue) and was in complete remission at end of treatment. Patient UPN44 had no interim PET scan performed after C2 because the patient was scheduled to receive C8 of ABVD and PET was performed after C4 (metabolic complete response) but we do not have any blood collection after C4. The metabolic complete remission rate (DS 1-3) after C2 of chemotherapy and at the end of treatment were 85.2% (not done [ND] =6) and 86% [ND=17]), respectively. The mean of DeltaSUV after C2 was -78.8%. Two patients (UPN21 and UPN58) died after the C1 of chemotherapy and two patients experienced an early relapse. The median follow-up of patients was 22.7 months (range: 11.4-38.9 max

haematologica | 2021; 106(1)

Table 2. Synthesis of mutational profiles at diagnosis of the patient cohort.

Gene

Number of mutated patients by gene (%)

Recurrent variants

ITPKB GNA13 SOCS1 PTPN1 TNFAIP3 XPO1 STAT6

14 (23.3) 8 (13.3) 31 (51.7) 3 (5) 19 (31.7) 6 (10) 14 (23.3)

NFKBIE B2M

8 (13.3) 20 (33.3)

NO NO NO NO NO 5x E571K, 1x E571G 6x N417Y, 2x N417S; 2x D419Y, 3x D419G, 8x Y254fs 2x M1T, 2x M1K, 2x L7X

A patient was considered mutated if the variant was found in the biopsy genomic DNA (gDNA) and/or plasma circulating tumor DNA (ctDNA).

157

V. Camus et al.

Figure 2. Prevalence of somatic mutations detected in DNA extracted from the 31 available biopsies and 60 plasma samples of the patients at time of diagnosis. The "genomic DNA (gDNA) and/or circulating tumor DNA (ctDNA) " column corresponds to the patients considered mutated in the biopsy and/or the plasma samples.

months). Finally, 1-year overall survival and progression-free survival rates were 94.9% (range: 89.3-100%) and 89.8 % (range: 82.4-97.9%) respectively.

ctDNA as a biomarker for therapeutic response assessment in addition to PET In this ongoing trial, plasma ctDNA after C2 of chemotherapy has only been analyzed for 55 of 60 patients, of which 18 presented no somatic mutations at diagnosis in plasma samples and were therefore not informative for therapeutic molecular response assessment after C2. The remaining five non-analyzed patients had either no available plasma sample after C2 (two patients died before C2, one patient with loss of follow-up and two patients without performed blood collection). Figure 4 schematically shows the somatic mutation VAF variation in the plasma ctDNA of the 41 mutated evaluable patients between diagnosis and after C2 of chemotherapy. No somatic variant was identified in the blood draw performed concomitantly to interim PET after C2 of chemotherapy in any patient, including patient UPN17 who displayed negative PET after C2 (DS 2) but who was then refractory at end of first line treatment (DS 4). In this patient, the plasma ctDNA was positive for ITPKB p.K232FS mutation at diagnosis, negative after C2 and again positive at relapse at end of first line treatment (2x BEACOPPescalated, 4x ABVD) with VAF of 10.6%, 0% and 2.58%, respectively. The kinetics of ctDNA results were concordant with the PET results for this patient. In the whole cohort, we found no statistically significant difference between the concentration of cfDNA (ng/mL of plasma) after C2 among DS 1-3 patients (35 patients, median 35 ng/mL [range: 20.4-260]) versus DS 4-5 patients (seven patients, median 36.2 ng/mL [range: 21.8-80], P=0.79), which suggests a possible nontumor DNA plasma release by inflammatory mechanisms, regardless of the persistence of an active tumor disease. In addition, Figure 5 illustrates the case of another patient 158

(UPN38) presenting a positive PET after C2 of ABVD with a residual fixation (DS 5) of a left supraclavi-cular lymph node which was surgically removed and corresponded to a reactive lymph node. Ten identical somatic variants had been identified in this patient at diagnosis in both biopsy and plasma samples. For this patient, we were unable to analyze the second biopsy that was performed after C2 of ABVD due to insufficient materiel. However, this patient no longer had a detectable somatic mutation in the plasma after C2 with a clearance of all baseline mutations that were present at diagnosis. The amount of ctDNA was 1434.85 hGE/mL at diagnosis and after C2, the level of ctDNA was measured at 42 ng/mL of plasma. Taken together, these results argue for a false posi-tive result of PET after C2 with a consistent negativity of concomitant ctDNA sample. The patient received irradiation of the left supraclavicular lymph node and is conside-red in complete remission.

Discussion Herein we confirm the feasibility of non-invasive somatic mutation analysis in a prospective and indepen-dent setting in a large cohort of patients with cHL, as identified by NGS experiments even when using a limited number of target genes. In this particular and potentially curable disease, highly sensitive techniques like targeted NGS are crucial to highlight low frequency mutations. Low-coverage whole genome sequencing,11 sequencing of circulating cfDNA to detect genomic imbalances in HRS cells12 or targeted exome sequencing of isolated HRS cells13 have recently contributed to investigate genetic lesions underlying cHL but sample sizes were very low. In the present work, we found recurrent mutations in biopsies and plasma ctDNA of a cohort of consecutive 60 patients with cHL. This observation is consistent with previously published data regarding liquid biopsy in cHL with a different set of targeted genes6 and could haematologica | 2021; 106(1)

Liquid biopsy for classical Hodgkin Lymphoma Table 3. Baseline plasma circulating tumor DNA concentration according to patients’ characteristics at diagnosis.

Characteristics Sex Female Male Age ≥45 years <45 years Sedimentation rate (mm) ≥50 <50 Serum albumin ≥40 g/L <40 g/L Hemoglobin ≥10,5 g/dL <10,5 g/dL Bulky mass ≥10 cm Yes No Ann Arbor stage I-II III-IV White blood cell count ≥15 G/L <15 G/L Lymphopenia ≥0,6 G/L <0,6 G/L Number of involved nodal areas ≥4 <4 B symptoms Yes No International Prognostic Score (Hasenclever) 0, 1, 2 3, 4, 5 Histological subtype Scleronodular Others Identification of somatic variants at diagnosis Yes No

Plasma ctDNA concentration (hGE/mL) Mean (SD) Median

Range

28 32

320.8 (558.5) 325.4 (596.4)

87.1 52.7

[0-2,473.1] [0-2,684.9]

0.69

19 41

137.2 (194.7) 409.6 (667.1)

38.2 89.6

[0-683.3] [0-2,684.9]

0.12

23 30

472.1 (644.1) 243.2 (548.7)

244.9 41.5

[0-2,473.1] [0-2,684.9]

0.039

27 32

128.6 (290.4) 497.7 (699.1)

35.2 213.7

[0-1,434.9] [0-2,684.9]

0.003

54 6

239.2 (466.3) 1,080.7 (905.6)

45.9 715.3

[0-2,473.1] [251.5-2,684.9]

0.001

9 51

566.5 (917.7) 280.4 (491.6)

165.3 51.5

[0-2,684.9] [0-2,473.1]

0.44

31 29

169.11 (393.1) 488.2 (688.8)

35.2 246.1

[0-1,677.9] [0-2,684.9]

0.002

12 48

268.5 (327.5) 337 (622.7)

150.5 49.3

[0-944.2] [0-2,684.9]

0.49

56 4

263.9 (488) 1,155.7 (1,061)

49.3 845.8

[0-2,473.1] [246.1-2,684.9]

0.009

15 45

523.7 (520.3) 256.5 (581)

271.4 37.6

[44.8-1,677.9] [0-2,684.9]

0.009

33 27

464.1 (682) 151.3 (345.8)

244.9 39.1

[0-2,684.9] [0-1,677.9]

0.016

39 21

241.3 (519.9) 475.6 (649.2)

39.1 251.5

[0-2,473.1] [0-2,684.9]

0.018

42 18

338.6 (516.8) 287.6 (705.6)

133.3 32.7

[0-2,473.1] [0-2,684.9]

0.068

41 19

472.8 (643.3) 0.7 (3.1)

210 0

[9.7-2,684.9] [0-13.6]

<0.001

ctDNA: circulating tumor DNA; SD : standard deviation ; hGE/mL: haploid genome equivalents per mL of plasma.

add new information on driver events and tumorigenesis in this disease. As previously described in the literature, cHL displays a singular mutational signature, close to the primary mediastinal large B-cell lymphoma (PMBL) profile, compared to diffuse large B-cell lymphoma (DLBCL). The mutated genes point to the molecular deregulation of specific programs in cHL: NF-kB signature, STAT6 and cytokine signaling pathway (SOCS1). In our patients, mutations of SOCS1 haematologica | 2021; 106(1)

are frequent and are probably involved with STAT6 in the oncogenesis of cHL. This observation is consistent with the literature in PMBL and cHL populations15–18 except for the recent study by Spina et al.6 that did not cover SOCS1.6 In total, 70% of our patients with cHL harbored detectable somatic mutations. It is remarkable that 52% (127 variants detected in plasma samples, 66 of which were 159

V. Camus et al.

not detected in the tumors) of all somatic variants were only found in the plasma but not in the tumor because of the well-known tumor cell sparsity in Hodgkin lymphoma (HL). This confirms that ctDNA assessment might play an important role to define somatic mutations in this disease at the time of diagnosis. The detection of somatic alterations in the patients’ ctDNA samples represents a major technological advance and a step towards routine liquid biopsies in cHL. Our ctDNA data are fairly consistent with the results of the landmark study by Spina et al.6 despite several methodological differences and different gene panels. We have highlighted in this study seve-ral associations between median concentration of ctDNA at diagnosis, and disease stage, albumin level, sedimentation rate, LDH level and International Prognostic Score (IPS) which shows a correlation between ctDNA level and tumor mass and the inflammation it generates. The clearance of the mutations in plasma ctDNA was described as a new prognostic marker for mutated patients in other works in DLBCL patients with undetectable ctDNA after C2 who show a superior PFS compared to patients with positive ctDNA.9 Nevertheless, in our study, the finding that all patients clear ctDNA after C2 of thera-

py needs to be validated with more sensitive approaches that have a limit of detection lower than 0.5% VAF for disease monitoring. It is unlikely that those patients with cleared ctDNA after C2 have all been cured with only C2 of chemotherapy but future studies will probably assess various therapeutic strategies for these patients including the number of chemotherapy cycles needed to ensure the cure of the disease with real-time and dynamic ctDNA analysis. Outcome of patients in our study was excellent and therefore it is not possible to definitely examine associations between ctDNA levels and events. No somatic variant was identified after C2 of chemotherapy suggesting a possible non tumoral source of cfDNA after chemotherapy or an inflammatory process. Unfortunately, we did not have complete blood cell count data to evaluate the inflammatory syndrome after C2 of chemotherapy. In our previously published ctDNA study in DLBCL patients, we had been able to detect somatic variants after C2 of chemotherapy.17 We can hypothesize that our blood sampling process after C2 of chemotherapy may have been too late compared to the kinetics of clearance of ctDNA after treatment in cHL patients, and that it may have been necessary to perform blood sampling after 15 days or after

Figure 3 Distribution of circulating tumor DNA median variant allelic frequency for the 41 mutated patients with positive plasma at diagnosis. ctDNA: circulating tumor DNA; VAF: variant allelic frequency

Figure 4. Longitudinal assessment of mutation abundance in plasma circulating tumor DNA upon treatment. Evolution of median circulating tumor DNA (ctDNA) variant allelic frequency (VAF) for each patient (with detectable ctDNA mutation at diagnosis) throughout treatment (at diagnosis [“diag”] n=41) and after two cycles (C2) of chemotherapy (“post C2”, n=31). UPN: universal patient identification number.

160

haematologica | 2021; 106(1)

Liquid biopsy for classical Hodgkin Lymphoma

only C1 of chemotherapy. Future studies will probably address this timing concern. One of the limitations of our study, in addition to its noninterventional nature with plasma samples analyzed retrospectively in an unblinded fashion with a freeze-thaw step, is the absence of detectable mutations at time of diagnosis for 18 (30%) patients which potentially led to underestimating the prevalence of somatic mutations in this prospective cohort. We also missed some blood collections and so, in our study, ctDNA may be used as a biomarker for therapeutic response assessment in 31 of 60 (51.7%) of patients. This might be due to insufficient technological sensitivity, along with the examination of a restricted group of genes. We might possibly not have been able to identify ctDNA in patients with a low tumor burden, which limits its utilization as a MRD tool. Determining the technical limit of detection is a complex equation, as it depends on every sample, every variant and on many factors such as sequencer error, total DNA input, sample nature and qua-lity, coverage depth and sequencing quality. In order to determine whether or not we had missed variants in the initial tumor biopsy of the patient, a microdissection approach of HRS cells followed by a specific allele PCR could be performed but unfortunately, we were unable to do this due to insufficient material. Although additional mutations were found in the plasma compared to the tumor (which remains the gold standard), the false negative rate of the ctDNA analysis remains moderate with 20 of 89 (22.5%) of somatic variants missed in the plasma samples at diagnosis. Nevertheless, the variants detected in the plasma might better reflect the entire genetic panorama of the tumor. The few discrepancies between the detection of the mutation in the plasma and the tumor can be explained partly by the poor quality of some of our biopsies with tumor cell scarcity, potentially rendering the mutation detection in biopsy-extracted DNA impossible, and secondly by the absence of tumor DNA release in plasma by certain tumors and the short half-life (10-15 minutes) of circulating DNA in plasma.18

Our panel of genes analyzed is possibly too targeted and insufficiently sensitive to detect certain subclonal mutations of low allelic frequency, close to the noise threshold of the NGS sequencer (VAF 0.1%). In addition, our panel is not sufficient enough to detect somatic alterations in all patients due to the limited amount of included genes. The knowledge of cHL biology is booming and it will be ne-cessary to include other genes in future NGS panels such as ATM, KMT2D, TP53, ARID1A and CIITA, as suggested by results of the study by Spina et al.,6 in order to approach a more clinically relevant degree of informativeness (90%). However, this panel of genes is simple, reliable and inexpensive and seems to offer a good compromise between cost, sensitivity and easy applicability. In our opinion, the principle of a focused panel analysis for all cHL patients is an appropriate method to depict the genetic features of cHL at diagnosis and relapse. Another limitation of our technique is the absence of coverage of the immunoglobulin heavy chain locus for clonotypic immunoglobulin-gene rearrangement detection in ctDNA.19 Considering the baseline PET, ctDNA was significantly correlated to the MTV. Moreover, for the 41 â&#x20AC;&#x153;positiveâ&#x20AC;? plasma at diagnosis, the baseline MTV and the median plasmatic VAF were also significantly correlated. Such a correlation has been found in other lymphomas such as diffuse large cell B lymphomas17 and follicular lymphomas.19 Although significant, the correlations remained moderate (maximal =0.57 between the baseline MTV and the median plasmatic VAF). Several factors could explain these modest correlations, such as the scarcity of tumor cells, the short half-life of circulating DNA in plasma18 and the fact that some mutations may not have been explored. However, it is important to note the moderate value of these correlations. Regarding other methods of MRD in cHL, it has been demonstrated that pretreatment levels of plasma Epstein Barr virus DNA (EBV-DNA), as determined by quantitative real-time PCR (qRT-PCR), are correlated with unfavorable outcomes among a large prospective cohort of 274 patients

Figure 5. Example of possible application of liquid biopsy analysis. Patient (UPN38) presenting a positive positron emission tomography (PET) scan with residual fixation of a left supraclavicular lymph node (Deauville Scale 5) after two cycles (C2) of adriamycin (doxorubicin), bleomycin, vinblastine, dacarbazine (ABVD): the biopsy showed a reactive lymph node, indicating a false positive PET scan. Plasma circulating tumor DNA (ctDNA) sample was negative at time of the PET post C2 of ABVD with a clearance of all baseline mutations that were present at diagnosis. UPN: universal patient identification number.

haematologica | 2021; 106(1)

161

V. Camus et al. with previously untreated, advanced-stage HL a.21 Moreover, patients with plasma EBV(+) at month 6 after treatment also had an inferior outcome compared with plasma EBV(-) patients. In order to improve the clinical relevance of liquid biopsy in cHL, an increase in the sensitivity of somatic variant detection appears to be crucial. Sensitivity of somatic mutation analysis might be increased by (i) implementing new targets like EBV viral load; (ii) supplementary targeted genes or non-coding regions commonly altered by mutations identified in future whole genome analysis; and (iii) technological progress with greater read depth and more advanced bioinformatic pipelines. Finally, our data point out that ctDNA is an easy method to genotype cHL and might be valuable for the clinical management of the patients. However, our panel is not sufficient to monitor MRD in all patients and could not replace a larger panel of genes in order to precisely evaluate MRD. Our limited gene panel may not perform well enough and should be further improved to better assess its clinical utility in cHL. Future works with upgraded cfDNA detection, using interventional designs and larger patient cohorts will help to affirm that ctDNA analysis can be decisive for immediate non-invasive detection of patients with primary refractory disease and in post-therapeutic follow-up. We may speculate that ctDNA will soon allow convenient whole exome sequencing, which would enable the identification of somatic variants likely to respond to targeted therapy in relapsing patients.

nostic aid and therapeutic response evaluation, in association with PET. ctDNA concentration levels and genotypes are correlated with clinical characteristics and disease presentation. These data have to be confirmed in a large dedicated interventional study based on ctDNA MRD detection compared to PET. Disclosures HT received honoraria from Gilead and Immunogen; HT acts as a consultant and received honoraria from Takeda; HT acts as a consultant, received honoraria and research funding from Celgene, Karypharm, Roche, Janssen and is a member on the board of directors or advisory committees of these entities. VC received honoraria from Gilead, Amgen and BMS. FJ received honoraria from Roche, Janssen and Celgen. Contributions Conception and design: VC, FJ, AS. Administrative support: all authors. Provision of study materials or patients: VC, JMP, ELV, MC, PL, SL, EL, HL, ALM, NC, HT, AS, FJ. Collection and assembly of data: VC, MV, PJV, EB, LB, BM, PE, JMP, ELV, MC, LBU, JLO, SB, PD, FJ. Data analysis and interpretation: VC, JLE, PJV, EB, LB, BM, PE, SD, JMP, ELV. Manuscript writing: VC, MV, JLE, PJV, FJ, PD, AS. Final approval of manuscript: all authors. Acknowledgments The authors would like to thank all the patients that provided samples for the study. The authors would like to thank Mrs Sorina-Dana Mihailescu for statistical analysis support.

Conclusion The assessment of somatic mutations in plasma ctDNA by NGS is suitable as a biomarker in cHL for both diag-

References 1. Hasenclever D, Diehl V. A prognostic score for advanced Hodgkin’s disease. International Prognostic Factors Project on Advanced Hodgkin’s Disease. N Engl J Med. 1998;339(21):1506-1514. 2. Hutchings M, Loft A, Hansen M, et al. FDGPET after two cycles of chemotherapy predicts treatment failure and progression-free survival in Hodgkin lymphoma. Blood. 2006;107(1):52-59. 3. Canellos GP, Rosenberg SA, Friedberg JW, Lister TA, Devita VT. Treatment of Hodgkin lymphoma: a 50-year perspective. J Clin Oncol. 2014;32(3):163-168. 4. Schmitz R, Stanelle J, Hansmann M-L, Küppers R. Pathogenesis of classical and lymphocyte-predominant Hodgkin lymphoma. Annu Rev Pathol. 2009;4151-174. 5. Diaz LA, Bardelli A. Liquid biopsies: genotyping circulating tumor DNA. J Clin Oncol. 2014;32(6):579-586. 6. Spina V, Bruscaggin A, Cuccaro A, et al. Circulating tumor DNA reveals genetics, clonal evolution, and residual disease in classical Hodgkin lymphoma. Blood. 2018; 131(22):2413-2425. 7. Camus V, Stamatoullas A, Mareschal S, et al. Detection and prognostic value of recurrent exportin 1 mutations in tumor and cell-free circulating DNA of patients with classical Hodgkin lymphoma. Haematologica. 2016; 101(9):1094-1101. 8. Bessi L, Viailly P-J, Bohers E, et al. Somatic

162

Funding This work was supported by grants from the Centre Henri Becquerel and the Canceropole nord ouest (CNO).

mutations of cell-free circulating DNA detected by targeted next-generation sequencing and digital droplet PCR in classical Hodgkin lymphoma. Leuk Lymphoma. 2019;60(2):498-502. 9. Scherer F, Kurtz DM, Newman AM, et al. Distinct biological subtypes and patterns of genome evolution in lymphoma revealed by circulating tumor DNA. Sci Transl Med. 2016;8(364):364ra155. 10. Bohers E, Viailly P-J, Becker S, et al. Noninvasive monitoring of diffuse large B-cell lymphoma by cell-free DNA high-throughput targeted sequencing: analysis of a prospective cohort. Blood Cancer J. 2018; 8(8):74. 11. Salipante SJ, Adey A, Thomas A, et al. Recurrent somatic loss of TNFRSF14 in classical Hodgkin lymphoma. Genes Chromosomes Cancer. 2016;55(3):278-287. 12. Vandenberghe P, Wlodarska I, Tousseyn T, et al. Non-invasive detection of genomic imbalances in Hodgkin/Reed-Sternberg cells in early and advanced stage Hodgkin’s lymphoma by sequencing of circulating cell-free DNA: a technical proof-of-principle study. Lancet Haematol. 2015;2(2):e55-e65. 13. Reichel J, Chadburn A, Rubinstein PG, et al. Flow sorting and exome sequencing reveal the oncogenome of primary Hodgkin and Reed-Sternberg cells. Blood. 2015; 125(7):1061-1072. 14. Mottok A, Renne C, Willenbrock K, Hansmann M-L, Brauninger A. Somatic hypermutation of SOCS1 in lymphocytepredominant Hodgkin lymphoma is accom-

panied by high JAK2 expression and activation of STAT6. Blood. 2007;110(9):33873390. 15. Ritz O, Guiter C, Dorsch K, et al. STAT6 activity is regulated by SOCS-1 and modulates BCL-XL expression in primary mediastinal B-cell lymphoma. Leukemia. 2008; 22(11):2106-2110. 16. Weniger MA, Melzner I, Menz CK, et al. Mutations of the tumor suppressor gene SOCS-1 in classical Hodgkin lymphoma are frequent and associated with nuclear phospho-STAT5 accumulation. Oncogene. 2006; 25(18):2679-2684. 17. Bohers E, Viailly P-J, Becker S, et al. Noninvasive monitoring of diffuse large B-cell lymphoma by cell-free DNA high-throughput targeted sequencing: analysis of a prospective cohort. Blood Cancer J. 2018; 8(8):74. 18. Elshimali Y, Khaddour H, Sarkissyan M, Wu Y, Vadgama J. The clinical utilization of circulating cell free DNA (CCFDNA) in blood of cancer patients. Int J Mol Sci. 2013; 14(9):18925-18958. 19. Delfau-Larue M-H, van der Gucht A, Dupuis J, et al. Total metabolic tumor volume, circulating tumor cells, cell-free DNA: distinct prognostic value in follicular lymphoma. Blood Adv. 2018;2(7):807-816. 20. Kanakry JA, Li H, Gellert LL, et al. Plasma Epstein-Barr virus DNA predicts outcome in advanced Hodgkin lymphoma: correlative analysis from a large North American cooperative group trial. Blood. 2013; 121(18): 3547-3553.

haematologica | 2021; 106(1)

ARTICLE

Non-Hodgkin Lymphoma

Tandem autologous hematopoietic stem cell transplantation for treatment of adult T-cell lymphoblastic lymphoma: a multiple center prospective study in China Yao Liu,1* Jun Rao,1,2* Jiali Li,1* Qin Wen,1,2 Sanbin Wang,3 Shifeng Lou,4 Tonghua Yang,5 Bin Li,6 Lei Gao,1,2 Cheng Zhang,1,2 Peiyan Kong,1,2 Li Gao,1,2 Maihong Wang,1,2 Lidan Zhu,1,2 Xixi Xiang,1,2 Sha Zhou,1,2 Xue Liu,1,2 Xiangui Peng,1,2 Jiangfan Zhong,1,2,7 and Xi Zhang1,2

Medical Center of Hematology, Xinqiao Hospital, Army Medical University, Chongqing, China; 2State Key Laboratory of Trauma, Burns and Combined Injury, Army Medical University, Chongqing, China; 3Department of Hematology, General Hospital of Kunming Military Region of People’s Liberation Army, Kunming, China; 4Department of Hematology, Second Affiliated Hospital of Chongqing Medical University, Chongqing, China; 5Department of Hematology, Yunan Provincial People’s Hospital, Kunming, China; 6 Department of Hematology, Second Yunnan Provincial Peoples Hospital, Yunnan, China and 7Department of Pathology, University of Southern California, Keck School of Medicine, Los Angeles, CA, USA 1

Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):163-172

*YL, JR and JL contributed equally as co-first author.

ABSTRACT

-cell lymphoblastic lymphoma (T-LBL) is a highly aggressive form of lymphoma with poor clinical outcomes and no standard treatment regimen. In this study, we assessed the safety and efficacy of tandem autologous hematopoietic stem cell transplantation (auto-HSCT) for adult T-LBL and evaluated prognostic factors affecting survival. A total of 181 newly-diagnosed adult T-LBL patients were enrolled: 89 patients were treated with chemotherapy alone, 46 were allocated to the single auto-HSCT group, 46 were treated with tandem auto-HSCT. Median follow-up time was 37 months; the 3-year progression/relapse rate of the tandem autoHSCT group was significantly lower than that of the single auto-HSCT and chemotherapy groups (26.5% vs. 53.1% and 54.8%). The 3-year progression-free survival (PFS) and overall survival (OS) rates of the tandem autoHSCT group (73.5% and 76.3%) were significantly higher than those of the single auto-HSCT group (46.9% and 58.3%) and the chemotherapy group (45.1% and 57.1%). In the tandem auto-HSCT group, age and disease status after the first transplant impacted OS and PFS. Multivariate analysis identified that disease status after the first transplant was the only independent prognostic factor for patients treated with tandem-HSCT. In addition, diagnostic models of the initial CD8+CD28+/CD8+CD28– T-cell ratio in predicting the disease status were found to be significant. Taken together, tandem autoHSCT can be considered an optimal strategy for adult T-LBL patients. (Study registered at: ChiCTR-ONN-16008480).

Correspondence: YAO LIU/XI ZHANG 648283926@qq.com/zhangxxi@sina.com Received: June 18, 2019. Accepted: November 26, 2019. Pre-published: November 28, 2019 https://doi.org/10.3324/haematol.2019.226985

Introduction Lymphoblastic lymphoma (LBL) accounts for 1-2% of all non-Hodgkin lymphomas (NHL) and is considered a highly aggressive lymphoma with poor clinical outcomes. T-cell LBL (T-LBL) accounts for approximately 90% of all LBL and occurs mostly in children and adolescents; it predominantly affects males.1 Pathologically, the typical characteristics presented by T-LBL patients are often large anteriormediastinal mass, hydropericardium and pleural effusion. In addition, some T-LBL patients may also present bone marrow (BM) involvement and central nervous system (CNS) infiltration.2,3 Currently there is still no standard clinical treatment regimen, with multi-agent high-dose chemotherapy regimens (that mimic the treathaematologica | 2021; 106(1)

Material published in Haematologica is covered by copyright. All rights are reserved to the Ferrata Storti Foundation. Use of published material is allowed under the following terms and conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode. Copies of published material are allowed for personal or internal use. Sharing published material for non-commercial purposes is subject to the following conditions: https://creativecommons.org/licenses/by-nc/4.0/legalcode, sect. 3. Reproducing and sharing published material for commercial purposes is not allowed without permission in writing from the publisher.

163

Y. Liu et al.

ment of other lymphoma types) representing first-line treatment for most patients.4 Intensive chemotherapy regimens (such as hyper-CVAD [cyclophosphamide, vincristine sulfate, adriamycin and dexamethasone] and acute lymphoblastic leukemia [ALL]-like regimens for children) have also been used in treating adult T-LBL patients. Subsequently, complete remission (CR) rates have been significantly improved to 90-95% along with a long-term disease-free survival rate of 50-55%.5,6 However, 5-year OS does not exceed 67%, and 30-40% of patients will relapse. Thus, additional therapeutic strategies are urgently needed in the treatment of T-LBL. Auto-hematopoietic stem cell transplantation (autoHSCT) was one potential treatment regimen for T-LBL patients; however, the therapeutic effect of auto-HSCT treatment for these patients is still controversial. A recent retrospective study indicated that auto-HSCT could improve OS,7 and demonstrated that auto-HSCT could reduce early relapse to some degree.8 However, the high 5year relapse rate was still the primary reason for restricting the application of this approach, as the cumulative relapse rate of T-LBL patients during the 5 years after receiving auto-HSCT in the CR1 stage was as high as 56%.9 Survival of patients with relapsed disease was generally <12 months, and only 10% of these patients could survive long term.10,11 Even after remission with second-line regimens, auto-HSCT for relapsed patients still yielded poor results and did not improve OS.12 Compared with autoHSCT, the overall effect of allogeneic hematopoietic stem cell transplantation (allo-HSCT) was associated with fewer relapses, but higher treatment-related mortality (TRM) offset any potential survival benefit.13 Thus, due to the high rates of relapse following auto-HSCT, and the increased TRM in patients treated with allo-HSCT, new treatment strategies must be designed, not only to reduce the relapse rate of T-LBL but also to improve OS and PFS. Consecutive or 'tandem' auto-HSCT might represent a novel regimen in the treatment of T-LBL. Tandem autoHSCT function as a second round of an in vivo 'reset' through repeated ultra-high-dose chemotherapy, and could eliminate residual tumor cells in the body more thoroughly, thus reducing the risk of post-transplant minimal residual disease (MRD) and relapse. Tandem-HSCT had been successfully applied in treating certain brain tumors (such as neuroblastoma and glioma), and achieved good clinical outcome.14-16 In recent years, several groups have indicated that tandem auto-HSCT could reduce the relapse rate of refractory, relapsed, aggressive lymphomas (such as diffuse large B-cell lymphoma, mantle cell lymphoma [MCL], and peripheral T-cell lymphoma) and improve long-term survival.17,18 Tandem auto-HSCT has been shown to be superior to salvage chemotherapy for relapsed patients; however, whether tandem auto-HSCT is effective for newly diagnosed T-LBL patients is still not known. In this scenario, a multi-center prospective clinical trial was designed to explore the therapeutic effect of tandem auto-HSCT treatment for adult T-LBL patients. Newly diagnosed adult T-LBL patients with stage III/IV disease were given 2-3 courses of induction therapy according to the Intensive Chemotherapy-Berlin-Frankfurt-Münster study group (IC-BFM) 2002 regimens. After achieving CR or a partial response (PR), the patients were divided into three groups, matched by age and gender: the chemotherapy (no-HSCT) group, the single auto-HSCT group, and 164

the tandem auto-HSCT group. The objectives of this study were to evaluate the therapeutic effect and safety of tandem auto-HSCT, and to assess factors affecting survival.

Methods Study design and patient selection This prospective phase II study was conducted between February 2005 and November 2013 at five centers in southwestern China. (Study registered at: ChiCTR-ONN-16008480.) Eligible adults (n=181) aged 18-59 years with stage III/IV, newly diagnosed T-LBL were enrolled: 87 patients from Xinqiao Hospital, 25 from the General Hospital of Kunming Military Region of PLA, 23 from the Second Affiliated Hospital of Chongqing Medical University, 23 from Yunan Provincial People’s Hospital, and 23 from the Second Yunnan Provincial Peoples’ Hospital. The study was performed in accordance with the Declaration of Helsinki and local laws. The protocol was approved by the ethics committee of each center, and written informed consent was obtained from all patients.

Treatment strategies Induction therapy - all patients were given induction therapy according to the IC-BFM 2002 regimens. After CR/PR was achieved, the patients were divided into three groups: 89 patients were treated with consolidation chemotherapy according to the IC-BFM 2002 regimens, 46 were treated with single auto-HSCT, and the other 46 patients were treated with tandem auto-HSCT. During the induction therapy, all the patients were given standard intrathecal injection to prevent CNS lymphoma. Patients with CNS involvement were given triple intrathecal therapy to return cerebrospinal fluid levels to normal. Patients with large mediastinal masses were not given local radiotherapy in this study. Conditioning regimens - stem cell mobilization and collection details can be found in the Online Supplementary Appendix. For all patients receiving auto-HSCT, as melphalan was not used in China, other myeloablation regimens were utilized during conditioning. The BEAC regimen (carmustine: 0.2 mg/m2 x 1 day [d]; etoposide: 100 mg/m2 x 4 d; cytosine arabinoside: 100 mg/m2, q12 hours [h] x 4 d; cyclophosphamide: 1.5 g/m2 x 4 d) was adopted as the conditioning chemotherapy regimen for the single auto-HSCT and the first transplant of the tandem auto-HSCT. The IAC regimen (idarubicin: 10 mg/m2 x 3 d; cyclophosphamide: 1 g/m2 x 2 d; cytosine arabinoside: 1 g/m2, q12 h x 4 d) was adopted as the conditioning treatment for the second transplant of the tandem autoHSCT. The interval between the two transplants was 3-6 months. Stem cells were infused 48 h after the final dose of each of the regimens. All patients received prophylaxes for bacterial infection, fungal infection, herpes simplex virus infection, and pneumocystis pneumonia, as in our previous work.19

Statistical analysis Treatment responses were classified according to uniform international standards. The c2 test was used to analyze the intergroup difference in clinical features, and the Mann-Whitney U test was used to analyze age difference. The Kaplan-Meier estimator was used to estimate the survival curves. The log-rank test was performed for intergroup comparison. Cox regression analysis was used to analyze risk factors. As baseline of chemotherapy group and HSCT group was imbalanced, to reduce the influence of potential confounders, propensity score matching was applied. P<0.05 was considered statistically significant. SPSS (version 22.0) was used to carry out all the above statistical analyses. haematologica | 2021; 106(1)

Tandem auto-HSCT for adult T-LBL

Table 1. Patientsâ&#x20AC;&#x2122; demographic and clinical characteristics in the chemotherapy (Chemo) and autologous hematopoietic stem cell transplantation (auto-HSCT) groups.

Characteristics

All (%)

Median age, years (range) 32 (18-53) Gender, Male 91 (50.2) ECOG score 0-1 84 (46.4) 2 97 (53.6) Stage III 67 (37) IV 114 (63) B symptom 75 (41.4) aaIPI score 2 65 (35.9) 3 116 (64.1) BM involvement 64 (35.3) Mediastinal masses 68 (37.6) CNS involvement 7 (3.8) Remission status CR PR Immunophenotypic classification based on flow cytometry Pre-T-LBL Pro-T-LBL Cortical T-LBL Medullary T-LBL Immunophenotypic classification based on immunohistochemical staining Early T-LBL Non-early T-LBL

Chemo (%)

Single HSCT (%)

Tandem HSCT (%)

34 (18-53) 47 (52.8)

28 (18-43) 19 (41.3)

29 (18-50) 25 (54.3)

0.803

44 (49.4) 45 (50.6)

22 (47.8) 24 (52.2)

18 (39.1) 28 (60.9)

39 (43.8) 50 (56.2) 40 (44.9)

17 (37.0) 29 (63.0) 17 (37.0)

11 (23.9) 35 (76.1) 18 (39.1)

36 (40.4) 53 (59.6) 32 (35.9) 26 (29.2) 4 (4.4)

15 (32.6) 31 (67.4) 13 (28.3) 20 (43.5) 2 (4.3)

14 (30.4) 32 (69.6) 19 (41.3) 22 (47.8) 1 (2.2)

47 (52.8) 42 (47.2)

22 (47.8) 24 (52.2)

24 (52.2) 22 (47.8)

11 13 2 6

3 5 2 3

5 8 1 5

0.914

17 47

9 37

17 29

0.17

0.51

0.076

0.627 0.446

0.419 0.067 0.788 0.853

ECOG: Eastern Cooperative Oncology Group; aaIPI: age-adjusted International Prognostic Index; BM: bone marrow; CNS: central nervous system; CR: complete remission; PR: partial response; T-LBL: T-cell lymphoblastic lymphoma.

Results Patients' characteristics Median age of the patients in this study was 32 years (range: 18-59 years). At disease onset, 75 patients (41.4%) exhibited symptom B, 68 patients (37.6%) had mediastinal masses, and 7 patients (3.8%) had CNS infiltration; 114 patients (63%) were stage IV (by Ann Arbor staging) at diagnosis. Sixty-four patients (35.3%) were diagnosed with BM and/or peripheral blood (PB) involvement, and nine of them were diagnosed through BM flow cytometry detection. No significant differences between the three groups (chemotherapy, single auto-HSCT, and tandem auto-HSCT) were found with respect to remission status, age, gender, Eastern Cooperative Oncology Group (ECOG) score, stage, symptom B or age-adjusted International Prognostic Index (aaIPI) score (Table 1). A comparative analysis between the chemotherapy group and the HSCT group showed that there was a statistically significant difference in the presence of mediastinal mass between the two groups when propensity score matching analysis was applied. Seventy-four pairs of patients were eligible for survival analysis. Patients' characteristics are summarized in Online Supplementary Table S1.

Progression/relapse rate Because CR/PR of all the patients in this study was haematologica | 2021; 106(1)

achieved through induction therapy prior to grouping, the progression/relapse rates of different consolidation treatments were analyzed. In the auto-HSCT groups (including both single auto-HSCT and tandem auto-HSCT group), 1-year and 3-year progression/relapse rates were 19.9% and 39.6%, respectively, both of which were significantly lower than those of the chemotherapy group (36.1% and 54.8%, respectively; P=0.04) (Figure 1A). With respect to the auto-HSCT groups, no statistical differences were found in 1-year progression/relapse rates between the tandem auto-HSCT group and the single auto-HSCT group (12.1% vs. 18.6%), but the 3-year progression/relapse rate of the tandem auto-HSCT group was significantly lower than that of the single auto-HSCT group (26.5% vs. 53.1%; P=0.003), indicating the advantage of tandem auto-HSCT treatment in long-term disease control (Figure 1B).

Survival analysis Median follow-up time was 37 months (range: 5-81 months). The 3-year PFS and OS rates of the auto-HSCT groups were 60.4% and 66.3%, respectively, which were significantly higher than those of the chemotherapy group (45.1% and 57.1%, respectively; P<0.05) (Figure 2A and B). The 3-year PFS and OS rates among the 46 patients treated with single auto-HSCT were 46.9% and 58.3%, respectively. The 3-year PFS and OS rates among the patients treated with tandem auto-HSCT were 73.5% and 76.3%, respectively, which were significantly higher than 165

Y. Liu et al.

those of the chemotherapy group and the single autoHSCT group (P=0.01 and P=0.02, respectively) (Figure 2C and D). Moreover, prognostic value of immunophenotyping on T-LBL patients was also evaluated when patients were divided into pre-T-LBL, pro-T-LBL, cortical T-LBL and medullary T-LBL groups according to the European Group for Immunological Characterization of Leukemias (EGIL) classification. Sixty-four patients were eligible for subgroup analysis based on flow cytometric results; survival analysis of patients showed that there was no difference in PFS between the four subgroups of T-LBL (Online Supplementary Figure S1A). When classification was divided into early T-cell subtype and non-early T-cell subtype according to immunohistochemical surface marker, 25 cases could not be classified due to the fact that some required markers were incomplete. PFS of the early T-cell group was much poorer compared to that of non-early Tcell group, but this difference was not significant (P=0.06) (Online Supplementary Figure S1B). In addition, when all the patients were divided into the chemotherapy group and the auto-HSCT group, in the chemotherapy group, patients with early T-cell subtype had a significantly poor survival compared with the non-early T-cell group, but a comparison with the auto-HSCT group showed no significant difference, suggesting that auto-HSCT might overcome the deleterious effect of the early T-cell subgroup on prognosis (Online Supplementary Figure S1C and D).

only the occurrence rate of gastrointestinal symptoms (66.7%) was significantly higher during the first transplant than during the second transplant (33.3%), while there were no significant differences between the two transplants in the occurrence rates of other complications (e.g., hemorrhage and fever). Two patients died from transplant-related causes due to intracranial hemorrhage and infectious shock. Major transplant-related complications during the two transplants are shown in Online Supplementary Table S2.

Univariate and multivariate analysis Information about the 46 patients in the tandem autoHSCT group was collected and analyzed to determine independent prognostic factors. The results showed that aaIPI score, disease stage, B symptom, mediastinal mass, CNS and BM involvement, and disease status before first transplant had no influence on PFS or OS (P>0.05) (Figure 3A-J). Univariate analysis showed that age and disease status after the first transplant had influences on OS and PFS. Multivariate analysis showed that only disease status after the first transplant had an influence on OS and PFS (Tables 2 and 3). For patients <32 years of age, the 3year OS rate and 3-year PFS rate were 94.4% and 88.1%, respectively; these rates were significantly higher than those among patients >32 years of age (58.5% and 50.4%, respectively; P<0.05) (Figure 4A and B). The 3year OS and PFS rates among patients who were considered to be in CR after the first transplant were 86.8% and 74.9%, respectively. These rates were significantly higher than those patients who achieved PR after the first transplant (28.1% and 28.1%, respectively; P<0.05) (Figure 4C and D).

Factors influencing toxicity related to tandem auto-hematopoietic stem cell transplantation and patient prognosis Because two high-dose pretreatments were carried out on patients in the tandem auto-HSCT group, treatmentrelated toxicity was one of the major concerns in this study. The median times for leukocyte and platelet reconstitution after the first transplant were 12 d (range: 9-21 d) and 14 d (range: 11-17 d) respectively, and the median times for leukocyte and platelet reconstitution after the second transplant were 14 d (range: 10-17 d) and 14 d (range: 8-40 d), respectively. There was no significant difference in the time for post-transplant hematopoietic reconstitution between the two transplants. Among the major transplant-related complications,

CD8+CD28+/CD8+CD28– equilibrium could predict the disease status of T-cell lymphoblastic lymphoma patients Recently, CD8+CD28– T cells have attracted interest because of their critical role in immunomodulation. They are recognized as a unique subpopulation of regulatory T cells that have an extensive and direct effect, such as by inhibiting T-cell activation and proliferation, and decreasing the secretion of proinflammatory cytokines by activated T cells. Here, flow cytometry showed that a lower

Figure 1. Cumulative relapse rate of the entire patients’ cohort stratified by treatment strategies. (A) Cumulative relapse rate between chemotherapy group and autologous hematopoietic stem cell transplantation (auto-HSCT) group, (B) cumulative relapse rate between chemotherapy group, single auto-HSCT group and tandem auto-HSCT. N: number.

166

haematologica | 2021; 106(1)

Tandem auto-HSCT for adult T-LBL

number of CD8+CD28– T cells (5.84±2.06%) were seen in the patients with newly diagnosed T-LBL, compared with the patients who achieved PR (11.19±2.13%) and CR (18.91±3.38%) (Figure 5A-C). During follow-up, when patients relapsed, the number of CD8+CD28– T cells (7.59±1.53%) decreased compared to that of patients with CR or PR. Meanwhile, the CD8+CD28+/CD8+CD28– ratio of the patients in CR or with PR was significantly lower than that of newly diagnosed patients; when patients in CR or with PR relapsed, the ratio would increase, suggesting that the number of CD8+CD28– T cells and the CD8+CD28+/CD8+CD28– ratio might reflect the disease

status of T-LBL patients. Interestingly, the number of CD8+CD28+ and CD8+CD28– T cells and the CD8+CD28+/CD8+CD28– ratio were found to be statistically significant in predicting disease status. The AUC and the 95% confidence intervals are shown in Figure 5D. ROC curves showed that the CD8+CD28+/CD8+CD28– ratio had the largest AUC (0.91) together with the best sensitivity (93%) and specificity (85%), suggesting that the CD8+CD28+/CD8+CD28– balance had the best prediction efficiency (Figure 5E). In addition, purified CD8+CD28+ and CD8+CD28– T cells were isolated to test their direct suppressive activity, results from two donors

Figure 2. Kaplan-Meier estimated overall survival (OS) and progression-free survival (PFS) of different treatment strategies. Kaplan-Meier estimated OS (A) and PFS (C) of the entire patient cohort stratified by chemotherapy group and autologous hematopoietic stem cell transplantation (auto-HSCT). Kaplan-Meier estimated OS (B) and PFS (D) of the entire patient cohort stratified by chemotherapy group, single auto-HSCT group and tandem auto-HSCT. N: number.

Table 2. Univariate and multivariate analyses of the overall survival of patients with tandem autologous hematopoietic stem cell transplantation (HSCT).

Prognostic variables aaIPI scores Clinical stage B symptoms BM involvement Mediastinum involvement Age Disease status before HSCT Disease status after 1st HSCT

Univariate analysis HR (95%CI)

Multivariate analysis HR (95%CI)

0.59 (0.293-1.19) 0.94 (0.19-4.664) 3.487 (0.829-14.664) 2.877 (0.685-12.07) 0.211 (0.045-1.078） 0.11 (0.14-0.903) 0.409 (0.097-1.718) 0.181 (0.045-0.727)

0.141 0.94 0.088 0.149 0.062 0.043 0.222 0.002

1.676 (0.204-13.773) 0.34 (0.033-3.491) 1.676 (0.204-13.773) 9.539 (0.636-13.02) 0.247 (0.019-3.246) 0.079 (0.003-2.181) 1.017 (0.119-8.695) 0.009 (0.00-0.364)

0.631 0.364 0.631 0.103 0.287 0.134 0.988 0.013

HR: hazard ratio; CI: confidence interval; aaIPI: age-adjusted International Prognostic Index; BM: bone marrow.

haematologica | 2021; 106(1)

167

Y. Liu et al.

(Patient 1: newly diagnosed; Patient 2: in CR) demonstrated that both CD8+CD28+ and CD8+CD28– T cells were unable to mediate suppressive activity (Figure 5F and H) after incubation with autologous monocytes, IL2 and

granulocyte-macrophage colony-stimulating factor CSF) for one week; only CD8+CD28– T cells could suppressive activity (Figure 5G and I). Cytotoxic demonstrated that stimulated CD8+CD28– T cells

(GMshow assay could

Figure 3. Overall survival (OS) and progression-free survival (PFS) of different clinical characteristics in tandem autologous hematopoietic stem cell transplantation (auto-HSCT) group. Kaplan-Meier estimated OS and PFS of patient with tandem auto-HSCT stratified by age-adjusted International Prognostic Index (aaIPI) score (A and B), symptom B (C and D), disease conditions before transplantation (E and F), bone marrow involvement (H and J) and clinical state (I and J). CR: complete remission; PR: partial response; N: number.

Table 3. Univariate and multivariate analyses of the progression-free survival of patients with tandem autologous hematopoietic stem cell transplantation (HSCT).

Prognostic variables aaIPI scores Clinical stage B symptoms BM involvement Mediastinum involvement Age Disease status before HSCT Disease status after 1st HSCT

Univariate analysis HR (95%CI) 0.704 (0.372-1.335) 1.294 (0.275-6.099) 2.274 (0.651-7.942) 2.506 (0.704-8.919) 0.257 (0.065-1.02） 0.204 (0.043-0.971) 0.445 (0.125-1.582) 0.256 (0.071-0.918)

P 0.282 0.744 0.198 0.156 0.053 0.046 0.211 0.009

Multivariate analysis HR (95%CI) P 0.57 (0.203-1.6) 1.181 (0.170-8.815) 1.345 (0.231-7.837) 3.177 (0.592-17.035) 0.32 (0.053-1.914) 0.194 (0.026-1.442) 0.582 (0.119-2.854) 0.116 (0.017-0.804)

0.286 0.86 6 0.742 0.177 0.212 0.109 0.505 0.029

HR: hazard ratio; CI: confidence interval; aaIPI: age-adjusted International Prognostic Index; BM: bone marrow.

168

haematologica | 2021; 106(1)

Tandem auto-HSCT for adult T-LBL

decrease the cytotoxic T lymphocyte (CTL)-mediated cytotoxic ability.

Discussion ALL-like treatment regimens for children are currently recommended as induction treatment for all T-LBL patients.20 Accordingly, all patients in this study were given induction therapy according to the ALL-like treatment regimens for children. Overall, 93 (51.4%) of the 181 patients achieved CR after 2-3 treatment courses. CR rates are lower than those reported by other centers,5 which might be due to the advanced disease stage of the patients selected in this study (all stage III/IV). Following induction therapy, all patients were randomly assigned to three groups: chemotherapy (no HSCT), auto-HSCT or tandem auto-HSCT. The chemotherapy group included 89 patients, and the single and tandem auto-HSCT groups included 92 patients in total. Overall, both 1-year and 3-year progression/relapse rates of the auto-HSCT groups were lower than those of the chemotherapy group, and the lower rates associated with better OS: the 3-year OS and 3-year PFS rates of the auto-HSCT groups were significantly higher than those of the chemotherapy group. Results from this study indicate that combining auto-HSCT and increasing intensity of chemotherapy for adult T-LBL patients can reduce the relapse rate and enhance T-LBL patient survival compared with routine chemotherapy. Tremendous progress has been made in targeting the lymphoma surface molecule, especially in B-cell NHL,

such as mAb directed against CD20, CD40, CD52 and CCR4, but little advance has been achieved in T-cell NHL.21 Tandem auto-HSCT had been shown to reduce the risk of post-transplant MRD as well as relapse ratio. For this reason, this approach had been tested in clinical trials in the treatment of myeloma.22,23 Recently, in the context of lymphoma, Le-Gouill et al.13 conducted tandem auto-HSCT for 15 refractory recurrent high-risk NHL patients with a median follow-up time of 20 months (range: 0-55 months) and achieved an OS rate of 67%. A study by Espigado et al. also indicated that the 12-year OS rate among malignant lymphoma patients treated with tandem auto-HSCT was as high as 71%, significantly higher than patients who did not undergo transplant.11 Furthermore, the disease status achieved before the second HSCT could impact patient prognosis; the 12-year OS rate among patients who achieved CR before the second transplant was significantly higher than that of patients who did not achieve CR. These studies demonstrated that tandem auto-HSCT treatment was effective in treating relapse / refractory lymphoma and myeloma; however, no large-scale study of the use of tandem auto-HSCT for treatment of T-LBL has thus far been reported. To evaluate the safety and efficacy of tandem autoHSCT treatment for adult T-LBL patients, the clinical outcome of tandem auto-HSCT and single auto-HSCT treatment were further compared. As revealed by this study, there was no statistical difference in 1-year progression/relapse rates between tandem auto-HSCT treatment and single auto-HSCT treatment, but the 3-

Figure 4. Overall survival (OS) and progression-free survival (PFS) of age and disease status after first hematopoietic stem cell transplantation (HSCT) in tandem autologous (auto)-HSCT group. Kaplan-Meier estimated OS and PFS of patients with tandem auto-HSCT stratified by year (A and B) and disease status after first autoHSCT group (C and D). CR: complete remission; PR: partial response; N: number.

haematologica | 2021; 106(1)

169

Y. Liu et al.

year progression/relapse rate of the tandem auto-HSCT group was significantly lower than that of the single autoHSCT group. We attributed this to the fact that tandem auto-HSCT was much more effective, removing residual tumor cells in the body, thereby deepening remission status. Compared to the single auto-HSCT group and the chemotherapy group, the 3-year progression/relapse rate was significantly reduced in the tandem auto-HSCT group. The increase in survival rates might also be related to the comparable safety profiles of the single auto-HSCT and tandem auto-HSCT treatment regimens. As indicated by this study, there was no significant difference in the rates of the common complications between the two transplants, and the occurrence rate of gastrointestinal symptoms during the second transplant was significantly lower than that during the first. This might be due to the fact that the intensity of conditioning regimens during the second transplant was lower than that during the first. Two patients (2.1%) in the tandem auto-HSCT group died of transplant-related causes (intracranial hemorrhage and infectious shock, respectively), but the death rate was not higher than the rates of transplant-related death in treating lymphomas with auto-HSCT reported by other centers. Our results suggest, therefore, that the safety

level of tandem auto-HSCT treatment for adult T-LBL patients is acceptable. Age is an important prognostic factor in T-LBL. For example, Hanne et al. used the combination of ALL-like high-dose chemotherapy and auto-HSCT to treat LBL patients. Median age of patients included in their study was 32 years, and the study found that the PFS and OS rates among patients <32 years of age were 54% and 69%, respectively; these rates were significantly higher than those among patients >32 years of age.24 We also found that patient age influenced OS and PFS: patients <33 years of age had longer OS and PFS than patients >33 years of age. This might be because BM of younger patients is more resilient and can tolerate a high enough dose of chemotherapy to eliminate residual tumor cells and achieve deeper remission before transplant. This indicated that tandem auto-HSCT might represent a novel treatment method to achieve longer survival in young patients. The finding that disease status after the first transplant had an influence on OS and PFS, which was consistent with reports from other centers.7 Multivariate Cox analysis showed that, among the factors influencing survival, only disease status after the first transplant had an influence on OS and PFS. This result was helpful in improving the induction treatment

Figure 5. CD8+CD28+/CD8+CD28– equilibrium could predict the disease status of T-cell lymphoblastic lymphoma patients. Average frequency of CD8+CD28+ T cells, CD8+CD28- T cells, and CD8+CD28+/CD8+CD28– ratio in patients (A-C). Receiver operating characteristic curves of T cells and their balance in predicting the disease stage; diagonal lines in (D) are the diagnostic reference lines. (E) Area under curve (AUC) details of each subgroup. (F-I) Suppressive activities of CD8+CD28+ and CD8+CD28– T cells were analyzed, (a and d) mononuclear cell proliferation plus anti-CD3 monoclonal antibodies in the presence of autologous irradiated monocytes, (b) as (a) plus freshly isolated CD8+CD28+ T cells, (c) as (a) plus freshly isolated CD8+CD28– T cells, (d and e) as (b and c) plus IL-2 and granulocyte-macrophage colony-stimulating factor stimulation. (J) Cytotoxic function of antigen-specific cytotoxic T lymphocytes on target cells, (a) T2 cells plus p540 peptide, (b) as (a) plus stimulated CD8+CD28+ T cells, (c) as (a) plus stimulated CD8+CD28– T cells. CR: complete remission; PR: partial response; RE: Pt: patient; CI: confidence intervals; N: number; ns: not significant.

170

haematologica | 2021; 106(1)

Tandem auto-HSCT for adult T-LBL

regimen, and suggested that striving to achieve CR after the first transplant will be an important factor that will influence prognosis. OS and PFS of patients who were in PR after the first transplant was only 28.1%, which is clinically unsatisfactory. For these patients, it is recommended changing the treatment strategy, and allo-HSCT may be effective in improving their survival. CD28 serves as co-stimulator for T-cell activation and survival, and is expressed on all naïve T cells in newborns.25 However, as T cells experience antigenmediated activation and differentiation, CD28 expression was gradually lost.26 There is growing evidence that CD8+CD28+ T cells play a central role in inducing immune responses against tumor.27 Evidence had been found that CD4+ and CD8+T cells in MCL nodes were higher, and higher peritumoral CD8+ T-cell count and CD4/CD8 ratio were correlated with longer OS;28,29 however, there has been no research into the prognostic value of PB T-lymphocyte subset counts in T-LBL. Our research had found that peripheral CD8+CD28– T cells of newly diagnosed T-LBL was significantly lower than that of patients achieving CR or PR, but when these patients relapsed, the number of CD8+CD28– T cells would decrease again. Meanwhile, the ratio of CD8+CD28+/CD8+CD28– possess the opposite tendency for change: ROC curves confirmed that the balance of CD8+CD28+/CD8+CD28– has the best sensitivity and specificity to predict disease status. These results suggest this non-invasive testing approach could help the clinical decision-making process when selecting the most appropriate treatment. In conclusion, compared to sequential chemotherapy, auto-HSCT could improve the OS and PFS of adult T-LBL patients. Importantly, a second auto-HSCT consolidation

References 1. Bassan R, Maino E, Cortelazzo S. Lymphoblastic lymphoma: an updated review on biology, diagnosis, and treatment. Eur J Haematol. 2016;96(5):447-460. 2. Lepretre S, Graux C, Touzart A, Macintyre E, Boissel N. Adult T-type lymphoblastic lymphoma: treatment advances and prognostic indicators. Exp Hematol. 2017;51:7-16. 3. Portell CA, Sweetenham JW. Adult lymphoblastic lymphoma. Cancer J. 2012;18(5): 432-438. 4. Abaza Y, Kantarjian HM, Faderl S, et al. HyperCVAD plus nelarabine in newly diagnosed adult T-cell acute lymphoblastic leukemia and T-lymphoblastic lymphoma. Am J Hematol. 2018;93(1):91-99. 5. Marks DI, Paietta EM, Moorman AV, et al. Tcell acute lymphoblastic leukemia in adults: clinical features, immunophenotype, cytogenetics, and outcome from the large randomized prospective trial (UKALL XII/ECOG 2993). Blood. 2009;114(25):5136-5145. 6. Thomas DA, O'Brien S, Cortes J, et al. Outcome with the hyper-CVAD regimens in lymphoblastic lymphoma. Blood. 2004; 104(6):1624-1630. 7. Song KW, Barnett MJ, Gascoyne RD, et al. Primary therapy for adults with T-cell lymphoblastic lymphoma with hematopoietic stem-cell transplantation results in favorable outcomes. Ann Oncol. 2007;18(3):535-540. 8. Shi Y, Zhou S, He X, et al. Autologous

haematologica | 2021; 106(1)

treatment after achieving CR through the first transplant was effective for young patients, and the safety level was comparable to patients receiving a single auto-HSCT. As far as we know, this study is the first prospective multicenter large-sample study of this treatment regimen for TLBL. The results proved the effectiveness and safety of tandem auto-HSCT treatment for T-LBL patients, and confirmed that tandem auto-HSCT could improve OS and PFS rates in adult T-LBL. Disclosures No conflicts of interest to disclose. Contributions YL and XZ conceived and designed the study; JL, SL, QW, JR, JZ, TY, SW, BL, LeG, LiG and CZ developed the methodology, operate data collection and analyzed and interpreted the data; YL, RJ wrote the manuscript; LG, YL, PK, MW, LZ, XX, SZ, XL, CZ, QW, XP, LG and XZ reviewed and revised the manuscript. Acknowledgments We thank Ms. Huan-feng Liu, Yunjing Zeng, and Yimei Feng for their assistance with the statistical analysis. The authors also thank their colleagues for helpful comments. Funding This work was supported by grants from the National Natural Science Foundation of China (n. 81370594, 81070388, 81270569, 81600166), the Scientific and Technological Innovation Program of Chongqing social undertakings and people's livelihood guarantee (cstc2016shmsztzx10003), Science and technology innovation improvement project of AMU (2019XLC3020, 2019XLC1007).

hematopoietic stem cell transplantation in chemotherapy-sensitive lymphoblastic lymphoma: treatment outcome and prognostic factor analysis. Chin J Cancer Res. 2015;27(1):6673. 9. Levine JE, Harris RE, Loberiza FR Jr, et al. A comparison of allogeneic and autologous bone marrow transplantation for lymphoblastic lymphoma. Blood. 2003;101(7):2476-2482. 10. Le Gouill S, Lepretre S, Briere J, et al. Adult lymphoblastic lymphoma: a retrospective analysis of 92 patients under 61 years included in the LNH87/93 trials. Leukemia. 2003;17(11):22202224. 11. Espigado I, Rios E, Marin-Niebla A, et al. High rate of long-term survival for high-risk lymphoma patients treated with hematopoietic stem cell transplantation as consolidation or salvage therapy. Transplant Proc. 2008;40(9):31043105. 12. Kang W, Hahn JS, Kim JS, Cheong JW, Yang WI. Nine-year survival of lymphoblastic lymphoma patients. Yonsei Med J. 2006;47(4):466-474. 13. Le Gouill S, Moreau P, Morineau N, Harousseau JL, Milpied N. Tandem high-dose therapy followed by autologous stem-cell transplantation for refractory or relapsed high grade nonHodgkin's lymphoma with poor prognosis factors: a prospective pilot study. Haematologica. 2002;87(3):333-334. 14. Lee JW, Lim DH, Sung KW, et al. Tandem highdose chemotherapy and autologous stem cell transplantation for high-grade gliomas in children and adolescents. J Korean Med Sci. 2017;32(2):195-203.

15. Lee JW, Lee S, Cho HW, et al. Incorporation of high-dose (131)I-metaiodobenzylguanidine treatment into tandem high-dose chemotherapy and autologous stem cell transplantation for high-risk neuroblastoma: results of the SMC NB-2009 study. J Hematol Oncol. 2017;10(1): 108. 16. Sung KW, Son MH, Lee SH, et al. Tandem high-dose chemotherapy and autologous stem cell transplantation in patients with high-risk neuroblastoma: results of SMC NB2004 study. Bone Marrow Transplant. 2013;48 (1):68-73. 17. Takaishi K, Muto T, Mimura N, et al. Long-term complete remission following tandem autologous stem cell transplantation and consolidative radiotherapy for refractory mediastinal grayzone lymphoma. Int J Hematol. 2018;108(4): 452-455. 18. Chen YB, Li S, Fisher DC, et al. Phase II trial of tandem high-dose chemotherapy with autologous stem cell transplantation followed by reduced-intensity allogeneic stem cell transplantation for patients with high-risk lymphoma. Biol Blood Marrow Transplant. 2015; 21(9):1583-1588. 19. Gao L, Zhang C, Liu Y, et al. Favorable outcome of haploidentical hematopoietic stem cell transplantation in Philadelphia chromosome-positive acute lymphoblastic leukemia: a multicenter study in Southwest China. J Hematol Oncol. 2015;8:90. 20. Sweetenham JW. Treatment of lymphoblastic lymphoma in adults. Oncology (Williston Park). 2009;23(12):1015-1020.

171

Y. Liu et al. 21. Suresh T, Lee LX, Joshi J, Barta SK. New antibody approaches to lymphoma therapy. J Hematol Oncol. 2014;7:58 22. Kharfan-Dabaja MA, Hamadani M, Reljic T, et al. Comparative efficacy of tandem autologous versus autologous followed by allogeneic hematopoietic cell transplantation in patients with newly diagnosed multiple myeloma: a systematic review and meta-analysis of randomized controlled trials. J Hematol Oncol. 2013;6:2. 23. Michallet M, Sobh M, El-Cheikh J, et al. Evolving strategies with immunomodulating drugs and tandem autologous/allogeneic

172

hematopoietic stem cell transplantation in first line high risk multiple myeloma patients. Exp Hematol. 2013;41(12):1008-1015. 24. Bersvendsen H, Kolstad A, Blystad AK, et al. Multimodal treatment with ALL-like chemotherapy, auto-SCT and radiotherapy for lymphoblastic lymphoma. Acta Oncol. 2014;53(5):680-687. 25. Weng NP, Akbar AN, Goronzy J. CD28(-) T cells: their role in the age-associated decline of immune function. Trends Immunol. 2009; 30(7):306-312. 26. Seyda M, Elkhal A, Quante M, Falk CS, Tullius SG. T cells going innate. Trends Immunol.

2016;37(8):546-556. 27. Strioga M, Pasukoniene V, Characiejus D. CD8+ CD28- and CD8+ CD57+ T cells and their role in health and disease. Immunology. 2011;134(1):17-32. 28. Zhang XY, Xu J, Zhu HY, et al. Negative prognostic impact of low absolute CD4(+) T cell

counts in peripheral blood in mantle cell lymphoma. Cancer Sci. 2016;107(10):14711476. 29. Nygren L, Wasik AM, BaumgartnerWennerholm S, et al. T-cell levels are prognostic in mantle cell lymphoma. Clin Cancer Res. 2014;20(23):6096-6104.

haematologica | 2021; 106(1)

ARTICLE

Plasma Cell Disorders

Preclinical development of a humanized chimeric antigen receptor against B-cell maturation antigen for multiple myeloma

Ferrata Storti Foundation

Lorena Perez-Amill,1 Guillermo Suñe,1 Asier Antoñana-Vildosola,1 Maria Castella,1 Amer Najjar,2 Jaume Bonet,3 Narcis Fernández-Fuentes,4 Susana Inogés,5 Ascensión López,5 Clara Bueno,6 Manel Juan,7 Alvaro UrbanoIspizua1,8,9 and Beatriz Martín-Antonio1,8

Department of Hematology, Hospital Clinic, IDIBAPS, Barcelona, Spain; 2Department of Pediatrics - Research, The University of Texas M. D. Anderson Cancer Center, Houston, TX, USA; 3Laboratory of Protein Design and Immunoengineering, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland; 4Department of Biosciences, Faculty of Sciences and Technology, Universitat de Vic - Universitat Central de Catalunya, Vic, Spain; 5 Department of Hematology and Cell Therapy Area and Department of Immunology and Immunotherapy, Clinic University of Navarra, Navarra, Spain; 6Josep Carreras Leukemia Research Institute and Cell Therapy Program of the School of Medicine, University of Barcelona. Barcelona, Spain; 7Department of Immunotherapy, Hospital Clinic, IDIBAPS, Barcelona, Spain; 8Josep Carreras Leukemia Research Institute, Barcelona, Spain and 9 Department of Hematology, University of Barcelona, Barcelona, Spain 1

Haematologica 2021 Volume 106(1):173-184

ABSTRACT

ultiple myeloma is a prevalent and incurable disease, despite the development of new and effective drugs. The recent development of chimeric antigen receptor (CAR)T cells has shown impressive results in the treatment of patients with relapsed or refractory hematologic B-cell malignancies. In recent years, B-cell maturation antigen (BCMA) has appeared as a promising antigen to target using a variety of immunotherapy treatments, including CART cells, for patients with multiple myeloma. To this end, we generated clinical-grade murine CART cells directed against BCMA, named ARI2m cells. Having demonstrated its efficacy, and in an attempt to avoid the immune rejection of CART cells by the patient, the single chain variable fragment was humanized, creating ARI2h cells. ARI2h cells showed comparable in vitro and in vivo efficacy to that of ARI2m cells, and superiority in cases of high tumor burden disease. In terms of inflammatory response, ARI2h cells produced less tumor necrosis factor-α and were associated with a milder in vivo toxicity profile. Large-scale expansion of both ARI2m and ARI2h cells was efficiently conducted following Good Manufacturing Practice guidelines, obtaining the target CART-cell dose required for treatment of multiple myeloma patients. Moreover, we demonstrated that soluble BCMA and BCMA released in vesicles both affect CAR-BCMA activity. In summary, this study sets the bases for the implementation of a clinical trial (EudraCT code: 2019-001472-11) to study the efficacy of ARI2h-cell treatment for patients with multiple myeloma.

Correspondence: BEATRIZ MARTIN-ANTONIO bmartina@clinic.cat Received: June 3, 2019. Accepted: January 3, 2020. Pre-published: January 9, 2020. https://doi.org/10.3324/haematol.2019.228577

Introduction Multiple myeloma (MM) remains an incurable hematologic malignancy responsible for 15-20% of all blood cancers1,2 and new cases have been increasing, on average, by 0.8% each year over the past decade.3 The natural history of MM is relapse until refractory disease without reaching a plateau of survival, with less than 10% of patients achieving sustained complete remission beyond 5-10 years after autologous stem-cell transplantation.4 Moreover, patients are rarely cured after highdose chemotherapy followed by autologous stem-cell transplantation, indicating that novel strategies are required to improve the survival of patients with relapsed/refractory MM. In recent years, chimeric antigen receptor (CAR)T-cell immunotherapy, based on the infusion of autologous T cells genetically modified to recognize an antigen haematologica | 2021; 106(1)

173

L. Perez-Amill et al.

expressed on the tumor cell, has changed the modality of treatment for certain hematologic malignancies. Specifically, in acute lymphoblastic leukemia and lymphomas targeting CD19 outstanding responses have been achieved from the use of CART cells.5-8 In MM, B-cell maturation antigen (BCMA)9-11 has appeared as the most promising target for CART-cell immunotherapy. Clinical studies in patients with relapsed/refractory MM receiving CART-BCMA cells have documented excellent responses.12 Unfortunately, however, patients end up relapsing.12,13 Interestingly, in comparison to CART19, a higher CART-cell dose is required to achieve responses, 150x106 CART cells being the the lowest dose required to obtain response in patients who have relapsed after or are refractory to a median of seven lines of treatment.12,14 Moreover, it has been shown that a deepening of the response is obtained over time.12 In addition, the use of humanized or human CAR instead of murine CAR is emerging as the current trend in CART-cell immunotherapy.15,16 Here, using the same structure developed for CAR19 (ARI-0001), a CAR that has already been used in a phase I clinical trial (NCT03144583) for B-cell malignancies at our institution,17 we generated a murine CAR against BCMA (ARI2m) and a humanized version (ARI2h) for academic use. The efficacy and inflammatory response of both ARI2m and ARI2h cells were compared. Both CART cells showed comparable anti-MM activity. However, a greater efficacy was observed for ARI2h cells in cases of high tumor burden. The feasibility of clinical-grade expansion was tested in parallel in two different institutions and successfully achieved for both CAR. Finally, the impact of soluble BCMA (sBCMA) on ARI2m cell activity was analyzed, demonstrating how sBCMA can negatively affect CAR-BCMA activity. Overall, the results of this study have set the bases for a multicenter clinical trial on the use of ARI2h cells in MM patients in Spain (EudraCT code: 2019-001472-11).

Methods Ethics statement Research involving human materials was approved by the Clinical Research Ethical Committee (Hospital Clinic, Barcelona, Spain). Peripheral blood T cells were obtained from healthy donors after informed consent. All animal work was performed with approval from the Animal Research Ethical Committee (Hospital Clínic, Barcelona, Spain).

Cloning and humanization strategy The anti-BCMA single chain variant fragment (scFv) was designed from the J22.9 antibody.18 Human CD8a domains, 4-1BB and CD3 domains were obtained from the CART19 used at our Institution.17 Anti-CD19 scFv was substituted for the anti-BCMA scFv of the J22.9 antibody. In order to obtain ARI2h, the scFv sequence of the J22.9 antibody was humanized using two predictive models (Blast and Germline). Selected amino acids (excluding complementarity-determining regions and Vernier zone) were substituted for their homologous sequence in humans.

Predictive in silico models Immunogenicity against MHC-I was predicted with NetMHC-4.019-21 as previously described.15 In detail, binding 174

affinities of every 9-mer peptides from both scFv that are not encoded by the human genome were evaluated for 12 HLA-I alleles (5 type A and 7 type B) to predict binding affinity. As humanized ARI2h scFv contains human framework regions, only sub-peptides from complementaritydetermining regions and Vernier zone were considered. The affinity threshold to select only strong binders was <100 nM. To determine binding affinities, structural models derived for each antibody-BCMA pair were built with M4T22 using the crystal structure of the J22.9xiBCMAhuman complex (4ZFO)18 from the Protein Data Bank23 as a template. The quality and stereochemistry of the model was assessed using Prosa-II24 and PROCHECK,25 respectively. For each structural complex antibody-antigen model a total of 100 minimization trajectories were performed, followed by an estimation of the binding affinity as the energy difference between the complex and its separated components (ddG). Minimization and ddG analysis were carried out with Rosetta Design Suite.26,27 The quality of the binding affinity model was very high28 because of the high level of sequence identity between target sequences and the template showing similar protein-protein interfaces (Online Supplementary Figure S1D).

Clinical-grade production of ARI2 cells Lentiviral particles were produced inside a clean room facility following Good Manufacturing Practice guidelines as previously described.17. Clinical-grade ARI2h/ARI2m cells were produced using CliniMACS Prodigy (Miltenyi, Biotec) as described elsewhere.17 Further details of the methods are provided in the Online Supplementary Methods.

Results ARI2m cells demonstrate potent anti-myeloma activity The ARI2m sequence (Online Supplementary Figure S1A) contains the scFv sequence of the anti-BCMA antibody J22.9,18 which has been shown to be effective against MM,29 and human CD8a, 4-1BB and CD3ζ as hinge, transmembrane, co-stimulatory and signaling domains (Figure 1A). CART-cell transfection efficiencies varied between 30-60% in different experiments. CAR expression was retained after cryopreservation (Online Supplementary Figure S1B). The efficacy of ARI2m cells was tested against ARP1 and U266 MM cells by co-culturing T cells and MM cells at an effector:target (E:T) ratio of 1:1 over 4 days. ARI2m cells efficiently eliminated MM cells in comparison to untransduced (UT) T cells (Figure 1B) while no cytotoxicity was observed against a BCMA-negative cell line (K562) demonstrating the specificity of ARI2m cells (Figure 1B). In addition, limiting dilution cytotoxicity assays with E:T ratios from 1:1 to 1:0.125 demonstrated a high efficacy of ARI2m cells at eliminating MM cells at a low E:T ratio at 36 h (Figure 1C), and the effect continued to increase to 72 h (Online Supplementary Figure S1C). The production of pro-inflammatory cytokines by ARI2m cells was analyzed after co-culturing ARI2m cells and MM cells at different E:T ratios, at 24 and 48 h. A high interferon (IFN)γ production was observed at 24 h which increased at 48 h (Figure 1D). Some IFNγ production was detected for UT T cells, as expected, since UT T cells are activated during in vitro expansion. Minimum levels of interleukin (IL)6 were detected at 24 h, with the levels increasing haematologica | 2021; 106(1)

CAR-BCMA for multiple myeloma

at 48 h (Figure 1D). Tumor necrosis factor (TNF)α production decreased at 48 h in comparison to that at 24 h, suggesting that TNFα is produced early during CART-cell activation (Figure 1D). The in vivo efficacy of ARI2m cells was analyzed in a murine model in which NSG mice received 1x106 ARP1 MM cells. Mice were treated 6 days later with either 10x106 UT cells or 10x106 T cells containing 2x106 ARI2m cells

(Figure 1E). ARI2m cells prevented disease progression and performed better than UT T cells (Figure 1E, F). As expected, this translated into increased survival (Figure 1G). Furthermore, analysis of mice tissues at the experimental endpoint showed an absence of MM cells in the bone marrow and spleen (Figure 1H). T cells were mainly in the spleen (Figure 1I), whereas CART cells proliferated mainly in bone marrow, as indicated by a higher percentage of

A C

H I

Figure 1. ARI2m cells demonstrate potent in vitro and in vivo anti-myeloma activity. (A) Design of ARI2m. (B) Cytotoxicity assays of ARI2m against ARP1 and U266 (multiple myeloma cell lines) and non-myeloma K562 cells performed from 24-96 h. (C) Limiting dilution cytotoxicity assay against ARP1 and U266 cells performed at ratios from 1:1 to 0.125:1 (T cell: tumor cell line) at 36 h. (D) Cytokine profile of interferon γ, interleukin 6 and tumor necrosis factor α after 24 h and 48 h of coculturing T cells and ARP1 cells. (E-J) In vivo efficacy of ARI2m cells. (E, F) Diagram of experimental design and quantification of disease progression by weekly bioluminescence imaging and overall survival of the different groups of mice (G). (H) Flow cytometry of bone marrow and spleen of mice at the end of the experiment. (I) Percentage of total T cells and chimeric antigen receptor T cells in bone marrow and spleen ofn mice treated with ARI2m. (J). Soluble B-cell maturation antigen from mice serum after being treated with ARI2m or UT T cells. *P<0.05. LTR: long terminal repeat; UT: untransduced; IFNγ: interferon γ; IL6: interleukin 6; TNFα: tumor necrosis factor α; MM: multiple myeloma; BM: bone marrow; GFP: green fluorescent protein; sBCMA: soluble B-cell maturation antigen.

haematologica | 2021; 106(1)

175

L. Perez-Amill et al.

CART cells from the whole T-cell population in bone marrow than in the spleen (Figure 1I). Moreover, sBCMA was analyzed as an additional marker for MM progression in mice serum. As expected, a high amount of sBCMA was detected in mice treated with UT T cells while no sBCMA was found in mice treated with ARI2m cells (Figure 1J).

Humanization of ARI2m does not change affinity binding against human B-cell maturation antigen and enhances cytotoxicity against highly advanced myeloma disease Early disappearance of CART cells15 is associated with xenorecognition of the murine component of the CAR scFV by the human immune system. To circumvent this problem, we designed a humanized sequence of the scFv of ARI2m (Online Supplementary Figure S1A). Two different humanized variants were created based on two different predictive algorithms (Blast and Germline) by substitution of amino acids of the variable regions of heavy and light chains. In vitro comparison of the efficacy of both variants demonstrated that the Germline variant had slightly higher anti-MM activity (Figure 2A) and equal specificity, since neither of them eliminated K562 cells (Figure 2A). Therefore, the Germline variant, which was termed ARI2h, was selected for additional characterization. Predictive in silico models for the immunogenicity of both scFv demonstrated higher immunogenicity for ARI2m than for ARI2h (Figure 2B). Binding affinity prediction of both scFv against murine and human BCMA showed that both scFv bind to human BCMA with similar affinity while being unable to bind murine BCMA (Figure 2C). A structural comparison between the two scFv showed that most structural changes during the humanization were concentrated in the heavy chain of the antibody. Complementarity-determining regions on the antibody included no mutations and presented almost no structural drift (Figure 2D), reinforcing the idea that both antibodies have an almost identical binding surface. Subsequently, we analyzed whether T-cell transfection with either ARI2m or ARI2h CAR constructs would lead to different phenotypes in ARI2 cells. We did not observe any difference in either the proportion of ARI2+CD4 and ARI2+CD8 cell populations (Figure 2E) or the proportion of memory T-cell subsets (Online Supplementary Figure S1E). Moreover, as it was noted that the in vitro efficacy of ARI2h was slightly lower than that of ARI2m cells (Figure 2A), a long-term cytotoxicity assay was performed in which tumor and ARI2 cells were cultured together at a low E:T ratio (0.125:1). This assay showed slower killing kinetics for ARI2h cells, but an equal anti-MM activity at later timepoints (Figure 2F). Accordingly, T-cell proliferation upon CAR-antigen binding was slower for ARI2h cells (Figure 2G). In addition, whereas the same level of IFNγ was produced by both CART cells, lower TNFα and similar IL6 production was detected for ARI2h cells (Figure 2H). The anti-MM activity of ARI2h and ARI2m cells was further evaluated using two different in vivo models of MM (early and advanced disease models). Mice received MM cells on day 0 and were treated with 5x106 CART cells on day 6 or 14 to create an early and advanced model of disease, respectively (Figure 3A, B). In the early disease model, ARI2h and ARI2m cells prevented MM progression equally (Figure 3A, C). Around day 50, mice started to show signs of toxicity, which were more severe in the ARI2m group and translated into a lower survival rate for this group (Figure 3D). These results suggested that toxicity was related to a 176

global higher number of T cells proliferating in the ARI2mtreated group, as previously observed.30 In the advanced disease model, whereas ARI2m abrogated disease progression completely, minimal levels of MM disease were detected in the ARI2h-treated group at certain time points (Figure 3B), although this difference was not statistically significant (Figure 3C). Interestingly, mice treated with ARI2h cells again showed increased survival, consistent with lower toxicity, in comparison with that of mice treated with ARI2m cells (Figure 3D). Accordingly, in both disease models, the global number of T cells was higher for ARI2m than for ARI2h cells (Figure 3E), which might explain the greater xeno-graft-versus-host disease observed in the ARI2m-treated group. Importantly, the majority of T cells in bone marrow corresponded to CART cells, for both ARI2m and ARI2h cells, and in both disease models (Figure 3E). Lastly, analysis of mice serum showed that both CAR secreted large amounts of IFNγ. However, and in agreement with previous observations of slower kinetics for ARI2h cells, IFNγ production was slower in the ARI2h group. Thus, in the early model, IFNγ could not be detected in the ARI2h group 3 days after CART-cell infusion but its levels was higher at day 31 (Figure 3F). Similarly, in the advanced disease model, no IFNγ was detected at day 5 for ARI2h cells, but levels increased at day 21 (Figure 3F). These results suggested a faster activity of ARI2m than of ARI2h cells, which in cases of high tumor burden, might lead to faster exhaustion of CART cells and their disappearance. To test this hypothesis, a third in vivo experiment with a lower CART-cell dose (3x106) was performed. Disease burden was higher at the time of CART-cell injection than in the previous experiments (Figure 3B vs. Figure 4A). In this model, neither ARI2m nor ARI2h cells could prevent disease progression (Figure 4B). However, the performance of ARI2h cells at slowing disease progression was better than that of ARI2m cells (Figure 4B). At the last time-point, when mice were euthanized due to disease progression and not to xeno-graft-versus-host disease, T cells were almost undetectable (Figure 4C). However, mice treated with ARI2h cells had higher numbers of ARI2h cells in the bone marrow (Figure 4C). These data suggest that slower CART-cell proliferation could lead to longer CART-cell persistence and superior antitumor activity in cases of high tumor burden. To further support these findings, we exposed both ARI2m and ARI2h cells to consecutive in vitro challenges with MM (Figure 4D). In agreement with the lower ARI2m efficacy observed in animals with a high tumor burden, these experiments demonstrated more durable persistence for CD4 and CD8 ARI2h cells, which was not consistent for ARI2m cells (Figure 4E).

ARI2h cells induce less tumor necrosis factor α production than do ARI2m cells

Previous studies have shown that macrophages, after being activated by CART cells, are the main producers of IL6, IL1b and TNFα31,32 which lead to the development of cytokine release syndrome (CRS) in patients. To further analyze the pro-inflammatory profile of both CAR, we therefore established an in vitro system adding macrophages. Thus, effector and target cells were co-culture in the presence of macrophages (Figure 5A). The addition of macrophages did not affect CART-cell cytotoxicity (Figure 5B). The level of IFNγ was only slightly increased but very significant increases were detected for IL6, TNFα and IL1b, the last one not being detected in the absence of haematologica | 2021; 106(1)

CAR-BCMA for multiple myeloma

haematologica | 2021; 106(1)

Figure 2. Humanization of ARI2m into ARI2h and comparison of the affinity and immunogenicity of ARI2m vs ARI2h. (see also Online Supplementary Figure S1). (A) Limiting dilution cytotoxicity assay of ARI2m versus both humanized versions (Blast and Germline). (B) Predicted affinities of 9-mer peptides derived from either the ARI2m or ARI2h single chain variant fragment (scFv) against HLA-I. The total number of sites with predicted affinity <100 nM is shown on the left, and specific interactions of HLA-I alleles with each 9-mer peptide of the scFv is shown on the right. (C) Estimation of the binding affinity of both scFv against human and murine B-cell maturation antigen (BCMA). The difference in Gibbs free energy between the complex and its separated components for each protein pair highlights the inability of any of the antibodies to target murine BCMA. (D). Local structural comparison between the two scFv performed through RMSD. Positions of the complementarity-determining regions are highlighted in green while sequence differences between the two antibodies are mapped in gray under the curve. The resolution of the template structure of the antibody is represented as a dashed line. The graphic shows low structural drift on the antibodies’ complementarity-determining regions between the two antibody variants. (E). Phenotype characterization of T cells transduced with either ARI2m or ARI2h chimeric antigen receptor construct and after being expanded for 7 days (n=3). (F) Long-term cytotoxic assay comparing ARI2m versus ARI2h cells against ARP1 (multiple myeloma cells) and K562 (non-myeloma cells) for proliferation (G) and interferon γ, tumor necrosis factor α and Interleukin 6 production (H). *P<0.05. UT: untransduced; ddg: difference in Gibbs free energy; RMSD: root mean square deviation; E:T; effector to target cell ratio; CSFE: carboxyfluorescein succinimidyl ester; IFNγ: interferon γ; TNFα: tumor necrosis factor α; IL6: interleukin 6.

177

L. Perez-Amill et al. A

178

Figure 3. In vivo efficacy of ARI2m cells and ARI2h cells. Mice received multiple myeloma cells on day 0, and were treated at the indicated days. Disease progression was followed by weekly bioluminescence in the early disease model (A) and advanced disease model (B) and quantified (C). (D) Kaplan-Meier curve representing the overall survival of the different groups of mice. (E) Total CD3+ T cells and percentage of ARI2 cells from the CD3+ T-cell populations found in the bone marrow and spleen in both the early and advanced disease models. (F) Enzyme-linked immunosorbent assay of interferon Îł from murine serum at days 3 and 31 for the early disease model and at days 5 and 21 for the advanced disease model. *P<0.05. MM: multiple myeloma; CAR: chimeric antigen receptor; BM: bone marrow; IFNÎł: interferon Îł; CART: chimeric antigen receptor T-cell treatment.

CAR-BCMA for multiple myeloma

macrophages (Figure 5C). Next, we compared the proinflammatory activity of ARI2m cells and ARI2h cells over 2 days using this setting. We observed similar IFNÎł, IL6 and IL1b production for both CAR (Figure 5D) but TNFÎą production was lower for ARI2h cells, consistent with the findings of the long-term in vitro assay previously performed (Figure 2H) and the lesser in vivo toxicity of ARI2h compared to ARI2m cells.

ARI2m and ARI2h cells were efficiently expanded and the required CART-cell dose (>150x106 CART cells)12 was achieved in all cases (Figure 6A, B). Anti-MM activity for both ARI2h and ARI2m cells was also demonstrated (Figure 6C). In both institutions, all productions achieved the minimum threshold required for product release (Figure 6D, E).

Efficient clinical production and activity of ARI2 cells Pre-clinical data presented here support the development of a phase I multicenter clinical trial for MM patients (EudraCT code: 2019-001472-11) to evaluate the efficacy of treatment with ARI2 cells. The feasibility of large-scale, clinical-grade production was tested both for ARI2m and ARI2h cells using T cells from healthy donors. A CART-cell production system has already been established at our institution and is being used in an ongoing phase I clinical trial with a CAR19 product (ARI-0001).17 Four expansions were conducted for each CAR in two different institutions. Both

Clinical studies with CART19 in acute lymphoblastic leukemia have shown that 1x106 CART cells/kg are sufficient to achieve complete remission.5 In MM however a higher dose is needed (>150x106).12 BCMA is cleaved and released as sBCMA into the extracellular milieu.33 We hypothesized that sBCMA can bind to CAR-BCMA, partially hampering its anti-MM activity, thereby explaining the high CAR-BCMA dose required to achieve complete remission in MM patients. We therefore measured the amount of sBCMA in serum from patients with monoclonal gammopathy of undetermined significance, patients with newly

Soluble and released B-cell maturation antigen affects ARI2 cells activity

Figure 4. Humanization of ARI2m enhances cytotoxicity against highly advanced myeloma disease. (A) Schematic representation and images of an in vivo experiment in mice receiving ARP1 multiple myeloma cells (MM) and treated with either ARI2m or ARI2h cells. (B) Disease progression was followed by weekly bioluminescence of the experiment in (A). (C) Total CD3+ T cells and percentage of ARI2 cells from the CD3+ T-cell population found in the bone marrow. (D) Graph showing the scheme of consecutive challenges of chimeric antigen receptor T (CART) cells to MM cells. (E) Percentage of CART cells in CD4 or CD8 T-cell subsets after each challenge. *P<0.05. **P<0.0001.

haematologica | 2021; 106(1)

179

L. Perez-Amill et al.

diagnosed MM and MM patients at relapse. As expected, we observed larger amounts of sBCMA in MM patients (Figure 7A). To test whether sBCMA inhibits CART-cell activity, MM cells were co-cultured with ARI2m cells in the presence of recombinant BCMA protein. The results confirmed that recombinant BCMA blocks the activity of ARI2m cells, in terms of cytotoxicity and IFNγ production (Figure 7B). MM patients had around 100 ng/mL of sBCMA and a titration assay with recombinant BCMA demonstrated inhibition of ARI2m-cell activity up to 32 ng/mL of BCMA (Figure 7C). BCMA shedding is mediated by γ-secretase, which directly cleaves and releases BCMA into the milieu, decreasing surface BCMA expression. BCMA shedding can, therefore, be blocked using γ-secretase inhibitors.33 To analyze the effect of a γ-secretase inhibitor (DAPT), we measured membranebound BCMA in MM cells and the amount of sBCMA before and after treating MM cells with DAPT. As expected, DAPT treatment increased BCMA surface expression in MM cells and decreased the release of sBCMA (Figure 7D, E). The increased membrane-bound BCMA and decreased sBCMA associated with DAPT treatment were also detected after co-culturing MM cells with UT T cells (Figure 7D, E). In the case of co-cultures of ARI2m and MM cells, the addition of DAPT decreased the amount of sBCMA, as expected. However, membrane-bound BCMA was poorly detected because of the high ARI2m-cell activity in vitro, eliminating MM cells (Figure 7D, E). Having demonstrated that DAPT treatment prevents BCMA shedding, we tested

whether this effect results in enhanced ARI2m-cell activity. This analysis was conducted in transwell plates, in which untouched MM cells were placed in the upper well and cocultured ARI2m and MM cells in the lower well (Figure 7F). In this setting, MM cells in the upper well released sBCMA continuously. A cytotoxicity assay showed reduced ARI2mcell cytotoxicity and IFNγ production in the presence of sBCMA (with MM cells in the upper well) than in the control without MM cells in the upper well (Figure 7G). Moreover, the addition of DAPT to this transwell assay enhanced the cytotoxicity and IFNγ production of ARI2m cells without affecting the proliferation of ARI2m cells (Figure 7G). In addition, using confocal fluorescence microscopy we observed that BCMA is released from MM cells in vesicles (Figure 7H) through a mechanism that could also reduce CART-cell activity. In order to confirm that these BCMA vesicles could temporarily affect CAR-BCMA activity, MM cells overexpressing BCMA fused to green fluorescent protein (MM-BCMA-GFP) were co-cultured with ARI2m cells for 3 h and time-lapse in vivo imaging was performed. We confirmed that BCMA released in vesicles bind to ARI2m cells, distracting ARI2m cells from their target MM cells (Figure 7I and Online Supplementary Movie S1). Moreover, we also observed that, after being in contact with MM cells, ARI2m cells could acquire BCMA in their membranes from the surface of the MM cells and, as a consequence, fratricide was observed between ARI2m cells (Figure 7J and Online Supplementary Movie S2). To further confirm this event,

Figure 5. ARI2h cells induce less tumor necrosis factor α production than do ARI2m cells. (A) Schematic representation of monocyte and T-cell isolation from the same buffy coat, differentiation of the monocytes into macrophages, in vitro expansion and co-culture of chimeric antigen receptor T (CART) cells with multiple myeloma (MM) cell lines. (B) Cytotoxicity and (C) production of the cytokines interferon γ, interleukin 6, tumor necrosis factor α and interleukin 1b by ARI2m cells co-cultured with ARP1 cells with or without macrophages. (D) Cytokine production over 48 h after co-culturing ARI2m/ARI2h with macrophages and ARP1 MM cells. Macrophages were added at a 1:3 macrophage:CART ratio. *P<0.05. **P<0.0001. M-CSF: macrophage colony-stimulating factor; UT: untransduced; Mac: macrophages; IFNγ: interferon γ; IL6: interleukin 6; TNFα: tumor necrosis factor α; IL1b: interleukin 1b.

180

haematologica | 2021; 106(1)

CAR-BCMA for multiple myeloma

Figure 6. Efficient clinical production and activities of ARI2m and ARI2h cells. (A, B) Clinical expansion of ARI2m and ARI2h cells showing the total T-cell (A) and total chimeric antigen receptor T-cell numbers (B) achieved at the end of the expansion. (C) Cytototoxicy assays against the U266 multiple myeloma cell line of both ARI2m and ARI2h cells at the end of the expansion. Results from (A-C) are the median of four clinical expansions of ARI2m and ARI2h cells. (D, E). Detailed comparison of the four clinical productions performed in two different institutions, showing the percentage of each cell population achieved (D) and the specified parameters for product release (E). *P<0.05. **P<0.0001. MM: multiple myeloma; UT: untransduced.

haematologica | 2021; 106(1)

181

L. Perez-Amill et al.

Figure 7. Soluble and released B-cell maturation antigen affects the activity of ARI2 cells. (A) Measurement of soluble B-cell maturation antigen (sBCMA) from patients with monoclonal gammopathy of undetermined significance, multiple myeloma (MM) at diagnosis and MM at relapse. (B) Cytotoxicity assay and interferon γ production of ARI2m cells co-cultured with ARP1 MM cells, adding recombinant BCMA protein (BCMA) at 10,000 ng/mL. (C) Cytotoxicity assays (at 24 h) of ARI2m cells with ARP1 MM cells adding recombinant BCMA at the indicated doses. (D, E) Mean fluorescence intensity of BCMA (D) and concentration of sBCMA (E) of ARP1 MM cells alone or in co-culture with either ARI2m or UT T cells with or without DAPT. (F) Design of the cytotoxic assays in transwell plates to analyze the impact of sBCMA and DAPT. ARI2m cells and ARP1 cells were co-cultured in the well, and additional ARP1 cells were added to the transwells as a source of continuous release of sBCMA. (G) Cytotoxicity, interferon γ production and T-cell proliferation in the experiment described in (F). (H) Confocal fluorescence image of MM cells stained with a cell tracker, CMAC, in which BCMA is visualized with a monoclonal anti-BCMA. (I, J) Time lapse images from two different in vivo time lapse experiments over 3 h of ARI2m cells stained with the cell tracker, CMAC, and co-cultured with either RPMI MM cells (I) or ARP1 MM cells (J) overexpressing BCMA in green fluorescent protein (GFP) (see also Online Supplementary Movies S1 and S2). (K) Co-culture assay of T cells with MM-RPMI cells overexpressing BCMA-GFP to which latrunculin A was added in parallel. Percentage of T cells acquiring BCMA-GFP on their surface is shown. (L). MM-RPMI cells and ARI2m cells were co-cultured for 2 h to allow transfer of BCMA-GFP to ARI2m cells, then ARI2m cells were sorted and added into a 48 h cytotoxicity assay against RPMI-MM cells. *P<0.05. **P<0.0001. MGUS: monoclonal gammopathy of undetermined significance; Dx: diagnosis; E:T: effector:target cell ratio; IFNγ: interferon γ; UT: untransduced; MFI: mean fluorescence intensity; DAPT; a γ-secretase inhibitor; TW: transwell; LatA: latrunculin A.

182

haematologica | 2021; 106(1)

CAR-BCMA for multiple myeloma

ARI2m and MM cells co-cultured in the presence of a trogocytosis inhibitor (latrunculin A) showed a decreased percentage of BCMA acquisition by ARI2m cells (Figure 7K). In addition, after acquiring BCMA in their membranes, ARI2m cells showed decreased anti-MM activity (Figure 7L).

Discussion BCMA was identified in 201310 as the most promising antigen for CART-cell immunotherapy for the treatment of patients with relapsed/refractory MM, a finding which was confirmed in different clinical studies in MM patients12,13 and led us to develop our murine CAR-BCMA cells with 4-1BB as a co-stimulatory domain (ARI2m cells). ARI2m cells demonstrated strong anti-MM activity which is retained in their humanized version (ARI2h cells). Additionally, a greater efficacy of ARI2h versus ARI2m cells was observed in vivo in mice with high tumor burden. Our results set the bases for a multicenter phase I/II clinical trial of treatment with ARI2h cells for patients with relapsed/refractory MM in Spain. The success of BCMA as a target for CART-cell immunotherapy was demonstrated for the first time in patients with a CAR-BCMA with CD28 as the co-stimulatory domain. However, this CAR displayed marked toxicity.14 The CD28 was therefore replaced by 4-1BB: the new CAR (bb2121) demonstrated manageable toxicity and it was found that a minimum dose of 150x106 CART cells was required to obtain responses.12 In parallel, two other studies,13,34 demonstrated that a lower number of previous treatments is associated with better responses, and that short-term CART-cell expansion is more consistent after lymphodepletion.34 Different factors influence CART-cell expansion and persistence, which enhance the long-term control of the disease.5,12,13,34 Work needs to be done to improve these aspects of CAR-BCMA therapy, because a high number of MM patients end up relapsing.12,13 One of the suggested reasons for this is the limited persistence of CART cells. In this regard, human or humanized CAR, by avoiding the immunological reaction of the human immune system against the murine components of the scFv of the CAR, might increase the persistence of CART cells.15,16,30 Based on previous studies, and supported by our results showing that both ARI2m and ARI2h cells prevented disease progression equally, we selected ARI2h cells, which are humanized CAR-BCMA cells, to be used for a clinical trial for MM patients (EudraCT code: 2019-001472-11). Long-term CART-cell persistence is also associated with durability of remission.5,12,13,34 Factors influencing long-term CART-cell persistence include the exhaustion profile of CART cells29,35 and the affinity for the target antigen in combination with the co-stimulatory domain of CART cells. In this regard, studies on CART19 with CD28 and 4-1BB demonstrated that strong activation of CART cells, due to high affinity or high expression of the target antigen combined with the CD28 domain, leads to faster CART-cell proliferation with increased exhaustion and shorter persistence. Contrariwise, slower CART-cell activation, due to a lower affinity to the target antigen or to the 4-1BB domain, reduces exhaustion, thereby improving persistence.35-39 Here, even though ARI2m and ARI2h cells had the 4-1BB co-stimulatory domain and presented the same affinity against human BCMA, ARI2h cells had slower kinetic activity, and demonstrated greater efficacy than ARI2m cells in a murine model haematologica | 2021; 106(1)

of high tumor burden, and longer persistence after consecutive challenges to tumor cells. CRS and neurotoxicity are common after the administration of CART cells.5,12,13,34,40-42 Although these problems are efficiently managed by following international guidelines,43,44 the ideal CART-cell treatment should try to minimize the development of CRS. Here, the use of ARI2h cells instead of ARI2m was further supported by the observation of lower TNFα production in vitro with the ARI2h cells. Whereas IL6 is the effector cytokine for CRS,31,45,46 and exponentially increases as CRS develops, other cytokines such as TNFα and IL1b31,32 are the main initiators of CRS, as they are produced at early time points by monocytes and macrophages once they are activated by IFNγ produced by CART cells. In fact, TNFα, acts as an initiator cytokine orchestrating the cytokine cascade in many inflammatory diseases, appearing as a therapeutic target in inflammatory diseases.46 Here, our in vitro model with macrophages mimicking a model more similar to the in vivo scenario, demonstrated that ARI2h cells induced less TNFα production by macrophages, a relevant finding, as CRS in MM patients after CAR-BCMA treatment associates with a higher peak of TNFα.12 Moreover, CAR with faster kinetics associate with higher CRS,37 suggesting that the slower kinetics of ARI2h cells in comparison to ARI2m cells might also explain the observed lower level of in vivo toxicity. Last, we analyzed the impact of sBCMA on CART-cell activity. Even though studies with CAR-BCMA in MM have not found any correlation between sBCMA and CART-cell activity,10,12,34,47 we observed that the high in vitro CART activity rapidly eliminating MM cells impeded a proper analysis of the role of sBCMA in preclinical studies. Our in vitro models performed at a low CAR-BCMA:MM ratio, with the creation of an environment with continuous release of sBCMA, and the addition of a γ-secretase inhibitor confirmed the negative impact of sBCMA on CAR-BCMA activity. Moreover, we observed that BCMA is also released in vesicle structures, distracting CART cells from their targets, and that BCMA can be transferred to CART cells through a mechanism which appears to be trogocytosis, a finding already made for CART19 cells,48 causing decreased antiMM activity of the ARI2m cells. In conclusion, we have developed CAR-BCMA cells with 4-1BB as a co-stimulatory domain which have been humanized, retaining high efficacy and demonstrating less toxicity than their murine counterparts. ARI2 cells can be efficiently expanded under Good Manufacturing Practice conditions for use in a clinical trial. Moreover, we demonstrated that sBCMA and BCMA released from MM cells can affect CART-cell activity. Disclosures No conflicts of interests to disclose. Contributions LPA and BMA designed the study, LPA performed in vitro and in vivo experiments. GS performed clonings, virus production and in vivo experiments. LPA and BMA analyzed data and wrote the manuscript. AN designed the scFv of ARI2m and provided viral vector for GFP-Ffluc. MC and AA provided advice for experiments. AA performed some of the CART in vitro expansions. MJ, SI and AL performed clinical grade ARI2 production. JB and NFF performed predictive models for bind-ing affinity. CB performed cell sortings. AUI provided funding and constructive ideas. All authors reviewed the manuscript. 183

L. Perez-Amill et al.

Acknowledgments We acknowledge the Multiple Myeloma Research Center (Little Rock, AK, USA) for providing the ARP1 cell line and the Confocal Fluorescence Microscopy facilities at the University of Barcelona.

References 1. Palumbo A, Anderson K. Multiple myeloma. N Engl J Med. 2011;364(11):1046-1060. 2. Kumar SK, Callander NS, Alsina M, et al. Multiple myeloma, version 3.2017, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw. 2017;15(2):230-269. 3. Cancer Stat Facts: Myeloma. National Cancer Institute Surveillane, Epidemiology, and End Results Program Web site. Available at http://seercancergov/statfacts/html/mulmyhtml. [Accessed January 2, 2020] 4. Martinez-Lopez J, Blade J, Mateos MV, et al. Long-term prognostic significance of response in multiple myeloma after stem cell transplantation. Blood. 2011;118(3):529-534. 5. Maude SL, Laetsch TW, Buechner J, et al. Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia. N Engl J Med. 2018;378(5):439-448. 6. Schuster SJ, Svoboda J, Chong EA, et al. Chimeric antigen receptor T cells in refractory B-cell lymphomas. N Engl J Med. 2017; 377(26):2545-2554. 7. Park JH, Riviere I, Gonen M, et al. Long-term follow-up of CD19 CAR therapy in acute lymphoblastic leukemia. N Engl J Med. 2018;378(5):449-459. 8. Neelapu SS, Locke FL, Bartlett NL, et al. Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N Engl J Med. 2017;377(26): 2531-2544. 9. O'Connor BP, Raman VS, Erickson LD, et al. BCMA is essential for the survival of longlived bone marrow plasma cells. J Exp Med. 2004;199(1):91-98. 10. Carpenter RO, Evbuomwan MO, Pittaluga S, et al. B-cell maturation antigen is a promising target for adoptive T-cell therapy of multiple myeloma. Clin Cancer Res. 2013;19(8):2048-2060. 11. Novak AJ, Darce JR, Arendt BK, et al. Expression of BCMA, TACI, and BAFF-R in multiple myeloma: a mechanism for growth and survival. Blood. 2004;103(2):689-694. 12. Raje N, Berdeja J, Lin Y, et al. Anti-BCMA CAR T-cell therapy bb2121 in relapsed or refractory multiple myeloma. N Engl J Med. 2019;380(18):1726-1737. 13. Zhao WH, Liu J, Wang BY, et al. A phase 1, open-label study of LCAR-B38M, a chimeric antigen receptor T cell therapy directed against B cell maturation antigen, in patients with relapsed or refractory multiple myeloma. J Hematol Oncol. 2018;11(1):141. 14. Ali SA, Shi V, Maric I, et al. T cells expressing an anti-B-cell maturation antigen chimeric antigen receptor cause remissions of multiple myeloma. Blood. 2016;128(13):16881700. 15. Sommermeyer D, Hill T, Shamah SM, et al. Fully human CD19-specific chimeric antigen receptors for T-cell therapy. Leukemia. 2017;31(10):2191-2199. 16. Turtle CJ, Hanafi LA, Berger C, et al. CD19 CAR-T cells of defined CD4+:CD8+ composition in adult B cell ALL patients. J Clin Invest. 2016;126(6):2123-2138. 17. Castella M, Boronat A, Martin-Ibanez R, et

184

Funding Celgene and the Carlos III Institute of Health (project: PI17/01043) provided funding for all in vitro and in vivo studies.

al. Development of a novel anti-CD19 chimeric antigen receptor: a paradigm for an affordable CAR T cell production at academic institutions. Mol Ther Methods Clin Dev. 2018;12:134-144. 18. Oden F, Marino SF, Brand J, et al. Potent antitumor response by targeting B cell maturation antigen (BCMA) in a mouse model of multiple myeloma. Mol Oncol. 2015;9(7): 1348-1358. 19. Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M. NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8-11. Nucleic Acids Res. 2008;36(Web Server issue):W509512. 20. Andreatta M, Nielsen M. Gapped sequence alignment using artificial neural networks: application to the MHC class I system. Bioinformatics. 2016;32(4):511-517. 21. Nielsen M, Andreatta M. NetMHCpan-3.0; improved prediction of binding to MHC class I molecules integrating information from multiple receptor and peptide length datasets. Genome Med. 2016;8(1):33. 22. Fernandez-Fuentes N, Madrid-Aliste CJ, Rai BK, Fajardo JE, Fiser A. M4T: a comparative protein structure modeling server. Nucleic Acids Res. 2007;35(Web Server issue):W363368. 23. Berman HM, Westbrook J, Feng Z, et al. The protein data bank. Nucleic Acids Res. 2000;28(1):235-242. 24. Sippl MJ. Recognition of errors in threedimensional structures of proteins. Proteins. 1993;17(4):355. 25. Laskowski RA, MacArthur MW, Moss DS, Thornton J. PROCHECK: a program to check the sterochemical quality of protein structures. J Appl Cryst. 1993;26:283-291. 26. Fleishman SJ, Leaver-Fay A, Corn JE, et al. RosettaScripts: a scripting language interface to the Rosetta macromolecular modeling suite. PLoS One. 2011;6(6):e20161. 27. Leaver-Fay A, Tyka M, Lewis SM, et al. ROSETTA3: an object-oriented software suite for the simulation and design of macromolecules. Methods Enzymol. 2011;487: 545-574. 28. Baker D, Sali A. Protein structure prediction and structural genomics. Science. 2001;294 (5540):93-96. 29. Bluhm J, Kieback E, Marino SF, et al. CAR T cells with enhanced sensitivity to B cell maturation antigen for the targeting of B cell non-Hodgkin's lymphoma and multiple myeloma. Mol Ther. 2018;26(8): 1906-1920. 30. Smith EL, Staehr M, Masakayan R, et al. Development and evaluation of an optimal human single-chain variable fragmentderived BCMA-targeted CAR T cell vector. Mol Ther. 2018;26(6):1447-1456. 31. Giavridis T, van der Stegen SJC, Eyquem J, Hamieh M, Piersigilli A, Sadelain M. CAR T cell-induced cytokine release syndrome is mediated by macrophages and abated by IL1 blockade. Nat Med. 2018;24(6):731-738. 32. Norelli M, Camisa B, Barbiera G, et al. Monocyte-derived IL-1 and IL-6 are differentially required for cytokine-release syn-

drome and neurotoxicity due to CAR T cells. Nat Med. 2018;24(6):739-748. 33. Laurent SA, Hoffmann FS, Kuhn PH, et al. ÎłSecretase directly sheds the survival receptor BCMA from plasma cells. Nat Commun. 2015;6:7333. 34. Cohen AD, Garfall AL, Stadtmauer EA, et al. B cell maturation antigen-specific CAR T cells are clinically active in multiple myeloma. J Clin Invest. 2019;129(6):2210-2221. 35. Salter AI, Ivey RG, Kennedy JJ, et al. Phosphoproteomic analysis of chimeric antigen receptor signaling reveals kinetic and quantitative differences that affect cell function. Sci Signal. 2018;11(544). 36. Long AH, Haso WM, Shern JF, et al. 4-1BB costimulation ameliorates T cell exhaustion induced by tonic signaling of chimeric antigen receptors. Nat Med. 2015;21(6):581-590. 37. Milone MC, Fish JD, Carpenito C, et al. Chimeric receptors containing CD137 signal transduction domains mediate enhanced survival of T cells and increased antileukemic efficacy in vivo. Mol Ther. 2009;17(8):1453-1464. 38. Wherry EJ, Kurachi M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol. 2015;15(8):486-499. 39. Smith-Garvin JE, Koretzky GA, Jordan MS. T cell activation. Ann Rev Immunol. 2009;27:591-619. 40. Schuster SJ, Svoboda J, Chong EA, et al. Chimeric antigen receptor T cells in refractory B-cell lymphomas. N Engl J Med. 2017;377(26):2545-2554. 41. Locke FL, Neelapu SS, Bartlett NL, et al. Phase 1 results of ZUMA-1: a multicenter study of KTE-C19 anti-CD19 CAR T cell therapy in refractory aggressive lymphoma. Mol Ther. 2017;25(1):285-295. 42. Shah N, Alsina M, Siegel DS, et al. Initial results from a phase 1 clinical study of bb21217, a next-generation anti BCMA CAR T therapy. Blood. 2018;132(Suppl 1):488. 43. Porter D, Frey N, Wood PA, Weng Y, Grupp SA. Grading of cytokine release syndrome associated with the CAR T cell therapy tisagenlecleucel. J Hematol Oncol. 2018;11 (1):35. 44. Lee DW, Santomasso BD, Locke FL, et al. ASTCT consensus grading for cytokine release syndrome and neurologic toxicity associated with immune effector cells. Biol Blood Marrow Transplant. 2019;25(4):625638. 45. Hunter CA, Jones SA. IL-6 as a keystone cytokine in health and disease. Nat Immunol. 2015;16(5):448-457. 46. Parameswaran N, Patial S. Tumor necrosis factor-Îą signaling in macrophages. Crit Rev Eukaryot Gene Expr. 2010;20(2):87-103. 47. Friedman KM, Garrett TE, Evans JW, et al. Effective targeting of multiple B-cell maturation antigen-expressing hematological malignances by anti-B-cell maturation antigen chimeric antigen receptor T Cells. Hum Gene Ther. 2018;29(5):585-601. 48. Hamieh M, Dobrin A, Cabriolu A, et al. CAR T cell trogocytosis and cooperative killing regulate tumour antigen escape. Nature. 2019;568(7750):112-116.

haematologica | 2021; 106(1)

ARTICLE

Plasma Cell Disorders

Exploiting MYC-induced PARPness to target genomic instability in multiple myeloma

Ferrata Storti Foundation

Daniele Caracciolo,1 Francesca Scionti,1 Giada Juli,1 Emanuela Altomare,1 Gaetanina Golino,1 Katia Todoerti,2,3 Katia Grillone,1 Caterina Riillo,1 Mariamena Arbitrio,4 Michelangelo Iannone,4 Eugenio Morelli,5 Nicola Amodio,1 Maria Teresa Di Martino,1 Marco Rossi,1 Antonino Neri,2,3 Pierosandro Tagliaferri1 and Pierfrancesco Tassone1,6

Department of Experimental and Clinical Medicine, Magna Græcia University, Campus Salvatore Venuta, Catanzaro, Italy; 2Department of Oncology and Hemato-oncology, University of Milan, Milan, Italy; 3Hematology, Fondazione Cà Granda IRCCS Policlinico, Milan, Italy; 4IRIB-CNR, Catanzaro, Italy; 5Jerome Lipper Multiple Myeloma Center, Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA and 6 Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, PA, USA 1

Haematologica 2021 Volume 106(1):185-195

ABSTRACT

ultiple myeloma (MM) is a hematologic malignancy strongly characterized by genomic instability, which promotes disease progression and drug resistance. Since we previously demonstrated that LIG3-dependent repair is involved in the genomic instability, drug resistance and survival of MM cells, we here investigated the biological relevance of PARP1, a driver component of the alternative nonhomologous end joining pathway, in MM. We found a significant correlation between higher PARP1 mRNA expression and poor prognosis of MM patients. PARP1 knockdown or its pharmacological inhibition by olaparib impaired MM cell viability in vitro and was effective against in vivo xenografts of human MM. Anti-proliferative effects induced by PARP1 inhibition were correlated with an increase of DNA double-strand breaks, activation of the DNA damage response and apoptosis. Importantly, by comparing a gene expression signature of PARP-inhibitor sensitivity to our plasma cell dyscrasia gene expression profiling, we identified a subset of MM patients who could benefit from PARP inhibitors. In particular, gene set enrichment analysis suggested that high MYC expression correlates with sensitivity to PARP inhibitors in MM. Indeed, we identified MYC as a promoter of PARP1-mediated repair in MM and, consistently, we demonstrated that cytotoxic effects induced by PARP inhibition are mostly detectable in MYC-proficient MM cells. Taken together, our findings indicate that MYC-driven MM cells are addicted to PARP1 alternative non-homologous end joining repair, which therefore represents a druggable target in this still incurable disease.

Correspondence: PIERFRANCESCO TASSONE tassone@unicz.it Received: October 21, 2019. Accepted: February 17, 2020. Pre-published: February 20, 2020. https://doi.org/10.3324/haematol.2019.240713

Introduction Genomic instability represents a key hallmark of cancer since it progressively promotes the acquisition of features that lead to tumorigenesis. Indeed, almost all cancers are characterized by the tendency to accumulate genetic aberrations, such as gene copy-number variations, translocations and mutations, which finally confer growth and survival advantages.1 Genomic instability is strongly fostered by alterations of DNA repair pathways, which compromise “genomic guardian” mechanisms involved in the prevention of neoplastic transformation. At the same time, the DNA damage response machinery represents a specific Achilles heel of tumors, since it could be therapeutically exploited by designing anticancer strategies targeting cancer DNA repair vulnerabilities with a synthetic lethality approach.2 The discovery that homologous recombination-deficient tumor cells are specifically sensitive to inhibition of poly(ADPhaematologica | 2021; 106(1)

185

D. Caracciolo et al.

ribose) polymerase 1 (PARP1)3-5 provides an example of how to therapeutically exploit these vulnerabilities and led to the rapid development of several other DNA damage response inhibitors such as ATM, ATR, CHK1 and CHK2, DNA-PK and WEE1 inhibitors.6 However, it remains unclear why a substantial number of patients who lack homologous recombination mutations still benefit from PARP inhibition,7 a condition defined as “PARPness”.8 Double-strand breaks in DNA are mainly repaired by classical non-homologous end joining (NHEJ)9 and homologous recombination.10 Recently, a third repair pathway, named alternative NHEJ11 has been demonstrated to function as an error-prone, back-up pathway when the two major mechanisms are defective, contributing to the pathogenesis of several tumors.12,15 Multiple myeloma (MM) is a hematologic malignancy characterized by the growth of malignant plasma cells harboring numerous karyotypic aberrations.16 Indeed, the presence of specific cytogenetic abnormalities defines subgroups of MM patients with different prognoses requiring risk-adapted treatment. Although new therapeutics have prolonged the survival of MM patients, a cure for this disease is still an unmet need, and the identification of novel and actionable molecular drivers might provide innovative therapeutic approaches. It is known that genomic instability is a hallmark of MM but, to date, specific Achilles heels have not been identified. PARP1 is the best-characterized member of the PARP family and plays a crucial role in the alternative NHEJ pathway.17,18 Indeed, it senses DNA damage via its DNA binding domain,19 and subsequently synthesizes poly (ADP-ribose) polymers which are added to itself and other acceptor proteins, thus recruiting other DNA repair proteins, including DNA ligase 3 (LIG3). We have provided evidence that upregulation of LIG3mediated DNA repair plays a pivotal role in genomic instability and survival of MM cells.20 Starting from these data, we here report that PARP1 is crucial for survival of MYC-addicted MM cells, and we provide the rational framework for the use of PARP inhibitors as a therapeutic strategy in this still incurable disease.

Methods A more detailed description of the methods used is provided in the Online Supplementary Methods.

Cell lines and primary tumor specimens Cell lines were obtained from the American Type Culture Collection or kindly provided by sources indicated in the Online Supplementary Methods.

Western blot analysis Whole cell protein extracts were prepared from MM cells and from peripheral blood mononuclear cells in NP40 Cell Lysis Buffer (Novex®) containing a cocktail of protease inhibitors (Sigma, Steinheim, Germany). Cell lysates were loaded and separated by polyacrylamide gel electrophoresis. Proteins were transferred onto membranes by a Trans-Blot® TurboTM Transfer Starter System for 7 min. After protein transfer, the membranes were blotted with antibodies listed in the Online Supplementary Table and visualized with a C-DiGit® Blot Scanner (LI-COR) using ECL Western Blotting Detection Reagents (Thermo Fisher Scientific). Images were captured using Image Studio® software (LI-COR, version 5.0).

Immunofluorescence Cells were harvested, centrifuged onto glass slides (Cytospin 4, Thermo Scientific), and fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, for 12 min at 22°C, followed by three 5 min washes in PBS. Cells were permeabilized (0.1% Triton X-100 in PBS, 15 min), washed in PBS (3 times, for 5 min each), and incubated for 1 h at 22°C with blocking buffer (1.5% bovine serum albumin in PBS). They were reacted for more than 12 h at 4°C with primary antibodies listed in the Online Supplementary Table, washed in PBS (3 times, for 5 min each), and incubated for 1 h at 22°C in the dark, with appropriate secondary antibodies. Cells were washed three times in PBS and mounted under coverslips with Vectashield with DAPI (Vector Laboratories). Images were acquired with an SP2 Leica Zeiss confocal laser-scanning microscope with a 63× oil objective.

Animals and in vivo models of human multiple myeloma Male CB-17 severe combined immunodeficient (SCID) mice (6to 8-weeks old; Harlan Laboratories, Inc., IN, USA) were housed and monitored in our Animal Research Facility. Experimental procedures and protocols had been approved by the Magna Graecia University Institutional Review Board and were conducted according to protocols approved by the National Directorate of Veterinary Services (Italy). Mice were subcutaneously inoculated with 5x106 H929 or ABZB cells and treatment started when palpable tumors became detectable (100-200 mm3). Animals were treated daily with olaparib (Selleckchem) (100 mg/kg) or vehicle (10% v/v dimethylsulfoxide in 10% w/v Kleptose (hydroxypropyl-bcyclodextrin) in purified de-ionized water) via oral gavage. Tumor sizes were measured as described elsewhere,21 and the investigator was blinded to group allocation.

Statistical analysis Each experiment was performed at least three times and values are reported as means ± standard deviations. Comparisons between groups were made with the Student t-test, while statistical significance of differences among multiple groups was determined by GraphPad software (www.graphpad.com). Graphs were obtained using Graphpad Prism version 6.0. P-values of less than 0.05 were accepted as statistically significant.

Analysis of cell viability and apoptosis Cell viability was analyzed by Cell Titer-Glo assay (Promega, Madison, WI, USA); apoptosis was evaluated by flow cytometric analysis (Attune NxT Flow cytometer, Thermo Fisher Scientific, IL, USA) following annexin V-7-aminoactinomycin D staining (BD Pharmingen).

Cell cycle analysis The cell cycle was analyzed by propidium iodide flow cytometry (BD Pharmingen), according to the manufacturer’s instructions, using an Attune NxT Flow cytometer (Thermo Fisher Scientific). 186

Results High PARP1 expression occurs in multiple myeloma and predicts for poor prognosis To understand the role of PARP1 in the pathophysiology of MM, the prognostic relevance of its mRNA expression was investigated. Analysis of public MM datasets revealed that higher PARP1 mRNA expression was significantly correlated with shorter overall survival and event-free surhaematologica | 2021; 106(1)

PARP1 addiction in multiple myeloma

Figure 1. High PARP1 levels are associated with poor prognosis in multiple myeloma patients. (A) Data obtained interrogating the GSE24080 dataset using PRECOG software (https://precog.stanford.edu/index.php). Prognostic relevance of PARP1 mRNA (NM_208644_at) expression, on overall survival and event-free survival of patients with multiple myeloma (MM). (B) PARP1 mRNA expression analysis from proprietary datasets. Left panel: PARP1 levels in healthy donors as compared to those in cases of MM and primary and secondary plasma cell leukemia. Right panel: PARP1 expression in MM patients according to translocations/cyclins classification. (C) Analysis of the GSE9782 dataset using GenomicScape software (www.genomicscape.com). Prognostic relevance of PARP1 levels on overall and progression-free survival of MM patients treated with bortezomib. *P<0.05; **P<0.01. EFS: event-free survival; pPCL: primary plasma cell leukemia; sPCL: secondary plasma cell leukemia; TC: translocations/cyclins; HR: hazard ratio; OS: overall survival.

haematologica | 2021; 106(1)

187

D. Caracciolo et al.

vival (Figure 1A), thus highlighting its pivotal contribution to disease pathogenesis. Indeed, PARP1 mRNA expression increased during disease progression and in high-risk MM subgroups harboring t(4;14) and t(14;16) translocations (translocations/cyclins classes TC4 and TC5, respectively) (Figure 1B, Online Supplementary Figure S1A).22 Of note, no significant influence of copy number on PARP1 gene expression was found (Online Supplementary Figure S1B), suggesting that high expression of PARP1 in MM patients could be derived from deregulation of transcriptional or post-transcriptional mechanisms that normally regulate PARP1 expression. Notably PARP1 mRNA was the most expressed among other PARP family members. Furthermore, mRNA expression of other PARP family members did not retain any prognostic relevance, except for PARP2 (Online Supplementary Figure S1C, D). Moreover analysis of the GSE9782 dataset revealed that higher levels of PARP1 mRNA predicted poor overall and progression-free survival for patients who received bortezomib-based therapy (Figure 1C). Overall, these findings strongly suggest the involvement of PARP1 in the genomic instability of MM, which promotes disease progression and drug resistance.

PARP1 is a therapeutic target in multiple myeloma Based on clinical findings, which suggested a relevant role of PARP1 in the pathogenesis of MM, protein levels were analyzed in a panel of MM cell lines, primary cells from MM patients and peripheral blood mononuclear cells from healthy donors. As shown in Figure 2A, there was upregulation of PARP1 protein expression with a nuclear distribution in MM cells, as compared to the level of expression in peripheral blood mononuclear cells from healthy donors. In particular PARP1 protein was undetectable in only one of four primary MM cells evaluated and in the U266 cell line. To investigate its biological relevance in MM cells, knockdown experiments were performed to evaluate whether PARP1 was required for MM cell viability. Importantly, PARP1 downregulation significantly reduced viability and increased apoptosis of MM cell lines (Figure 2B, Online Supplementary Figure S2A). The effects induced by a clinically available PARP1/2 inhibitor, olaparib, on MM cell survival were investigated. Importantly, MM cells were highly sensitive to olaparib with an median inhibitory concentration (IC ) <10 mM observed in seven of eight tested cell lines (Figure 2C). In particular, the IC was 0.5 mM in NCI-H929, KMS-12BM, KMS26 and INA-6 cell lines, as in the highly sensitive BRCA2-defective CAPAN1 cells, a pancreatic adenocarcinoma cell line that was used as a positive control cell line in this experiment.23 Moreover, olaparib induced an increase of apoptotic cell death and a reduction of clonogenic growth in a dose-dependent manner (Online Supplementary Figure SB, C). Importantly, olaparib impaired the viability of primary MM plasma cells co-cultured with stromal cells (Figure 2C), thus overcoming the protective role of the bone marrow microenvironment. To confirm the translational relevance of our in vitro findings, the in vivo anti-MM activity of olaparib was evaluated and was shown to result in significant inhibition of tumor growth (Figure 2E). Finally, the basal expression of the other olaparib target, PARP2, was evaluated. Notably, its protein expression was much lower than that of PARP1 (Online Supplementary 50

188

Figure S2D), thus suggesting a minor role of PARP2 in driving olaparib activity in MM cell lines. Overall, these findings strongly suggest that PARP1 could be a promising therapeutic target in MM.

PARP1 inhibition triggers the DNA damage response in multiple myeloma cells Since PARP1 plays a critical role in the repair of doublestrand breaks, the effect of PARP1 inhibition on the DNA damage response was investigated. Importantly, olaparib treatment induced a relevant increase of unrepaired DNA damage, along with a significant activation of the DNA damage response and apoptosis signaling, as demonstrated by increased phosphorylation of ATM, ATR, CHK1, CHK2, and H2AX, occurring together with PARP1 and caspase-3 cleavage (Figure 3A). Similar results were also obtained after PARP1 knockdown in MM cell lines or after olaparib treatment in primary MM cells (Online Supplementary Figure S3A, B). Cell cycle analysis revealed that there was also G2-arrest after olaparib treatment (Figure 3B), an effect that was abrogated by caffeine24 (Online Supplementary Figure S3C), thus suggesting that checkpoint activation induced by olaparib treatment depends on ATM/ATR signaling. To investigate molecular mechanisms underpinning the increase of the DNA double-strand breaks observed after olaparib treatment, functional experiments were performed. To this aim, the activity of alternative NHEJ repair was evaluated given the pivotal role exerted by PARP1 in this DNA repair pathway. Indeed, alternative NHEJ repair was clearly reduced in MM cells treated with olaparib as compared to that in cells treated with vehicle alone (Figure 3C, Online Supplementary Figure S3D). To investigate whether PARP trapping onto damaged DNA is responsible for the cytotoxic activity of PARP inhibitors, R8226 cell lysates were fractionated into nuclear-soluble and chromatin-bound fractions. Notably, although PARylation was reduced by treatment, olaparib alone did not significantly increase PARP1 chromatin binding as compared to the nuclear-soluble fraction (Figure 3D), suggesting that the effects of olaparib in MM are correlated to inhibition of double-strand break repair rather than to PARP-trapping activity.

MYC drives sensitivity to PARP inhibition To identify predictive biomarkers for PARP inhibition in MM, a published PARP inhibition response gene expression signature25 was evaluated in a proprietary dataset of plasma cell dyscrasias, previously analyzed by global gene expression profiling using microarray technology.26 A large proportion of secondary plasma cell leukemia (PCL) samples and human myeloma cell lines showed a PARP inhibitor-positive expression pattern, MM samples showed heterogeneous expression levels, whereas primary PCL cases mostly showed an opposite expression trend. Notably, among MM TC classes, half the TC2 cases were grouped together with the majority of secondary PCL and all human myeloma cell line PARP inhibitor-positive clusters (Figure 4A). Therefore, in order to identify concordantly modulated sets of genes that were potentially associated with the PARP inhibitor signature in MM, PARP inhibitor-positive and -negative TC2 cases were compared by gene set enrichment analysis.27 Interestingly, groups of genes regulated by MYC and involved in DNA repair resulted among the most significantly upregulated haematologica | 2021; 106(1)

PARP1 addiction in multiple myeloma

Figure 2. Addiction of multiple myeloma cells to PARP1. (A) Left panel: immunoblot of PARP1 performed on peripheral blood mononuclear cells collected from healthy donors, multiple myeloma (MM) cell lines and patients (Pts). GAPDH was used as a loading control. Right panel: immunofluorescence analysis showing subcellular distribution of PARP1 in R8226 cells and a patientâ&#x20AC;&#x2122;s (Pt#2) MM cells (magnification 20x): PARP1 (green); DAPI (blue) was used for nuclear staining. (B) Indicated cell lines were transfected with scramble control or PARP1-siRNA. Top: a Cell Titer-Glo assay was performed 4 days after transfection. Bottom: immunoblot analysis of PARP1. GAPDH was used as a loading control. Analysis was performed 48 h after transfection. (C) Indicated MM cells were treated with increasing dose of olaparib. Left panel: a Cell Titer-Glo assay was performed 7 days after treatment. Results are expressed as a percentage of vehicle-treated cells. (D) Cell viability of CD138+ cells from four different MM patients co-cultured with HS-5 stromal cells and treated with olaparib 5 uM or vehicle. The assay was performed 4 days after treatment. (E) In vivo growth of H929 subcutaneous xenograft daily treated with vehicle or olaparib (100 mg/kg) via oral gavage. Averaged tumor volume of each group Âą standard deviation is shown. Data are representative of at least three independent experiments. *P<0.05; **P<0.01.

haematologica | 2021; 106(1)

189

D. Caracciolo et al.

Figure 3. Olaparib activates the DNA damage response in multiple myeloma cells. (A) R8226 and H929 cells were treated with olaparib. Left panel: Immunoblot analysis was performed 24 h after treatment. Right panel: immunofluorescence evaluation by Îł-H2AX foci. Representative images of unrepaired double-strand breaks are shown. DAPI (blue) was used for nuclear staining. (B) Cell cycle analysis performed 24 h after treating H929 and R8226 cells with olaparib (2.5 mM). (C) Alternative non-homologous end junction repair was evaluated by EJ2-GFP assay on R8226 cells 48 h after treatment with olaparib (2.5 mM) treatment. (D) Western blot analysis of nuclear-soluble and chromatin-bound fractions prepared from R8226 cells. Cells were treated for 30 min with vehicle or olaparib as indicated. Histone H3 and LAMIN A/C were used as positive markers for chromatin and nuclear-soluble fractions, respectively. Data are representative of at least three independent experiments. *P<0.05; **P<0.01. Olap: olaparib; A-NHEJ: alternative non-homologous end junction; PAR: poly (ADP)-ribose; PARP: poly (ADP)-ribose polymerase; H3: histone H3.

190

haematologica | 2021; 106(1)

PARP1 addiction in multiple myeloma

Figure 4. MYC drives sensitivity to PARP inhibition. (A) Hierarchical agglomerative clustering of samples of plasma cell dyscrasias on PARP inhibitor gene expression signature. Types of samples (normal [N], multiple myeloma [MM], primary plasma cell leukemia [PPCL], secondary plasma cell leukemia [SPCL], human myeloma cell lines [HMCL]) and MM translocations/cyclins classes (TC1, TC2, TC3, TC4, TC5) are depicted in different colors. The color scale bar represents the relative gene expression changes normalized by the standard deviation. PPCL, SPCL and MM-TC2 significant clusters are highlighted: 16/24 PPCL, P=4x10-4; 7/12 SPCL P=8.4x103 ; 8/30 TC2, P=6.5x10-3; 7/30 TC2, P=9.4x10-3. (B) Enrichment plots of selected significant (nominal P<0.05) gene sets by gene set enrichment analysis on PARP inhibitor-positive versus PARP inhibitor-negative MM-TC2 cases. Normalized enrichment scores (NES) are indicated. (C) Left panel: cell viability analysis in P493-6 cells after 7 days of exposure to vehicle or olaparib (5 mM), in the presence of doxycycline (DOX: 500 ng/mL) (low MYC) or in its absence (high MYC). Right panel: cell viability analysis in U266 empty vector (EV) and U266 MYC-positive cells after 7 days of exposure to vehicle or olaparib (5 mM). Data are representative of at least three independent experiments. *P<0.05; **P<0.01.

haematologica | 2021; 106(1)

191

D. Caracciolo et al.

in PARP inhibitor-positive versus PARP inhibitor-negative cases (Figure 4B, Online Supplementary Figure S4A). Similar results were also obtained comparing PARP inhibtor--positive and -negative PCL cases. Accordingly, the highest median levels of expression of MYC transcripts were reached in secondary PCL and human myeloma cell lines across the groups of plasma cell dyscrasias and in MM class TC2 (Online Supplementary Figure S4B). To validate the bioinformatics results, basal levels of MYC protein were evaluated in olaparib-treated MM cell lines. Notably, olaparib-insensitive U266 cells did not express MYC protein (Online Supplementary Figure S4C).28

Next, olaparib was tested on P936 B cell-like lymphoma cells, a cellular model in which MYC levels can be turned on or off by adding doxycycline. Notably, olaparib treatment effectively induced cell death only in Myc-on conditions. Conversely, as formal proof of our hypothesis, overexpression of MYC in U266 cells restored cell death upon PARP1 knockdown or PARP1 inhibitor treatment (Figure 4C, Online Supplementary Figure S4D). Moreover, the lack of correlation between PARP inhibitor sensitivity and the evidence of homologous recombination deficiency (as evaluated by RAD51 protein levels), or the presence of MM characteristic cytogenetic abnormalities (13q dele-

Figure 5. Bortezomib-resistant cells are highly sensitive to PARP inhibitors. (A) Left: alternative non-homologous end junction repair (A-NHEJ) was evaluated by an EJ2-GFP assay on AMO1 and ABZB cells. Right: Affymetrix CytoScan HD Array analysis of copy-number variations, using genomic DNA from AMO1 and ABZB cells. (B) A Cell Titer-Glo Assay was performed on AMO1 cells transfected with PARP1 ORF or control ORF (EV) and then treated for 24 h with increasing doses of bortezomib. PARP1 overexpression was confirmed by western blot analysis. (C) In vivo growth of luciferase gene-marked ABZB xenografts treated daily with vehicle or olaparib (100 mg/kg) via oral gavage. Left: averaged tumor volume of each group Âą standard deviation is shown. *P<0.05; **P<0.01. Right: PAR levels and Ki67 and HE expression were evaluated respectively by western blot and immunohistochemical analysis (20x, 40x insets) from a representative ABZB xenograft per group. (D) Left: Immunoblot of PARP1 and MYC performed on AMO1 and ABZB cells. GAPDH was used as a loading control. Right: quantitative polymerase chain reaction for PARP1 promoter performed after chromatin immunoprecipitation with MYC antibody in AMO1 and ABZB cells compared with negative (Ch22) and IgG controls. (E) Left: ABZB cells were transfected with siRNA-normal control (NC) or siRNA-MYC. Immunoblot of PARP1 and MYC was performed 48 h after transfection. Right: promoter activity of transfected PARP1 and negative CTRL promoter constructs in ABZB cells co-transfected with either siRNA-NC or siRNA-MYC. Data are representative of at least three independent experiments. *P<0.05; **P<0.01.

192

haematologica | 2021; 106(1)

PARP1 addiction in multiple myeloma

tion, 1q gain, hyperdyploidy) or somatic mutations (BRAF, NRAS, KRAS, DIS3) (Online Supplementary Figure 4E), further support the hypothesis that PARP-inhibition triggers MYC-dependent synthetic lethality in MM.

Bortezomib-resistant cells are highly sensitive to PARP inhibition Based on clinical data, which suggest a potential role in bortezomib resistance (Figure 1C), the effects of PARP1inhibition in this context were also investigated. Notably, AMO1 bortezomib-resistant (ABZB) cells showed greater alternative NHEJ repair and copy-number variations as compared to their sensitive counterpart AMO1 cells (Figure 5A), thus suggesting an association between an increase of genomic instability and acquisition of bortezomib resistance. Interestingly, enforced expression of PARP1 in AMO1 cells significantly antagonized the effect induced by bortezomib on cell viability (Figure 5B). Importantly, olaparib showed significant growth inhibitory activity against ABZB cells in vitro and in vivo (Figure 5C, Online Supplementary Figure S5A, B). Since higher levels of PARP1 and MYC proteins were observed in ABZB cells than in AMO1 cells (Figure 5D), we next investigated a potential role of MYC in PARP1 transcription. Notably, analysis of a MM patientsâ&#x20AC;&#x2122; dataset (GSE24080), showed a significant positive correlation between PARP1 and MYC expression (Online Supplementary Figure S5C).

Indeed, bioinformatic screening (cistrome.org) revealed significant enrichment of MYC binding consensus sequences to PARP1 promoter (Online Supplementary Figure S5D), which was next validated by chromatin immunoprecipitation analysis (Figure 5E). Short interfering RNA-mediated knockdown of MYC induced downregulation of PARP1 protein and decreased promoter activity (Figure 5E), indicating that MYC binds to the PARP1 promoter, modeling its transcription.

Discussion The DNA damage response is an attractive area of investigation because of the opportunity to selectively kill cancer cells addicted to compensatory DNA repair pathways by synthetic lethality.29 There is experimental evidence that alternative NHEJ is an error-prone DNA repair pathway30-35 and master driver of genomic instability in cancer.36 In this study, we investigated the involvement of PARP1, a pivotal component of error-prone alternative NHEJ repair, in the pathogenesis of MM genomic instability.37 The alternative NHEJ repair pathway is involved in important processes of B-lineage differentiation such as VD-J recombination and class switch recombination,38 critical steps in which errors in the rearrangement of germline DNA could generate translocation of proto-

Figure 6. Cartoon describing the addiction of MYC-driven multiple myeloma cells to PARP1.

haematologica | 2021; 106(1)

193

D. Caracciolo et al.

oncogenes, such as MYC,39 into the immunoglobulin loci, thus potentially contributing to myelomagenesis. Consistently, we found that high PARP1 mRNA expression is significantly correlated with poor event-free and overall survival in MM patients, and increases during disease progression and in high-risk cases. Moreover, PARP1 mRNA is the most expressed among PARP family members, thus suggesting a critical role of PARP1 in MM pathogenesis. In fact, we found that PARP1 knockdown induces MM cell death. With the aim of translating these findings into the design of a new therapeutic strategy, we provide evidence that olaparib, an available PARP inhibitor, induces antiMM activity at clinical doses,40 even against bortezomibresistant MM cells. Growth inhibitory effects induced by olaparib were correlated to inhibition of alternative NHEJ41 which led to an increase of unrepaired DNA damage and apoptotic cell death. To investigate mechanisms leading to PARPi sensitivity in MM, we used a PARP inhibition sensitivity gene expression signature.25 Interestingly our analysis highlighted a pivotal role of MYC, a driver transcription factor hyperactivated in the majority of MM patients.42,43 Indeed, the signature was enriched in patients with TC2 MM and in subgroups with secondary PCL, which display the highest MYC expression.44 Consistently, we found that PARP inhibition induces cytotoxic effects against MM, which mostly occur in MYC-proficient cells. Our findings are consistent with recent reports showing that MYC-dependent Burkitt lymphoma45 and neuroblastoma46 are sensitive to PARP inhibition. Indeed, MYC drives genomic instability by deregulation of the DNA damage response, thus increasing the dependency of cancer cells on low-fidelity DNA repair pathways.47-49 Moreover, MYC is a strong inducer of replication stress, which requires activation of the DNA damage response to repair DNA damage and to sustain cell survival.50,51 Based on these findings, we hypothesize that MYC-driven MM cells switch their DNA repair machinery to errorprone PARP-mediated alternative NHEJ. This event allows them to overcome DNA damage overload from oncogenic stress, promoting cell survival and, at the same time, the

References 1. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646-674. 2. Murata S, Zhang C, Finch N, et al. Predictors and modulators of synthetic lethality: an update on PARP inhibitors and personalized medicine. Biomed Res Int. 2016;2016: 2346585. 3. Kim G, Ison G, McKee AE, et al. FDA approval summary: olaparib monotherapy in patients with deleterious germline BRCAmutated advanced ovarian cancer treated with three or more lines of chemotherapy. Clin Cancer Res. 2015;21(19):4257-4261. 4. Balasubramaniam S, Beaver JA, Horton S, et al. FDA approval summary: rucaparib for the treatment of patients with deleterious BRCA mutation-associated advanced ovarian cancer. Clin Cancer Res. 2017;23(23): 7165-7170. 5. [No authors listed]. Olaparib for metastatic breast cancer in patients with a germline

194

acquisition of new genetic changes leading to disease progression (Figure 6). Indeed, it could be hypothesized that MYC induces alternative NHEJ repair to balance the downregulation of homologous recombination induced by bortezomib,52,53 contributing to the development of drug resistance and, at the same time, making MM cells more dependent on PARP1-mediated DNA repair to survive. Consistently, we also report here that bortezomibresistant MM cells are extremely sensitive to PARP inhibition. This finding is of translational relevance, since it could support the design of a maintenance therapy with PARP inhibitors as a strategy to prevent disease progression and acquisition of drug resistance. In conclusion, our study demonstrates addiction of MYC-driven MM cells to PARP1, which could be exploited as a new opportunity for synthetic lethality in the clinical scenario of precision oncology. Moreover, our findings highlight the role of MYC as a driver of PARP1-mediated repair, identifying a novel mechanism of genome stability and survival regulation in MM and a potential biomarker of PARPness in this still incurable disease. Disclosures No conflicts of interests to disclose. Contributions DC, FS, GJ, EA, GG, KG, RC, MA and EM performed experiments and analyzed the data; KT and AN performed microarray experiments and provided biological samples; MI performed the immunohistochemical analysis; NA, MTDM, and MR critically evaluated of experimental data and the manuscript. DC, PT and PT conceived the study and wrote the manuscript. Acknowledgments This work was supported by the Italian Association for Cancer Research (AIRC) with “Special Program for Molecular Clinical Oncology–5 per mille”, 2010/15 and its Extension Program” n. 9980, 2016/18 (PI: PT); and also by “Innovative Immunotherapeutic Treatments of Human Cancer” Multi Unit Regional n. 16695 (cofinanced by AIRC and the CARICAL Foundation), 2015/18 (PI: PT). We thank Dr. Ivana Criniti for her study coordination support and editorial assistance.

BRCA mutation. N Engl J Med. 2017;377(17):1700. 6. Ashworth A, Lord CJ. Synthetic lethal therapies for cancer: what's next after PARP inhibitors? Nat Rev Clin Oncol. 2018;15 (9):564-576. 7. Pilie PG, Gay CM, Byers LA, O'Connor MJ, Yap TA. PARP inhibitors: extending benefit beyond BRCA-mutant cancers. Clin Cancer Res. 2019;25(13):3759-3771. 8. Pilie PG, Tang C, Mills GB, Yap TA. State-ofthe-art strategies for targeting the DNA damage response in cancer. Nat Rev Clin Oncol. 2019;16(2):81-104. 9. Pannunzio NR, Watanabe G, Lieber MR. Nonhomologous DNA end-joining for repair of DNA double-strand breaks. J Biol Chem. 2018;293(27):10512-10523. 10. San Filippo J, Sung P, Klein H. Mechanism of eukaryotic homologous recombination. Annu Rev Biochem. 2008;77:229-257. 11. Chang HHY, Pannunzio NR, Adachi N, Lieber MR. Non-homologous DNA end joining and alternative pathways to doublestrand break repair. Nat Rev Mol Cell Biol.

2017;18(8):495-506. 12. Fan J, Li L, Small D, Rassool F. Cells expressing FLT3/ITD mutations exhibit elevated repair errors generated through alternative NHEJ pathways: implications for genomic instability and therapy. Blood. 2010;116(24): 5298-5305. 13. Newman EA, Lu F, Bashllari D, et al. Alternative NHEJ pathway components are therapeutic targets in high-risk neuroblastoma. Mol Cancer Res. 2015;13(3):470-482. 14. Tobin LA, Robert C, Nagaria P, et al. Targeting abnormal DNA repair in therapyresistant breast cancers. Mol Cancer Res. 2012;10(1):96-107. 15. Ceccaldi R, Liu JC, Amunugama R, et al. Homologous-recombination-deficient tumours are dependent on Poltheta-mediated repair. Nature. 2015;518(7538):258-262. 16. Neri P, Bahlis NJ. Genomic instability in multiple myeloma: mechanisms and therapeutic implications. Expert Opin Biol Ther. 2013;13(Suppl 1):S69-82. 17. Krishnakumar R, Kraus WL. The PARP side of the nucleus: molecular actions, physiolog-

haematologica | 2021; 106(1)

PARP1 addiction in multiple myeloma

ical outcomes, and clinical targets. Mol Cell. 2010;39(1):8-24. 18. Paddock MN, Bauman AT, Higdon R, et al. Competition between PARP-1 and Ku70 control the decision between high-fidelity and mutagenic DNA repair. DNA Repair (Amst). 2011;10(3):338-343. 19. Yang G, Liu C, Chen SH, et al. Super-resolution imaging identifies PARP1 and the Ku complex acting as DNA double-strand break sensors. Nucleic Acids Res. 2018;46(7):34463457. 20. Caracciolo D, Di Martino MT, Amodio N, et al. miR-22 suppresses DNA ligase III addiction in multiple myeloma. Leukemia. 2019;33(2):487-498. 21. Leone E, Morelli E, Di Martino MT, et al. Targeting miR-21 inhibits in vitro and in vivo multiple myeloma cell growth. Clin Cancer Res. 2013;19(8):2096-2106. 22. Zhan F, Huang Y, Colla S, et al. The molecular classification of multiple myeloma. Blood. 2006;108(6):2020-2028. 23. McCabe N, Lord CJ, Tutt AN, et al. BRCA2deficient CAPAN-1 cells are extremely sensitive to the inhibition of poly (ADP-ribose) polymerase: an issue of potency. Cancer Biol Ther. 2005;4(9):934-936. 24. Sarkaria JN, Busby EC, Tibbetts RS, et al. Inhibition of ATM and ATR kinase activities by the radiosensitizing agent, caffeine. Cancer Res. 1999;59(17):4375-4382. 25. McGrail DJ, Lin CC, Garnett J, et al. Improved prediction of PARP inhibitor response and identification of synergizing agents through use of a novel gene expression signature generation algorithm. NPJ Syst Biol Appl. 2017;3:8. 26. Lionetti M, Barbieri M, Todoerti K, et al. Molecular spectrum of BRAF, NRAS and KRAS gene mutations in plasma cell dyscrasias: implication for MEK-ERK pathway activation. Oncotarget. 2015;6(27): 24205-24217. 27. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545-15550. 28. Dib A, Gabrea A, Glebov OK, Bergsagel PL, Kuehl WM. Characterization of MYC translocations in multiple myeloma cell lines. J Natl Cancer Inst Monogr. 2008;(39):25-31. 29. Hartwell LH, Szankasi P, Roberts CJ, Murray AW, Friend SH. Integrating genetic

haematologica | 2021; 106(1)

approaches into the discovery of anticancer drugs. Science. 1997;278(5340):1064-1068. 30. Iliakis G, Murmann T, Soni A. Alternative end-joining repair pathways are the ultimate backup for abrogated classical non-homologous end-joining and homologous recombination repair: Implications for the formation of chromosome translocations. Mutat Res Genet Toxicol Environ Mutagen. 2015;793: 166-175. 31. Arana ME, Seki M, Wood RD, Rogozin IB, Kunkel TA. Low-fidelity DNA synthesis by human DNA polymerase theta. Nucleic Acids Res. 2008;36(11):3847-3856. 32. Wood RD, Doublie S. DNA polymerase theta (POLQ), double-strand break repair, and cancer. DNA Repair (Amst). 2016;44:22-32. 33. Zhuang J, Jiang G, Willers H, Xia F. Exonuclease function of human Mre11 promotes deletional nonhomologous end joining. J Biol Chem. 2009;284(44):30565-30573. 34. Taylor RM, Whitehouse CJ, Caldecott KW. The DNA ligase III zinc finger stimulates binding to DNA secondary structure and promotes end joining. Nucleic Acids Res. 2000;28(18):3558-3563. 35. Simsek D, Brunet E, Wong SY, et al. DNA ligase III promotes alternative nonhomologous end-joining during chromosomal translocation formation. PLoS Genet. 2011;7(6):e1002080. 36. Caracciolo DM, M.; Tagliaferri, P.; Tassone,P. Alternative non-homologous end joining repair: a master regulator of genomic instability in cancer[Review]. Prec Cancer Med. 2019;2:8. 37. Morgan GJ, Walker BA, Davies FE. The genetic architecture of multiple myeloma. Nat Rev Cancer. 2012;12(5):335-348. 38. Boboila C, Jankovic M, Yan CT, et al. Alternative end-joining catalyzes robust IgH locus deletions and translocations in the combined absence of ligase 4 and Ku70. Proc Natl Acad Sci U S A. 2010;107(7):3034-3039. 39. Mikulasova A, Ashby C, Tytarenko RG, et al. Microhomology-mediated end joining drives complex rearrangements and over expression of MYC and PVT1 in multiple myeloma. Haematologica. 2020:105(4): 1055-1066. 40. Fong PC, Boss DS, Yap TA, et al. Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA mutation carriers. N Engl J Med. 2009;361(2):123-134. 41. Howard SM, Yanez DA, Stark JM. DNA damage response factors from diverse path-

ways, including DNA crosslink repair, mediate alternative end joining. PLoS Genet. 2015;11(1):e1004943. 42. Chesi M, Robbiani DF, Sebag M, et al. AIDdependent activation of a MYC transgene induces multiple myeloma in a conditional mouse model of post-germinal center malignancies. Cancer Cell. 2008;13(2):167-180. 43. Kuehl WM, Bergsagel PL. MYC addiction: a potential therapeutic target in MM. Blood. 2012;120(12):2351-2352. 44. Weinhold N, Kirn D, Seckinger A, et al. Concomitant gain of 1q21 and MYC translocation define a poor prognostic subgroup of hyperdiploid multiple myeloma. Haematologica. 2016;101(3):e116-119. 45. Maifrede S, Martin K, PodszywalowBartnicka P, et al. IGH/MYC translocation associates with BRCA2 deficiency and synthetic lethality to PARP1 inhibitors. Mol Cancer Res. 2017;15(8):967-972. 46. Colicchia V, Petroni M, Guarguaglini G, et al. PARP inhibitors enhance replication stress and cause mitotic catastrophe in MYCN-dependent neuroblastoma. Oncogene. 2017;36(33):4682-4691. 47. Li Z, Owonikoko TK, Sun SY, et al. c-Myc suppression of DNA double-strand break repair. Neoplasia. 2012;14(12):1190-1202. 48. Muvarak N, Kelley S, Robert C, et al. c-MYC generates repair errors via increased transcription of alternative-NHEJ factors, LIG3 and PARP1, in tyrosine kinase-activated leukemias. Mol Cancer Res. 2015;13(4):699712. 49. Zhang W, Liu B, Wu W, et al. Targeting the MYCN-PARP-DNA damage response pathway in neuroendocrine prostate cancer. Clin Cancer Res. 2018;24(3):696-707. 50. Cottini F, Hideshima T, Suzuki R, et al. Synthetic lethal approaches exploiting DNA damage in aggressive myeloma. Cancer Discov. 2015;5(9):972-987. 51. Kotsantis P, Petermann E, Boulton SJ. Mechanisms of oncogene-induced replication stress: jigsaw falling into place. Cancer Discov. 2018;8(5):537-555. 52. Neri P, Ren L, Gratton K, et al. Bortezomibinduced "BRCAness" sensitizes multiple myeloma cells to PARP inhibitors. Blood. 2011;118(24):6368-6379. 53. Pawlyn C, Loehr A, Ashby C, et al. Loss of heterozygosity as a marker of homologous repair deficiency in multiple myeloma: a role for PARP inhibition? Leukemia. 2018;32(7): 1561-1566.

195

ARTICLE Ferrata Storti Foundation

Platelet Biology & its Disorders

Autoantibody-mediated desialylation impairs human thrombopoiesis and platelet lifespan

Irene Marini,1 Jan Zlamal,1 Christoph Faul,2 Ursula Holzer,3 Stefanie Hammer,4 Lisann Pelzl,1 Wolfgang Bethge,2 Karina Althaus1,4 and Tamam Bakchoul1,4

Transfusion Medicine, Medical Faculty of Tübingen, University Hospital of Tübingen; Department of Internal Medicine II, Oncology, Hematology, Clinical Immunology, Rheumatology and Pneumology, University Hospital of Tübingen; 3Department of Pediatric Hematology and Oncology, University Children's Hospital of Tübingen and 4 Center for Clinical Transfusion Medicine, University Hospital of Tübingen, Tübingen, Germany 1 2

Haematologica 2021 Volume 106(1):196-207

ABSTRACT

mmune thrombocytopenia is a common bleeding disease caused by autoantibody-mediated accelerated platelet clearance and impaired thrombopoiesis. Accumulating evidence suggests that desialylation affects platelet lifespan in immune thrombocytopenia. Herein, we report on novel effector functions of autoantibodies from patients with immune thrombocytopenia, which might interfere with the clinical picture of the disease. Data from our study show that a subgroup of autoantibodies is able to induce cleavage of sialic acid residues from the surface of human platelets and megakaryocytes. Moreover, autoantibody-mediated desialylation interferes with the interaction between cells and extracellular matrix proteins leading to impaired platelet adhesion and megakaryocyte differentiation. Using a combination of an ex vivo model of thrombopoiesis, a humanized animal model, and a clinical cohort study, we demonstrate that cleavage of sialic acid induces significant impairment of the production, survival as well as function of human platelets. These data may indicate that prevention of desialylation should be investigated in the future in clinical studies as a potential therapeutic approach to treat bleeding in immune thrombocytopenia.

Correspondence: TAMAM BAKCHOUL tamam.bakchoul@med.uni-tuebingen.de Received: August 26, 2019. Accepted: December 11, 2019. Pre-published: December 19, 2019. https://doi.org/10.3324/haematol.2019.236117

196

Introduction Immune thrombocytopenia (ITP) is an autoimmune disorder characterized by bleeding due to isolated thrombocytopenia with a platelet count below 100x109/L.1,2 The incidence of ITP ranges between 3.3-3.9/100,000 per year in adults, and between 1.9-6.4/100,000 per year in children.3,4 While the disorder has a brief course with spontaneous remission in the majority of children, most adult patients develop chronic ITP which can be associated with clinically significant bleeding including hemorrhages in the skin or mucous membranes, giving rise to petechiae and purpura and, rarely, intracranial manifestations.5,6 The exact mechanism of autoimmunity leading to ITP is still unclear, but includes an alteration of the balance between effector and regulatory cells.5 This imbalance results in a breakdown of the immune tolerance in ITP patients, causing functional alterations in the humoral and cellular immune systems which are thought to contribute to the pathogenesis of ITP. For a long time, it was thought that the low platelet count is caused solely by enhanced destruction of opsonized platelets in the spleen upon binding of the antiplatelet autoantibodies (AAb) to the glycoproteins expressed on the surface of platelets.6-8 However, platelet desialylation has recently been suggested to be an alternative mechanism of platelet destruction, which occurs in the liver through AshwellMorell receptors.9 This may explain the observation that a subgroup of ITP patients (up to 20%) is refractory to splenectomy and treatment with intravenous immune globulin G (IgG).10 Furthermore, in recent years more attention has been drawn to impaired platelet generation as another major cause of the low platelet count in ITP.11,12 This is due to the fact that megakaryocytes, the cells that release platelets into the circulation,13 express the same glycoproteins (GPIIb/IIa and GPIb/IX), present on the platelet’s surface, that are targeted by AAb in ITP.12,14 haematologica | 2021; 106(1)

Antibody-mediated desialylation

Based on these considerations, we hypothesized that the AAb-mediated desialylation of the glycoproteins expressed on platelets and megakaryocytes may interfere with platelet lifespan, affecting the clinical picture in ITP. In this study, we investigated a cohort of ITP patients with a focus on the impact of AAb-mediated desialylation on the survival and function of human platelets and megakaryocytes.

Methods Study cohort Patients were enrolled into the study after a retrospective review of their clinical records by two physicians to confirm the diagnosis of ITP. Experiments were performed using leftover serum from the ITP patients. Control sera were obtained from non-ITP thrombocytopenic patients (with hypoproliferative thrombocytopenia) as well as from randomly chosen healthy blood donors, after obtaining written consent. Studies involving human material were approved by the Ethics Committee of the Medical Faculty, Eberhard-Karls Universität, University Hospital of Tübingen, Germany, and were conducted in accordance with the Declaration of Helsinki. Additional details are available in the Online Supplementary Methods.

Antibody characterization Free and platelet-bound AAb were detected by the monoclonal antibody-specific immobilization of platelet antigens (MAIPA) assay, as described previously.15-17 AAb-mediated changes in the sialylation status of platelets and megakaryocytes were analyzed using a flow cytometer. In order to investigate the AAb-induced apoptosis, the mitochondrial inner transmembrane potential was measured by flow cytometry, as previously described with minor modifications.18 Furthermore, the impact of desialylation on the ability of platelets to adhere to human serum albumin (Kedrion, Barga, Italy), fibrinogen (Sigma Aldrich, Munich, Germany) or von Willebrand factor (VWF: Baxalta, Vienna, Austria) was assessed using an adhesion test, as previously described.19 Additional information is provided in the Online Supplementary Methods and Online Supplementary Figure S1.

In vivo analysis The impact of desialylation on the survival of human platelets was analyzed using the NSG mouse model, as described previously.20,21 All animal studies were conducted in Tübingen, Germany, and were approved by the responsible authorities of the State of Baden-Württemberg.

In vitro generation of human megakaryocytes Megakaryocytes were produced from CD34+ cells as previously described with minor modifications.22,23 In brief, CD34+ cells were magnetically isolated and cultured for differentiation into megakaryocytes in cell medium supplemented with 50 ng/mL of thrombopoietin for 14 days. The megakaryocytes generated in vitro were identified and characterized using a flow cytometry gating strategy based on DNA content, expression of CD41 and CD42a as well as viable markers, as previously described.23,24 In order to investigate the impact of AAb-induced desialylation on thrombopoiesis, megakaryocytes were incubated with IgG from ITP patients or healthy donors and proplatelet generation as well as platelet release were assessed by microscopy and flow cytometry, respectively, as previously described with minor modifications.25 Furthermore, the ability of megakaryocytes to adhere to human serum albumin, fibrinogen, VWF or collagen (Collagenhaematologica | 2021; 106(1)

Horm, Takeda, Linz, Austria) was verified and quantified after 2 hours (h) of incubation with IgG fractions from ITP patients or healthy donors, taking images from seven different microscope fields (x40, Olympus IX73, Olympus GmbH, Germany). Additional details are provided in the Online Supplementary Methods and Online Supplementary Figure S2.

Statistical analysis Statistical analyses were performed using GraphPad Prism 7 (La Jolla, CA, USA). A t-test was used to analyze normally distributed data, whereas a nonparametric test (Mann-Whitney test) was used when data were not normally distributed, as assessed by the D’Agostino and Pearson omnibus normality test. Group comparisons were performed using the Wilcoxon rank-sum test and the Fisher exact test with categorical variables. A P-value <0.05 was assumed to represent statistical significance.

Results Patients’ characteristics The clinical records of 100 patients with ITP were independently reviewed by two physicians. Forty percent of the patients were female and the median age at the time that laboratory testing was performed was 54 years (interquartile range [IQR], 0-86 years) (Table 1). The median platelet count was 32x109/L (IQR, 7-74x109/L) and 48/100 (48%) patients reported bleeding events. With regards to therapy, 35/100 (35%) patients had received steroids and 32/100 (32%) had been treated with high doses of intravenous IgG. Rituximab and thrombopoietin analogs were given as second-line treatments to seven (7%) and eight (8%) patients, respectively. On the day of laboratory investigation, MAIPA assay revealed positive results for free or bound AAb in 51/100 (51%) patients, while no AAb were detectable in the sera of 49/100 (49%) patients (Table 1 and Online Supplementary Table S1).

Prevalence and clinical relevance of desialylating autoantibodies According to the calculated cutoffs (fold increases of 1.75 and 1.42, for b-galactose and N-acetylglucosamine, respectively) sera from 35/100 ITP patients of the study cohort induced desialylation. In particular, sera from 28/100 (28%) patients caused the exposure of b-galactose, sera from 21/100 (21%) caused the exposure of N-acetylglucosamine, and sera from 14/100 (14%) led to the exposure of both residues (Figure 1A, B). A comparable frequency of desialylating AAb was observed between pediatric and adult patients (39% vs. 34%, respectively; P=0.8037). Furthermore, more sera from patients with secondary ITP induced desialylation in comparison to sera from patients with primary ITP (60% vs. 31%, respectively; P=0.0395). Lower platelet counts were documented in patients with AAb inducing desialylation compared to those with no desialylating ability (median [IQR], 13x109/L [542x109/L] vs. 46x109/L [11-81x109/L], respectively; P=0.0286) (Table 1). Most importantly, bleeding events were more frequently reported by patients with desialylating AAb (72% vs. 28%; P=0.0008) (Table 1). In particular, a greater tendency to develop petechiae (46% vs. 19%; P=0.0044) and hematomas (49% vs. 17%; P=0.0009) was observed in ITP patients with desialylating AAb compared to that in patients who tested negative in the lectin-binding assay. 197

I. Marini et al. A

Figure 1. Platelet desialylation ability of autoantibodies from patients with immune thrombocytopenia and the impact of glycoprotein specificity. (A, B) Sera from thrombocytopenic patients without immune thrombocytopenia (ITP, empty diamonds) and patients with ITP (black diamonds) were incubated with washed platelets from healthy donors. Cleavage of sialic acid and exposure of b-galactose (A) and N-acetylglucosamine (B) on test platelets was detected by flow cytometry and determined as fold increase of the cleavage compared to that when the platelets were incubated with control sera. (C, D) Distribution of desialylating autoantibodies according to the glycoprotein specificity as determined in a monoclonal antibody-specific immobilization of platelet antigens (MAIPA) assay. Data are presented as the mean of the fold increase for each sample when tested with platelet suspensions from three healthy donors and compared to control sera. The number of ITP sera tested is reported in the graphs. The values presented are the mean and range. FI: fold increase; PLT: platelet; AAB: autoantibody; GP: glycoprotein.

The impact of glycoprotein specificity on the ability to cleave sialic acid Among the 51 patients with a positive test result in the MAIPA assay for at least one glycoprotein, desialylation was observed in 26 (51%). Not only did sera and IgG containing AAb against GPIb/IX cause platelet desialylation, but those containing anti-GPIIb/IIIa AAb were also able to induce cleavage of sialic acid and expose b-galactose or N-acetylglucosamine (60% vs. 48% and 20% vs. 48%, respectively) (Figure 1C, D). Of note, cleavage of sialic acid was also observed for sera without detectable AAb in the MAIPA assay, leading to increased expression of bgalactose and N-acetylglucosamine in 7/49 (14%) and 5/49 (10%) cases, respectively (Figure 1C, D). This may indicate the presence of low-avidity AAb in these sera, as has been reported previously for anti-HPA-1a alloantibodies26,27 and complement-fixing AAb.28

Autoantibodies against GPIIb/IIIa induce desialylation via crosslinking Fcγ RIIA It is well established that AAb against GPIb/IX cause 198

platelet desialylation via Fc-independent release of sialidase from α granules of platelets.9 To verify the mechanism of platelet desialylation by sera containing antiGPIIb/IIIa AAb, the expression of b-galactose and Nacytelglucosamine was analyzed after incubation of washed platelet with IgG fractions from these sera. No significant difference in the ability to induce cleavage of sialic acid from the platelet surface was observed between patients’ sera and their corresponding IgG fractions (fold-increase, [mean ± standard error of mean, SEM]: 5.96±1.31 vs. 5.69±0.99; P=0.8714 and 3.69±0.76 vs. 5.20±0.21, P=0.1028, for b-galactose and N-acetylglucosamine, respectively) (Figure 2A, B). These results indicate that IgG AAb initiate the cleavage of the sialic acid. Next, washed platelets were preincubated with sialidase inhibitor. A significant reduction of the IgG-mediated cleavage of sialic acid was observed (% desialylation [mean ± SEM]: 100±0% vs. 63±2%; P=0.0001 and 100±0% vs. 57±6%; P=0.0002, for b-galactose and Nacetylglucosamine, respectively) (Figure 2C, D). Since the experiments were performed with IgG fractions and haematologica | 2021; 106(1)

Antibody-mediated desialylation Table 1. Demographic characteristics and clinical manifestations of the study cohort.

P-value

All ITP patients included in the study (n=100)

Patients with desialylating autoantibodies (n=35)

Patients with non-desialylating autoantibodies (n=65)

F=40 (40%) M=60 (60%) 54 (0-86)

F=15 (43%) M=20 (57%) 49 (15-65)

F=25 (38%) M=40 (62%) 55 (23-68)

Laboratory findings Platelet count x109/L at testing, median (IQR) Anti-GP IIb/IIIa Anti-GP Ib/IX Anti-GP IIb/IIIa + Anti-GP Ib/IX No detectable AAb

32 (7-74) 23/100 (23%) 5/100 (5%) 23/100 (23%) 49/100 (49%)

13 (5-42) 15/35 (42%) 3/35 (9%) 8/35 (23%) 9/35 (26%)

46 (11-81) 8/65 (12%) 2/65 (3%) 15/65 (23%) 40/65 (62%)

0.0286 0.0010 0.3400 0.9999 0.0008

Bleeding symptoms, n (%) Any bleeding Petechiae Hematoma Mucosal Epistaxis Gastrointestinal Central nervous system Transfusion

48/100 (48%) 28/100 (28%) 28/100 (28%) 15/100 (15%) 12/100 (12%) 2/100 (2%) 1/100 (1%) 10/100 (10%)

25/35 (72%) 16/35 (46%) 17/35 (49%) 7/35 (20%) 8/35 (23%) 1/35 (3%) 1/35 (3%) 8/35 (23%)

23/65 (28%) 12/65 (19%) 11/65 (17%) 8/65 (12%) 4/65 (6%) 1/65 (2%) 0/65 (0%) 2/65 (3%)

0.0008 0.0044 0.0009 0.3763 0.0208 0.9999 0.3434 0.0033

Demographics Gender, n (%)

MAIPA

bound or free

Age in years, median (IQR)

0.8308 0.1893

*Results of antibody testing are reported as positive when either the direct or indirect monoclonal antibody-specific immobilization of platelet antigens (MAIPA) assay revealed a positive reaction, indicating the presence of platelet-bound and free autoantibodies, respectively. F: female; M: male; IQR: interquartile range; GP: glycoprotein; AAb: autoantibodies.

washed platelets, these data suggest that intracellular and not plasma sialidase is responsible for the AAb-mediated desialylation of platelets. Next, we blocked the Fcγ RIIA signaling using the monoclonal antibody IV.3. Significant decreases in the expression of b-galactose and N-acetylglucosamine were observed (% desialylation [mean ± SEM]: 100±0% vs. 52±11%; P=0.0110 and 100±0% vs. 60±9%; P=0.0190, for b-galactose and N-acetylglucosamine, respectively) (Figure 2E, F). Taken together, these results indicate that IgG AAb against GPIIb/IIIa are able to crosslink Fcγ RIIA leading to the release of sialidase from the intracellular compartments of platelets.

Desialylation contributes to antibody-mediated destruction of human platelets Patients with desialylating AAb had significantly lower platelet counts compared to patients with sera that failed to induce cleavage of sialic acid (median platelet count [range], 13x109/L [0-106x109/L] vs. 46x109/L [1-234x109/L]; P=0.0121) (Figure 3A). A NSG mouse model was used to analyze the impact of AAb inducing desialylation on human platelet survival. In this model, 80-100x106/mL human platelets (~8-10% of the total circulating platelets) are usually detectable 30 minutes (min) after injection into the murine circulation, and more than 70% of these platelets circulate for up to 5 h in the absence of plateletreactive antibodies (data not shown). Upon injection of IgG fractions containing desialylating AAb from ITP patients, a rapid reduction of platelet survival was observed compared to that following injection of control IgG (% human platelet survival after 5 h [mean ± SEM]: 28±5% vs. 79±5%, respectively; P=0.0005) (Figure 3B). AAbmediated platelet clearance was partially dependent on desialylation. In fact, in the presence of a sialidase inhibitor a remarkable rescue of platelet survival was observed (% human platelet survival after 5 h [mean ± haematologica | 2021; 106(1)

SEM]: 28±5% vs. 45±3%; P=0.0191) (Figure 3B). These results indicate that the ability to induce cleavage of sialic acid contributes to AAb-mediated platelet destruction in vivo.

Antibody-mediated desialylation is not associated with platelet apoptosis To verify a potential correlation between platelet desialylation and apoptosis, mitochondrial inner transmembrane potential depolarization was investigated after incubation with desialylating and non-desialylating AAb. No correlation was observed between the ability of AAb to induce desialylation and apoptosis (r2: -0.333, P=0.3510 vs. r2-0.344, P=0.3310, for b-galactose and N-acetylglucosamine, respectively) (Figure 3C, D).

Desialylation contributes to autoantibody-mediated platelet dysfunction By testing IgG fractions from ITP patients, it was seen that inhibition of platelet adhesion was greater with desialylating AAb than with non-desialylating ones (% adherent cells/field [mean ± SEM]: 74±8% vs. 34±7%; P=0.0021 and 67±5% vs. 26±2%; P=0.0027, for fibrinogen and VWF, respectively) (Figure 4A, B). When platelets were pretreated with a pharmaceutically available sialidase inhibitor, significant improvements of platelet adhesion were observed (% adherent cells/field [mean ± SEM]: 36±5% vs. 86±6%; P=0.0001 and 26±2% vs. 67±11%; P=0.0203, for fibrinogen and VWF, respectively) (Figure 4C, D).

Desialylating autoantibodies impair in vitro thrombopoiesis To determine the ability of AAb to induce desialylation of glycoproteins on the surface of megakaryocytes, IgG fractions from ten ITP sera that induced a change in the 199

I. Marini et al. A

sialylation pattern on platelets were incubated with megakaryocytes derived from peripheral blood CD34+ cells. As shown in Figure 5A, IgG fractions from eight of the ten sera were able to induce cleavage of sialic acid and lead to exposure of b-galactose and N-acetylglucosamine. More interestingly, the ability to initiate desialylation was not restricted to AAb against GPIb/IX. In fact, anti-GPIIb/IIIa AAb were also able to induce megakaryocyte desialylation (Figure 5B, C). Next, we investigated the impact of AAb-mediated desialylation on proplatelet formation. Megakaryocytes that were cultured in the presence of IgG fractions from desialylating AAb showed significantly less ability to form proplatelet extensions compared to those cultured in the presence of control IgG (% proplatelet-forming megakaryocytes, [mean ± SEM]: 100±0% vs. 42±12%;, P=0.0037) 200

Figure 2. Mechanism of antiGPIIb/IIIa-mediated platelet desialylation. Platelets from healthy donors were incubated with sera or immunoglobulin G (IgG) fractions from patients with immune thrombocytopenia (ITP) with non-desialylating (black columns) and desialylating (gray columns) autoantibodies (AAB). Cleavage of sialic acid and exposure of b-galactose (A) and N-acetylglucosamine (B) on test platelets was detected by flow cytometer. Note that there was no difference in the desialylation ability between sera and corresponding IgG fractions. (C-F) To verify the mechanism of platelet desialylation induced by antiGPIIb/IIIa AAB, platelets were preincubated with buffer or a sialidase inhibitor (C,D), or the monoclonal antibody IV.3 (E, F) prior to the addition of IgG fractions from sera containing desialylating AAB against GPIIb/IIIa complexes. IgG from three different ITP patients were tested in four independent experiments using washed platelets from healthy donors. Data are presented as the mean ± standard error of mean of the measured fold increase compared to the data obtained for IgG from three healthy donors used as controls. ns, not significant, *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001. FI: fold increase; PLT: platelet; GP: glycoprotein; mAB: monoclonal antibody.

(Figure 6A). More interestingly, the ability to form proplatelets was largely rescued when sialidase was inhibited (% proplatelet-forming megakaryocytes [mean ± SEM]: 42±12% vs. 90±9%; P=0.0326) (Figure 6A-C).

Autoantibody-mediated desialylation interferes with the interaction between megakaryocytes and extracellular matrix proteins Interaction between megakaryocytes and extracellular matrix proteins was reported to be essential for the differentiation of these cells and their ability to form proplatelet extensions.29 When megakaryocytes were incubated with desialylating AAb, it was found that there was a significant reduction of cells that adhered to fibrinogen and VWF, but not collagen, compared to when the megakaryocytes were incubated with non-desialylating haematologica | 2021; 106(1)

Antibody-mediated desialylation

Figure 3. The impact of desialylating autoantibodies on human platelet survival. (A) Platelet count was determined on the day of sample collection for laboratory investigations for patients with immune thrombocytopenia (ITP) with non-desialylating (black circles) and desialylating (gray circles) autoantibodies (AAB). Symbols represent the platelet count of each patient. Lines indicate the median and range of platelet counts in each group, tested in four different experiments using washed platelets from three different healthy donors. (B) The direct impact of desialylating AAB on platelet survival was analyzed using the NSG mouse model. Freshly isolated human platelets were treated with buffer (filled symbols) or a sialidase inhibitor (empty symbols) and injected into NGS mice via the lateral tail vein. After 30 minutes (min), blood was taken to determine the baseline value of circulating human platelets (100%). Subsequently, immunoglobulin G (IgG) fractions of desialylating ITP sera (gray lines) or sera from healthy donors (black lines) were injected and further blood samples were taken at 60, 120, and 300 min after baseline, as indicated. The data shown are the mean percentage ± standard error of mean of the remaining human platelets in the mouse circulation after injection of four different AAB. (C, D) The correlation between the ability of AAB to induced desialylation and apoptosis was investigated by incubating IgG fractions from non-desialylating (black circles) and desialylating (gray circles) AAB with washed platelets from healthy donors and exposure of b-galactose (C) and N-acetylglucosamine (D), and changes in mitochondrial inner transmembrane potential were analyzed. Spearman r : -0.333, P=0.3510 vs. r :-0.344, P=0.3310, for b-galactose and N-acetylglucosamine, respectively. Data are presented as mean of the fold increase evaluated in three different experiments using washed platelets from three healthy donors. *P<0.05 and ***P<0.001. PLT: platelet; FI: fold increase. 2

AAb (% adherent cells/field [mean ± SEM]: 88±8% vs. 58±8%, P=0.0479; 79±8% vs. 47±4%, P=0.0232 and 90±5% vs. 88±4%, P=0.8080, for fibrinogen, VWF and collagen, respectively) (Figure 7A-C). This effect was dependent on the ability of the AAb to induce desialylation. In fact, in the presence of a sialidase inhibitor there were significant improvements of the ability of megakaryocytes to adhere to both proteins (% adherent cells/field [mean ± SEM]: 50±6% vs. 88±5%, P=0.0022 and 47±4% vs. 82±7%, P=0.0020, for fibrinogen and VWF, respectively) (Figure 7D, E). In contrast, no difference was observed for cell adhesion to collagen (% adherent cells/field [mean ± SEM]: 90±4% vs. 88±6%; P=0.7083) (Figure 7F). To confirm that sialylation plays an important role in megakaryocyte differentiation, exogenous sialidase was haematologica | 2021; 106(1)

added to the culture medium at the beginning of the differentiation stage. As shown in Online Supplementary Figure S3, the numbers of total megakaryocytes as well as mature megakaryocytes were lower in the presence of sialidase than in the presence of buffer (CD41+ megakaryocytes/µL [mean ± SEM]: 69±11 cells/mL vs. 27±4 cells/mL, P=0.0233; CD41+/CD42a+ megakaryocytes/mL [mean ± SEM]: 63±11 cells/mL vs. 25±4 cells/mL, P=0.0312, respectively) (Online Supplementary Figure S3A, B). Interestingly, the ratio of mature cells was similar (ratio of mature/total megakaryocytes, [mean ± SEM]: 0.91±0.03 vs. 0.91±0.05; P=0.9540) (Online Supplementary Figure S3C). Importantly, fewer platelets were released into the supernatant after megakaryocyte desialylation (platelets/mL [mean ± SEM]: 11±2 cells/mL vs. 4±1 cells/mL; P=0.0384) (Online Supplementary Figure S3D). These results suggest that sialylation might be 201

I. Marini et al.

important for the differentiation of CD34+ cells into mature megakaryocytes.

Discussion Thrombocytopenia in ITP patients is thought to be due to multiple alterations of the immune system leading to

increased platelet destruction and impaired thrombopoiesis.2,25,30,31 In this report, we describe novel effector functions in which desialylating AAb from ITP patients interfere with platelet lifespan and function. First, we confirm that desialylation contributes to AAb-mediated destruction of human platelets and uncover a new mechanism, in which IgG anti-GPIIb/IIIa AAb induce crosslinking of FcÎł RIIA leading to the release of sialidase from

Figure 4. The impact of desialylating autoantibodies on platelet function. Washed platelets were preincubated with immunoglobulin G (IgG) fractions from healthy donors (white columns), patients with immune thrombocytopenia (ITP) with non-desialylating (black columns) and ITP patients with desialylating (gray columns) autoantibodies (AAB). Cells were then seeded on slides coated with fibrinogen (A) or von Willebrand factor (VWF, B). (C, D) To verify the role of desialylation in AABinduced impairment of platelet function, cells were preincubated with buffer (filled gray columns) or sialidase inhibitors (gray-white columns) prior to platelet adhesion to slides coated with fibrinogen (C) or VWF (D). Photographs of randomly chosen fields were captured to calculate the percentage of adherent cells compared to those in the control condition (considered as 100%). The number of adherent platelets in the sample that was incubated with IgG from healthy donor in each individual experiment was considered 100%. Data are shown as the mean of the percentage of adherent platelets Âą standard error of mean in the presence of AAB from three different ITP patients tested in three independent experiments with platelets from three different donors; ns, not significant, *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001. Images from representative fields of platelets adhered to fibrinogen or VWF, in the presence of control IgG (left panel) or ITP AAB (middle and right panels), respectively, are shown (scale bar: 10 mm). PLT: platelet.

202

haematologica | 2021; 106(1)

Antibody-mediated desialylation

human platelets. Second, our data demonstrate that desialylation induced by AAb causes significant impairment of platelet function. This finding seems to have considerable relevance, since bleeding was observed more frequently in ITP patients with desialylating AAb in our clinical study. Most importantly, and to the best of our knowledge, we provide for the first time evidence that AAb-mediated desialylation inhibits megakaryocyte adhesion to bone marrow extracellular matrix proteins, leading to limited cell differentiation and reduced ability to form proplatelets. The central role of AAb produced by B cells in ITP has been clearly established.15,31 In contrast, the exact mechanisms by which AAb interfere with platelet lifespan are still a matter of research. According to the current under-

standing of ITP, AAb-mediated platelet destruction can be caused in one or more ways, such as phagocytosis, apoptosis, or complement activation.28,32,33 While these mechanisms are related to the Fc-mediated effector functions of the AAb, recent studies proposed desialylation as an Fcindependent pathway of platelet elimination.9,34,35 Although AAb against GPIb/IX are generally considered to be the major mediators of platelet desialylation,36 systematic clinical studies to analyze the impact of glycoprotein specificity are rare.37,38 In contrast to previous studies, in our study desialylation was frequently observed in association with AAb against GPIIb/IIIa. Our data show that anti-GPIIb/IIIa-induced desialylation is mediated by FcÎł RIIA crosslinking. F(abâ&#x20AC;&#x2122;) fragments lacking the Fc-domain 2

C Figure 5. The ability of autoantibodies to desialylate CD34+-derived megakaryocytes. (A) CD34+-derived megakaryocytes were treated with immunoglobulin G (IgG) fractions from ten sera containing platelet-desialylating autoantibodies (AAB). Cleavage of sialic acid and exposure of b-galactose and N-acetylglucosamine on megakaryocytes were measured by flow cytometer as fold increase compared to that produced with control IgG. (B, C) The distribution of desialylating AAB according to the glycoprotein specificity, as determined in a monoclonal antibody-specific immobilization of platelet antigens (MAIPA) assay. Each IgG sample was evaluated in three independent experiments using three different cell lines. The values presented are the mean and range. FI: fold increase; GP: glycoprotein.

haematologica | 2021; 106(1)

203

I. Marini et al.

showed no desialylation ability (data not shown). Of note, monoclonal antibodies against GPIIb/IIIa failed to cause desialylation of murine platelets. One possible explanation for this discrepancy is that murine platelets, in contrast to human platelets, do not express the FcÎł RIIA on their surface. In order to investigate the biological relevance of human platelet desialylation, we first used the NSG animal model of passive immune thrombocytopenia. In this model, all tested desialylating AAb were efficient in removing human platelets from mouse circulation, regardless of their glycoprotein specificity. Importantly, when sialidase was inhibited, platelet survival was significantly enhanced. Considering that ITP AAb have been shown to be capable of inducing platelet apoptosis and a subgroup of them concomitantly desialylation,39-41 we decided to analyze the impact of desialylating AAb on platelet apoptosis. In our study, we observed that 40% of desialylating ITP AAb induced apoptosis signals in test platelets. However, our data showed no statistically significant correlation between the ability of desialylation and apoptosis, suggesting that ITP AAb-mediated desialylation may contribute to the impaired lifespan of platelets in an apoptosis-independent way. In addition to accelerated platelet clearance, accumulat-

ing evidence suggests that AAb are responsible for abnormal platelet production from megakaryocytes in ITP patients.30 Recent studies have shown that ITP patients have significantly fewer proplatelet-forming megakaryocytes compared to healthy controls,42 and that AAb have the ability to inhibit proplatelet formation by megakaryocytes.12 Moreover, animal studies indicate that antiplatelet AAb impair migration and adhesion of megakaryocytes without affecting these cellsâ&#x20AC;&#x2122; viability and proliferation,43 or their number in the bone marrow.44 In our study, desialylating AAb from ITP patients were able to induce cleavage of sialic acid from the surface of megakaryocytes leading to a significant reduction in their ability to adhere to extracellular matrix proteins, fibrinogen and VWF, in a sialylation-dependent manner. Abnormal interactions between these glycoprotein receptors of the megakaryocytic lineage and their corresponding extracellular matrix ligands are thought to be responsible for impaired megakaryocyte development and reduced platelet production as, for example, shown in patients carrying functional mutations in GPIIb/IIIa45 and GPIb/IX.45,46 In addition, data from ex vivo models of thrombopoiesis demonstrated that the binding of fibrinogen and VWF to GPIIb/IIIa and GPIb/IX, respectively, enhances proplatelet production.29

204

Figure 6. The impact of desialylating autoantibodies on proplatelet formation. (A) After 14 days of differentiation, megakaryocytes were cultured for 24 hours in the presence of immunoglobulin fractions from healthy donors (white column) or desialylating autoantibodies (AAB, gray column). To test the impact of desialylation on proplatelet formation, cells were treated with a sialidase inhibitor prior to their incubation with desialylating AAB (graywhite column). The percentage of proplatelet-forming megakaryocytes was determined as the number of megakaryocytes presenting at least one cytoplasmic projection with respect to the total number of megakaryocytes. Results were normalized to control immunoglobulin G (IgG) as 100%. IgG samples from three different patients with immune thrombocytopenia were evaluated in three independent experiments using three different megakaryocyte cell lines and the mean values are shown. (B, C) Representative bright field (B) and immune fluorescence microscopic images (C) of megakaryocyte-forming proplatelet extensions in the presence of IgG fractions from healthy donors (left panel), desialylating AAB (middle) or in the presence of a sialidase inhibitor (right panel). Arrowheads indicate proplatelet extensions generated from megakaryocytes. Green staining, filamin; blue staining, nuclei. Scale bar: 20 mm. The data shown are the mean Âą standard error of mean; ns, not significant, *P<0.05 and **P<0.01. proPLT: proplatelet; MK: megakaryocyte.

haematologica | 2021; 106(1)

Antibody-mediated desialylation

A recently published report described that AAb from ITP patients interfere with the interactions of megakaryocytes with different components of the bone marrow extracellular microenvironment, leading to reduced proplatelet formation.47 Our study confirms this observation and uncovers a new mechanism for the impaired thrombopoiesis in ITP. Desialylating AAb caused a significant reduction of proplatelets by inducing the cleavage of sialic acid from the megakaryocyte surface. We also found that sialylation is important for the differentiation of megakaryocytes. Along the same line, lower platelet counts were found in ITP patients with desialylating AAb, suggesting that not only may platelet elimination be increased in this group of patients, but thrombopoiesis might also be more severely affected. Of note, data from a recent clinical study showed that oseltamivir, a sialidase inhibitor, improved the response to thrombopoietin agonists in more than half of ITP patients with AAb against GPIb/IX.36 The concept of preventing desialylation in order to improve thrombopoiesis in ITP should be verified in the future. Of note, reduced sialylation of serum IgG has been observed in autoimmune diseases such as rheumatoid arthritis and systemic lupus erythematosus.48 In our study, the exposure of b-galactose and N-acetylglucosamine was reduced when platelets were preincubated with a sialidase inhibitor, indicating that the cleavage of sialic acid through

endogenous sialidase is responsible for the increased lectin binding in our test system. This finding makes it very unlikely that our observation was modulated by the sialylation status of the AAb. Although thrombocytopenia is considered to be a general feature in ITP, the clinical manifestations of the disease are very heterogeneous, with no biomarker available to predict the risk of bleeding. The presence of AAb might influence the clinical picture in ITP patients not only by causing a lower platelet count but also by reducing platelet responsiveness. In fact, functional platelet defects were observed in ITP patients with a bleeding tendency.49,50 In our study, desialylating AAb caused impairment of platelet adhesion to fibrinogen as well as VWF suggesting a possible impact on the hemostatic function of platelets. It is worth noting that a greater bleeding tendency was observed in ITP patients with desialylating AAb than in those with AAb that failed to induce cleavage of sialic acid from the surface of platelets. Although prospective studies are still needed to verify a potential correlation, this finding may indicate a clinical relevance of our laboratory observations, suggesting that the treatment defined for each patient must take into consideration all of these aspects. Furthermore, our data as well as those from other recent studies36,41 suggest that a combination of therapies targeting the different pathological mechanisms, including platelet desialylation, by using, for example, a sialidase

Figure 7. Interference of desialylating autoantibodies with the interaction between megakaryocytes and bone marrow extracellular matrix proteins. (A-C) Megakaryocytes were seeded on fibrinogen (A), von Willebrand factor (VWF, B) or collagen (C) -coated coverslips in the presence of immunoglobulin G fractions from healthy donors (white columns), non-desialylating (black columns) or desialylating (gray columns) autoantibodies (AAB). Photographs of randomly chosen fields were captured to calculate the percentage of adherent cells compared to that in the control condition (considered as 100%). (D-F) To verify the role of desialylation in the impairment of megakaryocyte adhesion, a sialidase inhibitor (gray-white columns) was added to the cell culture prior to the incubation with AAB. Cells were then seeded on fibrinogen (D), VWF (E) or collagen (F)-coated coverslips. Data shown are the mean of the percentage adherent megakaryocytes Âą standard error of mean for AAB from three different patients with immune thrombocytopenia tested in three independent experiments with three different cell lines; ns, not significant, *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001. MK: megakaryocyte; IgG: immunoglobulin G.

haematologica | 2021; 106(1)

205

I. Marini et al.

inhibitor, might be a more efficient clinical approach producing better, prolonged outcomes. Since our study was retrospective and no follow-up data were available, we cannot draw definitive conclusions on the impact of desialylating AAb on response to therapy and disease progression or on the explicit effect of megakaryocyte desialylation on thrombopoiesis in ITP. The final proof of our concept regarding the contribution of AAb-mediated desialyation to the clinical picture in patients with ITP can only be provided by larger prospective clinical studies. In conclusion, our results indicate that desialylation plays an important role in platelet destruction and impaired thrombopoiesis in ITP; and demonstrate that receptor sialyation status is of considerable functional relevance, since its changes not only affect platelet survival but also interfere with megakaryocyte differentiation and platelet generation. These novel findings highlight the multiple effects of AAb in ITP and add to the existing evidence that ITP is a group of disorders rather than one disease sharing common characteristics, namely loss of

References 1. Provan D, Stasi R, Newland AC, et al. International consensus report on the investigation and management of primary immune thrombocytopenia. Blood. 2010;115(2):168-186. 2. Neunert C, Lim W, Crowther M, et al. The American Society of Hematology 2011 evidence-based practice guideline for immune thrombocytopenia. Blood. 2011;117(16): 4190-4207. 3. Moulis G, Palmaro A, Montastruc JL, Godeau B, Lapeyre-Mestre M, Sailler L. Epidemiology of incident immune thrombocytopenia: a nationwide population-based study in France. Blood. 2014;124(22):33083315. 4. Bennett CM, Neunert C, Grace RF, et al. Predictors of remission in children with newly diagnosed immune thrombocytopenia: data from the Intercontinental Cooperative ITP Study Group Registry II participants. Pediatr Blood Cancer. 2018;65(1). 5. McKenzie CG, Guo L, Freedman J, Semple JW. Cellular immune dysfunction in immune thrombocytopenia (ITP). Br J Haematol. 2013;163(1):10-23. 6. Shulman NR, Marder VJ, Weinrach RS. Similarities between known antiplatelet antibodies and the factor responsible for thrombocytopenia in idiopathic purpura. Physiologic, serologic and isotopic studies. Ann N Y Acad Sci. 1965;124(2):499-542. 7. Rank A, Weigert O, Ostermann H. Management of chronic immune thrombocytopenic purpura: targeting insufficient megakaryopoiesis as a novel therapeutic principle. Biologics. 2010;4:139-145. 8. Ku FC, Tsai CR, Der Wang J, Wang CH, Chang TK, Hwang WL. Stromal-derived factor-1 gene variations in pediatric patients with primary immune thrombocytopenia. Eur J Haematol. 2013;90(1):25-30. 9. Li J, van der Wal DE, Zhu G, et al. Desialylation is a mechanism of Fc-independent platelet clearance and a therapeutic target in immune thrombocytopenia. Nat Commun. 2015;6:7737.

206

immune tolerance toward platelet and megakaryocyte antigens and impaired platelet lifespan, leading to thrombocytopenia with increased bleeding tendency. Disclosures No conflicts of interests to disclose. Contributions I.M. and TB designed the study; IM, JZ, KA, LP, and TB performed the experiments; IM, JZ, SH, WB, CF UH KA and TB collected and analyzed the data, interpreted the results and wrote the manuscript. All authors read and approved the manuscript. Acknowledgments The authors would like to thank Flavianna Rigoni for her excellent technical assistance and Stephen Bosher a native English speaker, for editing the manuscript. This work was supported by a research grant from the Medical Faculty of Tübingen, University Hospital of Tübingen, Germany, to IM and TB as well as one from Karina Althaus (TÜFF-Program 2563-0-0).

10. Vianelli N, Galli M, de Vivo A, et al. Efficacy and safety of splenectomy in immune thrombocytopenic purpura: long-term results of 402 cases. Haematologica. 2005;90(1):72-77. 11. Lev PR, Grodzielski M, Goette NP, et al. Impaired proplatelet formation in immune thrombocytopenia: a novel mechanism contributing to decreased platelet count. Br J Haematol. 2014;165(6):854-864. 12. Iraqi M, Perdomo J, Yan F, Choi PY, Chong BH. Immune thrombocytopenia: antiplatelet autoantibodies inhibit proplatelet formation by megakaryocytes and impair platelet production in vitro. Haematologica. 2015;100(5):623-632. 13. Thon JN, Montalvo A, Patel-Hett S, et al. Cytoskeletal mechanics of proplatelet maturation and platelet release. J Cell Biol. 2010;191(4):861-874. 14. McMillan R, Wang L, Tomer A, Nichol J, Pistillo J. Suppression of in vitro megakaryocyte production by antiplatelet autoantibodies from adult patients with chronic ITP. Blood. 2004;103(4):1364-1369. 15. Kiefel V, Freitag E, Kroll H, Santoso S, Mueller-Eckhardt C. Platelet autoantibodies (IgG, IgM, IgA) against glycoproteins IIb/IIIa and Ib/IX in patients with thrombocytopenia. Ann Hematol. 1996;72(4):280-285. 16. Kiefel V, Scheld S, Freitag E, Kroll H, Mueller-Eckhardt C. [Two modifications of MAIPA assays for the demonstration of platelet antibodies]. Beitr Infusionsther Transfusionsmed. 1994;32:237-239. 17. Kiefel V. Transfusionsmedizin und Immunhämatologie. Springer. 2010:p624. 18. Chen M, Yan R, Zhou K, et al. Akt-mediated platelet apoptosis and its therapeutic implications in immune thrombocytopenia. Proc Natl Acad Sci U S A. 2018;115(45):E10682E10691. 19. Marini I, Aurich K, Jouni R, et al. Cold storage of platelets in additive solution: the impact of residual plasma in apheresis platelet concentrates. Haematologica. 2019;104(1):207-214. 20. Bakchoul T, Fuhrmann J, Chong BH, Bougie D, Aster R, Subcommittee on Platelet Immunology. Recommendations for the use of the non-obese diabetic/severe combined

immunodeficiency mouse model in autoimmune and drug-induced thrombocytopenia: communication from the SSC of the ISTH. J Thromb Haemost. 2015;13(5):872-875. 21. Fuhrmann J, Jouni R, Alex J, et al. Assessment of human platelet survival in the NOD/SCID mouse model: technical considerations. Transfusion. 2016;56(6):1370-1376. 22. Mattia G, Vulcano F, Milazzo L, et al. Different ploidy levels of megakaryocytes generated from peripheral or cord blood CD34+ cells are correlated with different levels of platelet release. Blood. 2002;99(3):888-897. 23. Marini I, Rigoni F, Zlamal J, et al. Blood donor-derived buffy coat to produce platelets in vitro. Vox Sang. 2020;115(1):94-102. 24. Liu Y, Wang Y, Gao Y, et al. Efficient generation of megakaryocytes from human induced pluripotent stem cells using food and drug administration-approved pharmacological reagents. Stem Cells Transl Med. 2015;4(4):309-319. 25. Di Buduo CA, Alberelli MA, Glembostky AC, et al. Abnormal proplatelet formation and emperipolesis in cultured human megakaryocytes from gray platelet syndrome patients. Sci Rep. 2016;6:23213. 26. Peterson JA, Kanack A, Nayak D, et al. Prevalence and clinical significance of lowavidity HPA-1a antibodies in women exposed to HPA-1a during pregnancy. Transfusion. 2013;53(6):1309-1318. 27. Bakchoul T, Kubiak S, Krautwurst A, et al. Low-avidity anti-HPA-1a alloantibodies are capable of antigen-positive platelet destruction in the NOD/SCID mouse model of alloimmune thrombocytopenia. Transfusion. 2011;51(11):2455-2461. 28. Najaoui A, Bakchoul T, Stoy J, et al. Autoantibody-mediated complement activation on platelets is a common finding in patients with immune thrombocytopenic purpura (ITP). Eur J Haematol. 2012;88 (2):167-174. 29. Balduini A, Pallotta I, Malara A, et al. Adhesive receptors, extracellular proteins and myosin IIA orchestrate proplatelet formation by human megakaryocytes. J Thromb Haemost. 2008;6(11):1900-1907. 30. Marini I, Bakchoul T. Pathophysiology of

haematologica | 2021; 106(1)

Antibody-mediated desialylation

autoimmune thrombocytopenia: current insight with a focus on thrombopoiesis. Hamostaseologie. 2019;39(3):227-237. 31. Cines DB, Liebman HA. The immune thrombocytopenia syndrome: a disorder of diverse pathogenesis and clinical presentation. Hematol Oncol Clin North Am. 2009;23(6):1155-1161. 32. Tsubakio T, Tani P, Curd JG, McMillan R. Complement activation in vitro by antiplatelet antibodies in chronic immune thrombocytopenic purpura. Br J Haematol. 1986;63(2):293-300. 33. Peerschke EI, Andemariam B, Yin W, Bussel JB. Complement activation on platelets correlates with a decrease in circulating immature platelets in patients with immune thrombocytopenic purpura. Br J Haematol. 2010;148(4):638-645. 34. Yan R, Chen M, Ma N, et al. Glycoprotein Ibα clustering induces macrophage-mediated platelet clearance in the liver. Thromb Haemost. 2015;113(1):107-117. 35. Quach ME, Dragovich MA, Chen W, et al. Fc-independent immune thrombocytopenia via mechanomolecular signaling in platelets. Blood. 2018;131(7):787-796. 36. Revilla N, Corral J, Minano A, et al. Multirefractory primary immune thrombocytopenia; targeting the decreased sialic acid content. Platelets. 2019;30(6):743-751. 37. Tao L, Zeng Q, Li J, et al. Platelet desialylation correlates with efficacy of first-line therapies for immune thrombocytopenia. J Hematol Oncol. 2017;10(1):46.

haematologica | 2021; 106(1)

38. Peng J, Ma SH, Liu J, et al. Association of autoantibody specificity and response to intravenous immunoglobulin G therapy in immune thrombocytopenia: a multicenter cohort study. J Thromb Haemost. 2014;12(4):497-504. 39. Mason KD, Carpinelli MR, Fletcher JI, et al. Programmed anuclear cell death delimits platelet life span. Cell. 2007;128(6):11731186. 40. Deng G, Yu S, Li Q, et al. Investigation of platelet apoptosis in adult patients with chronic immune thrombocytopenia. Hematology. 2017;22(3):155-161. 41. Grodzielski M, Goette NP, Glembotsky AC, et al. Multiple concomitant mechanisms contribute to low platelet count in patients with immune thrombocytopenia. Sci Rep. 2019;9(1):2208. 42. Riviere E, Viallard JF, Guy A, et al. Intrinsically impaired platelet production in some patients with persistent or chronic immune thrombocytopenia. Br J Haematol. 2015;170(3):408-415. 43. Zeng DF, Chen F, Wang S, et al. Autoantibody against integrin αvb3 contributes to thrombocytopenia by blocking the migration and adhesion of megakaryocytes. J Thromb Haemost. 2018;16(9): 1843-1856. 44. Guo L, Kapur R, Aslam R, et al. Antiplatelet antibody-induced thrombocytopenia does not correlate with megakaryocyte abnormalities in murine immune thrombocytopenia. Scand J Immunol. 2018;88(1):e12678.

45. Kunishima S, Kashiwagi H, Otsu M, et al. Heterozygous ITGA2B R995W mutation inducing constitutive activation of the αIIbb3 receptor affects proplatelet formation and causes congenital macrothrombocytopenia. Blood. 2011;117 (20):5479-5484. 46. Strassel C, Eckly A, Leon C, et al. Intrinsic impaired proplatelet formation and microtubule coil assembly of megakaryocytes in a mouse model of Bernard-Soulier syndrome. Haematologica. 2009;94(6):800-810. 47. Grodzielski M, Di Buduo CA, Goette NP, et al. Autoantibodies in immune thrombocytopenia affect the physiological interaction between megakaryocytes and bone marrow extracellular matrix proteins. Br J Haematol. 2018;183(2):319-323. 48. Seeling M, Bruckner C, Nimmerjahn F. Differential antibody glycosylation in autoimmunity: sweet biomarker or modulator of disease activity? Nat Rev Rheumatol. 2017;13(10):621-630. 49. van Bladel ER, Laarhoven AG, van der Heijden LB, et al. Functional platelet defects in children with severe chronic ITP as tested with 2 novel assays applicable for low platelet counts. Blood. 2014;123(10):15561563. 50. Frelinger AL 3rd, Grace RF, Gerrits AJ, Carmichael SL, Forde EE, Michelson AD. Platelet function in ITP, independent of platelet count, is consistent over time and is associated with both current and subsequent bleeding severity. Thromb Haemost. 2018;118(1):143-151.

207

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):208-219

Platelet Biology & Its Disorder

Low-dose Btk inhibitors selectively block platelet activation by CLEC-2

Phillip L.R. Nicolson,1,2 Sophie H. Nock,3 Joshua Hinds,1 Lourdes Garcia-Quintanilla,1 Christopher W. Smith,1 Joana Campos,1 Alexander Brill,1,4 Jeremy A. Pike,1,5 Abdullah O. Khan,1 Natalie S. Poulter,1,5 Deidre M. Kavanagh,1,5 Stephanie Watson,1 Callum N. Watson,1 Hayley Clifford,6 Aarnoud P. Huissoon,6 Alice Y. Pollitt,3 Johannes A. Eble,7 Guy Pratt,2 Steve P. Watson1,5 and Craig E. Hughes3

Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK; 2Department of Haematology, Queen Elizabeth Hospital, Birmingham, UK; 3Institute for Cardiovascular and Metabolic Research, Harborne Building, University of Reading, Reading, UK; 4Department of Pathophysiology, Sechenov First Moscow State Medical University (Sechenov University), Moscow, Russia; 5Centre of Membrane Proteins and Receptors (COMPARE), Universities of Birmingham and Nottingham, Midlands, UK; 6Department of Immunology, Heartlands Hospital, Birmingham, UK and 7Institute for Physiological Chemistry and Pathobiochemistry, University of Münster, Müz nster, Germany 1

ABSTRACT

Correspondence: PHILLIP LR NICOLSON p.nicolson@bham.ac.uk

nhibitors of Bruton tyrosine kinase (Btk) have been proposed as novel antiplatelet agents. In this study we show that low concentrations of the Btk inhibitor ibrutinib block CLEC-2-mediated activation and tyrosine phosphorylation including Syk and PLCγ2 in human platelets. Activation is also blocked in patients with X-linked agammaglobulinemia (XLA) caused by a deficiency or absence of Btk. In contrast, the response to GPVI is delayed in the presence of low concentrations of ibrutinib or in patients with XLA, and tyrosine phosphorylation of Syk is preserved. A similar set of results is seen with the second-generation inhibitor, acalabrutinib. The differential effect of Btk inhibition in CLEC2 relative to GPVI signaling is explained by the positive feedback role involving Btk itself, as well as ADP- and thromboxane A -mediated activation of P2Y and TP receptors, respectively. This feedback role is not seen in mouse platelets and, consistent with this, CLEC-2-mediated activation is blocked by high but not by low concentrations of ibrutinib. Nevertheless, thrombosis was absent in eight out of 13 mice treated with ibrutinib. These results show that Btk inhibitors selectively block activation of human platelets by CLEC-2 relative to GPVI suggesting that they can be used at low doses in patients to target CLEC-2 in thrombo-inflammatory disease. 2

STEVE P WATSON s.p.watson@bham.ac.uk CRAIG E HUGHES c.e.hughes@reading.ac.uk Received: February 8, 2019. Accepted: January 15, 2020. Pre-published: January 16, 2020. https://doi.org/10.3324/haematol.2019.218545

208

Introduction The C-type lectin receptor CLEC-2 is expressed at high levels on platelets and at low levels on a subpopulation of hematopoietic cells.1 Its only known physiological ligand is podoplanin.2 Recent studies have demonstrated a critical role for CLEC-2 in inflammation-driven venous thrombosis,3,4 generating interest in drugs that target CLEC-2 in thrombo-inflammatory disorders while preserving hemostasis. This interest has been further fueled by the demonstration that CLEC-2 plays a minimal or negligible role in hemostasis.5 CLEC-2 has a cytoplasmic tail containing a single tyrosine in a conserved YxxL sequence in a motif that represents half of an immunoreceptor tyrosine-based activation motif (ITAM), known as a hemITAM. The conserved tyrosine is phosphorylated by the tyrosine kinase Syk.6 This leads to recruitment of Syk through binding of its tandem SH2 domains and initiation of a downstream signaling cascade involving Src, Syk and Tec (including Bruton tyrosine kinase [Btk]) kinases, various adapter proteins including LAT, Gads, Grb-2 and SLP-76, the Vav family haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

GTP exchange proteins and the effector protein, PLCγ2.79 This hemITAM signaling pathway is similar to that used by the YxxL-containing platelet ITAM receptors such as the collagen receptor GPVI-FcRγ complex and the low affinity immune receptor FcγRIIA10. A fundamental difference between these pathways is, however, that Syk binds to two phosphorylated tyrosines within a single ITAM in GPVI-FcRγ and FcγRIIA, and to two phosphorylated hemITAM in separate cytosolic receptor chains of CLEC2.11 Another difference between CLEC-2 and GPVI signaling in human platelets is the critical dependence of CLEC2 on positive feedback signaling through ADP and thromboxane A (TxA ), actin polymerization, and the small GTPase Rac.8 In contrast these positive feedback pathways play a relatively minor role in CLEC-2 signaling in mouse platelets.12 Recently, irreversible inhibitors of the Tec kinase Btk have entered clinical use for the treatment of B-cell malignancies.13,14 These include the first-generation inhibitor ibrutinib and the second-generation inhibitor acalabrutinib. Ibrutinib is associated with a significant increase in major bleeding which is much reduced with the secondgeneration inhibitor acalabrutinib.14,15 Patients with Xlinked agammaglobulinemia (XLA) who have functiondisrupting mutations or loss of Btk do not, however, have increased bleeding. This observation, coupled with protein biochemistry measurements, have led us to conclude that the bleeding induced by ibrutinib is due to the high dosing strategy and off-target effects.16 Platelets express two members of the Tec family of tyrosine kinases, namely Btk and Tec. Btk has an approximate 10- to 20fold greater level of expression in both human and mouse platelets.17,18 Various studies have shown redundancy between the two kinases downstream of GPVI, with Tec able to partially compensate for the absence of Btk in human platelets from patients with XLA and in mouse platelets, which are deficient in Btk.16,19,20 Recently, Btk inhibitors have been proposed to represent a new class of antithrombotic drug21 based on the observation that they delay or inhibit ex vivo platelet activation by GPVI16,22-24 and by immobilized atherosclerotic plaque at arterial rates of flow.25 In addition, Btk inhibitors block platelet activation by CLEC-2. However, based on studies using ibrutinib-treated human platelets, Manne et al. have proposed that Btk, in contrast to GPVI, lies upstream of Syk in the CLEC-2 signaling cascade.26 In the present study, we demonstrate that the results of Manne et al. are explained by the positive feedback role of ADP and TxA and that concentrations of ibrutinib that have little or no effect on the response to GPVI stimulation selectively block activation by CLEC-2. This observation, together with the pivotal role of CLEC-2 in thrombo-inflammation, suggests that inhibitors of Btk represent new antiplatelet agents for thrombo-inflammatory disorders with minimal effect on hemostasis. 2

Methods

Light transmission aggregometry Aggregation was monitored by light transmission using a Model 700 aggregometer (Chronolog, Havertown, PA, USA) as previously described.16

Protein phosphorylation Eptifibatide-treated, washed platelets were stimulated in a Model 700 aggregometer in the presence of ibrutinib or vehicle as described previously.16 Activation was terminated after 300 seconds by addition of reducing sample buffer. This was followed by lysate separation by sodium dodecylsulfate polyacrylamide gel electrophoresis, electro-transfer and western blot as described previously.16

Inferior vena cava stenosis assay All mouse experiments were performed using wild-type (WT) mice on a C57Bl/6 genetic background under Home Office project licenses P0E98D513 and PC427E5DD. Mice were sourced from Charles River UK Ltd. (Margate, UK). For ex vivo platelet function assays and in vivo thrombosis assays 8-week old C57Bl/6 WT male mice were treated by intraperitoneal injection with a total dose of 140 mg/kg of ibrutinib or vehicle (5% dimethylsulfoxide, 30% polyethylene glycol 300, 5% Tween 20, 60% deionized water) in divided doses once daily over 2 – 4 days. Blood was taken at the stated times and platelet function analysis was performed as described above using flow cytometry. The IVC stenosis model was performed as described by Payne et al.3 In brief, mice were anesthetized using isoflurane and then a laparotomy was performed. Side branches of the IVC were identified and tied off before the IVC itself was stenosed with a ligature and a 30-gauge spacer to maintain a small degree of vessel patency. The incision was then closed and mice were allowed to recover from surgery. Buprenorphine was used preand post-operatively for analgesia. Mice were then culled 48 h after the surgery and the IVC was examined for the presence and size of thrombus.

Other methods Details of reagents and methods used for flow cytometry, flow adhesion, imaging, image analysis and cell transfection are detailed in the Online Supplementary Material.

Statistical analysis All data are presented as mean ± standard error of the mean with statistical significance taken as P<0.05 unless otherwise stated. Statistical analyses were performed using the MannWhitney test, Fisher exact test, Welch t-test or one or two-way analysis of variance (ANOVA) with corrections for multiple comparisons, as stated. Statistical analysis of the half maximal inhibitory concentration (IC ) values was performed using the Welch t-test. All statistical analyses were performed using GraphPad Prism 7 (GraphPad Software Inc. La Jolla, CA, USA). 50

Platelet preparation Blood was taken from consenting patients or healthy, drugfree volunteers, into 4% sodium citrate as previously described.16 Mouse blood was drawn from mice asphyxiated with CO following isoflurane anesthesia by inferior vena cava (IVC) puncture and collected into acid-citrate-dextrose. Platelet-rich plasma 2

haematologica | 2021; 106(1)

was obtained by centrifugation. Washed human and mouse platelets were obtained by further centrifugation of the plateletrich plasma in the presence of prostacyclin and resuspended in modified Tyrode buffer as previously described.16 Platelets were used at a cell density of 4x108/mL for aggregometry and biochemical studies.

Ethical approval Ethical approval for collecting blood from patients and healthy volunteers was granted by the National Research Ethics Service (10/H1206/58) and Birmingham University Internal 209

P.L.R. Nicolson et al.

Ethical Review (ERN_11-0175), respectively. Work on patients with XLA has ethical approval via the University of Birmingham Human Biomaterials Resource Centre (16-251 Amendment 3).

Results CLEC-2-mediated human platelet aggregation is inhibited by low concentrations of Btk inhibitors To examine the role of Btk downstream of platelet CLEC-2 in humans we took washed platelets from healthy donors and added increasing concentrations of ibrutinib before stimulating the platelets with the CLEC2 agonist rhodocytin. As shown in Figure 1A, B, ibrutinib completely inhibited CLEC-2-mediated aggregation at concentrations as low as 70 nM. Inhibition was marked after 5 min incubation with ibrutinib and only increased slightly with times up to 60 min (Figure 1A, B and Online Supplementary Figure S1). Strikingly, the IC for inhibition was over 20-fold lower than that required to block GPVImediated platelet aggregation16 (Table 1). This selective 50

inhibition of the response to CLEC-2 activation is consistent with the results of Manne et al.26 We further studied the effect of in vivo Btk inhibition on platelet-rich plasma obtained from patients with chronic lymphocytic leukemia (CLL) treated with ibrutinib or the more specific second-generation Btk inhibitor acalabrutinib. Unlike platelets from patients treated with a control chemotherapy regimen (fludarabine, cyclophosphamide and rituximab; FCR), platelets from Btk inhibitor-treated patients did not aggregate even at high levels of CLEC-2 stimulation (Figure 1C). We have previously shown that platelet-rich plasma from Btk inhibitor-treated patients does, however, aggregate in response to GPVI and G protein-coupled receptor (GPCR) stimulation.16

Adhesion to podoplanin under venous flow conditions is abrogated by Btk inhibition Given the role of platelet CLEC-2 and podoplanin in the thrombo-inflammatory and venous thrombosis models used by Hitchcock et al.4 and Payne et al.,3 we investigated whether inhibition of CLEC-2-mediated platelet function

C ii

Figure 1. Btk inhibitors completely block CLEC-2-mediated platelet aggregation in vitro and ex vivo. Washed platelets at 4x108/mL isolated from healthy donors were stimulated with 300 nM rhodocytin after incubation with ibrutinib or vehicle at the stated concentrations for 5 min. Light transmission aggregometry measurements were then undertaken for 5 min. (A) Representative traces of three identical aggregation experiments. (B) Dose response curves (n=3). (C) Platlet-rich plasma was isolated from the blood of patients with chronic lymphocytic leukemia following treatment with fludarabine, cyclophosphamide and rituximab (FCR), ibrutinib or acalabrutinib based chemotherapy regimens. (i) Representative trace and (ii) mean Âą standard error of mean (SEM) (post-FCR, n=5; post-ibrutinib, n=12; post-acalabrutinib, n=3) of optical density 5 min after stimulation with rhodocytin 10 - 300 nM. All results are shown as mean Âą SEM. Statistical analysis was performed with a two-way analysis of variance with the Tukey correction for multiple comparisons. *P<0.05, ns=not significant. HD: healthy donor; WP: washed platelets; OD: optical density; CLL: chronic lymphocytic leukemia; PRP: platelet-rich plasma.

210

haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

by ibrutinib reduced or prevented adhesion to the endogenous ligand under flow. Blood from healthy donors, patients with XLA or those taking ibrutinib or acalabrutinib was flowed over podoplanin at a venous shear. Strikingly we found that both the adhesion and aggregate size of these platelets were markedly reduced when compared to those of platelets from healthy donor blood regardless of the method of Btk inhibition (Figure 2).

Btk and Syk inhibition block CLEC-2-mediated Btk phosphorylation To examine the effect of ibrutinib on CLEC-2-mediated protein phosphorylation, ibrutinib-treated washed human platelets were lysed after stimulation with rhodocytin and lysates probed with phosphospecific antibodies. As shown in Figure 3A, phosphorylation of Syk Y525/6, SLP-76 Y145, LAT Y200, Btk Y223 and Y551, and PLCγ2 Y1217 was blocked by 70 nM ibrutinib with IC values similar to those for aggregation (Table 1). This concentration of ibrutinib reduced total phosphotyrosine to basal levels and inhibited phosphorylation on all measured tyrosine sites except the constitutive phosphorylation of Src Y418. Loss of Src pY418 was seen at higher concentrations of ibrutinib (Figure 3Aiv) as in our previous study.16 We performed similar experiments with acalabrutinib and found that Btk Y223 phosphorylation was lost at a concentration (3 mM) that also partially inhibited phosphorylation of Syk (Online Supplementary Figure S2). This concentration caused a marked reduction in platelet activation by CLEC-2 but no change in platelet aggregation in response to GPVI ligation,16 confirming the selectivity for 50

Table 1. Half maximal inhibitory concentration (IC ) values for all doseresponse curves shown in the figures in this publication. 50

Figure

Dose-response curve

1B 3Aii 3Aii 3Aiii 3Aiii 3Aiii 3Aiii 3Aiv 3Bii 3Biii 3Biii 6B 6Bii 6Bii 6Bii 6Bii 6Biii 6Biii 6Biii 6Biii 6Biv

HD Aggregation HD Btk pY223 HD PLCγ2 pY1217 HD Syk pY525/6 HD LAT pY200 HD SLP76 pY145 HD Btk pY551 HD Src pY418 HD Aggregation (added 10 mM ADP) HD Btk pY223 (added 10 mM ADP) HD PLCγ2 pY1217 (added 10 mM ADP) WT Mouse WP Aggregation WT Mouse Btk pY223 WT Mouse PLCγ2 Y753 WT Mouse PLCγ2 Y759 WT Mouse PLCγ2 Y1217 WT Mouse Syk pY525/6 WT Mouse LAT pY132 WT Mouse LAT pY200 WT Mouse Btk pY551 WT Mouse Src pY418

IC (mM)

95% CI

0.023 0.024 0.034 0.052 0.036 0.032 0.059 0.034 0.006 0.023 13.28 0.076 0.074 0.065 0.122 0.35

0.009 – 0.053 0.002 – 0.133 0.010 – 0.119 0.020 – 0.150 0.028 – 0.189 0.015 – 0.066 0.023 – 0.179 0.017 – 0.07 0.003 – 0.009 0.014 – 0.037 4.650 – 129.2 0.011 – 0.559 0.026 – 0.200 0.017 – 0.232 0.044 – 0.363 0.126 – 1.036

Where no value is shown this indicates that the IC could not be calculated. 95% CI: 95% confidence interval; HD: healthy donor; WT: wild-type. 50

haematologica | 2021; 106(1)

CLEC-2. The observation, however, of partial phosphorylation of Syk and complete abrogation of phosphorylation of Btk in the presence of acalabrutinib suggests that Btk does not lie upstream of Syk, in contrast to the conclusion of Manne et al.26 A further prediction in the model of Manne et al.,26 in which Syk lies downstream of Btk, is that inhibition of Syk should not block phosphorylation of Btk. To test this, we incubated human platelets with increasing concentrations of the Syk inhibitor PRT 060318 prior to stimulation with rhodocytin. Lysates were then probed using phosphospecific antibodies to Syk pY525/6, LAT pY200, Btk pY551 and PLCγ2 pY1217. Online Supplementary Figure S3 shows that phosphorylation of Syk, LAT, Btk and PLCγ2 were lost in tandem, providing further evidence against the model. This result is in keeping with our previous results examining CLEC-2-mediated phosphorylation events in mice with a Syk mutation rendering it unable to undergo tyrosine phosphorylation.9 In summary, the above results show that low concentrations of ibrutinib and acalabrutinib selectively block platelet activation by CLEC-2 over GPVI and provide evidence against the model of Manne et al. in which Btk lies upstream of Syk.

ADP rescues Syk phosphorylation in the presence of Btk inhibition We have previously shown that CLEC-2 signaling in human platelets is critically dependent on positive feedback signals from thromboxane TP and ADP P2Y receptors.8 Our standard platelet preparation method leads to partial desensitization of the ADP P2Y receptor27 and loss of aggregation in response to ADP as a consequence of this (Online Supplementary Figure S4). This may have accentuated the previously described dependency on P2Y receptors. To investigate this, we repeated the studies on aggregation and protein phosphorylation but in the presence of excess ADP (10 mM). Similar to the results with no added ADP, 70 nM ibrutinib inhibited aggregation (Figure 3Bii) and phosphorylation of Btk Y223 and PLCγ2 Y1217 (Figure 3Bi-ii). In contrast, however, it did not inhibit phosphorylation of Syk Y525/6, LAT Y200 or Btk on its Src kinase phosphorylation site Y551 (Figure 3Bi and iii). The demonstration of phosphorylation of Syk, LAT and Btk on Y551 but absence of Btk autophosphorylation on Y223 in ibrutinib-treated human platelets confirms that Btk lies downstream of Syk in the CLEC-2 signaling cascade in human platelets. 12

CLEC-2-mediated platelet aggregation and tyrosine phosphorylation are blocked in patients with Btk mutations Our previous study found off-target effects of ibrutinib even when used at concentrations as low as 70 nM.16 To investigate whether the potent inhibition of CLEC-2mediated platelet activation by ibrutinib was mediated by its blockade of Btk or an off-target effect, we studied platelet aggregation in response to rhodocytin in patients with XLA. CLEC-2-mediated aggregation and phosphorylation of PLCγ2 Y1217 were blocked, whereas phosphorylation of Syk Y525/6 and LAT Y200 was only partially reduced (Figure 4). These findings demonstrate that the initial phosphorylation of Syk and LAT is independent of Btk and that this is then increased by a Btk-dependent pathway. 211

P.L.R. Nicolson et al.

A ii

B ii

Figure 2. Btk inhibition blocks platelet adhesion to podoplanin under venous flow conditions. (A, B) Whole blood from healthy donors, patients treated with ibrutinib or acalabrutinib or those with X-linked agammaglobulinemia (XLA) was incubated with DiOC dye for 5 min before being flowed across a capillary coated with podoplanin-Fc (100 mg/mL) at 125 s-1 for 5 min. Images were taken every second using fluorescent channels on a Zeiss Axio inverted microscope at 20X magnification. Representative images from healthy donors (Ai), patients treated with ibrutinib 420 mg once daily (Aii), patients treated with acalabrutinib 100 mg twice daily (Bi) or patients with XLA (Bii). Ilastik 1.1.2 machine learning software was used to automatically and reproducibly identify platelets. Data on platelet surface area coverage and cluster size were measured using the KNIME 3.4 analytics platform. (C, D) Mean Âą standard error of mean (SEM) showing the increases in total platelet aggregate area over time of healthy donors (n=4) and ibrutinib-treated patients (n=5) (C) and in alabrutinib-treated patients (n=2) and patients with XLA (n=2) (D). (E,F) Mean Âą SEM showing increases in mean aggregate size over time of healthy donors and ibrutinib-treated patients (E) and of acalabrutinib-treated patients and patients with XLA, with the curve from healthy donor platelets included for comparison (F). The statistical analysis was performed with two-way analysis of variance. *P<0.05, ns=not significant. 6

212

haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

A ii

A iii

A iv

B ii

B iii

Figure 3. Low concentrations of ibrutinib block all phosphorylation events downstream of platelet CLEC-2 ligation. Blockade upstream of Btk pY223 is rescued by addition of excess ADP. (A, B) Healthy donor washed human platelets at 4x108/mL were incubated in the presence of eptifibatide 9 mM with ibrutinib or vehicle for 5 min without (A) or with (B) ADP at 10 mM prior to stimulation with 300 nM rhodocytin. Platelets were then lysed with 5X reducing sample buffer 5 min after addition of agonist. Whole cell lysates were then separated by sodium dodecylsulfate polyacrylamide gel electrophoresis and western blots were probed for whole cell phosphorylation or kinase phosphorylation with the stated antibodies downstream of the platelet CLEC-2 receptor. A representative blot from four identical experiments (i). Mean tyrosine phosphorylation levels Âą standard error of mean (SEM) of four identical experiments for phosphorylation events downstream (ii) and upstream (iii) of Btk pY223 as well as Src pY418 (iv). The mean aggregation trace from Figure 1B is included as a dotted line to aid comparison. Mean Âą SEM of three identical experiments for light transmission aggregometry measurements in the presence of ADP 5 min after stimulation with 300 nM rhodocytin are shown as a solid line in (ii) and as a dotted line in (iii). Rho: rhodocytin.

haematologica | 2021; 106(1)

213

P.L.R. Nicolson et al. Ai

A ii

Bi B ii

Figure 4. Platelets from patients with X-linked agammaglobulinemia do not aggregate in response to ligation of CLEC-2, but signaling events upstream of Btk are partially preserved. (A) Platelet-rich plasma was isolated from the blood of patients with X-linked agammaglobulinemia (XLA) and stimulated with rhodocytin. (i) Representative trace and (ii) mean Âą standard error of mean (SEM) (healthy donors, n=6; XLA patients, n=4) of optical density 5 min after stimulation with rhodocytin 300 nM. Statistical analysis was performed with a two-tailed t-test. (B) Eptifibatide (9 mM)-treated washed platelets at 4x108/mL from patients with XLA were lysed with 5X reducing sample buffer 5 min after addition of rhodocytin 300 nM or phosphate-buffered saline as shown. Whole cell lysates were then separated by sodium dodecylsulfate polyacrylamide gel electrophoresis and western blots were probed for whole cell or tyrosine phosphorylation with the stated antibodies. (i) Blot and (ii) mean Âą SEM (n=3) for phosphorylation events downstream of the CLEC-2 receptor. Results were analyzed statistically using a one-way analysis of variance with Sidak correction for multiple comparisons. OD: optical density.

The kinase domain of Btk is required for signaling downstream of CLEC-2 We have previously shown that Btk supports platelet activation by GPVI through acting both as a kinase and as an adapter protein.16 However, the observation of complete loss of aggregation in the presence of ibrutinib and acalabrutinib suggests that the kinase activity of Btk is critical for activation by CLEC-2. Consistent with this, wild-type but not kinase-dead Btk restored NFAT signaling in a CLEC2-transfected cell line model, which was blocked by ibrutinib and acalabrutinib (Figure 5A-C). CLEC-2 and Btk expression was similar regardless of the Btk construct used (Figure 5D, E). The lack of signaling with kinase-dead Btk confirms a fundamental difference between the role of Btk in the CLEC-2 and GPVI signaling pathways.

Btk does not lie upstream of Syk in CLEC-2-stimulated mouse platelets We extended the studies to investigate the role of Btk in CLEC-2 signaling in mouse platelets. In agreement with the results of Lee et al.,28 we found that only high concentrations of ibrutinib, which abrogated platelet activation by 214

GPVI, blocked CLEC-2-mediated platelet aggregation (Figure 6A, Table 1 and data not shown). Phosphorylation of Syk Y519/520 (equivalent to Y525/526 in human platelets), and LAT Y132 and Y200 was preserved in the presence of ibrutinib (Figure 6B). This is consistent with the results in human platelets stimulated by CLEC-2 in the presence of ADP and further confirms that Btk does not lie upstream of Syk and LAT in mouse platelets.

The effect of ibrutinib on deep vein thrombosis We sought to establish whether ibrutinib would prevent IVC thrombus formation in a mouse model of deep vein thrombosis (DVT) that is known to be dependent on CLEC-2. Mice were dosed with ibrutinib via intraperitoneal injection for up to 4 days. This dosing strategy blocked CLEC-2-mediated but not PAR4-mediated platelet activation (Figure 7A). Mice dosed with ibrutinib had a reduction in thrombus prevalence relative to controls (Figure 7Bi) but this was not statistically significant. In the small number of ibrutinib-treated mice which still formed IVC thrombi, thrombus size was not significantly altered relative to that in controls (Figure 7Bii). haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

Figure 5. NFAT luciferase activity in Btk deficient DT40 cells is restored with transfection of wild-type but not kinase-dead Btk. Btk-deficient DT40 cells were transfected with either wild-type or kinase-dead Btk with or without CLEC-2. All cells were transfected with a NFAT-luciferase reporter plasmid. Cells were stimulated with rhodocytin 300 nM in the presence of serum. (A) Luciferase activity was measured and is shown as mean Âą standard error of mean (SEM) of five identical experiments. (B, C) Cells were stimulated with rhodocytin 50 nM in the presence or absence of ibrutinib (0.5 - 10 mM) (B) and acalabrutinib (0.5 - 10 mM) (C). Serum was excluded during stimulation to avoid plasma binding of the drugs. Luciferase activity of vehicle- and drug-treated samples was measured and the difference is shown as the mean Âą SEM of three independent experiments. (D) CLEC-2 expression in the different transfection conditions. (E) Representative western blot showing Btk expression levels in cells transfected with different Btk constructs (n=5). *P<0.05, ns = not significant. WT: wild-type; KD: kinase-dead; WB: western blot.

Discussion The main findings from the above studies into the role of Btk downstream of CLEC-2 receptor ligation in platelets are: (i) Btk lies downstream of Syk in the CLEC2 signaling cascade in human and mouse platelets, in contrast to the conclusion of Manne et al.; and (ii) Btk inhibitors selectively block platelet activation by CLEC-2 relatively to GPVI. This suggests that Btk inhibitors can be used to selectively block CLEC-2-mediated platelet activation in thrombo-inflammation without any increase in bleeding or other side effects due to blockade of GPVI and ITAM receptors in other cells. The results described here do not support the conclusions of Manne et al. that Btk lies upstream of Syk in CLEC-2 signaling in both human and mouse platelets.26 While we were able to replicate the complete loss of whole cell tyrosine, Syk and Btk phosphorylation with ibrutinib in human platelets reported by Manne et al., we haematologica | 2021; 106(1)

show that ibrutinib blocks Btk phosphorylation at its Src phosphorylation site (Y551) which is not in keeping with the known mechanism of action of ibrutinib: Btk is phosphorylated by Src at Y551 but ibrutinib then blocks subsequent autophosphorylation at Y223. We also report loss of Btk phosphorylation using the Syk inhibitor PRT 060318. Both of these findings are inconsistent with a model in which Btk lies upstream of Syk. This is further supported by the observation that phosphorylation of Syk and Btk Y551 is restored in the presence of the secondary messenger ADP. These results demonstrate that the phosphorylation of Syk is initially independent of Btk and is then increased by a pathway that involves Btk and ADP. Our results show that human CLEC-2-mediated platelet activation is critically dependent on the kinase function of Btk. Platelets exposed to low concentrations of ibrutinib and the more Btk-selective acalabrutinib, as well as platelets from patients treated with these inhibitors and patients with genetic loss-of-function mutations of Btk, do 215

P.L.R. Nicolson et al. A

B ii

B iii

B iv

Figure 6. Ibrutinib blocks CLEC-2-mediated tyrosine phosphorylation of Btk pY223 at low concentrations in mouse platelets. Aggregation is only blocked at ~250-fold higher concentrations. (A) Wild-type mouse washed platelets at 4x108/mL were incubated with vehicle or ibrutinib for 5 min prior to stimulation with rhodocytin 300 nM for 5 min. A representative light transmission aggregometry trace from three identical experiments is shown. (B) Eptifibatide (9 mM)-treated wild-type mouse washed platelets at 4x108/mL were incubated with ibrutinib or vehicle for 5 min before being stimulated with rhodocytin 300 nM. Platelets were then lysed with 5X reducing sample buffer 5 min after addition of an agonist. Whole cell lysates were then separated by sodium dodecylsulfate polyacrylamide gel electrophoresis and western blots were probed for whole cell phosphorylation or kinase phosphorylation with the stated antibodies downstream of the platelet CLEC-2 receptor. (i) Representative blots and (iiiv) mean results of three identical experiments examining degree of aggregation and tyrosine phosphorylation levels of the proteins shown. Mean aggregation (n=3) results are shown in (ii) and are included as dotted lines in (iii-iv) to aid comparison. All results are shown as mean Âą standard error of mean. OD: optical density.

216

haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

A ii

B ii

Figure 7. Mice dosed with ibrutinib have a trend towards reduction of thrombus formation in an in vivo venous thrombosis model. Wild-type mice were dosed with ibrutinib (35 - 70 mg/kg) or vehicle for 1 - 3 days before and 2 days following inferior vena cava (IVC) stenosis surgery to achieve consistent CLEC-2 inhibition throughout the post-surgical period. (A) Heparinized whole blood was collected via tail bleeding from mice undergoing the same dosing schedule but not undergoing surgery at the stated time points. The blood was incubated with FITC-conjugated anti-P-selectin and PE-conjugated anti-activated integrin ÎąIIbb3 antibodies for 30 min before undergoing red cell lysis and fixation and analysis using flow cytometry. The effect on platelet activation after stimulation with PAR4 peptide (500 nM) and rhodocytin (300 nM) is shown as (i) representative plots, (ii) mean data Âą standard error of mean (ibrutinib, n=6; vehicle, n=3). (B) Two hours following the pre-surgical dose of ibrutinib or vehicle, IVC stenosis was induced under general anesthesia with a ligature. Mice were allowed to recover and were then culled 48 h later and examined for the size of any IVC thrombus. (i) Thrombus prevalence and (ii) thrombus weight are shown (ibrutinib, n=13; vehicle, n=12). The horizontal line represents the median thrombus weight. A Fisher exact test was used for the statistical analysis of thrombus prevalence, whereas a Mann-Whitney test was used for the analysis of weight. ns=not significant. SSC: side scatter; FSC: forward scatter.

haematologica | 2021; 106(1)

217

P.L.R. Nicolson et al.

not demonstrate any CLEC-2-mediated activation, even when stimulated with very high levels of agonist. We have previously shown that genetic or pharmacological Btk inhibition of platelets does not result in loss of activation in response to GPVI in human platelets because: (i) Btk was able to function as an adapter downstream of GPVI to preserve platelet activation; and (ii) Tec may be able to compensate for the lack of Btk kinase or adapter activity.16 Here we show that Tec is not able to compensate for a lack of Btk in the CLEC-2 signaling cascade through studies on patients with XLA and that the kinase domain of Btk is critical for CLEC-2 signaling in a transfected cell line model. The present results suggest that Btk inhibitors have the potential to be used at very low concentrations as selective inhibitors of CLEC-2 in thrombo-inflammatory disorders with minimal off-target effects and no increased bleeding. Not only does Btk inhibition prevent platelet activation in response to CLEC-2 but we have shown that it also results in abrogation of platelet adhesion to podoplanin at venous rates of shear and a reduction in DVT formation in ibrutinib-treated mice. The in vivo DVT study was powered based on the rates of thrombosis seen in CLEC-2-deficient and WT mice in the studies by Payne et al.3 The lack of statistical significance in the trend towards inhibition of DVT formation may reflect the fact that pharmacological blockade of Btk in mice did not result in full inhibition of CLEC-2 signaling throughout the 48 h required for venous thrombosis in this model and the fact that, unlike for human CLEC-2 signaling, mouse CLEC-2 is not critically dependent on Btk. Interestingly the thrombi that did form in the ibrutinibtreated mice were the same size as those in the control mice. This is consistent with a model in which CLEC-2 activation is required to initiate formation of a thrombus, but not its propagation. In conclusion, the present study shows that Btk lies downstream of Syk in human and mouse platelets and that Btk inhibitors block human CLEC-2-mediated platelet function at concentrations more than 20-fold lower than those that block GPVI signaling16 and mouse CLEC-2. The mechanism underlying this selectivity is that Btk kinase activity is critical for human platelet CLEC-2 function, but not for GPVI. This means that low-

References 1. Lowe K, Watson SP, Lax S, Frampton J. CLEC-2 is not required to maintain separation of intact blood and lymphatic vasculature. Blood. 2015;123(20):3200-3207. 2. Astarita JL, Acton SE, Turley SJ. Podoplanin: emerging functions in development, the immune system, and cancer. Front Immunol. 2012;3(283):1-11. 3. Payne H, Ponomaryov T, Watson SP, Brill A. Mice with a deficiency in CLEC-2 are protected against deep vein thrombosis. Blood. 2017;129(14):2013-2020. 4. Hitchcock JR, Cook CN, Bobat S, et al. Inflammation drives thrombosis after Salmonella infection via CLEC-2 on platelets. J Clin Invest. 2015;125(12):4429-4446. 5. Rayes J, Watson SP, Nieswandt B. Functional significance of the platelet immune receptors GPVI and CLEC-2. J Clin Invest.

218

dose Btk inhibitors could be used therapeutically to inhibit CLEC-2 in platelets without any off-target effects on GPVI and ITAM receptors in other hematopoietic cells. This provides a mechanism for selective targeting of CLEC-2 in thrombo-inflammatory disorders, such as DVT or infection-driven thrombosis, while preserving hemostasis.3,4 Disclosures This work was supported by British Heart Foundation (BHF) Programme grant (RG/13/18/30563), a BHF Clinical Fellowship to PLRN (FS/17/20/32738), a BHF Senior Basic Science Research Fellowship to AB (FS/19/30/34173), an AMS springboard grant to AYP (SBF002\1099) and the University of Birmingham’s Institute of Translation Medicine and Institute of Cardiovascular Sciences; SPW holds a BHF Chair (CH03/003). JAE is supported by the Deutsche Forschungsgemeinschaft (DFG grant: Eb177/13-1). Btk deficient DT40 cells, plasmid constructs and rabbit anti-Btk antibody were a kind gift from Dr Mike Tomlinson (University of Birmingham, UK). PLR Nicolson has received research grants from Rigel Pharmaceuticals, Novartis Pharmaceuticals and materials from Principia Biopharma. Contributions PLRN, CEH, SHN and SPW conceived the study and wrote the manuscript. PLRN, SHN, JH, LG-Q, CWS, JC, AB, AOK, NSP, DNK, SW and CNW performed experiments. PLRN, SHN, JH and JAP performed data analysis. HC and APH consented and provided access to patients with XLA. JAE provided key reagents. GP consented and provided access to patients with CLL. CEH, AYP and SPW supervised the study. All authors critically appraised the manuscript. Acknowledgments This work was supported by a British Heart Foundation (BHF) Programme grant (RG/13/18/30563), a BHF Clinical Fellowship to PLRN (FS/17/20/32738), a BHF Senior Basic Science Research Fellowship to AB (FS/19/30/34173), an AMS springboard grant to AYP (SBF002\1099) and the University of Birmingham’s Institute of Translation Medicine and Institute of Cardiovascular Sciences; SPW holds a BHF Chair (CH03/003). JAE is supported by the Deutsche Forschungsgemeinschaft (DFG grant: Eb177/13-1).

2019;129(1):12-23. 6. Severin S, Pollitt AY, Navarro-Núñez L, et al. Syk-dependent phosphorylation of CLEC-2: a novel mechanism of hem-immunoreceptor tyrosine-based activation motif signaling. J Biol Chem. 201;286(6):4107-4116. 7. Spalton JC, Mori J, Pollitt AY, Hughes CE, Eble JA, Watson SP. The novel Syk inhibitor R406 reveals mechanistic differences in the initiation of GPVI and CLEC-2 signaling in platelets. J Thromb Haemost. 2009;7(7): 1192-1199. 8. Pollitt AY, Grygielska B, Leblond B, Desire L, Eble JA, Watson SP. Phosphorylation of CLEC-2 is dependent on lipid rafts, actin polymerization, secondary mediators, and Rac. Blood. 2010;115(14):2938-2946. 9. Hughes CE, Finney BA, Koentgen F, Lowe KL, Watson SP. The N-terminal SH2 domain of Syk is required for (hem)ITAM, but not integrin, signaling in mouse platelets. Blood. 2015;125(1):144-154.

10. Fuller G, Williams J, Tomlinson MG, et al. The C-type lectin receptors CLEC-2 and Dectin-1, but not DC-SIGN, signal via a novel YXXL-dependent signaling cascade. J Biol Chem. 2007;282(17):12397-12409. 11. Hughes CE, Pollitt AY, Mori J, et al. CLEC-2 activates Syk through dimerization. Blood. 2010;115(14):2947-2955. 12. Borgognone A, Navarro-Núñez L, Correia JN, et al. CLEC-2-dependent activation of mouse platelets is weakly inhibited by cAMP but not by cGMP. J Thromb Haemost. 2014;12(4):550-559. 13. Byrd JC, Furman RR, Coutre SE, et al. Targeting BTK with ibrutinib in relapsed chronic lymphocytic leukemia. N Engl J Med. 2013;369(1):32-42. 14. Byrd JC, Harrington B, O'Brien S, et al. Acalabrutinib (ACP-196) in relapsed chronic lymphocytic leukemia. N Engl J Med. 2016; 374(4):323-332. 15. Wang M, Rule S, Zinzani PL, et al.

haematologica | 2021; 106(1)

Btk in CLEC-2-mediated platelet function and signaling

Acalabrutinib in relapsed or refractory mantle cell lymphoma (ACE-LY-004): a singlearm, multicentre, phase 2 trial. Lancet. 2018;391(10121):659-667. 16. Nicolson PLR, Hughes CE, Watson S, et al. Inhibition of Btk by Btk-specific concentrations of ibrutinib and acalabrutinib delays but does not block platelet aggregation to GPVI. Haematologica. 2018;103(12):2097-2108. 17. Burkhart JM, Vaudel M, Gambaryan S, et al. The first comprehensive and quantitative analysis of human platelet protein composition allows the comparative analysis of structural and functional pathways. Blood. 2012;120(15):e73-82. 18. Zeiler M, Moser M, Mann M. Copy number analysis of the murine platelet proteome spanning the complete abundance range. Mol Cell Proteomics. 2014;13(12):3435-3445. 19. Quek LS, Bolen J, Watson SP. A role for Bruton’s tyrosine kinase (Btk) in platelet activation by collagen. Curr Biol. 1998;8(20): 1137-1140. 20. Atkinson BT, Ellmeier W, Watson SP. Tec regulates platelet activation by GPVI in the absence of Btk. Blood. 2003;102(10):35923599. 21. Busygina K, Denzinger V, Bernlochner I, Weber C, Lorenz R, Siess W. Btk inhibitors as first oral atherothrombosis-selective

haematologica | 2021; 106(1)

antiplatelet drugs? Thromb Haemost. 2019;119(08):1212-1221. 22. Levade M, David E, Garcia C, et al. Ibrutinib treatment affects collagen and von Willebrand factor-dependent platelet functions. Blood. 2014;124(26):3991-3995. 23. Kamel S, Horton L, Ysebaert L, et al. Ibrutinib inhibits collagen-mediated but not ADP-mediated platelet aggregation. Leukemia. 2015;29(4):783-787. 24. Bye AP, Unsworth AJ, Desborough MJ, et al. Severe platelet dysfunction in NHL patients receiving ibrutinib is absent in patients receiving acalabrutinib. Blood Adv. 2017;1(26):2610-2623. 25. Busygina K, Jamasbi J, Seiler T, et al. Oral Bruton tyrosine kinase inhibitors selectively block atherosclerotic plaque–triggered thrombus formation in humans. Blood. 2018;131(24):2605-2616. 26. Manne BK, Badolia R, Dangelmaier C, et al. Distinct pathways regulate syk protein activation downstream of immune tyrosine activation motif (ITAM) and hemITAM receptors in platelets. J Biol Chem. 2015;290(18):11557-11568. 27. Koessler J, Hermann S, Wener K, et al. Role of purinergic receptor expression and function for reduced responsiveness to adenosine diphosphate in washed human

platelets. PLoS One. 2016; 11(1):e0147370. 28. Lee RH, Piatt R, Conley PB, Bergmeier W. Effects of ibrutinib treatment on murine platelet function during inflammation and in primary hemostasis. Haematologica. 2017;102(3):e89-e92. 29. Gitz E, Pollitt AY, Gitz-Francois JJ, et al. CLEC-2 expression is maintained on activated platelets and on platelet microparticles. Blood. 2014;124(14):2262-2270. 30. Suzuki-Inoue K, Fuller G, Garcia A, et al. A novel Syk-dependent mechanism of platelet activation by the C-type lectin receptor CLEC-2. Blood. 2006;107(2):542-549. 31. Navarro-Núñez L, Pollitt AY, Lowe K, Latif A, Nash GB, Watson SP. Platelet adhesion to podoplanin under flow is mediated by the receptor CLEC-2 and stabilised by Src/Sykdependent platelet signalling. Thromb Haemost. 2015;113(5):1109-1120. 32. Fillbrunn A, Dietz C, Pfeuffer J, Rahn R, Landrum GA, Berthold MR. KNIME for reproducible cross-domain analysis of life science data. J Biotechnol. 2017;261:149156. 33. Sommer C, Straehle C, Kothe U, Hamprecht FA. Ilastik: interactive learning and segmentation toolkit. 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro. IEEE. 2011:230-233.

219

ARTICLE Ferrata Storti Foundation

Platelet Biology & Its Disorder

Desialylation of O-glycans on glycoprotein Ibα drives receptor signaling and platelet clearance Yingchun Wang,1 Wenchun Chen,1 Wei Zhang,2 Melissa M. Lee-Sundlov,3 Caterina Casari,4 Michael C. Berndt,5 Francois Lanza,6 Wolfgang Bergmeier,4,7 Karin M. Hoffmeister,3 X. Frank Zhang2 and Renhao Li1

Aflac Cancer and Blood Disorders Center, Children’s Healthcare of Atlanta, Department of Pediatrics, Emory University School of Medicine, Atlanta, GA, USA; 2Department of Bioengineering, Lehigh University, Bethlehem, PA, USA; 3Blood Research Institute, BloodCenter of Wisconsin, Milwaukee, WI, USA; 4McAllister Heart Institute, University of North Carolina, Chapel Hill, NC, USA; 5Faculty of Health Sciences, Curtin University, Perth, Western Australia, Australia; 6UMR_S949 INSERM, Université de Strasbourg, EFS-Alsace, Strasbourg 67065, France and 7Department of Biochemistry and Biophysics, University of North Carolina, Chapel Hill, NC, USA 1

Haematologica 2021 Volume 106(1):220-229

ABSTRACT

Correspondence: RENHAO LI renhao.li@emory.edu Received: October 13, 2019. Accepted: January 22, 2020. Pre-published: January 23, 2020. https://doi.org/10.3324/haematol.2019.240440

220

uring infection neuraminidase desialylates platelets and induces their rapid clearance from circulation. The underlying molecular basis, particularly the role of platelet glycoprotein (GP)Ibα therein, is not clear. Utilizing genetically altered mice, we report that the extracellular domain of GPIbα, but neither von Willebrand factor nor ADAM17 (a disintegrin and metalloprotease 17), is required for platelet clearance induced by intravenous injection of neuraminidase. Lectin binding to platelet following neuraminidase injection over time revealed that the extent of desialylation of O-glycans correlates with the decrease of platelet count in mice. Injection of α2,3-neuraminidase reduces platelet counts in wild-type but not in transgenic mice expressing only a chimeric GPIbα that misses most of its extracellular domain. Neuraminidase treatment induces unfolding of the O-glycosylated mechanosensory domain in GPIbα as monitored by single-molecule force spectroscopy, increases the exposure of the ADAM17 shedding cleavage site in the mechanosensory domain on the platelet surface, and induces ligand-independent GPIb-IX signaling in human and murine platelets. These results suggest that desialylation of O-glycans of GPIbα induces unfolding of the mechanosensory domain, subsequent GPIb-IX signaling including amplified desialylation of N-glycans, and eventually rapid platelet clearance. This new molecular mechanism of GPIbα-facilitated clearance could potentially resolve many puzzling and seemingly contradicting observations associated with clearance of desialylated or hyposialylated platelet.

Introduction More than 100 billion platelets are cleared every day from a human body through a highly efficient and tightly regulated process. Exogenous agents could commandeer the process and accelerate platelet clearance, leading to thrombocytopenia and hemorrhage. One such agent is neuraminidase, which hydrolyzes the glycosidic linkages of sialic acids to the host glycoprotein and as a consequence exposes the penultimate galactoses. Injection of neuraminidase causes thrombocytopenia in mice or rats within a few hours, followed by a gradual rise in the platelet count back to the normal level in 4-5 days due to continuous thrombopoiesis in the body.1,2 Certain bacterial infection involves the release of bacterial neuraminidase in the blood and thrombocytopenia, often before the onset of septic shock.3,4 Moreover, there is accumulating evidence to support the involvement of endogenous neuraminidase in platelet clearance. For example, cold storage of murine platelets induces presentation of lysosomal neuraminidases on the plasma membrane, accelerating their clearance upon infusion into a recipient mouse.5,6 Many antibodies targeting the N-terminal ligand-binding domain (LBD) of platelet glycoprotein (GP)Ibα induce platelet signaling and surface presentation of lysosomal neuraminidase 1 (Neu1), leading to thrombocytopenia.7-9 Binding of plasma von haematologica | 2021; 106(1)

Desialylating GPIbα O-glycans causes PLT clearance

Willebrand factor (VWF) to GPIbα on the platelet produces similar signaling events including desialylation.10-12 Therefore, it appears that neuraminidase is critically involved in platelet clearance. It has been suggested that after platelet desialylation the exposed galactose residues on the platelet are recognized by the Ashwell-Morell receptor (AMR), also known as the asialoglycoprotein receptor, on the surface of liver macrophages and hepatocytes, thereby inducing internalization of the desialylated platelet by these cells and its clearance from the circulation. In support of this model, injection of neuraminidase does not reduce platelet counts in mice lacking AMR and fast clearance of desialylated platelets is significantly reduced in mice lacking AMR.3,5,13 Consistently, genetic ablation of certain sialyltransferases such as ST3Gal-IV in mice results in constitutive exposure of galactoses on the platelet, accelerated platelet clearance, and a significantly lower platelet count.14 Upon transfusion, St3gal4-/- or desialylated platelets are cleared at a faster rate than wild-type (WT), unless the recipient mice were pretreated with asialofetuin, a competitive inhibitor of AMR.5,15 While the involvement of the AMR in mediating clearance of desialylated platelets has been established, the underlying molecular mechanism remains controversial, as several studies reported seemingly contradictory or confusing observations. AMR is a multi-subunit receptor complex that contains several lectin domains for binding of galactose or galactosamine.16-18 It exhibits a much higher binding affinity and ligand preference for tetra- or triantennary galactoses than di- or mono-antennary ones.17,19,20 In other words, AMR binds primarily exposed galactose residue on N-glycans instead of O-glycans because only the former supports a tetra- or tri-antennary sugar structure. It was observed that proteolytic removal of the N-terminal ligand-binding domain (LBD) of GPIbα enhanced survival of transfused St3gal4-/- platelets or coldstored WT platelets.5,15 Since the LBD of human GPIbα contains 2 N-glycans,21-23 desialylated N-glycans on the LBD were suggested as the ligands for AMR.5,15 However, these experiments were performed on murine platelets, yet murine GPIbα does not contain any N-glycosylation sites (i.e., Asn-X-Ser/Thr) and therefore should have no Nglycans for AMR binding (Online Supplementary Figure S1). The involvement of the LBD in clearance of cold-stored platelets was later attributed to its cold-induced interaction with VWF that does not require N-glycans.11,24 Nonetheless, injection of neuraminidase causes thrombocytopenia in GPIbα-/- mice as in WT ones, but neuraminidase-treated GPIbα-/- platelets are cleared at a slower pace than similarly treated WT ones.3 How GPIbα is involved in neuraminidase-induced platelet clearance remains unresolved. As a major subunit of the GPIb-IX receptor complex, GPIbα is specifically expressed in the platelet.25 It was recently reported that the juxtamembrane portion of GPIbα contains a structured domain that responds to tensile force.26,27 Mechanical unfolding of this mechanosensory domain (MSD), as a result of VWF or antibody binding under shear, induces GPIb-IX signaling in the platelet that leads to its clearance.9-11 The MSD in both human and murine GPIbα contains several O-glycosylation sites (Online Supplementary Figure S1).10 Here we report that neuraminidase-mediated desialylation of O-glycans in GPIbα induces MSD unfolding and subsequently platelet haematologica | 2021; 106(1)

signaling. Unlike the model of AMR recognizing GPIbα N-glycans, our new mechanism does not require the presence of N-glycans in GPIbα. The novel involvement of GPIbα O-glycans provides a platelet-specific element and explains its facilitating role in mediating clearance of desialylated platelets.

Methods Human subjects Citrated whole blood was drawn from healthy volunteers according to a protocol approved by Emory University Institutional Review Board (IRB# 00006228), in which all volunteers gave written informed consent. The collected whole blood was used to prepare platelet-rich plasma (PRP) and plasma. Experiments involving fresh human platelets were performed in accordance with approved IRB protocols.

Materials and mice Neuraminidase from Arthrobacter ureafaciens was purchased from Sigma-Aldrich (St. Louis, MO, USA), and α2,3-neuraminidase cloned from Streptococcus pneumoniae was from New England Biolabs (Ipswich, MA, USA). C57BL/6J and VWF-/- mice were purchased from Jackson Laboratories (stock ns. 000664 and 003795, respectively). Transgenic hTg, GPIbα-/-, and IL4R-IbαTg mice have been described previously.28,29 St3gal1fl/fl mice carrying LoxP sites on exon 2 (stock n. 006897) were backcrossed in to the C57BL/6J background and paired with Pf4-Cre mice to delete St3gal1 in the megakaryocyte lineage (St3Gal1MK-/-). Similarly, Adam17fl/fl mice (stock n. 009597) were paired with Pf4-Cre mice to delete Adam17 in the megakaryocyte lineage (Adam17MK-/-). Six- to 8-week-old mice of both sexes were used in all experiments. All experiments involving mice were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee of Emory University, Blood Center of Wisconsin, and University of North Carolina.

Intravenous injection and blood count

Each mouse was injected intravenously with 100 mL saline or 0.5 mU/g bacterial neuraminidase, 0.6 U/g α2,3-neuraminidase in 100 mL saline. Before injection and at indicated time after, approximately 60 mL of whole blood was taken via the facial vein using heparinized micro-hematocrit capillary tube (Fisher Scientific, Pittsburgh, PA, USA) and mixed 9:1 (v/v) with 3.8% sodium citrate in a tube. Blood cell counts were determined on a Sysmex XP300 automated hematology analyzer. Platelet counts of Adam17MK-/- mice were obtained by flow cytometry as described.30

In vitro neuraminidase treatment Citrated human or murine whole blood was collected as described above. 10 mU α2,3,6,8-neuraminidase or equal activity of α2,3-neuraminidase was added to 60 mL citrated PRP, which were mixed and incubated at 37˚C for 15, 60 and 180 minutes (min) before being analyzed for glycans.

Laser optical tweezer measurement Single-molecule force measurement of GPIb-IX was performed as described.26,27 Recombinant biotinylated GPIb-IX, and WM23 coupled to one end of the DNA handle, were prepared as described.27 Before the pulling on desialylated GPIb-IX, 1 U of α2,3-neuraminidase was added to streptavidin beads with immobilized GPIb-IX and incubated at room temperature for 30 min. Data were analyzed as described.26 221

Y. Wang et al.

Statistical analysis Data are expressed as mean±standard deviation. An unpaired two-tailed t-test analysis was performed for statistical analyses. P<0.05 was considered statistically significant. All analyses were performed using Prism software.

Results Neuraminidase induces platelet clearance via a GPIbα-dependent but von Willebrand factor-independent and ADAM17-independent mechanism Neuraminidase from A. ureafaciens, with broad substrate Platelets

Erythrocytes

Relative cell count (%)

specificity (i.e., α2,3,6,8-neuraminidase), was injected intravenously into WT C57BL/6 mice, followed by periodic blood draw for 4 days and counting of platelets and erythrocytes therein. Compared to those prior to neuraminidase injection (t=0), counts of platelets, but not erythrocytes, became markedly lowered upon neuraminidase injection, reaching a minimum at 24 hours (h) after injection (Figure 1A and B). This result indicates that neuraminidase injection induces platelet clearance, consistent with earlier reports.3 Neuraminidase can desialylate VWF in the plasma and induce its clearance through AMR.13 Desialylated VWF binds the platelet spontaneously.31 Thus, it is possible that desialylated VWF may bind GPIbα and induce platelet

ECL binding (MFI/103)

SNA binding (MFI)

RNA binding (MFI/103)

Time after injection (d) 222

Figure 1. Effects of infused bacterial α2,3,6,8-neuraminidase on platelets and erythrocytes in mice. A. ureafaciens neuraminidase (0.5 mU/g of body weight) was injected intravenously into wild-type (WT) (), VWF-/(), or IL4R-IbαTg (▲) mice. Injection of saline into wild-type mice () was performed for comparison. Blood were collected from each mouse via facial vein immediately before (t=0) or days after the injection. Counts of (A) platelets and (B) erythrocytes were performed on a cell counter and normalized with the count before the injection being 100% (mean±standard deviation [SD], n=4). (C, E, G) Platelets and (D, F, H) erythrocytes in the collected whole blood were labeled with cellspecific antibodies and noted lectins (mean±SD, n=3). Each cell was identified through the bound antibody and concurrently the bound lectin detected by flow cytometry. The mean fluorescence intensity (MFI) was calculated for the entire cell population and plotted over the course of days after neuraminidase injection. Comparison to the saline result was performed by unpaired twotailed Student’s t-test. *P<0.05; **P<0.01; ***P<0.001.

haematologica | 2021; 106(1)

Desialylating GPIbα O-glycans causes PLT clearance

signaling and platelet clearance.10,11 To ascertain the involvement of VWF and GPIbα, neuraminidase injection and blood cell count were performed on VWF-/- and IL4RIbαTg mice. In IL4R-IbαTg platelets most of the GPIbα extracellular domain is replaced with that of the α-subunit of interleukin-4 receptor.10,29 Injection of α2,3,6,8neuraminidase significantly reduced the platelet count in VWF-/- mice to the same extent as it did the WT, but did not change platelet count in IL4R-IbαTg mice (Figure 1A). Neuraminidase injection did not affect the erythrocyte count in either strain (Figure 1B). Overall, these results suggest that GPIbα, but not VWF, is required for efficient clearance of desialylated platelets. GPIbα is proteolyzed, or shed, by ADAM17 from the platelet surface.32,33 Inhibition of GPIbα shedding by targeting ADAM17 or GPIbα can impede platelet clearance.34-36 Neuraminidase treatment of platelets induces ADAM17-mediated shedding of GPIbα.6 Thus, it is possible that desialylation accelerates platelet clearance by inducing GPIbα shedding. To ascertain the involvement of GPIbα shedding, neuraminidase injection and blood cell count were performed on Adam17MK-/- mice, which do not express ADAM17 in their platelets. Injection of α2,3,6,8-neuraminidase induced the same extent of thrombocytopenia in Adam17MK-/- mice as in WT (Online Supplementary Figure S2), which indicates that GPIbα shedding is not required for efficient clearance of desialylated platelets.

Desialylation of O-glycans of GPIbα causes platelet clearance To characterize the effects of neuraminidase, platelets

and erythrocytes were collected from each blood draw before and following neuraminidase injection and their glycan contents were assessed through binding of Erythrina cristagalli lectin (ECL), Sambucus nigra agglutinin (SNA), and Arachis hypogaea agglutinin (peanut agglutinin, PNA) (Figure 1C-H). ECL binds to the unsialylated terminal galactose mono saccharide with the highest affinity for galactosyl(b-1,4) N-acetylglucosamine found in both Nand O-glycans.37,38 SNA specifically binds α2,6-linked sialic acid, which occurs primarily in N-glycans, whereas PNA binds specifically unsialylated core 1 and core 2 O-glycans.39,40 For all three lectins, although actual binding levels were different for platelets from three mouse strains at some time points, the trend was similar through all strains. ECL binding increased significantly to a maximum at 24 h after neuraminidase injection, but returned to the pre-injection level at 48 h and remained essentially unaltered thereafter (Figure 1C). SNA binding changed slightly at 24 h, decreased significantly to a minimum at 48 h, and increased again at 72 h (Figure 1E). Like ECL binding, PNA binding increased significantly to a maximum at 24 h; unlike ECL binding, it decreased gradually thereafter to the pre-injection level (Figure 1G). For erythrocytes from all three strains, the ECL and PNA binding levels increased significantly at 24 h, as expected, while SNA binding decreased (Figure 1D, F, H), verifying that neuraminidase desialylated erythrocytes as well.3 However, the changes in lectin binding levels in erythrocytes thereafter were not similar to those in platelets. Overall, the changes of ECL and SNA binding over time do not correlate with those of platelet count, but that of PNA binding does inversely.

Relative PLT counts (%)

C Relative RBC counts (%)

Time after injection (d) haematologica | 2021; 106(1)

Figure 2. Efficacy of α2,3neuraminidase injection in mediating platelet clearance in mice. (A, C) α2,3-neuraminidase (0.6 U/g of body weight) was injected intravenously into wild-type (WT) (), GPIbα-/- (), or IL4RIbαTg (▲) mice. According to the information sheet supplied by the manufacturer, the activity of 1U α2,3-neuraminidase is equivalent of 1 mU A. ureafaciens neuraminidase. For comparison, saline was concurrently injected into the same strains (corresponding open symbols). (B, D) α2,3-neuraminidase was injected intravenously into St3gal1MK-/(◆) and its littermate WT () mice. Blood was collected from each mouse via facial vein immediately before (t=0) or days following the injection. Counts of (A,B) platelets and (C,D) erythrocytes were performed on a CBC counter and normalized with the count before the injection being 100% (mean±standard deviation, n=4-7).

223

Y. Wang et al.

Change of lectin binding (relative ratio to untreated)

Human platelets

Time (min) These results suggest that desialylation of O-glycans may be involved in platelet clearance. In O-glycans, most sialic acids are typically added via an α2,3 linkage by ST3Gal-I.41,42 The GPIbα extracellular domain is heavily decorated with O-glycans.25 To test whether desialylation of O-glycans of GPIbα induces platelet clearance, a recombinant α2,3-neuraminidase was injected into WT, GPIbα-/- and IL4R-IbαTg mice at a dose equivalent to that of the bacterial neuraminidase. Daily blood counts following the injection showed that, compared to the injection of saline, injection of α2,3-neuraminidase significantly decreased platelet counts in wildtype but not IL4R-IbαTg mice (Figure 2A). The extent of decrease in WT was markedly larger than that in GPIbα-/(Figure 2A). In St3gal1MK-/- mice, lack of ST3Gal-I results in hypo-sialylation of O-glycans in platelets. The extent of platelet count decrease in WT mice induced by α2,3-neuraminidase injection was markedly greater than that in St3gal1MK-/- mice (Figure 2B). For comparison, counts of erythrocytes in all strains were not significantly altered by the injection of α2,3-neuraminidase (Figure 2C and D). Overall, these results indicate that desialylation of O-glycans in GPIbα induces platelet clearance. T-antigen (Galb1–3GalNAcα1-Ser/Thr) is the most common O-glycan core structure and is synthesized by a single enzyme termed core 1 b3galactosyltransferase.43 The T-antigen is generally masked by further monosaccharide addition, in particular, sialylation. Thus, desialylation of O-glycans in GPIbα may induce presentation of T224

Murine platelets

Figure 3. Changes in lectin bindings to platelets following treatment of α2,3,6,8- or α2,3-neuraminidase in vitro. Citrated (A, C) human or (B, D) wild-type (WT) murine platelets were collected and treated with (A, B) 10 mU/ml α2,3,6,8-neuraminidase or (C, D) α2,3-neuraminidase at the equivalent activity level at 37˚C for 15, 60, and 180 minutes (min) before being analyzed for lectin binding. The glycan content was detected by flow cytometry using Fluorescein isothiocyanate (FITC)conjugated PNA (), FITCECL(), FITC-SNA() and biotinconjugated MAL-ΙΙ combined with FITC-streptavidin (). The mean fluorescence intensity was quantified for the entire cell population and normalized to untreated sample (0 min). Data are shown as mean ± standard deviation, n=4. Results at 60 min are compared to those of untreated sample by unpaired t-test. *P<0.05; **P<0.01.

antigens on the platelet surface and subsequent platelet opsonization by anti-T-antigen antibodies. To test this possibility, IgG binding was measured following in vitro treatment of α2,3-neuraminidase of citrated human or murine platelet-rich plasma (PRP). Online Supplementary Figure S3 shows that no significant binding of plasma IgG to human or murine platelets was observed following neuraminidase treatment, thus ruling out the possibility that desialylation of O-glycans induces platelet clearance through its opsonization by anti-T-antigen antibodies.

Desialylation induces unfolding of mechanosensory domain of GPIbα in vitro To further characterize the effects of neuraminidase treatment in vitro, citrated PRP was collected from healthy human donors and WT mice, followed by treatment with bacterial neuraminidase or α2,3-neuraminidase for a time period of up to 3 h. Mice receiving neuraminidase at the equivalent dose (unit per mL of blood) exhibited 50% decrease in platelet count at 2 h, and reached <10% platelet count in 8 h (Online Supplementary Figure S4). Before and after the treatment, platelets were probed by flow cytometry for binding of PNA, ECL, SNA and MALII. At 1-3 h of treatment, both neuraminidases induced significant increases in PNA and ECL binding (Figure 3). Binding of MAL-II, which has high affinity for α2,3-sialic acids, was also significantly reduced, confirming desialylation of O-glycans. However, binding of SNA showed mixed trends, likely reflecting a species difference haematologica | 2021; 106(1)

Desialylating GPIbα O-glycans causes PLT clearance

Figure 4. Desialylation of O-glycans induces unfolding of the mechanosensory domain (MSD) in recombinant GPIb-IX. (A) Overlaid consecutive force-distance traces of pulling MSD under a speed of 200 nm/s without (left) and with (right) prior treatment of α2,3-neuraminidase. (B) Averaged MSD unfolding force without (n=49) or with (n=15) treatment of α2,3-neuraminidase. Error bars are standard error of the mean. *P<0.05 by unpaired t-test.

between human and mouse. GPIbα has been implicated in accelerating platelet clearance in several circumstances.44 Several recent reports have linked the step of unfolding of the MSD in GPIbα to the transduction of platelet signaling by GPIb-IX to the induction of platelet clearance.9-11 Located between the sialomucin region and the transmembrane domain of GPIbα, the MSD contains many Ser and Thr residues (Online Supplementary Figure S1). Deletion of approximately 30 residues of MSD reduced the molecular weight of GPIbα by more than 10 kDa,26 indicating that MSD is glycosylated. Recent characterization of a recombinant, unglycosylated form of MSD suggests that it is less stable and more flexible than the native, glycosylated MSD in the GPIb-IX complex.26,27 Thus, we reasoned that desialylation of O-glycans may destabilize the MSD and cause its unfolding. To directly monitor the effect of desialylation on MSD unfolding, single-molecule force spectroscopy was performed. We first verified that purified GPIb-IX complex could be efficiently desialylated by in vitro treatment of α2,3-neuraminidase (Online Supplementary Figure S5). Next, recombinant biotinylated GPIb-IX was immobilized on a streptavidin bead held by a fixed micropipette, and the Fab fragment of MAb WM23 was coupled to a DNA handle-attached bead that was controlled by an optical laser trap.26 The strong tethered bond between WM23 and its epitope in the sialomucin region of GPIbα enabled mulhaematologica | 2021; 106(1)

tiple cycles of stretching and relaxation without the detachment of WM23 from GPIb-IX.26,27 The event of MSD unfolding was observed in 84% of recorded forceextension pulling traces (Figure 4A). In comparison, after treating the immobilized GPIb–IX beads with α2,3-neuraminidase, MSD unfolding was observed in only 24% of pulling traces, indicating that MSD had become unfolded prior to force pulling as observed previously for GPIb-IX mutants that contain already unfolded MSD.26 In addition, analysis of the force-extension traces that contained the unfolding event revealed a significant decrease of the unfolding force upon neuraminidase treatment (Figure 4B). Since neither GPIbb nor GPIX are significantly O-glycosylated,45 these results suggest that desialylation of GPIbα O-glycans in the GPIb-IX complex induces MSD unfolding.

Neuraminidase induces mechanosensory domain unfolding and GPIb-IX signaling in platelets The GPIbα MSD contains the ADAM17 shedding cleavage site therein (Online Supplementary Figure S1). Increased binding of MAb 5G6, which binds the ADAM17 shedding cleavage site in a conformation-insensitive manner,46,47 has been utilized to detect the exposure of this cleavage site in GPIbα, and by extension unfolding of the MSD, on the platelet surface.9-11 Here it was used to probe MSD unfolding on neuraminidase-treated platelets. Since 5G6 recognizes only human but not murine GPIbα,46 transgenic 225

Y. Wang et al. mice expressing only human GPIbα (hTg)28 were used. Injection of neuraminidase induced platelet clearance in hTg mice, although the extent of platelet reduction in hTg mice is less than that in the WT (Online Supplementary Figure S6). Citrated human and murine hTg platelets were treated with α2,3,6,8- or α2,3-neuraminidase in the presence of EDTA for 30 min at 37˚C, after which binding of monoclonal antibodies SZ2 and 5G6 was measured by flow cytometry (Figure 5A-D). Ethylenediamine tetraacetic acid (EDTA) was added to block shedding of GPIbα during desialylation and to keep constant the GPIbα level as

226

shown by the unaltered binding of SZ2. In contrast, binding of 5G6 was significantly increased in platelets treated by either neuraminidase, indicating that the shedding cleavage site in the MSD becomes more accessible,9,10 supporting the notion that the MSD becomes unfolded in both desialylated human and murine platelets. Furthermore, expression of P-selectin and PS lipids on the platelet surface, as well as intracellular calcium influx, all of which are known indicators of GPIb-IX signaling, were significantly increased in platelets after neuraminidase treatment (Figure 5A-D). Previous studies have demonstrated that GPIb-IX signal-

Figure 5. Desialylation induces unfolding of the mechanosensory domain (MSD) and GPIb-IX signaling in platelets. Citrated washed (A,C) hTg or (B,D) human platelets (1x106 cells, with 2mM EDTA) were treated with or without α2,3,6,8-neuraminidase or α2,3-neuraminidase at 37˚C for 30 minutes (min). Monoclonal anti-GPIbα antibodies 5G6 and SZ2, anti-P-selectin antibody, GFP-LactC2, or Fura-2 were then added for 30 min, and the samples were fixed and analyzed by flow cytometry. The mean fluorescence intensity (MFI) was measured for each cell population and normalized against that of untreated platelets (mean±standard deviation [SD], n=4). *P<0.05; **P<0.01 by unpaired t-test. (E-H) Confocal fluorescence images of washed (E) hTg and (G) human platelets that had been treated with PBS buffer (-Neu), α2,3,6,8-neuraminidase (+Neu), or neuraminidase plus 1.5 mg/mL antiGPIbb antibody RAM.1 (+Neu+RAM.1). White arrowheads mark some filopodia in platelets. Scale bar, 10 µm. (F, H) Quantification of filopodia in (F) hTg and (H) human platelets (mean±SD, n=11-12). Images of filopodia from 11-12 view fields (~80-110 platelets per view field) in two independent experiments were visually examined, and counted for comparison.10 **P<0.01; ***P<0.001 by unpaired Student’s t-test.

haematologica | 2021; 106(1)

Desialylating GPIbα O-glycans causes PLT clearance

ing induces formation of filopodia in platelets and transfected mammalian cells.10,48-50 To test whether neuraminidase treatment induces GPIb-IX signaling in platelets, fresh human or hTg platelets were collected, treated with neuraminidase as described above, fixed and stained with phalloidin. Microscopic images of platelets before and after treatment revealed that approximately 40% of desialylated platelets but <10% of sialylated ones exhibited filopodia (Figure 5E and F). The increased filopodia formation was inhibited by anti-GPIbb antibody RAM.110,49 (Figure 5E and F), confirming the role of GPIb-IX in mediating filopodia formation in desialylated platelets. Consistently, St3gal1MK-/- platelets, with hyposialylated O-glycans, display more filopodia than their littermate WT ones (Online Supplementary Figure S7). Overall, these results demonstrate that desialylation induces MSD unfolding and GPIb-IX signaling in both human and murine platelets.

Discussion This study provides a new molecular mechanism of neuraminidase-induced platelet clearance. Earlier studies have identified GPIbα as critical in facilitating clearance of desialylated mouse platelets.3,5 It was suggested that the N-glycans on the LBD of GPIbα are critically involved.13 However, mouse GPIbα does not contain any N-glycosylation sites (Online Supplementary Figure S1). Utilizing genetically modified mice, we report here that GPIbα, but neither VWF nor ADAM17, is required for the efficient clearance of desialylated platelets (Figure 1 and Online Supplementary Figure S2). Following injection of neuraminidase, only the extent of O-glycan desialylation is inversely correlated with the platelet count in mice (Figure 1), leading to the additional discovery that specific desialylation of O-glycans in GPIbα induces platelet clearance

(Figure 2). At the molecular level, desialylation of O-glycans in GPIbα significantly lowers the force threshold to unfold the MSD in GPIb-IX and exposes the shedding cleavage site therein (Figures 4 and 5). Moreover, desialylation of O-glycans activates GPIb-IX signaling in both human and murine platelets (Figure 5 and Online Supplementary Figure S5). Overall, these results suggest that desialylation of O-glycans, instead of N-glycans, in GPIbα induces MSD unfolding and GPIb-IX signaling, and subsequently platelet clearance. Understanding the mechanism of neuraminidase-mediated platelet clearance and the roles of involved factors (Figure 6) in the process may help design and develop effective therapeutics to treat related thrombocytopenic conditions. It has been reported previously that binding of VWF or most anti-LBD antibodies causes unfolding of the MSD, which activates GPIb-IX and induces platelet signaling.9-11 GPIb-IX signaling induced by anti-LBD antibodies results in surface presentation of Neu1 and desialylation of the platelet.7 Consistently, GPIb-IX signaling induced by VWF binding also leads to desialylation of the platelet.10,11 Therefore, we postulate that desialylation of GPIbα Oglycans induces GPIb-IX signaling and leads to surface expression of Neu1. Since Neu1 can remove sialic acids from both N- and O-glycans,51 it should induce exposure of b-galactose on N-glycans of platelet glycoproteins other than GPIbα, which may be recognized by AMR and possibly other receptors (Figure 6). In this model, desialylation of GPIbα O-glycans results in amplified desialylation of N-glycans of other glycoproteins, which reconciles the requirement of GPIbα for rapid platelet clearance after neuraminidase treatment, the preferred recognition of galactoses on N-glycans by AMR, and the lack of N-glycans in mouse GPIbα. The involvement of Neu1 or other lysosomal neuraminidases in this process should be verified by further studies in the future. In this study, we show for the first time that desialyla-

Figure 6. A model for the desialylation-mediated GPIb-IX signaling and platelet clearance. In this model, exogenous neuraminidase removes sialic acids from O-glycans in GPIbα, thereby inducing unfolding of the mechanosensory (MSD), exposure of the Trigger sequence therein, GPIb-IX signaling into the platelet, surface expression of Neu1, and exposure of b-galactoses (b-Gal) on N-glycans of many glycoproteins, which are recognized by the Ashwell-Morell receptor and other receptors for clearance. Unfolding of the MSD also increases the accessibility of the ADAM17 shedding cleavage site in the MSD, facilitates shedding of GPIbα, which results in the exposure of the Trigger sequence.

haematologica | 2021; 106(1)

227

Y. Wang et al.

tion of GPIbα O-glycans destabilizes MSD and increases exposure of the ADAM17 shedding cleavage site therein, which explains an earlier observation that neuraminidase treatment induces ADAM17-mediated shedding of GPIbα.6,52 Shedding of GPIbα has been linked to accelerated platelet clearance.34-36 However, ADAM17, and by extension shedding of GPIbα, is not required for the neu(Online raminidase-induced platelet clearance6 Supplementary Figure S2). The reason for this has not been reported. Our results suggest that desialylation of GPIbα O-glycans induces MSD unfolding, the exposure of the juxtamembrane Trigger sequence therein, and consequently GPIb-IX signaling across the membrane (Figure 6). Since the ADAM17 shedding cleavage site32 is located Nterminal to the Trigger sequence (Online Supplementary Figure S1), shedding of GPIbα would also lead to the exposure of the Trigger sequence and induce platelet clearance. Although inactive ADAM17 may prevent shedding of GPIbα, it should not block desialylation of GPIbα, MSD unfolding and the subsequent exposure of the Trigger sequence. Upon MSD unfolding, sequences in the MSD in addition to the ADAM17 shedding cleavage site may also become more exposed and prone to more proteolysis. Consistent with this notion, GPIbα is fragmented and its expression significantly reduced in murine platelets that lack the Cosmc protein, functional T-synthase and intact core 1 O-glycans.53 In other words, proper sialylation of Oglycans may protect platelets from clearance54 by stabilizing GPIbα. GPIbα-/- and IL4R-IbαTg mice respond differently to neuraminidase injection. Due to the defect in thrombopoiesis, GPIbα-/- mice are severely thrombocytopenic, with less than 10% of the normal platelet count.28,55 Nevertheless, injection of α2,3,6,8-neuraminidase caused a significant reduction in platelet count in GPIbα-/- mice, suggesting that it can desialylate N-glycans on glycoproteins other than GPIbα that are subsequently recognized by AMR.3 In comparison, due to accelerated platelet clearance, IL4R-IbαTg mice are mildly

References 1. Gasic GJ, Gasic TB, Stewart CC. Antimetastatic effects associated with platelet reduction. Proc Natl Acad Sci U S A. 1968;61(1):46-52. 2. Choi SI, Simone JV, Jorney LJ. Neuraminidase-induced thrombocytopenia in rats. Br J Haematol. 1972;22(1):93-101. 3. Grewal PK, Aziz PV, Uchiyama S, et al. Inducing host protection in pneumococcal sepsis by preactivation of the AshwellMorell receptor. Proc Natl Acad Sci U S A. 2013;110(50):20218-20223. 4. Xiang B, Zhang G, Guo L, et al. Platelets protect from septic shock by inhibiting macrophage-dependent inflammation via the cyclooxygenase 1 signalling pathway. Nat Commun. 2013;4:2657. 5. Rumjantseva V, Grewal PK, Wandall HH, et al. Dual roles for hepatic lectin receptors in the clearance of chilled platelets. Nat Med. 2009;15(11):1273-1280. 6. Jansen AJ, Josefsson EC, Rumjantseva V, et al. Desialylation accelerates platelet clearance after refrigeration and initiates GPIbα metalloproteinase-mediated cleavage in mice. Blood. 2012;119(5):1263-1273. 7. Li J, van der Wal DE, Zhu G, et al.

228

10.

11.

12.

13.

14.

thrombocytopenic, with about 50% of the normal platelet count.10,29 Injection of either α2,3,6,8- or α2,3-neuraminidase did not cause significant reduction of platelet count in IL4R-IbαTg mice (Figures 1 and 2). In the IL4RIbαTg platelet, the Trigger sequence in GPIb is exposed due to the removal of GPIbα extracellular residues including the rest of MSD, leading to constitutive GPIb-IX signaling that includes elevated P-selectin expression and significant filopodia formation.10 Considering that desialylation of GPIbα leads to unfolding of MSD and thus the same kind of GPIb-IX signaling (Figures 4 and 5), it is conceivable that neuraminidase cannot cause further GPIb-IX signaling in the IL4R-IbαTg platelet, and thus further reduction of the platelet count in IL4R-IbαTg mice. This is a critical observation, as no mechanisms other than the one described in this paper have been postulated to account for the accelerated clearance rate of IL4R-IbαTg platelets as well as the inability of neuraminidase to reduce platelet count in IL4R-IbαTg mice. Disclosures No conflicts of interest to disclose. Contributions YW, WC, MML-S, CC, WB and KMH designed, performed research, analyzed results, prepared figures and edited the manuscript; WZ performed research and analyzed results; MCB and FL provided critical reagents; XFZ and RL designed research, analyzed results, prepared figures and wrote the paper. Acknowledgments We thank the Emory Children's Pediatric Research Center Flow Cytometry Core and Emory Integrated Cellular Imaging Core for technical support. Funding This work was supported in part by NIH grants HL082808, HL123984 (RL, XFZ) and HL144976 (WB).

Desialylation is a mechanism of Fc-independent platelet clearance and a therapeutic target in immune thrombocytopenia. Nat Commun. 2015;6:7737. Yan R, Chen M, Ma N, et al. Glycoprotein Ibα clustering induces macrophage-mediated platelet clearance in the liver. Thromb Haemost. 2015;113(1):107-117. Quach ME, Dragovich MA, Chen W, et al. Fc-independent immune thrombocytopenia via mechanomolecular signaling in platelets. Blood. 2018;131(7):787-796. Deng W, Xu Y, Chen W, et al. Platelet clearance via shear-induced unfolding of a membrane mechanoreceptor. Nat Commun. 2016;7:12863. Chen W, Druzak SA, Wang Y, et al. Refrigeration-induced binding of von Willebrand factor facilitates fast clearance of refrigerated platelets. Arterioscler Thromb Vasc Biol. 2017;37(12):2271-2279. Riswari SF, Tunjungputri RN, Kullaya V, et al. Desialylation of platelets induced by Von Willebrand Factor is a novel mechanism of platelet clearance in dengue. PLoS Pathog. 2019;15(3):e1007500. Grewal PK, Uchiyama S, Ditto D, et al. The Ashwell receptor mitigates the lethal coagulopathy of sepsis. Nat Med. 2008;14(6):648655. Ellies LG, Ditto D, Levy GG, et al.

15.

16. 17. 18.

19.

20. 21.

Sialyltransferase ST3Gal-IV operates as a dominant modifier of hemostasis by concealing asialoglycoprotein receptor ligands. Proc Natl Acad Sci U S A. 2002; 99(15):10042-10047. Sorensen AL, Rumjantseva V, NayebHashemi S, et al. Role of sialic acid for platelet life span: exposure of β-galactose results in the rapid clearance of platelets from the circulation by asialoglycoprotein receptor-expressing liver macrophages and hepatocytes. Blood. 2009;114(8):1645-1654. Ashwell G, Harford J. Carbohydrate-specific receptors of the liver. Annu Rev Biochem. 1982;51:531-554. Grewal PK. The Ashwell-Morell receptor. Methods Enzymol. 2010;479:223-241. Park EI, Mi Y, Unverzagt C, Gabius HJ, Baenziger JU. The asialoglycoprotein receptor clears glycoconjugates terminating with sialic acid alpha 2,6GalNAc. Proc Natl Acad Sci U S A. 2005;102(47):17125-17129. Lodish HF. Recognition of complex oligosaccharides by the multi-subunit asialoglycoprotein receptor. Trends Biochem Sci. 1991; 16(10):374-377. Spiess M. The asialoglycoprotein receptor: a model for endocytic transport receptors. Biochemistry. 1990;29(43):10009-10018. Lopez JA, Chung DW, Fujikawa K, Hagen FS, Papayannopoulou T, Roth GJ. Cloning of

haematologica | 2021; 106(1)

Desialylating GPIbα O-glycans causes PLT clearance

22.

23.

24.

25. 26.

27.

28.

29.

30.

31.

the α chain of human platelet glycoprotein Ib: a transmembrane protein with homology to leucine-rich α2-glycoprotein. Proc Natl Acad Sci U S A. 1987;84(16):5615-5619. Titani K, Takio K, Handa M, Ruggeri ZM. Amino acid sequence of the von Willebrand factor-binding domain of platelet membrane glycoprotein Ib. Proc Natl Acad Sci U S A. 1987;84(16):5610-5614. Lewandrowski U, Moebius J, Walter U, Sickmann A. Elucidation of N-glycosylation sites on human platelet proteins: a glycoproteomic approach. Mol Cell Proteomics. 2006;5(2):226-233. Chen W, Voos KM, Josephson CD, Li R. Short-acting anti-VWF (von Willebrand Factor) aptamer improves the recovery, survival, and hemostatic functions of refrigerated platelets. Arterioscler Thromb Vasc Biol. 2019;39(10):2028-2037. Li R, Emsley J. The organizing principle of the platelet glycoprotein Ib-IX-V complex. J Thromb Haemost. 2013;11(4):605-614. Zhang W, Deng W, Zhou L, et al. Identification of a juxtamembrane mechanosensitive domain in the platelet mechanosensor glycoprotein Ib-IX complex. Blood. 2015;125(3):562-569. Zhang XF, Zhang W, Quach ME, Deng W, Li R. Force-regulated refolding of the mechanosensory domain in the platelet glycoprotein Ib-IX complex. Biophys J. 2019; 116(10):1960-1969. Ware J, Russell S, Ruggeri ZM. Generation and rescue of a murine model of platelet dysfunction: the Bernard-Soulier syndrome. Proc Natl Acad Sci U S A. 2000;97(6):28032808. Kanaji T, Russell S, Ware J. Amelioration of the macrothrombocytopenia associated with the murine Bernard-Soulier syndrome. Blood. 2002;100(6):2102-2107. Stefanini L, Paul DS, Robledo RF, et al. RASA3 is a critical inhibitor of RAP1-dependent platelet activation. J Clin Invest. 2015; 125(4):1419-1432. De Marco L, Girolami A, Russell S, Ruggeri ZM. Interaction of asialo von Willebrand factor with glycoprotein Ib induces fibrinogen binding to the glycoprotein IIb/IIIa complex and mediates platelet aggregation. J Clin Invest. 1985;75(4):1198-1203.

haematologica | 2021; 106(1)

32. Gardiner EE, Karunakaran D, Shen Y, Arthur JF, Andrews RK, Berndt MC. Controlled shedding of platelet glycoprotein (GP)VI and GPIb-IX-V by ADAM family metalloproteinases. J Thromb Haemost. 2007; 5(7):1530-1537. 33. Bergmeier W, Piffath CL, Cheng G, et al. Tumor necrosis factor-α-converting enzyme (ADAM17) mediates GPIbα shedding from platelets in vitro and in vivo. Circ Res. 2004; 95(7):677-683. 34. Bergmeier W, Burger PC, Piffath CL, et al. Metalloproteinase inhibitors improve the recovery and hemostatic function of in vitro-aged or -injured mouse platelets. Blood. 2003;102(12):4229-4235. 35. Canault M, Duerschmied D, Brill A, et al. p38 mitogen-activated protein kinase activation during platelet storage: consequences for platelet recovery and hemostatic function in vivo. Blood. 2010;115(9):1835-1842. 36. Chen W, Liang X, Syed AK, et al. Inhibiting GPIbα shedding preserves post-transfusion recovery and hemostatic function of platelets after prolonged storage. Arterioscler Thromb Vasc Biol. 2016; 36(9):1821-1828. 37. Debray H, Montreuil J, Lis H, Sharon N. Affinity of four immobilized Erythrina lectins toward various N-linked glycopeptides and related oligosaccharides. Carbohydr Res. 1986;151:359-370. 38. Iglesias JL, Lis H, Sharon N. Purification and properties of a D-galactose/N-acetyl-Dgalactosamine-specific lectin from Erythrina cristagalli. Eur J Biochem. 1982;123(2):247252. 39. Novogrodsky A, Lotan R, Ravid A, Sharon N. Peanut agglutinin, a new mitogen that binds to galactosyl sites exposed after neuraminidase treatment. J Immunol. 1975; 115(5):1243-1248. 40. Shibuya N, Goldstein IJ, Broekaert WF, Nsimba-Lubaki M, Peeters B, Peumans WJ. The elderberry (Sambucus nigra L.) bark lectin recognizes the Neu5Ac(alpha 26)Gal/GalNAc sequence. J Biol Chem. 1987; 262(4):1596-1601. 41. Reboul P, George P, Geoffroy J, Louisot P, Broquet P. Study of O-glycan sialylation in C6 cultured glioma cells: regulation of a betagalactoside alpha 2,3 sialyltransferase activity

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

by Ca2+/calmodulin antagonists and phosphatase inhibitors. Biochem Biophys Res Commun. 1992;186(3):1575-1581. Dall'Olio F, Malagolini N, Trinchera M, Chiricolo M. Mechanisms of cancer-associated glycosylation changes. Front Biosci (Landmark Ed). 2012;17:670-699. Ju T, Brewer K, D'Souza A, Cummings RD, Canfield WM. Cloning and expression of human core 1 beta1,3-galactosyltransferase. J Biol Chem. 2002;277(1):178-186. Quach ME, Chen W, Li R. Mechanisms of platelet clearance and translation to improve platelet storage. Blood. 2018;131(14):15121521. King SL, Joshi HJ, Schjoldager KT, et al. Characterizing the O-glycosylation landscape of human plasma, platelets, and endothelial cells. Blood Adv. 2017;1(7):429442. Liang X, Russell SR, Estelle S, et al. Specific inhibition of ectodomain shedding of glycoprotein Ibα by targeting its juxtamembrane shedding cleavage site. J Thromb Haemost. 2013;z11(12):2155-2162. Tao Y, Zhang X, Liang X, Zang J, Mo X, Li R. Structural basis for the specific inhibition of glycoprotein Ibα shedding by an inhibitory antibody. Sci Rep. 2016;6:24789. Yuan Y, Kulkarni S, Ulsemer P, et al. The von Willebrand factor-glycoprotein Ib/V/IX interaction induces actin polymerization and cytoskeletal reorganization in rolling platelets and glycoprotein Ib/V/IX-transfected cells. J Biol Chem. 1999;274(51):36241-36251. Maurer E, Tang C, Schaff M, et al. Targeting platelet GPIbb reduces platelet adhesion, GPIb signaling and thrombin generation and prevents arterial thrombosis. Arterioscler Thromb Vasc Biol. 2013;33(6):1221-1229. McCarty OJ, Calaminus SD, Berndt MC, Machesky LM, Watson SP. von Willebrand factor mediates platelet spreading through glycoprotein Ib and alpha(IIb)beta3 in the presence of botrocetin and ristocetin, respectively. J Thromb Haemost. 2006;4(6):13671378. Pshezhetsky AV, Richard C, Michaud L, et al. Cloning, expression and chromosomal mapping of human lysosomal sialidase and characterization of mutations in sialidosis. Nat Genet. 1997;15(3):316-320.

229

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):230-237

Platelet Biology & Its Disorder

Characterization of breakthrough hemolysis events observed in the phase III randomized studies of ravulizumab versus eculizumab in adults with paroxysmal nocturnal hemoglobinuria Robert A. Brodsky,1 Régis Peffault de Latour,2,3 Scott T. Rottinghaus,4 Alexander Röth,5 Antonio M. Risitano,6 Ilene C. Weitz,7 Peter Hillmen,8 Jaroslaw P. Maciejewski,9 Jeff Szer,10 Jong Wook Lee,11 Austin G. Kulasekararaj,12 Lori Volles,4a Andrew I. Damokosh,4 Stephan Ortiz,4 Lori Shafner,4b Peng Liu,4 Anita Hill8c and Hubert Schrezenmeier13,14

Division of Hematology, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA; 2Université Paris Diderot, Paris, France; 3French Reference Center for Aplastic Anemia and PNH Hematology-Bone Marrow Transplantation, Hôpital Saint-Louis AP-HP, Paris, France; 4Alexion Pharmaceuticals Inc., Boston, MA, USA; 5Department of Hematology, West German Cancer Center, University Hospital Essen, University of Duisburg-Essen, Essen, Germany; 6Hematology, Department of Clinical Medicine and Surgery, Federico II University of Naples, Naples, Italy; 7Jane Anne Nohl Division of Hematology, Keck-USC School of Medicine, Los Angeles, CA, USA; 8Department of Haematology, St James's University Hospital, Leeds, UK; 9Department of Translational Hematology and Oncology Research, Taussig Cancer Institute, Cleveland Clinic Foundation, Cleveland, OH, USA; 10Clinical Haematology, Royal Melbourne Hospital, Melbourne, Victoria, Australia; 11Department of Hematology, Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; 12Department of Haematological Medicine, King’s College Hospital, NIHR/Wellcome King’s Clinical Research Facility, London, UK; 13Institute for Clinical Transfusion Medicine and Immunogenetics, German Red Cross Blood Transfusion Service Baden-Württemberg-Hessen, Germany and University Hospital Ulm, Ulm, Germany and 14Institute of Transfusion Medicine, University of Ulm, Ulm, Germany 1

Current address: Principal, Volles Consulting Group, Old Lyme, CT, USA. Current address: Vista Life Innovations, Madison, CT, USA. c Current address: Alexion Pharmaceuticals, Inc., Boston, MA, USA. a b

Correspondence: ROBERT A. BRODSKY rbrodsky@jhmi.edu Received: September 3, 2019. Accepted: January 9, 2020. Pre-published: January 16, 2020.

ABSTRACT

culizumab is first-line treatment for paroxysmal nocturnal hemoglobinuria (PNH); however, approximately 11-27% of patients may experience breakthrough hemolysis (BTH) on approved doses of eculizumab. Ravulizumab, a new long-acting C5 inhibitor with a four times longer mean half-life than eculizumab, provides immediate, complete, and sustained C5 inhibition over 8-week dosing intervals. In two phase III studies, ravulizumab was non-inferior to eculizumab (P ≤0.0004) for the BTH endpoint; fewer patients experienced BTH with ravulizumab versus eculizumab in both studies (301 [complement inhibitor−naïve patients], 4.0% vs. 10.7%; 302 [patients stabilized on eculizumab at baseline], 0% vs. 5.1%). In the current analysis, patientlevel data were evaluated to assess causes and clinical parameters associated with incidents of BTH reported during the 26-week treatment periods in the ravulizumab phase III PNH studies. Of the five BTH events occurring in ravulizumab-treated patients across the studies, none were temporally associated with suboptimal C5 inhibition (free C5 ≥0.5 mg/mL); four (80%) were temporally associated with complement-amplifying conditions (CAC). Of the 22 events occurring in eculizumab-treated patients, 11 were temporally associated with suboptimal C5 inhibition, including three events also associated with concomitant infection. Six events were associated with CAC only. Five events were unrelated to free C5 elevation or reported CAC. These results suggest that the immediate, inf

https://doi.org/10.3324/haematol.2019.236877

230

haematologica | 2021; 106(1)

Breakthrough hemolysis: ravulizumab vs. eculizumab

complete, and sustained C5 inhibition achieved through weight-based dosing of ravulizumab reduces the risk of BTH by eliminating BTH associated with suboptimal C5 inhibition in patients with PNH. (Registered at clinicaltrials.gov identifiers: Study 301, NCT02946463; Study 302, NCT03056040.)

Introduction Until recently, eculizumab was the only approved treatment for paroxysmal nocturnal hemoglobinuria (PNH),1,2 and since its regulatory approval in 2007, it has changed the paradigm for the treatment of patients with PNH.3,4 Longterm experience has established that eculizumab is efficacious and well tolerated; however, approximately 11-27% of patients may experience breakthrough hemolysis (BTH) on approved dosages of eculizumab during long-term treatment.5-7 Breakthrough hemolysis, characterized by the return of intravascular hemolysis and reappearance of classical PNH symptoms,5,8-11 may occur due to suboptimal C5 inhibition1 and/or CAC such as infection, surgery, or pregnancy that may lead to increased complement activation resulting from higher C3b density.12-14 In some patients with suboptimal C5 inhibition or CAC, BTH may be ameliorated by shortening the 2-week dosing interval and/or increasing the dose of eculizumab.5,7 Ravulizumab is a new long-acting C5 inhibitor developed to reduce the treatment burden associated with eculizumab through an improved dosing regimen and is now approved for treatment of adult patients with PNH in the USA, Japan, and Europe.11,15-17 It is notable that the mean terminal halflife of ravulizumab is approximately four times longer than that of eculizumab, allowing ravulizumab to provide immediate, complete, and sustained terminal C5 inhibition with an 8-week dosing interval.11,15-17 In the two largest international phase III clinical studies conducted to date in PNH patients, ravulizumab was noninferior to eculizumab across all key efficacy endpoints in patients naїve to complement inhibitor therapy (study 301) as well as in those on stable eculizumab therapy (study 302).16,17 BTH, a key secondary endpoint in both studies, was defined as one or more new or worsening symptoms or signs of intravascular hemolysis (fatigue, hemoglobinuria, abdominal pain, dyspnea, anemia [hemoglobin <10 g/dL], major adverse vascular event [MAVE] including thrombosis, dysphagia, or erectile dysfunction) in the presence of elevated lactate dehydrogenase (LDH) ≥2 times the upper limit of normal (ULN) after prior LDH reduction to <1.5xULN while on therapy. Point estimates for proportions of patients with BTH favored ravulizumab in study 301 (4.0% vs. 10.7%, difference, −6.7% [95% confidence interval (CI): −14.21, 0.18]; P <0.0001) and study 302 (0% vs. 5.1%, difference, −5.1% [95%CI: −8.89, 18.99; P <0.0004).16,17 The purpose of the current analysis was to investigate the causes of and clinical parameters associated with incidents of BTH at the patient level in both studies. Post hoc analyses were also performed to further evaluate BTH in ravulizumab- and eculizumab-treated patients.

In study 301 (clinicaltrials.gov identifier: NCT02946463), adult patients naїve to complement inhibitor with LDH ≥1.5 times the ULN and at least one PNH symptom (consistent with high disease activity as described in the Summary of Product Characteristics2) were randomized 1:1 to receive ravulizumab or eculizumab for 183 days.17 In study 302 (clinicaltrials.gov identifier: NCT03056040), adult patients with PNH stable on eculizumab therapy (LDH <1.5 times ULN) for at least 6 months were randomized 1:1 to ravulizumab or eculizumab for 183 days.16 Patients randomized to ravulizumab received loading followed by weight-based dosing every 8 weeks.16,17 Patients randomized to eculizumab received 900 mg every two weeks.16,17 In study 301, patients with a hemoglobin level ≤7 g/dL or ≤9 g/dL in the presence of anemia-related signs or symptoms warranting transfusion received red blood cell transfusion.17 Data from patients who experienced BTH in the 26week study period were subject to detailed investigation of BTH events. Both studies were approved by the institutional review board or independent ethics committee at participating centers and were conducted in accordance with the Declaration of Helsinki and the Council for International Organizations of Medical Sciences International Ethical Guidelines.

Derivation of breakthrough hemolysis definition Because there is no consensus in the medical literature regarding the definition of BTH, the definition was derived prospectively for use in both studies based on a literature review and interviews of eight study investigators. The resulting definition18 was reviewed with and agreed upon by the US Food and Drug Administration, the European Medicines Agency, and the Japan Pharmaceuticals and Medical Device Agency before studies started. Literature review - the literature review revealed that several biomarkers have been used to describe BTH (Online Supplementary Table S1). Elevated LDH level was most commonly used;6,14,19-22 however, variables such as eculizumab level,20-23 other markers of serum hemolytic activity,5,7 aspartate aminotransferase level,6 reticulocyte count,24 and bilirubin level24 have also been used. There was no consensus on whether hemoglobin levels and/or transfusion requirements should be considered when defining BTH.7,14,19,21,24 Physician interviews - parameters used by participating investigators to measure BTH included symptomatic manifestations (e.g., hemoglobinuria, abdominal pain, and other PNH-related symptoms) and laboratory indicators (e.g., decreased hemoglobin levels, elevated LDH levels, increased total complement levels, and increased bilirubin levels) (Online Supplementary Table S2).

inf

Methods Patients and treatment

Both studies (study 301 and 302) included patients aged ≥18 years with a confirmed diagnosis of PNH by flow cytometry.16,17 haematologica | 2021; 106(1)

Outcomes The key outcome of interest in this study was BTH causality. BTH events were categorized as the following: (i) temporal association that is free C5–related, defined as BTH associated with time-matched occurrence of free C5 ≥0.5 mg/mL;1 (ii) CAC-related, defined as BTH due to an inciting event (e.g., infection, trauma, or surgery); or (iii) BTH unrelated to elevated C5 and without a reported time-matched CAC. BTH causality was also analyzed by assessment of the correlation between serum free C5 concentrations <0.5 mg/mL or ≥0.5 mg/mL and BTH.

Additional analyses Additional post hoc analyses were conducted as follows: (i) 231

R.A. Brodsky et al. Table 1. Incidence of breakthrough hemolysis and overall temporal association.

Study 301 (Naïve patients) Ravulizumab Eculizumab n=125 n=121 Patients with BTH, n (%) BTH events, n BTH events with free C5 ≥0.5 mg/mL BTH events with infection (with no free C5 elevation) BTH events unrelated to elevated free C5 or infectionc

5 (4.0) 5 0 4 1

13 (10.7) 15 7a 4 4

Study 302 (Patients stable on eculizumab) Ravulizumab Eculizumab n=97 n=98 0 (0.0) 0 0 0 0

5 (5.1) 7 4b 2 1

n: total number of patients in treatment group; BTH: breakthrough hemolysis. aTwo patients in the eculizumab group with suboptimal C5 inhibition also had concomitant infection. bOne patient in the eculizumab group with suboptimal C5 inhibition also had concomitant infection. cThese cases had neither suboptimal C5 inhibition nor concomitant infection identified to explain cause of breakthrough hemolysis.

estimated prevalence of LDH excursion as reflected by elevations in LDH to ≥2xULN without regard for signs or symptoms of hemolysis; (ii) estimated exposure-adjusted incidence of BTH per 100 patient-years of study drug exposure; (iii) estimated time to first BTH events temporally associated with suboptimal C5 inhibition with adjustment for competing risk due to CAC or undetermined causality; and (iv) correlation between serum free C5 concentrations <0.5 mg/mL or ≥0.5 mg/mL and BTH events. (See Online Supplementary Appendix for further details of the methods used.)

Results Patients Study 301 included 125 patients treated with weightbased dosing of ravulizumab and 121 patients treated with approved eculizumab dose (900 mg every 2 weeks, q2w), and study 302 included 97 patients treated with weight-based dosing of ravulizumab and 98 patients treated with eculizumab (900 mg, q2w).16,17 We have previously reported patient demographics and clinical characteristics at baseline for both studies. Briefly, the percentage of male patients was 54.5% and 50.3% and the mean (standard deviation [SD]) age at first infusion of study drug was 45.5 (15.7) years and 47.7 (14.2) years in studies 301 and 302, respectively.16,17 Mean LDH (SD) was 1,606.4 (752.7) U/L and 231.6 (49.2) U/L in study 301 and study 302, respectively.16,17 In study 301, 13.8% of patients had baseline LDH 1.5 to <3xULN and 86.2% of patients had baseline LDH ≥3xULN.17 Mean (SD) glycosylphosphatidylinositol-deficient granulocyte clone size was 84.7% (20%) in study 301 and 83.3% (22.5%) in study 302.16,17 Mean (SD) total PNH red blood cell clone size was 38.6% (23.4%) in study 301 and 60.1% (31.9%) in study 302.16,17 Mean (SD) hemoglobin levels in study 301 were 9.4 (1.5) g/dL in the ravulizumab arm and 9.6 (1.7) g/dL in the eculizumab arm, and in study 302 were 11.1 (1.8) g/dL in the ravulizumab arm and 10.9 (1.8) g/dL in the eculizumab arm.

Incidence of breakthrough hemolysis As previously reported, there were numerically fewer occurrences of BTH in patients treated with ravulizumab compared with eculizumab in both studies.16,17 BTH recurred in the eculizumab but not the ravulizumab arms of both studies: two of the 13 patients experiencing BTH suffered two events each in the 301 study, and in study 302 one of five patients experiencing BTH suffered three 232

events. In study 301, it was noted that the mean (SD) baseline LDH was numerically higher in patients who experienced BTH (1,764.1 [809.7]) as compared with those who did not (1,593.9 [748.5]).

Breakthrough hemolysis causality: temporal association In study 301, none of the five BTH events in the ravulizumab group were temporally associated with suboptimal C5 inhibition (free C5 ≥0.5 mg/mL) (Table 1); four were temporally associated with CAC (all infections), and one did not have concomitant free C5 ≥0.5 mg/mL or a time-matched CAC reported. In the eculizumab group, seven of the 15 BTH events were temporally associated with suboptimal C5 inhibition; CAC (all infections) were associated with six of the 15 events, including two events in patients who also had suboptimal C5 inhibition, and four patients did not have concomitant free C5 ≥0.5 mg/mL or a time-matched CAC reported. In study 302, no BTH events were observed in patients treated with ravulizumab. Four of the seven events that occurred in the eculizumab group were temporally associated with suboptimal C5 inhibition, and three events were associated with CAC (all infections), including one event in a patient who also had suboptimal C5 inhibition (Table 1). This patient had three BTH events spanning 113 days of treatment. On day 113, the patient was hospitalized for BTH with symptoms of vomiting, flu-like symptoms, and cola-colored urine; the vomiting was thought to have been caused by a viral infection. This patient discontinued treatment and left the study due to lack of efficacy. One event was neither associated with incomplete C5 inhibition nor a reported CAC.

Breakthrough hemolysis: patient narratives Patient narratives on BTH in study 301 are shown in Table 2 for the ravulizumab group and in Table 3 for the eculizumab group. BTH events occurring in the eculizumab group in study 302 are summarized in Table 4. Overall, the majority of patients with BTH (70% [16 of 23]) exhibited one or two PNH-related signs or symptoms; the most commonly reported PNH-related signs and symptoms were anemia, dyspnea, hemoglobinuria, and fatigue. With respect to red blood cell transfusion, 8 patients (3 in the ravulizumab group and 5 in the eculizumab group) experiencing BTH required transfusion in study 301 (Tables 2 and 3) and 3 patients experiencing BTH in the eculizumab group required transfusion in study 302 (Table 4). haematologica | 2021; 106(1)

Breakthrough hemolysis: ravulizumab vs. eculizumab

Table 2. Ravulizumab breakthrough hemolysis events and narratives: study 301.

Patients’ characteristics (sex, age, body weight)

Breakthrough hemolysis event; symptoms

Study day

LDHa (U/L)

Female; 34 y; 115 kg

155 169

593 511

0.105 0.101

Female; 30 y; 57 kg Female; 24 y; 57 kg

1st; fatigue abdominal pain, dyspnea 1st; hemoglobinuria

687

1st; hemoglobinuria, anemia

113 127 155 169 183 71 85 99 99

517 773 513 926 555 544 525 827 615

Male; 37 y; 66 kg

1st; anemia

Male; 43 y; 70 kg

1st; anemia

Free C5b RBC (mg/mL) transfusion (U)

Possible CAC

Association

None

Giardiasis

CAC

0.0787

None

Viral infection

CAC

0.0602 0.0768 0.0428 0.0895 0.0909 0.0623 0.0414 0.0505 0.0766

Influenza, CAC upper respiratory infection

Gum infection

CAC

None

Unexplained

C5: complement component 5; CAC: complement-amplifying condition; LDH: lactate dehydrogenase; Pt: patient; RBC: red blood cell; U: number of units transfused; y: years.. a The upper limit of normal for LDH is 246 U/L. bFree C5 concentrations were quantified using a Gyros-based fluorescence assay.

Additional analyses Lactate dehydrogenase excursion only - an analysis assessing the prevalence of excursions of LDH ≥2xULN after reduction below 1.5xULN without regard to symptoms (i.e., the LDH portion of the BTH definition) showed treatment with ravulizumab to be superior to eculizumab with significantly fewer ravulizumab-treated patients experiencing LDH excursions compared with eculizumab-treated patients in both study 301 (8.8% [95%CI: 3.8, 13.8] vs. 20.7% [95%CI: 13.5, 27.9], respectively; treatment difference −11.7% [95%CI: −20.7, −2.7]; P=0.012) and study 302 (5.2% [95%CI: 0.8, 9.6] vs. 16.3% [9.0, 23.6], respectively; treatment difference, −11.2% [95%CI: −20.3, −2.4]; P=0.015). Exposure-adjusted BTH event rate - an analysis to evaluate the estimated incidence of BTH events per 100-patient years of exposure showed rates to be approximately 3-fold higher (incidence rate ratio, 0.32 [95%CI: 0.11, 0.92]; P=0.034) in the eculizumab group (21.5 [95%CI: 8.9, 51.7] events per 100 patient-years) versus the ravulizumab group (6.8 [95%CI: 2.2, 21.5] events per 100 patient-years) in study 301. In study 302, the estimated incidence rate per 100-patient years of exposure was 19.9 (95%CI: 7.2, 54.9) in the eculizumab group, whereas the exposureadjusted incidence rate was non-estimable in the ravulizumab group because no events were observed. Time to first event with adjustment for competing risk - in study 301, the analysis of time to first BTH event due to any cause showed significantly lower hazard in the ravulizumab group compared with the eculizumab group (hazard ratio [HR] 0.36 [95%CI: 0.13, 1.0]; P=0.049) (Online Supplementary Figure S1A). In the analysis to assess BTH events temporally associated with suboptimal C5 inhibition, after adjustment for competing risk due to a CAC or undetermined causality, the probability of a BTH event due to suboptimal C5 inhibition was shown to be significantly reduced with the ravulizumab group compared with the eculizumab group (HR 0; P<0.001) (Online Supplementary haematologica | 2021; 106(1)

Figure S2A). In a conservative evaluation, when the same analysis was applied to BTH events temporally associated with suboptimal C5 inhibition or undetermined causality after adjustment for competing risk due to a CAC, the hazard was still significantly lower in the ravulizumab group (HR [95%CI]: 0.10 [0.01, 0.81]; P=0.031) (Online Supplementary Figure S3A). In study 302, the hazard ratio estimates in the time-to-event analyses were all zero due to no BTH events occurring in the ravulizumab group (Online Supplementary Figures S1B, 2B and 3B). Correlation of BTH with serum free C5 levels - all post-baseline free C5 values in ravulizumab-treated patients were <0.5 mg/mL in both the 301 and 302 studies throughout 26 weeks of treatment. In study 301, the overall percentage of patients with BTH among patients who had any free C5 concentration ≥0.5 mg/mL across both treatment groups was 40% compared with 5.2% among those who had free C5 concentrations <0.5 mg/mL for every assessment (relative risk, 7.7) (Online Supplementary Table S3). In study 302, the percentage of patients with BTH among patients who had any free C5 concentration ≥0.5 mg/mL was 28.6% compared with 1.6% in patients with all free C5 concentrations <0.5 mg/mL (relative risk, 17.9) (Online Supplementary Table S3).

Discussion In patients with PNH receiving complement inhibitor therapy, a BTH event represents loss of disease control. BTH is manifested by classical PNH symptoms and can require blood transfusion,12,14,22 but more critically it can be associated with the return of the morbidity associated with PNH, including potentially life-threatening thromboembolic events.25-27 A consensus definition of BTH derived after a literature review and interviews with PNH experts, and prospectively accepted by regulatory authorities, was used in the two phase III studies of ravulizumab 233

R.A. Brodsky et al. Table 3. Eculizumab breakthrough hemolysis events and narratives: study 301.

Patients’ characteristics (sex, age, body weight)

Breakthrough hemolysis event; symptoms

Study day

LDHa (U/L)

Free C5b (µg/mL)

RBC transfusion (U)

Possible CAC

Male; 25 y; 92 kg Male; 45 y; 74 kg Female; 35 y; 88 kg

1st; anemia

866

55.9

None

Free C5 ≥0.5 mmg/mL

1st; hemoglobinuria

155

933

34.4

None

Free C5 ≥0.5 mmg/mL

57 71 169 183

571 1164 890 865

24.2 80.0 58.6 64.4

None

Free C5 ≥0.5 mg/mL

None

183

3720

86.1

Upper respiratory infection None

506

0.0445

None

Common cold

529

0.0644

700

0.148

None

Upper respiratory infection Non-specific infection

141

524

0.189

None

1242 1088 1172 653 4080 >4200 524

18.2 1.46 17.6 1.41 90.9 NA NA

1st; dyspnea

99 113 155 169 183 Unscheduled 57

None

Upper respiratory infection, acalculous cholecystitis None

1st; anemia

597

0.03

None

Unexplained

1st; fatigue, anemia st 1 ; abdominal pain

141

520

0.0748

None

Unexplained

169

579

0.0411

None

Unexplained

2 3

4 5

6 7 8 9

10 11 12 13

1st; fatigue, dyspnea, anemia 2nd; fatigue, hemoglobinuria, dyspnea, anemia Male; 39 y; 1st; fatigue, 94 kg anemia Male; 39 y; 1st; fatigue, 70 kg hemoglobinuria, abdominal pain, dyspnea, anemia, erectile dysfunction Female; 49 y; 1st; fatigue, 56 kg dyspnea, anemia Male; 35 y; 1st; anemia 93 kg Male; 57 y; 1st; dyspnea, 89 kg anemia Male; 52 y; 1st; fatigue, 73 kg anemia, hemoglobinuria 2nd; hemoglobinuria, anemia

Female; 50 y; 72 kg Male; 28 y; 55 kg Female; 64 y; 82 kg Female; 29 y; 48 kg

Influenza Bronchitis None

Association

Free C5 ≥0.5 mmg/mL + CAC Free C5 ≥0.5 mg/mL (missed day 169 dose) CAC

CAC CAC CAC Free C5 ≥0.5 mg/mL Free C5 ≥0.5 mg/mL + CAC

Unexplained

C5: complement component 5; CAC: complement-amplifying condition;LDH: lactate dehydrogenase; NA: not available; Pt: patient; RBC: red blood cell; U: number of units transfused; y: years. aThe upper limit of normal for LDH is 246 U/L. bFree C5 concentrations were quantified using electrochemiluminescence ligand binding assay.

to evaluate the incidence and causality of BTH in patients who were naїve to complement inhibitor therapy as well as those who were previously stabilized on eculizumab.18 In both studies, ravulizumab was shown to be non-inferior to eculizumab on the assessment of BTH,16,17 and treatment with ravulizumab was associated with numerically fewer episodes of BTH compared with eculizumab in each of these studies. The current analysis showed that no BTH events in patients treated with weight-based dosing of ravulizumab were associated with elevations in free C5 levels. In contrast, several patients treated with eculizumab across both studies had BTH events that were temporally associated 234

with elevations in free C5 levels (C5 ≥0.5 mg/mL), suggesting these patients may have had suboptimal C5 control. In such cases, it has been shown that BTH can be successfully managed in some patients by adjusting the dosing amount and/or frequency of eculizumab administration that differs from the approved regimen of eculizumab.5,7,12 However, in both studies (301 and 302), patients required maintenance with the approved eculizumab dosing (900 mg every 2 weeks); no changes to the dose levels or dosing regimen were permitted during the study, and if any such changes were made, these patients were excluded from the studies. Correspondingly, regardless of the treatment group, risk haematologica | 2021; 106(1)

Breakthrough hemolysis: ravulizumab vs. eculizumab

Table 4. Eculizumab breakthrough hemolysis events and narratives: study 302.

Patients’ characteristics (sex, age, body weight)

Breakthrough hemolysis event; symptoms

Study day

LDHa (U/L)

Male; 29 y; 95 kg

1st; hemoglobinuria 2nd; hemoglobinuria 3rd; hemoglobinuria 1st hemoglobinuria

24.1 24.8 19.3 91.9 0.1

None None None

Male; 34 y; 71 kg Female; 47 y; 75 kg

1257 1037 811 3846 618

None None 1

29 57 99 113 141

None

176

515

0.1

Flu-like symptoms Acute pyelonephritis

127

1846

2.1

155

799

0.1

None

Male; 60 y; 79 kg

Male; 60 y; 84 kg

1st; fatigue, hemoglobinuria, dyspnea, anemia 1st; fatigue, hemoglobinuria, anemia 1st; fatigue, dyspnea

Free C5b RBC (mg/mL) transfusion (U)

Possible CAC

Association

Free C5 ≥0.5 mg/mL Free C5 ≥0.5 mg/mL Free C5 ≥0.5 mg/mL CAC CAC

Gastroenteritis Free C5 ≥0.5 mg/mL + CAC

None

Unexplained

C5: complement component 5; CAC: complement-amplifying condition; LDH: lactate dehydrogenase; Pt: patient; RBC: red blood cell; U: number of units transfused; y: year. a The upper limit of normal for LDH is 246 U/L. bFree C5 concentrations were quantified using electrochemiluminescence ligand binding assay.

of BTH was approximately 8-fold (40% vs. 5% in study 301) and 18-fold (29% vs. 2% in study 302) higher in patients with free C5 ≥0.5 mg/mL compared with those who had free C5 concentrations <0.5 mg/mL for every assessment. These results suggest that reduction of free C5 may be associated with reduced risk of BTH in patients with PNH. In study 301 (complement inhibitor–naїve patients), similar proportions of patients in each treatment group experienced infection-related BTH (ravulizumab, 3.2% [4 of 125]; eculizumab, 3.3% [4 of 121]), possibly due to proximal complement activation. There was one patient in the ravulizumab group and four patients in the eculizumab group who had adequate C5 inhibition and no apparent infection or other CAC reported by the investigators. In study 302 (patients stabilized on eculizumab at baseline), no ravulizumab-treated patients and 2.0% (2 of 98) of eculizumab-treated patients experienced infectionrelated BTH. One eculizumab-treated patient had a BTH event that was not attributed to suboptimal C5 inhibition or an identifiable CAC. Observations from these studies suggest that BTH can occur due to infections despite the fact that the terminal complement activity is suppressed. The causes of hemolysis in the setting of infection/sepsis have not been fully elucidated.28 It has been shown that exposure of host red blood cells to infectious pathogen cells can cause hemolysis independent of complement activity, suggesting that the complement system may not be the sole cause of infection-triggered hemolysis.13,28 Results from the additional analyses of BTH supported and extended observations from the primary analyses. If only the objective LDH component of the BTH definition was considered, a significantly lower proportion of ravulizumab-treated patients experienced BTH compared with eculizumab in both studies. Treatment with ravulizumab was also shown to be associated with a 3-fold lower exposure-adjusted incidence of BTH events than eculizumab (6.8 vs. 21.5 events per 100 patient-years haematologica | 2021; 106(1)

of exposure), and a 90% lower risk of BTH due to suboptimal C5 inhibition compared with eculizumab (HR, 0.10; P=0.031) in study 301. Data in study 302 also support lower risk of BTH with no event occurring in the ravulizumab group. A unique aspect of these two studies is the consensus definition of BTH. Although not all potential biomarkers and/or causative factors were utilized in the definition of BTH that was used in these phase III studies (e.g., reduction of hemoglobin levels, elevation in reticulocyte count, subtherapeutic serum levels of complement inhibitor), the definition is conservatively based on objective criteria (LDH levels) and well-known, easily identifiable PNHrelated signs and symptoms (e.g., anemia, hemoglobinuria, fatigue, dyspnea), which may facilitate early recognition and treatment of BTH in patients with PNH receiving complement inhibitor therapy. Some limitations to this analysis are worthy of note. Approximately 15% of the observed BTH events were due to unexplained causes (unrelated to insufficient C5 inhibition and without a recognized/reported CAC). These events may have been caused by unrecognized or unreported CAC or other etiological factors. Notably, none of the BTH events in these studies were associated with major adverse vascular events (MAVE). However, this is not surprising since the incidence of MAVE was low (1.6% [n=2] and 0.8% [n=1] for ravulizumab and eculizumab, respectively, in study 301, and no patients experienced MAVE in study 302). Given the importance of BTH in the potential development of associated thrombosis, rigorous clinical interrogation may reveal other and/or new factors associated with manifestations of BTH. Considering the conservative approach for the additional analyses of the BTH data from these studies, where events due to suboptimal C5 inhibition were pooled with undetermined causality and then adjusted for competing risks of CAC, it is likely that the significant between-group differences observed in BTH were driven by treatment-related effects on inhibition of free C5 concentrations. Due to 235

R.A. Brodsky et al.

the strict definition of BTH applied in these studies, patients were required to have both high LDH levels and one or more new or worsening signs/symptoms of intravascular hemolysis. Because symptom reporting is at least in part subjective, it is possible the results reported herein could be confounded by ascertainment bias associated with under-reporting of BTH. For example, PNHrelated symptoms may have been unreported due to individual patient or cultural norms, or the symptoms were not associated with severe anemia or hemoglobinuria. This hypothesis is strengthened by the higher number of patients in both treatment arms who experienced LDH elevations but did not have symptoms of BTH reported by investigators. On the other hand, it is possible that patients with infection might have elevated LDH due to the infection29-31 rather than intravascular hemolysis, thereby overestimating the frequency of hemolysis associated with infections. In summary, weight-based dosing of ravulizumab administered every 8 weeks was associated with numerically fewer episodes of BTH versus eculizumab administered 900 mg every 2 weeks over 26 weeks of complement inhibitor therapy in PNH patients with high disease activity. The observed differences in BTH rates for ravulizumab versus eculizumab may be attributable to the ability of ravulizumab to completely inhibit free C5 over the entire 8-week dosing interval. Furthermore, no BTH events in the ravulizumab group were associated with free C5 concentrations ≥0.5 mg/mL. In contrast, some patients treated with eculizumab experienced multiple BTH events that were temporally associated with elevations in free C5 (one resulting in hospitalization). Similar numbers of patients receiving ravulizumab or eculizumab experienced CAC-related BTH, possibly due to proximal complement activation. Overall, results from these two studies demonstrate that ravulizumab achieved immediate and complete inhibition of free C5 over the entire 26-week treatment period, reducing the overall risk of BTH by eliminating free C5-associated BTH. Disclosures RAB is a member of the scientific advisory board for and receives grant funding from Alexion Pharmaceuticals, Inc.; RPdL has received honoraria, consulting fees, and research support from Alexion Pharmaceuticals, Inc., Pfizer, and Novartis, and has received research support from Amgen; STR is an employee and stockholder of Alexion Pharmaceuticals, Inc.; AR has received honoraria and consulting fees from Alexion Pharmaceuticals, Inc., Apellis Pharmaceuticals, Novartis, and Roche; AMR has received research support, honoraria, and consulting fees from Alexion Pharmaceuticals, Inc., Novartis, Alnylam, and Ra Pharma, lecture fees from Alexion Pharmaceuticals, Inc., Novartis, Pfizer, and Jazz, and served as an advisory board member for Alexion Pharmaceuticals, Inc.,

References ®

1. SOLIRIS (eculizumab) injection, for intravenous use. Boston, MA, USA: Alexion Pharmaceuticals Inc. 06/2019. 2. Soliris (eculizumab). [Summary of product characteristics] Paris, France: Alexion Europe SAS, 2018. 3. Griffin M, Munir T. Management of throm-

236

Novartis, Pfizer, and Jazz, as well as consultant for Amyndas; ICW has received honoraria and consulting fees from Alexion Pharmaceuticals, Inc.; PH has received honoraria from and has been a consultant for Alexion Pharmaceuticals, Inc., and has received research support (to the University of Leeds) from Apellis Pharmaceuticals; JPM has received consulting fees from Alexion Pharmaceuticals, Inc., Apellis Pharmaceuticals, and Ra Pharma; has also received speaker fees from and is a member of the Executive Committee of the International PNH Registry for Alexion Pharmaceuticals, Inc.; JS has received research support (to Royal Melbourne Hospital), honoraria, consulting fees, and travel support from Alexion Pharmaceuticals, Inc.; JWL has received honoraria, consulting fees, and research support (to Seoul St. Mary’s Hospital) from Alexion Pharmaceuticals, Inc.; AGK has received honoraria, travel support, and consulting fees from Alexion Pharmaceuticals, Inc.; LV is an employee and stockholder of Alexion Pharmaceuticals, Inc.; AID was an employee and stockholder of Alexion Pharmaceuticals, Inc. at the time of the study analysis and during the development of this manuscript; SO is an employee and stockholder of Alexion Pharmaceuticals, Inc.; LS was an employee and stockholder of Alexion Pharmaceuticals, Inc. at the time of this analysis and during development of the manuscript; PLis an employee and stockholder of Alexion Pharmaceuticals, Inc.; AH has received honoraria and consulting fees from Alexion Pharmaceuticals, Inc.; HS has received honoraria and research support (both to University of Ulm) from Alexion Pharmaceuticals, Inc. Contributions Study design: AH, AID, AR, PH, HS, RPL, JWL, AK, LS, LV, STR. Study investigator: AH, AR, RAB, RPL, AMR, ICW, PH, JPM, JS, JWL, AK, HS. Enrolled patients: AH, AR, RAB, RPL, AMR, ICW, PH, JPM, JS, JWL, AK, HS. Collection and assembly of data: All authors. Data analysis: AR, AID, HS, PL, SO, STR, RAB. Data interpretation: All authors. Manuscript review and revisions: All authors. Final approval of manuscript: All authors. Acknowledgments The sponsor and investigators would like to thank the patients and their families for their participation in and support for these clinical studies. The authors thank Rodrigo Pavani and Masayo Ogawa of Alexion Pharmaceuticals, Inc. for their contribution to the implementation of the 301 and 302 studies. The authors would like to acknowledge Michael D. Morren, RPh, MBA, of Peloton Advantage, LLC, an OPEN Health Company, and Jennifer M. Kulak, PhD, of ApotheCom (Yardley, PA, USA), who provided editorial and medical writing support funded by Alexion Pharmaceuticals, Inc. Editorial review was also provided by Kenneth Pomerantz, PhD, Shweta Rane, PhD, CMPP, and Radha Narayan, PhD, of Alexion Pharmaceuticals, Inc. Funding This study was supported by Alexion Pharmaceuticals, Inc.

bosis in paroxysmal nocturnal hemoglobinuria: a clinician's guide. Ther Adv Hematol. 2017;8(3):119-126. 4. Hill A, DeZern AE, Kinoshita T, Brodsky RA. Paroxysmal nocturnal haemoglobinuria. Nat Rev Dis Primers. 2017;3:17028. 5. Hillmen P, Muus P, Röth A, et al. Long-term safety and efficacy of sustained eculizumab treatment in patients with paroxysmal nocturnal haemoglobinuria. Br J Haematol.

2013;162(1):62-73. 6. Nakayama H, Usuki K, Echizen H, Ogawa R, Orii T. Eculizumab dosing intervals longer than 17 days may be associated with greater risk of breakthrough hemolysis in patients with paroxysmal nocturnal hemoglobinuria. Biol Pharm Bull. 2016;39(2):285-288. 7. Peffault de Latour R, Fremeaux-Bacchi V, Porcher R, et al. Assessing complement blockade in patients with paroxysmal noc-

haematologica | 2021; 106(1)

Breakthrough hemolysis: ravulizumab vs. eculizumab

10.

11.

12. 13.

14.

15.

turnal hemoglobinuria receiving eculizumab. Blood. 2015;125(5):775-783. Almeida AM, Bedrosian C, Cole A, et al. Clinical benefit of eculizumab in patients with no transfusion history in the International Paroxysmal Nocturnal Haemoglobinuria Registry. Intern Med J. 2017;47(9):1026-1034. Hรถchsmann B, Sicre de Fontbrune F, Lee JW, et al. Effect of eculizumab in paroxysmal nocturnal hemoglobinuria (PNH) patients with or without high disease activity: results from the International PNH Registry. Haematologica. 2017;102(Suppl 2):188-189. Lee JW, Peffault de Latour R, Brodsky RA, et al. Efficacy of eculizumab in patients with paroxysmal nocturnal hemoglobinuria (PNH) and high disease activity with or without history of aplastic anemia in the International PNH Registry. Blood. 2017; 130(Suppl 1):3487. Rรถth A, Rottinghaus ST, Hill A, et al. Ravulizumab (ALXN1210) in patients with paroxysmal nocturnal hemoglobinuria: results of 2 phase 1b/2 studies. Blood Adv. 2018;2(17):2176-2185. Brodsky RA. Eculizumab: another breakthrough. Blood. 2017;129(8):922-923. Harder MJ, Kuhn N, Schrezenmeier H, et al. Incomplete inhibition by eculizumab: mechanistic evidence for residual C5 activity during strong complement activation. Blood. 2017;129(8):970-980. Miyasaka N, Miura O, Kawaguchi T, et al. Pregnancy outcomes of patients with paroxysmal nocturnal hemoglobinuria treated with eculizumab: a Japanese experience and updated review. Int J Hematol. 2016; 103(6):703-712. ULTOMIRISโ ข (ravulizumab-cwvz) injec-

haematologica | 2021; 106(1)

16.

17.

18.

19.

20.

21.

22.

23.

tion, for intravenous use. Boston, MA, USA: Alexion Pharmaceuticals Inc., 12/2018. Kulasekararaj AG, Hill A, Rottinghaus ST, et al. Ravulizumab (ALXN1210) vs eculizumab in C5-inhibitor-experienced adult patients with PNH: the 302 study. Blood. 2019;133(6):540-549. Lee JW, Sicre de Fontbrune F, Lee LWL, et al. Ravulizumab (ALXN1210) vs eculizumab in adult patients with PNH naive to complement inhibitors: the 301 study. Blood. 2019; 133(6):530-539. Brodsky RA, De Latour RP, Rottinghaus ST, et al. A prospective analysis of breakthrough hemolysis in 2 phase 3 randomized studies of ravulizumab (ALXN1210) versus eculizumab in adults with paroxysmal nocturnal hemoglobinuria. Blood. 2018; 132(Suppl 1):2330. DeZern AE, Dorr D, Brodsky RA. Predictors of hemoglobin response to eculizumab therapy in paroxysmal nocturnal hemoglobinuria. Eur J Haematol. 2013;90(1):16-24. Hill A, Hillmen P, Richards SJ, et al. Sustained response and long-term safety of eculizumab in paroxysmal nocturnal hemoglobinuria. Blood. 2005;106(7):2559-2565. Kelly R, Arnold L, Richards S, et al. The management of pregnancy in paroxysmal nocturnal haemoglobinuria on long term eculizumab. Br J Haematol. 2010;149(3):446450. Sharma R, Keyzner A, Liu J, Bradley T, Allen SL. Successful pregnancy outcome in paroxysmal nocturnal hemoglobinuria (PNH) following escalated eculizumab dosing to control breakthrough hemolysis. Leuk Res Rep. 2015;4(1):36-38. Kelly R, Richards S, Hillmen P, Hill A. The pathophysiology of paroxysmal nocturnal

24.

25. 26.

27.

28.

29.

30.

31.

hemoglobinuria and treatment with eculizumab. Ther Clin Risk Manag. 2009; 5:911-921. Risitano AM, Notaro R, Marando L, et al. Complement fraction 3 binding on erythrocytes as additional mechanism of disease in paroxysmal nocturnal hemoglobinuria patients treated by eculizumab. Blood. 2009; 113(17):4094-4100. Hill A, Kelly RJ, Hillmen P. Thrombosis in paroxysmal nocturnal hemoglobinuria. Blood. 2013;121(25):4985-4996. Lee JW, Jang JH, Kim JS, et al. Clinical signs and symptoms associated with increased risk for thrombosis in patients with paroxysmal nocturnal hemoglobinuria from a Korean Registry. Int J Hematol. 2013; 97(6):749-757. Yenerel MN, Muus P, Wilson A, Szer J. Clinical course and disease burden in patients with paroxysmal nocturnal hemoglobinuria by hemolytic status. Blood Cells Mol Dis. 2017;65:29-34. Brauckmann S, Effenberger-Neidnicht K, de Groot H, et al. Lipopolysaccharide-induced hemolysis: evidence for direct membrane interactions. Sci Rep. 2016;6:35508. Quist J, Hill AR. Serum lactate dehydrogenase (LDH) in Pneumocystis carinii pneumonia, tuberculosis, and bacterial pneumonia. Chest. 1995;108(2):415-418. Oner AF, Bay A, Arslan S, et al. Avian influenza A (H5N1) infection in eastern Turkey in 2006. N Engl J Med. 2006;355(21):2179-2185. Erez A, Shental O, Tchebiner JZ, et al. Diagnostic and prognostic value of very high serum lactate dehydrogenase in admitted medical patients. Isr Med Assoc J. 2014; 16(7):439-443.

237

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(1):238-249

Platelet Biology & Its Disorder

AG-348 (mitapivat), an allosteric activator of red blood cell pyruvate kinase, increases enzymatic activity, protein stability, and adenosine triphosphate levels over a broad range of PKLR genotypes Minke A.E. Rab,1,2* Brigitte A. van Oirschot,1* Penelope A. Kosinski,3 Jeffrey Hixon,3° Kendall Johnson,3 Victor Chubukov,3 Lenny Dang,3 Gerard Pasterkamp,1 Stephanie van Straaten,1 Wouter W. van Solinge,1 Eduard J. van Beers,2 Charles Kung3# and Richard van Wijk1#

Laboratory of Clinical Chemistry and Haematology, University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands; 2Van Creveldkliniek, University Medical Center Utrecht, Utrecht University, Utrecht, the Netherlands and 3Agios Pharmaceuticals Inc., Cambridge, MA, USA

°Current address: KSQ Therapeutics, Cambridge, MA, USA. *MAER and BAvO contributed equally as co-first author. #CK and RvW contributed equally as co-senior authors

ABSTRACT

Correspondence: RICHARD VAN WIJK r.vanwijk@umcutrecht.nl Received: September 25, 2019. Accepted: January 23, 2020. Pre-published: January 23, 2020. https://doi.org/10.3324/haematol.2019.238865

238

yruvate kinase (PK) deficiency is a rare hereditary disorder affecting red blood cell (RBC) glycolysis, causing changes in metabolism including a deficiency in adenosine triphosphate (ATP). This affects red cell homeostasis, promoting premature removal of RBC from the circulation. In this study, we characterized and evaluated the effect of AG-348, an allosteric activator of PK that is currently in clinical trials for treatment of PK deficiency, on RBC and erythroid precursors from PK-deficient patients. In 15 patients, ex vivo treatment with AG-348 resulted in increased enzymatic activity in all patients' cells after 24 hours (h) (mean increase: 1.8-fold; range: 1.2-3.4). ATP levels increased (mean increase: 1.5-fold; range: 1.0-2.2) similar to control cells (mean increase: 1.6-fold; range: 1.4-1.8). Generally, PK thermostability was strongly reduced in PK-deficient RBC. Ex vivo treatment with AG-348 increased residual activity from 1.4- to >10-fold more than residual activity of vehicle-treated samples. Protein analyses suggest that a sufficient level of PK protein is required for cells to respond to AG348 treatment ex vivo, as treatment effects were minimal in patient cells with very low or undetectable levels of PK-R. In half of the patients, ex vivo treatment with AG-348 was associated with an increase in RBC deformability. These data support the hypothesis that drug intervention with AG348 effectively up-regulates PK enzymatic activity and increases stability in PK-deficient RBC over a broad range of PKLR genotypes. The concomitant increase in ATP levels suggests that glycolytic pathway activity may be restored. AG-348 treatment may represent an attractive way to correct the underlying pathologies of PK deficiency. (AG-348 is currently in clinical trials for the treatment of PK deficiency. Registered at clinicaltrials.gov identifiers: NCT02476916, NCT03853798, NCT03548220, NCT03559699).

Introduction Pyruvate kinase (PK) deficiency is a rare hereditary disorder affecting red blood cell (RBC) glycolysis, resulting in life long chronic hemolytic anemia.1 It is the most common cause of chronic hereditary non-spherocytic hemolytic anemia, although its exact prevalence is unknown.2-4 PK deficiency is an autosomal recessive disease that occurs worldwide. It is caused by mutations in PKLR, the gene encoding the RBC and liver specific isoforms PK-R and PK-L, respectively.5,6 Humans express two additional PK isoforms, PK-M1 and PK-M2, both encoded by the PKLR gene. PKM2 is expressed in early fetal tissues and most adult tissues, including early stage haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

erythroblasts, whereas expression of PK-M1 is confined to muscle. To date, more than 370 mutations have been reported in PKLR associated with PK deficiency (van Wijk, manuscript in preparation). Most of these mutations are missense mutations encoding single amino acid changes that affect the enzyme’s structure, stability or catalytic function.7,8 Only a few mutations occur relatively frequently, implying that the majority of patients harbor a unique combination of mutations.9 Pyruvate kinase is an allosterically regulated homotetrameric enzyme that catalyzes the conversion of phosphoenol pyruvate (PEP) to pyruvate in the final step of glycolysis, thereby producing energy in the form of adenosine triphosphate (ATP). PK-deficient RBC are characterized by changes in metabolism associated with defective glycolysis, including a build-up of the upstream metabolite 2,3-disphosphoglycerate (2,3-DPG) and deficiency in the PK product ATP. Since RBC rely totally on glycolysis for ATP production it is hypothesized that insufficient energy production affects red cell homeostasis, promoting premature removal of PK-deficient RBC from the circulation by the spleen. Increased RBC dehydration resulting from increased activity of the Gardos channel, and consequent potassium efflux;10-12 as well as ATP, depletion-mediated changes in RBC membrane integrity have been proposed as mechanisms triggering this premature clearance.13,14 In addition, ineffective erythropoiesis may contribute to the pathophysiological consequences of PK deficiency.15-17 Clinically, PK deficiency manifests as chronic hemolytic anemia of highly variable clinical severity. Patients may present with severe neonatal anemia requiring exchange transfusion, whereas at the other end of the spectrum patients are diagnosed later in life with fully compensated hemolysis. The ongoing Pyruvate Kinase Deficiency Natural History Study showed that, apart from anemia, the most frequently reported complications were iron overload and gallstones. However, other complications such as aplastic crises, osteoporosis, extramedullary hematopoiesis, post-splenectomy sepsis, pulmonary hypertension, and leg ulcers are not uncommon. Typically, worsening of hemolysis occurs during infections, stress, and pregnancy.18 Current treatment for PK deficiency is generally supportive, with splenectomy, blood transfusion and iron chelation as the main therapies, focusing on the anemia and iron overload state.19 Splenectomy is effective for the majority of PK-deficient patients, in that it reduces transfusion burden and improves baseline hemoglobin levels post splenectomy by an average of 1.6 g/dL. Fourteen percent of patients who underwent splenectomy, however, continued to require regular transfusions.18 Curative treatment possibilities include stem cell transplantation (SCT) or experimental gene therapy.20,21 Recently, it was shown that AG-348 is an allosteric activator of PK-R. This small molecule directly targets PK-R by binding in a pocket at the dimer-dimer interface, distinct from the allosteric activator fructose-1,6-bisphosphate binding domain (FBP). This induces the active R-state conformation of the PK-R tetramer, resulting in enhanced activity of both wild type and mutant PK-R.22 Phase I studies in healthy volunteers demonstrated glycolytic pathway activation upon treatment with AG-348 demonstrating its potential for treating PK deficiency.23 haematologica | 2021; 106(1)

Data from a phase II study in PK deficiency patients treated with AG-348 showed that approximately half of treated subjects experienced a rise in hemoglobin (Hb).24 In this study, we further evaluated the effect of ex vivo treatment with AG-348 on PK deficient red cells with a broad range of patient genotypes, including effects on enzyme activity and ATP levels. In particular, for the first time we report the effects of AG-348 on PK thermostability in red cell lysates, PK-R protein level, RBC deformability and the effect of AG-348 during in vitro PK-deficient erythropoiesis.

Methods This study was approved by the ethical committee of the University Medical Center Utrecht (Protocol 14-571/M).

Metabolic profiling Frozen whole blood samples were extracted with hot 70% ethanol, dried and resuspended in water. Relative abundance of central carbon metabolites was performed by ultra performance liquid chromatography-mass spectrometry (UPLC-MS) with high resolution accurate mass detection on a QExactive™ Orbitrap mass spectrometer (Thermo Fisher Scientific).25 Peak identification and integration was carried out with Maven software.26 Quantitative analysis of ATP and 2,3-DPG was performed as described previously.27

Pyruvate kinase activity, protein levels, and thermostability

Red blood cells were purified using microcrystalline α-cellulose columns.28,29 PK and hexokinase activity was measured as described.28,29 PK thermostability was measured as described30 after 0, 5, 10, 20, 40 and 60 minutes (min) of incubation at 53°C. Pyruvate kinase activity after ex vivo treatment with AG-348 was measured using low substrate (0.5 mM). PK-R protein levels were determined by Mesoscale Assay (MesoScale Discovery) as described22 using goat anti-PKLR antibody (Aviva) and mouse antiPKLR antibody (Abcam). SULFO-TAG goat anti-mouse (Mesoscale Discovery) was used as detection antibody. Protein level was determined by normalizing light intensity of the SULFOTAG electrochemiluminescence to lysate protein concentration. Western blot of PK-R and PK-M2 was performed as described31 using polyclonal antibodies against PK-L32 and monoclonal antibodies against PK-M2 (PodiCeps). Alexa Fluor® 680-conjugated goat anti-mouse and goat anti-rabbit antibodies (Li-Cor, Invitrogen) were used as detection antibodies.

Ex vivo treatment with AG-348 Purified RBC or whole blood of patients and controls were incubated for 3, 6 and 24 h at 37°C in presence or absence of AG-348 in phosphate-buffered saline containing 1% glucose, 170 mg/L adenine, and 5.25 g/L mannitol (AGAM, pH 7.40). RBC were incubated with different dosages of AG-348 (up to 10 mM) for up to 24 h, at 37°C. After 6 and 24 h, PK-R activity and ATP levels (CellTiterGlo; Promega, Madison, WI, USA) were measured. For thermostability, RBC lysates were incubated for 2 h with AG-348.

Osmoscan and deformability Osmoscan and deformability measurements were performed using the Lorrca (RRMechatronics).33,34 AGAM buffer was added to whole blood (1:10), and supplemented with either AG-348 (20 M) or DMSO (untreated). Samples were incubated for 24 h at 37°C before measurements. 239

M.A.E. Rab et al.

AG-348 treatment during in vitro erythropoiesis Erythroid cells were produced from peripheral blood mononuclear cells (PBMC).35 For proliferation, cells were cultured 10 days in CellQuin medium (Sanquin), supplemented with Stem Cell Factor (100 ng/mL, Amgen), erythropoietin (EPO) (2 U/mL, EPREX), dexamethasone (1 mM, Sigma), and 1 ng/mL IL-3 (R&D Systems). After 10 days, cells were reseeded (2E6 cells/mL) in CellQuin medium supplemented with EPO (10 U/mL, Eprex), heparine (5 U/mL, Leo Pharmacy), human AB plasma (3%, Sigma). At day 5, CellQuin was replaced by RetQuin medium (Sanquin). AG-348 (2 mmol/L) was added at days 0, 1, 3-10 during proliferation and days 0, 2, and 4 during differentiation.

Statistical analysis Statistical analysis was performed using Graphpad Prism (v7.04). t-test, Mann-Whitney test or Wilcoxon test was used when appropriate. One-way ANOVA was carried out or a Kruskal-Wallis test, followed by Dunn’s tests for post-hoc analysis. Pearson’s correlation was used to determine correlations of laboratory parameters with enzyme activity measurements.

Results Fifteen adult, transfusion independent, PK-deficient patients were enrolled for this study (median age: 44.0 years; range: 20.0-51.5). PK deficiency was confirmed by demonstrating compound heterozygosity or homozygosity for mutations in PKLR (Sanger sequencing). Whole blood from 15 healthy volunteers was used as control samples. Baseline patients' characteristics displayed varying degrees of anemia (Table 1). In addition, most patients had strongly elevated reticulocyte counts, a prominent feature of PKD patients, in particular after splenectomy.11 Most patients showed reduced red cell PK activity (Figure 1A) and reduced PK thermal stability (Figure 1C). Since the activity of many red cell enzymes is red cell age-dependent,36 we also compared PK activity to hexokinase (HK) activity (Table 1 and Figure 1B). This PK/HK ratio was clearly decreased in all patients, indicating a relatively strong decrease in activity of PK (r=-0.677, P<0.01) (Figure 1B). When causative mutations where classified as missense (M) or non-missense (NM) a genotype to phenotype correlation was identified.18 All four patients without splenectomy were of the M/M genotype, which is in line with the lower likelihood of splenectomy in this group.18 Similarly, these patients had generally higher Hb levels and lower reticulocyte counts (Table 1).

PK-R protein levels correlate with pyruvate kinase activity Protein levels of PK-R were measured by both western blot and Meso Scale Discovery (MSD). PK-R protein levels were variable between patients (Figure 1C and D). In general, PK-activity correlated well positively with PK-R protein levels. The most notable exception was patient 4. RBC from this patient contained very low levels of protein, as measured by both western blot and ELISA, but repeated measurements showed high levels of PK activity (Table 1 and Figure 1A, D and E). Low levels of PK-R protein in this patient could result from combined instability of the p.Arg510Gln37 mutant as well as the truncated p.(Arg488*) PK-R nonsense mutant. In line with this, PK thermostability was also strongly reduced in this patient 240

(Figure 1C). However, the reason for the high PK activity in this patient remains unknown. Increased PK-M2 expression, that could perhaps compensate for the loss in PK-R38 activity, was not detected except for patient 3 and, to a lesser degree, in RBC from patient 1 (Figure 1D).

Pyruvate kinase-deficient red blood cells show glycolytic intermediates levels that are consistent with decreased pyruvate kinase activity Metabolic profiling of PKD patients was performed by liquid chromatography-mass spectrometry (LC-MS/MS) on whole blood. Consistent with a decrease in PK activity at the final step of glycolysis, levels of PEP were significantly increased (P<0.0001) whereas levels of pyruvate and ATP were significantly decreased (both P<0.0001) (Figure 2). In addition, dihydroxyacetone phosphate (DHAP) and 2,3-DPG levels were increased (P<0.001 and P<0.0001), indicative of retrograde accumulation of upstream glycolytic intermediates.

Ex vivo incubation with AG-348 increases pyruvate kinase activity and adenosine triphosphate in a dose-dependent manner Ex vivo treatment of PK-deficient and healthy control RBC with increasing dose of AG-348 showed an increase in PK-activity and ATP levels. Illustrative PK activity dose response curves of two patients are shown in Figure 3A and B. Incubations with 10 mM AG-348 showed a mean 1.3-fold increase (range: 1.0-2.0) in PK activity in PK-deficient RBC after 6 h (P<0.0001) and of 1.8 (range: 1.2-3.4) after 24 h (P<0.0001) (Figure 3E). Similar values were obtained from control cells: 1.4 after 6 h (range: 0.9-1.9; P<0.01) and 2.3 (range: 1.2-7.1) after 24 h (P<0.0001). Adenosine triphosphate levels increased accordingly in a dose-dependent manner (Figure 3C and D). AG-348 treatment significantly increased mean ATP levels in PKdeficient RBC to 1.4-fold (range: 1.0-2.3) after 6 h (P<0.01) and 1.5-fold (range: 1.0-2.2) after 24 h (P<0.0001). The mean increase in ATP levels in control cells was 1.5-fold (range: 1.3-1.7) after 6 h (P<0.001) and 1.6-fold (range: 1.41.8) after 24 h (P<0.0001). We observed no differences in PK activity and ATP response between PK-deficient patients with M/M and M/NM genotypes (Table 1 and Figure 3E and F). Since glycolysis in RBC is known to be regulated by band 3,39 we verified RBC band 3 expression levels by western blot analysis. We observed no differences in band 3 expression levels of PK-deficient patients compared to healthy controls (Online Supplementary Figure S1). Collectively, these results are consistent with those of previous reports,22 and illustrate the ability of AG-348 to stimulate PK activity in patient samples with a range of genotypes.

Ex vivo treatment with AG-348 restores pyruvate kinase thermostability At baseline, PK-R thermostability was strongly reduced in all PK-deficient patients (Figure 4C). Whereas RBC from healthy controls showed a mean residual activity of 81% (range: 64-108%) after incubation at 53°C for 60 min, the mean residual PK-R activity in PK-deficient patients was significantly decreased under these conditions to only 30% (range: 6-54%) (P<0.0001). This confirms earlier observations30 and shows that PK thermostability is a haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

Table 1. Genotypes, amino acid change, splenectomy status and laboratory parameters are shown for all 15 pyruvate kinase (PK)-deficient patients included in this study.

Sex

PKLR mutation

P1 P2 P3 P4 P5 P6 P7

F M M F F F M

c.1121T>C/c.1456C>T c.331G>A/c.1456C>T c.1178A>G/c.1456C>T c.1462C>T/c.1529G>A c.1121T>C/c.1706G>A c.507+1G>A/c.1436G>A c.376-2A>C/c.1529G>A

P8 P9 P10 P11 P12 P13

F M M M M M

c.331G>A/c.1456C>T c.331G>A/c.1492C>T c.1154G>A/c.1529G>A c.721G>T/c.1529G>A c.142_159del/c.494G>T c.1269G>A/c.1654G>A

P14 F P15 M Contr.

c.401T>A/c.1529G>A c.1529G>A/c.1529G>A

Amino acid change

Genotype Splenectomy

p.(Leu374Pro)/p.(Arg486Trp) M/M p.(Gly111Arg)/p.(Arg486Trp) M/M p.(Asn393Ser)/p.(Arg486Trp) M/M p.(Arg488*)/p.(Arg510Gln) M/NM p.Leu374Pro)/p.(Arg569Gln) M/M p.[=; 0]/p.(Arg479His) M/NM p.(Glu125_Tyr126ins32; M/NM Tyr126Alafs*83)/p.(Arg510Gln) p.(Gly111Arg/p.(Arg486Trp) M/M p.(Gly111Arg)/p.(Arg498Cys) M/M p.(Arg385Lys)/p.(Arg510Gln) M/M p.(Glu241*)/p.(Arg510Gln) M/NM p.(Thr48_Thr53del)/p.(Gly165Val) M/NM p.([Met373_Ala423del; 0]) M/NM /p.(Val552Met) p.(Val134Asp)/p.(Arg510Gln) M/M p.(Arg510Gln)/p.(Arg510Gln) M/M

Reference range values:

Hb (g/dL)

Ret. PK HK PK/HK (%) (U/gHb)(U/gHb) ratio

Yes No Yes Yes No Yes Yes

9.7 11.2 9.4 6.6 10.9 7.7 8.0

10.0 7.0 >35.0 >34.0 7.0 75.0 55.0

9.0 3.1 16.9 15.9 2.1 1.9 1.7

5.0 1.7 6.2 7.4 1.5 6.0 5.6

1.8 1.8 2.7 2.1 1.4 0.3 0.3

No Yes No Yes Yes Yes

11.9 10.3 13.4 8.2 12.9 9.7

5.0 40.0 15.0 85.0 20.0 55.0

2.2 2.5 2.7 2.2 10.0 1.8

1.4 4.1 2.0 6.1 3.5 5.8

1.6 0.6 1.3 0.4 2.8 0.3

Yes Yes No

9.6 60.0 1.2 5.2 0.2 2.8 40.0 1.6 2.8 0.6 14.1 1.18 12.7 1.1 11.9 (1.3) (0.3) (2.3) (0.2) (2.1) M: 13.8-17.2 1.0-2.0 9.2-16.9 0.9-1.4 9.0-16.6 F: 11.9-15.5

Enzyme activity of PK was decreased or normal. Hexokinase (HK) activity was almost always increased due to reticulocytosis. PK/HK ratio was strongly decreased in all cases, indicative of PK deficiency. Mean values (standard deviation) of 15 healthy controls are shown. Reference range values of the UMC Utrecht population are also included. M: male; F: female; Hb: hemoglobin; Ret.: reticulocytes; PK: pyruvate kinase; U: units; H:, hexokinase; M/M: missense/missense; M/NM: missense/nonmissense.

common feature of many different mutant forms of PK-R, and not just restricted to known unstable variants like p.Arg510Gln. Treating PK-deficient RBC ex vivo with AG-348 (2 mM) prior to this assay showed a clear effect on stability of the enzyme (Figure 4C; illustrative examples of RBC from patients 12 and 9 are shown in Figure 4A and B). Ex vivo treatment with AG-348 lead to a mean residual PK activity of 58% (range: 0-115%; P<0.001) in RBC from PK-deficient patients and 106% in controls (range: 99-111%; P=0.13). We observed no differences in PK thermostability response between PK-deficient patients with M/M and M/NM genotypes (Table 1 and Figure 4C).

Effect of ex vivo AG-348 treatment on deformability of pyruvate kinase-deficient red blood cells Because RBC deformability is partly ATP dependent,40,41 we investigated the effect of ex vivo AG-348 treatment on this important RBC functional property by osmotic gradient ektacytometry (osmoscan) and deformability measurements. Figure 5A shows osmoscan curves of the 15 healthy control samples (gray) and that of patient 7 (red) as an example. At baseline, maximum deformability (EI ) of PK-deficient RBC was significantly decreased compared to controls, although the difference was small (P<0.01) (Figure 5D). Similarly, Omin and Ohyper, reflecting RBC surface-to-volume ratio and hydration status,42 respectively, were significantly different in PK-deficient patients (P=0.039 and 0.047) (Figure 5C and E). This implies that PK-deficient RBC are somewhat less deformable than normax

haematologica | 2021; 106(1)

mal RBC, possibly due to a slightly lower surface-to-volume ratio and a slight increase in RBC hydration. When PK-deficient patients were divided according to genotype (Table 1), there was a significant difference between M/M and M/NM genotypes, indicating that a more severe genotype results in lower maximal RBC deformability (EImax, P=0.039) (Figure 5B). After 24 h of ex vivo treatment with AG-348, an improvement in RBC deformability was observed in about half of the PK-deficient patients (Figure 5F, patients 1, 3, 4, 6, 7, 8, and 9) compared to their untreated RBC. The effect of these changes were similar in magnitude to the baseline difference in deformability observed between PK-deficient RBC and control RBC (Figure 5C). Taken in aggregate, however, there was no overall significant improvement of deformability.

Effect of ex vivo AG-348 treatment on PK-R-deficient erythroid development During erythropoiesis, PK-M2 levels are gradually replaced by PK-R43,44 and AG-348 has been described to also activate PK-M2.22 Because PK deficiency has been associated with ineffective erythropoiesis,15-17 we next investigated the effect of ex vivo AG-348 treatment on erythroid development from four PK-deficient patients and controls. Morphologically there were no major abnormalities in erythroid development between healthy controls or PK-deficient patients (Figure 6A). This was supported by there being no difference in erythroid surface marker expression35 (Online Supplementary Figure S2A). There was 241

M.A.E. Rab et al.

no difference in erythroid differentiation with or without AG-348 treatment. This was observed in both healthy control cells and in PK-deficient cells (Online Supplementary Figure S2B and C). In presence of AG-348 (2 mM), we did note, however, a small increase in cell number in both normal controls and patients after proliferation for 10 days, although the difference was not significant (P=0.059) (Figure 6B). Western blot analysis of PK-R and PK-M2 protein levels of healthy control samples clearly showed an increase in expression of PK-R and a decrease in PK-M2 during differ-

entiation (Online Supplementary Figure S3). The decrease in PK-M2 was also observed in all patient samples, whereas a clear increase in PK-R expression was seen only in samples from patients 7 and 15 (Online Supplementary Figure S3). PK-R protein levels were low or inconclusive in patients 3 and 10. Presence of AG-348 during ex vivo erythropoiesis was associated with higher levels of PK-R in patients 7 and 15. AG-348 was not seen to have any effect on PK-M2 expression. On the level of enzymatic activity, the PK/HK ratio from cultured red cells was higher in presence of AG-348 in

242

Figure 1. Baseline levels of pyruvate kinase (PK) activity, PK/Hexokinase (HK) ratio, PK thermostability, and PK-R protein levels in PK-deficient patients and controls. (A) PK activity of PK deficient patients (red) and healthy controls (gray). (B) PK/HK ratio in patients (red) and healthy controls (gray), as a means to evaluate PK activity in the presence of high number of reticulocytes, reflected by increased HK activity. (C) PK thermostability of patients (red) and controls (gray). (D) PK-R protein levels of patients (P#) and controls (HC) measured by western blot. (E) PK-R protein levels of patients (red) and controls (gray) measured by Meso Scale Discovery. ECL: electrochemiluminescence.

haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

both patients and healthy controls, although this effect did not reach statistical significance (P=0.064) (Figure 6C).

Discussion In this study, we characterized and evaluated the effect of AG-348, a recently reported first-in-class allosteric activator of PK-R22,23 that is currently in clinical trials for the treatment of PK deficiency (clinicaltrials.gov identifiers: NCT02476916, NCT03853798, NCT03548220, NCT03559699). Our find-

ings indicate that ex vivo treatment with AG-348 effectively increases PK-R activity over a broad range of PKLR genotypes as assessed by enzymatic activity as well as cellular ATP levels. Our findings are consistent with those previously reported by Kung et al.,22 and further extend those findings through the evaluation of a substantially larger cohort of PK-deficient blood samples, including a comprehensive baseline metabolomics analysis, and assessment for the first time of the effects of PK deficiency on parameters such as red cell deformability and ex vivo expansion of erythroid progenitors.

Figure 2. Pyruvate kinase (PK)deficient patients show glycolytic intermediates levels that are consistent with decreased PK activity. Upstream intermediates of PK of PK-deficient patients (red) are significantly increased compared to values of health controls (gray), with, for example, an almost 2-fold increase in 2,3-DPG. Downstream targets, pyruvate and adenosine triphosphate (ATP), are significantly decreased compared to healthy controls. PKD: pyruvate kinase deficiency; G6P: glucose-6-phosphate; DHAP: dihydroxyacetone phosphate; 2,3DPG: 2,3-diphosphoglycerate; PEP: phosphoenol pyruvate; FBP: fructose biphosphate; 2PG: 2-phosphoglycerate; 3PG: 3-phosphoglycerate; GAP: glyceraldehyde phosphate; 1,3DPG: 1,3-diphosphoglycerate; HK: hexokinase; PGI: phosphogluco isomerase; PFK: phosphofructokinase; FBA: fructose biphosphate aldolase; TPI: triosephosphate isomerase; GAPDH: glyceraldehyde phosphate dehydrogenase; BPGM: bisphosphoglycerate mutase; PGK: phosphoglycerate kinase; BPKG: bisphosphoglycerate kinase; PGM: phosphoglycerate mutase; ENO: enolase; LDH: lactate dehydrogenase.

haematologica | 2021; 106(1)

243

M.A.E. Rab et al.

Our cohort shows an overall mean 1.8-fold increase in PK activity in response to ex vivo treatment with AG-348 (range: 1.2-3.4) (Figure 3E). This is very similar to the previously reported mean fold increase of 2.1 (range: 1.3-3.4). Similarly, the 1.5-fold increase in ATP levels we observed in PK-deficient RBC (range: 1.0-2.2) was comparable to previously reported results (mean fold increase: 1.7; range: 1.3-2.4).

PK thermostability was found to be significantly improved upon ex vivo treatment with AG-348 even though there was a high variability in response among the different genotypes (Figure 4). An interesting subset was composed of six PK-deficient patients (patients 4, 7, 10, 11, 14, and 15) who had the common PK-R mutation c.1529G>A p.(Arg510Gln), previously shown to have little defect in catalytic activity but to be highly unstable. Five of these

244

Figure 3. Ex vivo AG-348 treatment of red blood cells (RBC) increase pyruvate kinase (PK) activity and adenosine triphosphate (ATP) levels in a dose dependent manner. (A and B) Representative patient dose response curves of PK activity, expressed as fold activation compared to incubation without AG-348, as measured in RBC of patients 12 (A) and 9 (B) after 6(turquoise) and 24-hour (h) (dark green) incubation with AG-348. (C and D) representative patient ATP dose response curves, expressed as fold activation compared to incubation without AG348, as measured in RBC of patients 12 (C) and 9 (D) after 6- (turquoise) and 24h (dark green) incubation with AG-348. (E) PK activation in RBC of 15 PK deficient patients after 6- (turquoise) and 24-h (dark green) incubation with 10 mM AG-348. Mean PK fold activation of healthy controls after 6- (light gray) and 24-h (dark gray) incubation with AG-348 10ÎźM is shown on the left of the graph. (F) ATP response (fold activation) of 15 PK deficient patients after 6(turquoise) and 24-h (dark green) incubation with 10 mM AG-348. Mean ATP response of healthy controls is shown on the left of the graph (gray bars). Error bars represent standard deviation. Dashed turquoise line represents 6-h incubation with AG-348, solid dark green line represents 24-h incubation.

haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

patients had unmeasurably low PK activity in the thermostability test, and the activity of four could be restored by pre-incubating RBC with AG-348 (Figure 4, patients 7, 10, 11 and 14). These four patients carried the p.(Arg510Gln) mutation in trans to a splice site mutation (patient 7), a missense mutation (patients 10 and 14), and a nonsense mutation (patient 11). A significant outlier was patient 15, who was homozygous for p.(Arg510Gln), suggesting the possibility that it may be more difficult to stabilize this mutation in the homozygous condition. Another outlier was patient 4, who had p.(Arg510Gln) paired with the nonsense mutation p.(Arg488*) yet had a relatively modest loss of activity in the thermostability test. Collectively these data illustrate that AG-348 has the potential to stabilize a diverse range of different mutant PK-R molecules in vitro, but also demonstrate the complex variability that is introduced by the compound heterozygous state of most patients. One surprising result was that the effects of AG-348 were variable across the different patient samples, and that overall increases in PK activity, ATP and thermostability were not observed together in some patients. For example, patient 13 shows a clear response in terms of PK activity after AG-348 treatment for 24 h whereas the ATP response was present at 6 h and back to baseline by 24 h. However, we did observe that residual PK-R protein levels

at baseline are important for the response to AG-348. When grouping patients according to their response in PK activity or ATP levels (i.e., >1.5-fold vs. <1.5-fold response), PK-R protein levels at baseline (Figure 1E) were found to be higher in the patients that showed a >1.5-fold increase upon ex vivo treatment with AG-348 (Figure 7). This suggests that the amount of PK-R protein is an important determinant for the in vitro activity of AG-348, and is consistent with its mechanism of action via direct binding and stimulation of mutant enzyme. While the in vitro assessments as presented in this study provide clear evidence that AG-348 can stimulate the activity of mutant PKR enzymes in a variety of contexts, the central question is whether changes in these parameters would be associated with increased likelihood of clinical response to AG-348. Recently, data from the phase II study of AG-348 in PKD patients have been published.24 Analysis of those results suggests some parallels between our data and the response observed in treated patients. For example, the clinical manuscript demonstrates that higher baseline PK-R protein levels in patients is associated with Hb increases, which parallels the findings described here. Similarly to the effects on thermostability noted above in patients with the p.(Arg510Gln) mutation, such patients in phase II had variable Hb responses.24

Figure 4. Pyruvate kinase (PK) thermostability is restored upon ex vivo treatment with AG-348. (A and B) Representative PK-deficient (PKD) patients' PK thermostability of red blood cell (RBC) lysates of patients 12 (A) and 9 (B) pre-incubated in absence (dashed line) or presence (solid line) of 2 mM AG-348 for 2 hours (h) at 37Â°C. For PK thermostability, residual PK activity (%) is measured after 5, 10, 20, 40 and 60 minutes (min) of heat treatment at 53Â°C. (C) Residual PK activity (%) after 60 min of heat treatment of 15 PKD pre-incubated in absence (light turquoise) and presence (dark turquoise bars) of 2 mM AG-348. Levels are normalized to baseline levels of PK-activity before heat treatment. Absence of bars indicate that PK residual activity was below the detection limit. Error bars represent standard deviation.

haematologica | 2021; 106(1)

245

M.A.E. Rab et al.

Here, for the first time, we report on the effects of AG348 on ex vivo produced PK-R deficient RBC. In presence of AG-348 we observed a trend towards a slightly improved proliferation of erythroid progenitor cells in both healthy controls and PK-deficient patients of various PKLR genotypes (Figure 5B). Further studies are warranted to establish this more firmly, but it could indicate that, apart from acting on mature RBC in the circulation, AG-348 could have a potential beneficial effect on both normal and PK-deficient early stage erythroid development. In addition, we show that ex vivo produced (mutant) PK-R can be activated to a similar degree as (mutant) PK-R from RBC from patients and healthy controls (Figure 6C). This was particularly evident for

246

patients 7 and 15, who also showed the most pronounced effect of AG-348 on PK-R protein levels (Online Supplementary Figure S3). At the same time, based on cell count, morphology, and erythroid differentiation markers, we did not observe major differences between normal and PK-deficient ex vivo erythroid proliferation. This is in contrast with previous reports that suggest ineffective erythropoiesis is a pathophysiological feature of PK deficiency, based on increased numbers of apoptotic early erythroid progenitor cells in the spleen of one PK-deficient patient,16 and similar findings in the spleen from a mouse model of PK deficiency.15 Regarding this observed discrepancy between our findings and previous studies, a possible confounder is that we

Figure 5. Red blood cell (RBC) deformability in pyruvate kinase (PK)-deficient patients is slightly decreased. (A) Osmoscan curve of patient 7 (red) and 15 healthy controls (HC) (gray). (B) Maximum deformability (EImax) is slightly decreased in PK deficiency compared to HC. When grouped according to genotype, i.e., missense/missense M/M versus missense/non-missense (M/NM), there is a significant difference, indicating that a more severe genotype results in lower maximal RBC deformability. (C) Omin is slightly increased in PKD patients compared to HC. (D) EImax is slightly decreased in PKD patients compared to HC. (E) Ohyper is slightly increased in PKD patients compared to HC. (F) After 24 hours of ex vivo treatment with 20 mM AG348 in about half of the PKD patients an improvement in RBC deformability was observed (patients 1, 3, 4, 6, 7, 8, and 9) compared to their untreated RBC. Error bars represent standard deviation. **P<0.01, *P<0.05.

haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

grew erythroid cells from a fixed number of PBMC. PBMC composition will likely vary between patients and controls, in particular the number of CD34+ cells. This may influence the eventual outcome of cellular proliferation. More importantly, genotype-phenotype correlations could explain these conflicting results as different genotypes result in different complications,18 indicating that different

mutations could result in a large variability in the occurrence of ineffective hematopoiesis. A limitation of this study is the absence of a functional assay that could directly measure the effect of ATP increase. We did observe an improvement in RBC deformability as measured by osmotic gradient ektacytometry after ex vivo treatment with AG-348 in half of the

Figure 6. Ex vivo treatment with AG-348 does not evidently affect pyruvate kinase (PKD) ex vivo erythroid proliferation and differentiation. (A) Morphology of ex vivo cultured erythroid cells of a PKD patient (P3) and a healthy control (HC) at the final stage of proliferation (day 10) and on various days during differentiation (days 0, 4 and 10). (B) Ex vivo erythroid proliferation (cell numbers in %) was slightly higher when cells were cultured in presence of 2 mM AG-348 (dark purple) compared to 0 mM AG-348 (light purple). This was also observed for HC. (C) PK/HK ratio of erythroid cells of PKD patients (purple) and HC (gray) cultured in presence (dark) or absence (light) of 2 mM AG-348 respectively was increased compared to cells cultured without AG-348. W/o: with/without. Error bars represent standard deviation.

haematologica | 2021; 106(1)

247

M.A.E. Rab et al.

Figure 7. Pyruvate kinase (PK)-deficient red cells with higher PK-R protein levels are more likely to show an increase in adenosine triphosphate (ATP) levels and/or PK-R activity after ex vivo treatment with AG-348. Base line PK-R protein levels in patients showing a response in PK activity or ATP levels (i.e., >1.5fold response vs. <1.5-fold response, red bars) upon ex vivo treatment with AG348 PK-R levels of healthy controls are shown in gray. ECL: electrochemiluminescense; gHb: gram hemoglobin. Error bars represent standard deviation.

patients (Figure 5F). However, this technique likely underestimates the effect of ATP increase by AG-348. Although ektacytometry is an important and established method in the diagnosis of RBC membrane and hydration disorders,33,34,45,46 it does have some limitations. Ektacytometry reflects the passive deformation of RBC in contrast to filter-based assays or microfluidics which forces RBC to actively deform; the latter is thought to be more dependent on ATP.14,41 This could explain the finding of only a small correlation with ATP levels and deformability measured in RBC concentrates intended for transfusion.40 Our results show only a slight difference in RBC deformability between PK-deficient patients and healthy controls, with the lowest deformability observed in M/NM patients (Figure 5B). This is in line with recent findings45 using the same technique, but contrasts with earlier findings based on a filtration-based assay.47 We also

References 1. Koralkova P, van Solinge WW, van Wijk R. Rare hereditary red blood cell enzymopathies associated with hemolytic anemia - pathophysiology, clinical aspects, and laboratory diagnosis. Int J Lab Hematol. 2014;36(3):388-397. 2. Beutler E, Gelbart T. Estimating the prevalence of pyruvate kinase deficiency from the gene frequency in the general white population. Blood. 2000;95(11):3585-3588. 3. Carey PJ, Chandler J, Hendrick A, et al. Prevalence of pyruvate kinase deficiency in a northern European population in the north of England. Blood. 2000;96(12):4005-4007. 4. de Medicis E, Ross P, Friedman R, et al. Hereditary nonspherocytic hemolytic anemia due to pyruvate kinase deficiency: a prevalence study in Quebec (Canada). Hum Hered. 1992;42(3):179-183. 5. Kanno H, Fujii H, Miwa S. Structural analysis of human pyruvate kinase L-gene and

248

observed normal deformability in the splenectomized PKdeficient patients compared to a small decrease in EImax in non-splenectomized patients, which correlated inversely to reticulocyte count (r=-0.740, P=0.0016). These results are in line with previous studies that show that reticulocytes of PK-deficient patients are preferentially removed by the spleen,11 probably due to their compromised deformability as a result of insufficient ATP levels.47 It has been suggested that such a decreased deformability could be due to the decreased hydration as a result of K+ efflux through PIEZO1-mediated Ca2+ influx and activation of Gardos channels.12 However, the osmoscan data show no signs of dehydration of RBC (i.e., decreased Ohyper), which is clearly measurable on RBC of patients with hereditary xerocytosis, where dehydration is thought to occur through the same mechanism.48 Collectively, the data presented in this study support the hypothesis that drug intervention with AG-348 effectively up-regulates PK-R enzymatic activity and increases stability in PK-deficient RBC over a broad range of PKLR genotypes. The concomitant increase in ATP levels suggests that the glycolytic pathway activity may be restored. AG-348 treatment may, therefore, represent an attractive way to correct the underlying pathologies of PK deficiency. Disclosures RW and EJB are consultant for Agios Pharmaceuticals. Contributions CK, BAO, PAK and RW designed the experiments; BAO, SS, KJ, VC and PAK recruited, and collected samples and/or performed experiments; PAK, JH, LD and CK provided AG-348; MAER and RW analyzed the data and wrote the manuscript with input and revisions of BAO, PAK, GP, VC, WWS, EJB and CK. Acknowledgments The authors would like to thank all the PK-deficient patients who participated in this study. Funding This research has been supported by AGIOS Pharmaceuticals, and in part by an unrestricted grant provided by RR Mechatronics.

identification of the promoter activity in erythroid cells. Biochem Biophys Res Commun. 1992;188(2):516-523. 6. Kanno H, Fujii H, Hirono A, Miwa S. cDNA cloning of human R-type pyruvate kinase and identification of a single amino acid substitution (Thr384 --> Met) affecting enzymatic stability in a pyruvate kinase variant (PK Tokyo) associated with hereditary hemolytic anemia. Proc Natl Acad Sci U S A. 1991;88(18):8218-8221. 7. Zanella A, Fermo E, Bianchi P, Chiarelli LR, Valentini G. Pyruvate kinase deficiency: the genotype-phenotype association. Blood Rev. 2007;21(4):217-231. 8. van Wijk R, Huizinga EG, van Wesel ACW, van Oirschot BA, Hadders MA, van Solinge WW. Fifteen novel mutations in PKLR associated with pyruvate kinase (PK) deficiency: structural implications of amino acid substitutions in PK. Hum Mutat. 2009;30(3):446453. 9. Canu G, De Bonis M, Minucci A, Capoluongo E. Red blood cell PK deficiency:

an update of PK-LR gene mutation database. Blood Cells Mol Dis. 2016;57:100-109. 10. Glader B. Salicylate-induced injury of pyruvate-kinase-deficient erythrocytes. N Engl J Med. 1976;294(17):916-918. 11. Mentzer WC, Baehner RL, SchmidtSchĂśnbein H, Robinson SH, Nathan DG. Selective reticulocyte destruction in erythrocyte pyruvate kinase deficiency. J Clin Invest. 1971;50(3):688-699. 12. Koller CA, Orringer EP, Parker JC. Quinine protects pyruvateâ&#x20AC;?kinase deficient red cells from dehydration. Am J Hematol. 1979; 7(3):193-199. 13. Park Y, Best CA, Badizadegan K, et al. Measurement of red blood cell mechanics during morphological changes. Proc Natl Acad Sci U S A. 2010;107(15):6731-6736. 14. Weed RI, LaCelle PL, Merrill EW. Metabolic dependence of red cell deformability. J Clin Invest. 1969;48(5):795-809. 15. Aizawa S, Harada T, Kanbe E, et al. Ineffective erythropoiesis in mutant mice with deficient pyruvate kinase activity. Exp

haematologica | 2021; 106(1)

Ex vivo treatment of PK deficiency with AG-348

Hematol. 2005;33(11):1292-1298. 16. Aizawa S, Kohdera U, Hiramoto M, et al. Ineffective erythropoiesis in the spleen of a patient with pyruvate kinase deficiency. Am J Hematol. 2003;74(1):68-72. 17. Aisaki K, Aizawa S, Fujii H, Kanno J, Kanno H. Glycolytic inhibition by mutation of pyruvate kinase gene increases oxidative stress and causes apoptosis of a pyruvate kinase deficient cell line. Exp Hematol. 2007; 35(8):1190-1200. 18. Grace RF, Bianchi P, van Beers EJ, et al. The clinical spectrum of pyruvate kinase deficiency: data from the Pyruvate Kinase Deficiency Natural History Study. Blood. 2018;131(20):2183-2192. 19. Grace RF, Layton DM, Barcellini W. How we manage patients with pyruvate kinase deficiency. Br J Haematol. 2019;184(5):721734. 20. van Straaten S, Bierings M, Bianchi P, et al. Worldwide study of hematopoietic allogeneic stem cell transplantation in pyruvate kinase deficiency. Haematologica. 2018; 103(2):e82-e86. 21. Garcia-Gomez M, Calabria A, Garcia-Bravo M, et al. Safe and efficient gene therapy for pyruvate kinase deficiency. Mol Ther. 2016;24(7):1187-1198. 22. Kung C, Hixon J, Kosinski PA, et al. AG-348 enhances pyruvate kinase activity in red blood cells from patients with pyruvate kinase deficiency. Blood. 2017;130(11):13471356. 23. Yang H, Merica E, Chen Y, et al. Phase 1 single- and multiple-ascending-dose randomized studies of the safety, pharmacokinetics, and pharmacodynamics of AG-348, a firstin-class allosteric activator of pyruvate kinase R, in healthy volunteers. Clin Pharmacol Drug Dev. 2019;8(2):246-259. 24. Grace RF, Rose C, Layton M, et al. Safety and efficacy of Mitapivat in pyruvate kinase deficiency. N Engl J Med. 2019;381(10):933944. 25. Allen EL, Ulanet DB, Pirman D, et al. Differential aspartate usage identifies a subset of cancer cells particularly dependent on OGDH. Cell Rep. 2016;17(3):876-890. 26. Clasquin MF, Melamud E, Rabinowitz JD. LC-MS data processing with MAVEN: a metabolomic analysis and visualization engine. Curr Protoc Bioinformatic. 2012; 14:14.11. 27. Kim H, Kosinski P, Kung C, et al. A fit-forpurpose LC–MS/MS method for the simultaneous quantitation of ATP and 2,3-DPG in

haematologica | 2021; 106(1)

human K2EDTA whole blood. J Chromatogr B Analyt Biomed Life Sci. 2017; 1061-1062:89-96. 28. Beutler E, Blume K, Kaplan J, Lohr G, Ramot B, Valentine W. International Committee for Standardization in Haematology: recommended methods for red-cell enzyme analysis. Br J Haematol. 1977;35(2):331-340. 29. Beutler E. Red cell metabolism. A manual of biochemical methods. 3rd edition. Grune & Stratton Inc., Orlando, FL, 1984. 30. Blume KG, Arnold H, Lohr GW, Beutler E. Additional diagnostic procedures for the detection of abnormal red cell pyruvate kinase. Clin Chim Acta. 1973;43(3):443-446. 31. van Oirschot BA, Francois JJJM, van Solinge WW, et al. Novel type of red blood cell pyruvate kinase hyperactivity predicts a remote regulatory locus involved in PKLR gene expression. Am J Hematol. 2014;89(4):380384. 32. Rijksen G, Veerman AJP, Schipper-Kester GPM, Staal GEJ. Diagnosis of pyruvate kinase deficiency in a transfusion-dependent patient with severe hemolytic anemia. Am J Hematol. 1990;35(3):187-193. 33. Lazarova E, Gulbis B, van Oirschot B, van Wijk R. Next-generation osmotic gradient ektacytometry for the diagnosis of hereditary spherocytosis: interlaboratory method validation and experience. Clin Chem Lab Med. 2017;55(3):394-402. 34. Da Costa L, Suner L, Galimand J, et al. Diagnostic tool for red blood cell membrane disorders: assessment of a new generation ektacytometer. Blood Cells Mol Dis. 2016;56(1):9-22. 35. van den Akker E, Satchwell TJ, Pellegrin S, Daniels G, Toye AM. The majority of the in vitro erythroid expansion potential resides in CD34- cells, outweighing the contribution of CD34+ cells and significantly increasing the erythroblast yield from peripheral blood samples. Haematologica. 2010;95(9):1594-1598. 36. Jansen G, Koenderman L, Rijksen G, Cats BP, Staal GEJ. Characteristics of hexokinase, pyruvate kinase, and glucose‐6‐phosphate dehydrogenase during adult and neonatal reticulocyte maturation. Am J Hematol. 1985;20(3):203-215. 37. Wang C, Chiarelli LR, Bianchi P, et al. Human erythrocyte pyruvate kinase: characterization of the recombinant enzyme and a mutant form (R510Q) causing nonspherocytic hemolytic anemia. Blood. 2001; 98(10):3113-3120.

38. Diez A, Gilsanz F, Martinez J, PérezBenavente S, Meza NW, Bautista JM. Lifethreatening nonspherocytic hemolytic anemia in a patient with a null mutation in the PKLR gene and no compensatory PKM gene expression. Blood. 2005;106(5):1851-1856. 39. Rogers SC, Ross JGC, D’Avignon A, et al. Sickle hemoglobin disturbs normal coupling among erythrocyte O2 content, glycolysis, and antioxidant capacity. Blood. 2013; 121(9):1651-1662. 40. Karger R, Lukow C, Kretschmer V. Deformability of red blood cells and correlation with atp content during storage as leukocyte-depleted whole blood. Transfus Med Hemother. 2012;39(4):277-282. 41. Fischer DJ, Torrence NJ, Sprung RJ, Spence DM. Determination of erythrocyte deformability and its correlation to cellular ATP release using microbore tubing with diameters that approximate resistance vessels in vivo. Analyst. 2003;128(9):11631168. 42. Clark R, Rossi E. Osmotic gradient ektacytometry: comprehensive characterization of red cell volume and surface maintenance. Blood. 1983;61(5):899-911. 43. Nijhof W, Wierenga PK, Staal GEJ, Jansen G. Changes in activities and isozyme patterns of glycolytic enzymes during erythroid differentiation in vitro. Blood. 1984;64(3):607613. 44. Takegawa S, Fujii H, Miwa S. Change of pyruvate kinase isozymes from M2- to Ltype during development of the red cell. Br J Haematol. 1983;54(3):467-474. 45. Zaninoni A, Fermo E, Vercellati C, et al. Use of laser assisted optical rotational cell analyzer (LoRRca MaxSis) in the diagnosis of RBC membrane disorders, enzyme defects, and congenital dyserythropoietic anemias: a monocentric study on 202 patients. Front Physiol. 2018;9:451. 46. Llaudet-Planas E, Vives-Corrons JL, Rizzuto V, et al. Osmotic gradient ektacytometry: a valuable screening test for hereditary spherocytosis and other red blood cell membrane disorders. Int J Lab Hematol. 2018;40(1):94102. 47. Leblond PF, Coulombe L, Lyonnais J. Erythrocyte populations in pyruvate kinase deficiency anaemia following splenectomy. Br J Haematol. 1978;39(1):63-70. 48. Cahalan SM, Lukacs V, Ranade SS, Chien S, Bandell M, Patapoutian A. Piezo1 links mechanical forces to red blood cell volume. Elife. 2015;4.

249

LETTERS TO THE EDITOR FcγRI and FcγRIII on splenic macrophages mediate phagocytosis of anti-glycoprotein IIb/IIIa autoantibody-opsonized platelets in immune thrombocytopenia Immunoglobulin G (IgG) anti-platelet autoantibodies are thought to play a central role in platelet destruction in immune thrombocytopenia (ITP).1 IgG autoantibodies are detected in up to 81% of patients with ITP by the direct monoclonal antibody immobilization of platelet antigens (MAIPA) assay,2 and target several platelet glycoproteins (GP) including GPIIb/IIIa, GPIb/IX, and GPV.2,3 While anti-GPIb antibodies can mediate FcγR-independent modes of platelet clearance,4 anti-GPIIb/IIIa autoantibodies are presumed to drive FcγR-dependent platelet clearance through mononuclear phagocytes in the spleen (splenic macrophages). However, the role of specific types of FcγR in the phagocytosis of autoantibodyopsonized platelets is unknown. Here, we purified macrophages from the spleens of ITP patients and incubated them with platelets opsonized with anti-GPIIb/IIIa ITP sera to induce phagocytosis. The role of specific FcγR types was investigated by treating macrophages with individual or combined blocking antibodies against specific FcγR. Anti-GPIIb/IIIa-specific ITP sera mediated significant phagocytosis of platelets relative to platelets incubated with sera from healthy donors. Targeting all FcγR by combining blocking antibodies led to near complete inhibition of splenic macrophage phagocytosis. Blockade of single FcγR types revealed that FcγRI and FcγRIII, but not FcγRIIA, were responsible for phagocytosis. Furthermore, we compared macrophages from ITP and control (trauma) spleens and determined that they had similar phagocytic activity, FcγR expression, and used the same types of FcγR in the phagocytosis of an unbiased target (anti-D-opsonized erythrocytes). Our results indicate that anti-GPIIb/IIIa ITP autoantibodies mediate FcγR-dependent splenic macrophage phagocytosis through FcγRI and FcγRIII. Despite the prevailing hypothesis that anti-GPIIb/IIIa autoantibodies clear platelets through splenic macrophage FcγR, direct demonstrations of this are lacking. McMillan et al. first observed that splenic leukocytes from patients with ITP mediated uptake and/or binding of healthy donor platelets without prior incubation with ITP serum, implicating the spleen as a site of both autoantibody production and platelet clearance.5 Kuwana et al. demonstrated in vitro that peripheral blood monocyte-derived macrophages acquire antigen from ITP patients’ platelets through FcγRI for presentation to GPIIb/IIIa-specific T cells.6 Nakar et al. successfully treated a small cohort of ITP patients with a blocking antibody against FcγRIII, implicating FcγRIII in the clearance of platelets.7 However, the contribution of specific splenic macrophage FcγR to the clearance of platelets has not been directly established and the involvement of other FcγR cannot be excluded. To identify which FcγR are involved in splenic macrophage phagocytosis, macrophages were purified from ITP patients’ spleen cell suspensions by CD14-positive selection (Online Supplementary Figure S1). Sera from five patients with ITP who were positive for GPIIb/IIIa autoantibodies but negative for antibodies against GPIb/IX and GPV (by indirect MAIPA assay) were used individually to opsonize healthy donors’ platelets for phagocytosis. Available characteristics of the ITP patients 250

who provided spleen samples are summarized in Online Supplementary Table S1 and while those of ITP patients who gave sera are summarized in Online Supplementary Table S2. Platelets were fluorescently labeled with 5chloromethylfluorescein diacetate (CMFDA) and phagocytosis was evaluated by confocal microscopy. Nonphagocytosed platelets were detected with an anti-GPIX fluorescent antibody after phagocytosis. Incubation of platelets with ITP serum led to a significant increase in splenic macrophage phagocytosis compared to the phagocytosis following incubation in normal human serum (NHS) (P=0.0015) (Figure 1A, B). To evaluate the specific FcγR types involved, blocking antibodies against FcγRI, FcγRIIA, FcγRIIA/B/C, and FcγRIII were used. Antibodies were deglycosylated using PNGase-F to reduce non-specific blockade, and each antibody was dose-dependently examined for its ability to bind and block phagocytosis (not shown). Two representative ITP sera were selected to evaluate FcγR utilization. Platelet uptake was reduced significantly by the combination of all FcγR blocking antibodies compared to the isotype control (P<0.0001) (Figure 1C). Using single blocking antibodies, blockade of FcγRI inhibited ITP splenic macrophage phagocytosis by 42% (P<0.0001), while blockade of FcγRIII inhibited phagocytosis by 38% (P<0.0001). Surprisingly, minimal, non-significant inhibition was achieved with blockade of FcγRIIA (10%, P=0.056) or all FcγRII isoforms (7%, P=0.15). Although antibody blockade of FcγRIII has been used to successfully treat ITP patients, unfavorable adverse events limited this approach.7 As a monovalent approach can overcome toxicity associated with bivalent FcγR blocking antibodies,8 we evaluated whether a monovalent FcγRIII blocking antibody inhibits phagocytosis with equal efficacy as a bivalent antibody. We generated a monovalent FcγRIII-blocking IgG1-humanized duobody composed of an anti-FcγRIII (3G8) Fab paired with an irrelevant anti-2,4,6-trinitrophenyl Fab. The construct also encoded N297A (which prevents Fc glycosylation) and PG-LALA (P329G, L234A, and L235A) mutations to completely abrogate Fc-FcγR binding.9 The inhibition of ITP splenic macrophage phagocytosis achieved by the monovalent anti-FcγRIII duobody was not significantly different from that achieved by the bivalent and deglycosylated blocking antibody (P=0.87) (Figure 1D). We next compared the leukocyte composition and macrophage FcγR expression in spleens from ITP patients with those in spleens from healthy controls (splenic samples available because of trauma). Both ITP and control spleens contained similar percentages of B cells (CD19+), T cells (CD3+), monocyte/macrophages (CD14+), and granulocytes (CD66b+) (Figure 2A), as evaluated by flow cytometry. Macrophage FcγR expression was also similar between the two types of spleen, with a non-significant trend to increased FcγR expression being observed for ITP macrophages relative to controls (Figure 2B, C). Lastly, we compared the phagocytic activity between splenic macrophages from ITP patients and controls using healthy donor erythrocytes opsonized with a commercial preparation of anti-D (Cangene) as an unbiased phagocytic target. Phagocytosis was evaluated by brightfield microscopy and non-phagocytosed erythrocytes were removed by hypotonic lysis. Erythrocytes were phagocytosed in an antibody-dependent manner and no significant difference in the phagocytic index was observed between the two spleen types (Figure 3A, B). Blocking antibodies were next used alone or in combinahaematologica | 2021; 106(1)

Letters to the Editor

Figure 1. Splenic macrophages from patients with immune thrombocytopenia phagocytose GPIIb/IIIa autoantibody-opsonized platelets through FcγRI and FcγRIII. (A) Splenic macrophages were isolated from spleens from patients with immune thrombocytopenia (ITP) by CD14-positive selection. Healthy donor platelets were opsonized with one of five different ITP sera positive for autoantibodies to GPIIb/IIIa but negative for GPIb/IX and GPV autoantibodies (identified by symbols , , , ■, and ● as patients 1-5, respectively, in Online Supplementary Table S2) (n=4 experiments each). A total of eight unique ITP spleens were used to perform the phagocytosis studies. Four normal human sera (NHS, allogeneic to the platelet donor) specimens were used to opsonize platelets as controls (n=7 experiments). Non-ops: non-opsonized (phosphate-buffered saline only). Phagocytic index: the number of phagocytosed platelets per 100 macrophages. (B) ITP splenic macrophage (left-most panel) with a phagocytosed anti-GPIIb/IIIa ITP serum-opsonized platelet as imaged by spinning disc confocal microscopy (63x objective). Platelets were labeled with the cytoplasmic dye 5-chloromethylfluorescein diacetate (CMFDA) (green, middle panel). External (non-phagocytosed) platelets were identified after phagocytosis using an AlexaFluor 647 (AF647)-conjugated anti-CD42a antibody (red, right panel). Platelets were additionally defined by size (1.5 μm to 3.5 mm) to distinguish them from internalized microparticles or platelet aggregates. Arrow: one phagocytosed platelet. (C) Splenic macrophage FcγR were blocked using deglycosylated antibodies to FcγRI (clone 10.1), FcγRIIA (IV.3), FcγRIIA/B/C (AT10), or FcγRIII (3G8), as indicated. Healthy donor platelets were opsonized with one of two representative anti-GPIIb/IIIa ITP sera (represented by ● and ■) (n=3 experiments; different spleen per experiment). Isotype control: 30 mg/mL deglycosylated mouse IgG1, 10 mg/mL deglycosylated mouse IgG2b (respective to combined blocking antibodies). (D) Inhibition of ITP splenic macrophage phagocytosis of anti-GPIIb/IIIa ITP serum-opsonized platelets by a deglycosylated blocking antibody to FcγRIII (clone 3G8, “FcγRIII”) or a monovalent FcγRIII-blocking IgG1-humanized duobody (“FcγRIII duobody)”. The duobody was bispecific (3G8 Fab, paired with anti-2,4,6-trinitrophenyl as an irrelevant Fab) and possessed PG-LALA (P G, L A, and L A) and N A mutations. Two different ITP sera (represented by ● and ■ ) were evaluated (n=3 experiments; different spleen per experiment). Significance for (A): Kruskal-Wallis test (non-parametric one-way analysis of variance [ANOVA]) with multiple comparisons against all means with a Dunn post-hoc test. Significance for panels (C, D): one-way analysis of variance with multiple comparisons with a Dunnett post-hoc test (C) or Tukey post-hoc test (D). P values: ****P<0.0001, **P=0.0015, ns=not significant. Percent phagocytosis was calculated relative to an untreated group (untreated splenic macrophages with opsonized platelets). Data error: mean ± standard deviation. 329

haematologica | 2021; 106(1)

234

235

297

251

Letters to the Editor

Figure 2. Leukocyte composition and splenic macrophage FcγR expression in spleens from patients with immune thrombocytopenia and from controls. (A) Splenic cells from patients with immune thrombocytopenia (ITP) or splenic trauma (controls) were analyzed by flow cytometry using fluorescent antibodies to determine the percentage of B cells (anti-CD19), T cells (anti-CD3), granulocytes (anti-CD66b), and monocyte/macrophages (anti-CD14). Four different ITP and five control spleens were assessed. Statistical significance was calculated by unpaired t-tests without assuming equal standard deviations. (B) FcγR expression by ITP splenic macrophages was detected using fluorescent antibodies to FcγRI (clone 10.1), FcγRIIA (IV.3), FcγRII isoforms (FcγRIIA/B/C) (AT10), and FcγRIII (3G8). Representative histograms from a single patient with ITP are shown. Blue shaded histograms represent macrophages stained with fluorescent antibody; red shaded histograms represent unstained controls. Y-axis, counts; X-axis, log10 fluorescence intensity. (C) The mean fluorescence intensity (MFI) for each FcγR was determined for ITP and control splenic macrophages (n=5, different spleens in each experiment). Statistical significance was calculated by unpaired two-tailed t-tests; ns = not significant. Data error: mean ± standard deviation.

tion to determine whether FcγR utilization may differ between ITP and control splenic macrophages for anti-Dopsonized erythrocytes. Blockade of all FcγR types led to a significant decrease in the phagocytosis of anti-Dopsonized erythrocytes, down to non-opsonized levels (Figure 3C, D). For macrophages from ITP spleens, FcγRI blockade inhibited phagocytosis of anti-D-opsonized erythrocytes by 58% (P<0.0001), while blockade of FcγRIII inhibited phagocytosis by 29% (P<0.0001). Blockade of FcγRIIA or all FcγRII isoforms did not significantly inhibit phagocytosis (4%, P=0.62 for FcγRIIA; and 6%, P=0.20 for all FcγRII isoforms). FcγRI was also the major phagocytic receptor in control splenic macrophages, as FcγRI blockade inhibited phagocytosis by 51% (P<0.0001) while FcγRIII blockade inhibited phagocytosis by 20% (P<0.05). No significant effect was observed following blockade of FcγRIIA (3%, P=0.99) or all FcγRII isoforms (4%, P=0.97). 252

While both Fc-dependent and Fc-independent autoantibody-mediated platelet clearance mechanisms in ITP have been explored,4 the dominant pathophysiological mechanism in most patients is thought to involve FcγRmediated platelet clearance by splenic macrophages.1 Therapies suggested to block FcγR-dependent processes such as anti-D and IVIg are effective in many patients with ITP, and an FcγRIII-specific blocking antibody was able to rapidly raise platelet counts in a small cohort of ITP patients.7 Furthermore, inhibition of spleen tyrosine kinase (Syk) by treatment with fostamatinib is clinically effective in ITP.10 The success of these therapies suggests that interfering with FcγR function is an effective strategy to increase platelet counts in patients with ITP. While involvement of splenic macrophage FcγRIII in anti-GPIIb/IIIa-autoantibody-opsonized platelet uptake is consistent with reports that selective blockade of FcγRIII is clinically effective,7 significant involvement of haematologica | 2021; 106(1)

Letters to the Editor

Figure 3. Splenic macrophages from patients with immune thrombocytopenia and from controls primarily utilize FcγRI for the phagocytosis of anti-Dopsonized erythrocytes. (A) Phagocytic activity of immune thrombocytopenia (ITP) and control splenic macrophages (MΦ) for erythrocytes opsonized with antiD (Cangene). Non-ops: non-opsonized (phosphate-buffered saline only). The phagocytic index was calculated as the number of erythrocytes engulfed per 100 macrophages. (B) Brightfield microscopy image of ITP splenic macrophages with phagocytosed erythrocytes. The arrow indicates a splenic macrophage with one phagocytosed anti-D-opsonized erythrocyte. (C, D) Splenic macrophages (MΦ) from control (C) and ITP (D) spleens were assessed for FcγR utilization in the phagocytosis of anti-D-opsonized erythrocytes (n=5 experiments; different spleens in each experiment). Macrophage FcγR were blocked using deglycosylated antibodies to FcγRI (clone 10.1), FcγRIIA (IV.3), FcγRIIA/B/C (AT10), or FcγRIII (3G8), as indicated. Isotype control: 30 mg/mL deglycosylated mouse IgG1, 10 mg/mL deglycosylated mouse IgG2b (with respect to combined blocking antibodies). Percent phagocytosis was calculated relative to an untreated group (untreated splenic macrophages with opsonized erythrocytes). Significance for panel (A): one-way analysis of variance (ANOVA) with multiple comparisons against all means with the Tukey post-hoc test. Significance for panels (C, D): one-way ANOVA with multiple comparisons against isotype control with the Dunnett post-hoc test. P values: ****P<0.0001; *P<0.05. Data error: mean ± standard deviation.

FcγRI was unexpected. Although the high-affinity nature of FcγRI for monomeric IgG suggests that FcγRI is saturated with IgG in vivo, FcγRI can engage effectively with immune complexes despite saturation under conditions such as cytokine stimulation.11 Antibody targeting of FcγRI as a therapeutic strategy for ITP has been reported for a single ITP patient and led to downmodulation of FcγRI in circulating monocytes and transient monocytopenia but not to an improvement in the platelet count.12 However, as the antibody used did not directly target the IgG binding region of FcγRI and was reported for a single patient,12 it remains difficult to make conclusions about this approach. We found that control and ITP splenic macrophage levels of FcγR expression, phagocytic activity, and the specific types of FcγR utilized in the phagocytosis of haematologica | 2021; 106(1)

anti-D-opsonized erythrocytes were not significantly different. The dominance of splenic macrophage FcγRI for anti-D-opsonized erythrocytes supports the findings of Nagelkerke et al., who also observed that FcγRI was the primary FcγR mediating red pulp splenic macrophage phagocytosis of anti-D-opsonized erythrocytes.13 A previous study by Audia et al. found similar FcγR expression between ITP and control splenic macrophages;14 we observed a trend of increased FcγR expression on ITP splenic macrophages relative to control macrophages although the difference did not reach statistical significance. Splenic macrophages from patients with ITP have been previously identified to have an M1-type polarization bias.15 Although we did not investigate macrophage polarization, our results indicate at least that splenic macrophage FcγR expression and phagocytic activity in 253

Letters to the Editor

patients with ITP are similar to those in healthy individuals. Combined FcγR blockade may be particularly useful in patients who are refractory to splenectomy, as it may block platelet clearance mediated by macrophages residing outside the spleen such as those in the liver, marrow, and lungs. Our results indicate that, to the extent that FcγR-dependent phagocytosis contributes to platelet clearance in ITP, the individual or combined blockade of FcγRI and FcγRIII is likely the most effective strategy for targeted FcγR blockade as a therapeutic modality. Peter A. A. Norris,1-4 George B. Segel,5 W. Richard Burack,6 Ulrich J. Sachs,7 Suzanne N. Lissenberg-Thunnissen,8 Gestur Vidarsson,8 Behnaz Bayat,7 Christine M. CsertiGazdewich,2,9 Jeannie Callum,2,10 Yulia Lin,2,10 Donald R. Branch,1,2,4,11,12 Rick Kapur,4,8 John W. Semple2-4,11,13,14 and Alan H. Lazarus1-4,11 1 Centre for Innovation, Canadian Blood Services, Ottawa, Ontario, Canada; 2Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada; 3Keenan Research Centre for Biomedical Science, Li Ka Shing Knowledge Institute, St. Michael's Hospital, Toronto, Ontario, Canada; 4Toronto Platelet Immunobiology Group, Toronto, Ontario, Canada; 5Department of Medicine, University of Rochester School of Medicine, Rochester, NY, USA; 6Department of Pathology and Laboratory Medicine, University of Rochester Medical Center, Rochester, NY, USA; 7Institute for Clinical Immunology and Transfusion Medicine, Justus Liebig University, Giessen, Germany; 8Department of Experimental Immunohematology, Sanquin Research and Landsteiner Laboratory, Amsterdam UMC, University of Amsterdam, Amsterdam, the Netherlands; 9Laboratory Medicine Program, University Health Network, Toronto, Ontario, Canada; 10Department of Laboratory Medicine and Molecular Diagnostics, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada; 11Department of Medicine, University of Toronto, Toronto, Ontario, Canada; 12Division of Advanced Therapeutics, Toronto General Hospital Research Institute, Toronto, Ontario, Canada; 13Division of Haematology and Transfusion Medicine, Lund University, Lund, Sweden and 14Department of Pharmacology, University of Toronto, Toronto, Ontario, Canada Correspondence: ALAN H. LAZARUS lazarusa@smh.ca/alan.lazarus@unityhealth.to doi:10.3324/haematol.2020.248385 Disclosures: Yulia Lin has received research funding from Novartis and Octapharma and consulting fees from Amgen and Pfizer. Alan H. Lazarus has patents on the use of monoclonal antibodies as replacements for IVIg and has received research funding from Rigel Pharmaceuticals Incorporated and CSL Behring. GBS, WRB, UJS, BB, SNL-T, CMC-G, GV, JC, DB, RK, and JWS do not have relevant conflicts on interest to disclose. Contributions: AHL, GBS and PAAN conceived the study; PAAN, AHL, GBS, and R.K. designed experiments; WRB, UJS, SNL-T, GV, BB, CMC-G JC, YL, Dand JWS provided reagents; PAAN performed experiments; AHL, GB and PAAN analyzed and interpreted data; PAAN and GBS wrote the manuscript; AHL, GBS, PAAN, WRB, UJS, SNL-T, GV, JC, RK and JWS edited the manuscript. Funding: this research is supported by the University of Toronto, Alexandra Yeo Chair Grant in Benign Hematology (to CMC-G) and

254

received funding support from the Canadian Blood Services Intramural Research Grant Program (to AHL) and Graduate Fellowship Program (to PAAN). It was also funded by the federal government (Health Canada) and provincial and territorial ministries of health. The views herein do not necessarily reflect the views of Canadian Blood Services or the federal, provincial, or territorial governments of Canada. We are grateful to Canadian Blood Services’ blood donors who made this research possible. WRB received funding, to support specimen acquisition, from grant award P50-CA130805 from the National Institutes of Health (USA) and an “Upstate New York Translational Research Grant” administered through UL1 RR024160 from the National Center for Advancing Translational Sciences of the National Institutes of Health.

References 1. Zufferey A, Kapur R, Semple JW. Pathogenesis and therapeutic mechanisms in immune thrombocytopenia (ITP). J Clin Med. 2017; 6(2):16. 2. Porcelijn L, Huiskes E, Oldert G, Schipperus M, Zwaginga JJ, de Haas M. Detection of platelet autoantibodies to identify immune thrombocytopenia: state of the art. Br J Haematol. 2018;182(3):423426. 3. Vollenberg R, Jouni R, Norris PAA, et al. Glycoprotein V is a relevant immune target in patients with immune thrombocytopenia. Haematologica. 2019;104(6):1237-1243. 4. Li J, van der Wal DE, Zhu G, et al. Desialylation is a mechanism of Fc-independent platelet clearance and a therapeutic target in immune thrombocytopenia. Nat Commun. 2015;6:7737. 5. McMillan R, Longmire RL, Tavassoli M, Armstrong S, Yelenosky R. In vitro platelet phagocytosis by splenic leukocytes in idiopathic thrombocytopenic purpura. N Engl J Med. 1974;290(5):249-251. 6. Kuwana M, Okazaki Y, Ikeda Y. Splenic macrophages maintain the anti-platelet autoimmune response via uptake of opsonized platelets in patients with immune thrombocytopenic purpura. J Thromb Haemost. 2009;7(2):322-329. 7. Nakar CT, Bussel JB. 3G8 and GMA161, anti FcγRIII inhibitory monoclonal antibodies in the treatment of chronic refractory ITP. (Summary of 2 pilot studies). Blood. 2009;114(22):2404. 8. Yu X, Menard M, Prechl J, Bhakta V, Sheffield WP, Lazarus AH. Monovalent Fc receptor blockade by an anti-Fcγ receptor/albumin fusion protein ameliorates murine ITP with abrogated toxicity. Blood. 2016;127(1):132-138. 9. Schlothauer T, Herter S, Koller CF, et al. Novel human IgG1 and IgG4 Fc-engineered antibodies with completely abolished immune effector functions. Protein Eng Des Sel. 2016;29(10):457-466. 10. Bussel JB, Arnold DM, Boxer MA, et al. Long-term fostamatinib treatment of adults with immune thrombocytopenia during the phase 3 clinical trial program. Am J Hematol. 2019;94(5):546-553. 11. van der Poel CE, Karssemeijer RA, Boross P, et al. Cytokine-induced immune complex binding to the high-affinity IgG receptor, FcγRI, in the presence of monomeric IgG. Blood. 2010;116(24):5327-5333. 12. Ericson SG, Coleman KD, Wardwell K, et al. Monoclonal antibody 197 (anti-FcγRI) infusion in a patient with immune thrombocytopenia purpura (ITP) results in down-modulation of FcγRI on circulating monocytes. Br J Haematol. 1996;92(3):718-724. 13. Nagelkerke SQ, Bruggeman CW, den Haan JMM, et al. Red pulp macrophages in the human spleen are a distinct cell population with a unique expression of Fc-γ receptors. Blood Adv. 2018;2(8):941-953. 14. Audia S, Santegoets K, Laarhoven AG, et al. Fcγ receptor expression on splenic macrophages in adult immune thrombocytopenia. Clin Exp Immunol. 2017;188(2):275-282. 15. Feng Q, Xu M, Yu YY, et al. High-dose dexamethasone or all-trans retinoic acid restores the balance of macrophages towards M2 in immune thrombocytopenia. J Thromb Haemost. 2017;15(9):18451858.

haematologica | 2021; 106(1)

Letters to the Editor

Impact and safety of chimeric antigen receptor T-cell therapy in older, vulnerable patients with relapsed/refractory large B-cell lymphoma Large B-cell lymphoma (LBCL) predominantly affects older adults with generally suboptimal outcomes even when treated with curative intent.1 In the relapsed/refractory setting, the outcome is dismal and effective treatment for these cases remains an unmet medical need.2 The development of chimeric antigen receptor T-cell therapy (CAR T) has revolutionized the treatment of relapsed/refractory LBCL.3 However, the impact and safety of this treatment in vulnerable, older patients with lymphoma, especially those with multiple comorbid conditions and functional limitation, has not been explored. These geriatric vulnerabilities have been associated with poor survival and increased treatment-related toxicities in older lymphoma patients receiving anthracycline-based chemoimmunotherapy.1 The Centers for Medicare & Medicaid Services have recently proposed complete coverage for CAR T in Medicare beneficiaries, highlighting the significant need for older patients including those with geriatric deficits and frailty. In this study, we examined the outcomes of older LBCL patients referred for treatment with commercial CAR T products, axicabtagene ciloleucel (Yescarta, Kite-Gilead) and tisagenlecleucel (Kymriah, Norvartis), at our institution, and explored the prevalence and impact of prospectively collected, baseline geriatric vulnerabilities. Importantly, we compared toxicities and outcomes of younger versus older patients who received CAR T using the Medicare coverage policy cutoff of 65 years. A multidimensional geriatric assessment of comorbidity burden, function, mobility, nutrition, mood, and medication was prospectively performed prior to CAR T by a geriatrician or by an interdisciplinary clinical provider, as previously described.4 Specifically, the comorbidity burden was defined using the Deyo/Charlson Comorbidity Index (DCI/CCI).5 Functional limitations were defined by deficits in basic or instrumental activities of daily living.4 Cognitive impairment was defined by a score <26 on the Montreal Cognitive Assessment.4 A waiver of authorization for retrospective collection of demographic, treatment, and survival data was obtained from the Institutional Review and Privacy Board. Patientsâ&#x20AC;&#x2122; referral data were collected from the central intake database from January 2018 to March 2019. All patients had a diagnosis of relapsed/refractory LBCL after two or more lines of systemic therapy. CAR T eligibility criteria, clinical care, and disease monitoring were in accordance with the Yescarta and Kymriah product inserts and standard institutional guidelines. Cytokine release syndrome and immune effector cell-associated neurotoxicity syndrome (ICANS) were graded according to the American Society of Transplantation and Cellular Therapy consensus.6,7 Standard descriptive statistics were used, and comparisons were performed using the Fisher exact test or Wilcoxon rank-sum test where appropriate. The association between receipt of CAR T and post-relapse overall survival, defined as the time survived from last biopsyproven relapsed/refractory state, was examined using univariable Cox proportional hazards regression with a time-dependent covariate accounting for time of initiation of CAR T. Overall survival and progression-free survival of the treated group were estimated from the date of CAR T infusion using the Kaplan-Meier method. All survival comparisons across groups were based on the haematologica | 2021; 106(1)

log-rank test, except for the comparison between patients treated with CAR T and those who received other therapies, which was based on a Wald test. All statistical analyses were performed using SAS software version 9.4 (The SAS Institute, Cary, NC, USA). Forty-two consecutive patients aged 65 years or older with relapsed/refractory LBCL were included in the analysis of post-relapse overall survival, including 24 patients who received CAR T and 18 who did not either because of clinical ineligibility as judged by physicians and/or death during the pre-requisite clinical evaluation. None of these 18 patients had undergone apheresis for a CAR T product. Instead, they received salvage chemotherapy or supportive care only. Age, gender, prior lines of therapy, relapse stage, comorbidity burden, and Karnofsky Performance Status were comparable in the two groups (Online Supplementary Table S1). With a median follow-up of 291 days (range, 162-572) for survivors, the group of older patients who had received CAR T had a lower risk of death compared to those who had not received CAR T (hazard ratio [HR] 0.31, 95% confidence interval [95% CI]: 0.10-0.93, P=0.04, for post-relapse overall survival). During the same period in which the above 24 patients â&#x2030;Ľ65 years old were treated, an additional 25 patients <65 years old received CAR T for relapsed/refractory LBCL. All 49 patients met eligibility criteria for treatment with a commercial product and the use of bridging therapy was at the discretion of treating physicians. The median number of lines of prior therapy was 3 (range, 2-9); the median time from last relapse/disease progression to CAR T was 86 days (range, 33-272); and 42/49 (86%) of the patients were in relapse or had progressive disease prior to CAR T. Most patients had a pre-CAR T assessment of function, comorbidity, cognition, mobility, mood, and nutrition. Overall, the comorbidity burden was moderate (median DCI/CCI 2; range, 2-7); 27% of patients had functional limitation; 44% had cognitive impairment; 29% had had prior falls; 27% had weight loss; and 10% had depression at baseline (Table 1 and data not shown). Baseline characteristics were similar in the two age groups including: Karnofsky Performance Status, functional limitations, cognition, mobility (prior falls), weight loss, and disease characteristics including prior lines of therapy, lactate dehydrogenase concentration, stage, and time to CAR T (Table 1). The older group included more females (P<0.001) and the patients had higher DCI/CCI values (P=0.04) (Table 1). Numerically more younger patients (84%) received axicabtagene ciloleucel than tisagenlecleucel (63%; P=0.11). We compared the safety and toxicity profiles between older and younger patients and found that the two groups had similar incidences of all grade and grade 3-4 cytokine release syndrome and ICANS (Table 1). Incidences of grade 3-4 hematologic and non-hematologic toxicities, as defined by Common Terminology Criteria for Adverse Events v5.0, were also similar between the two groups, although older patients appeared to have numerically fewer infections and cytopenia, and more metabolic and other toxicities (Table 1). The rate of intensive care unit admission was similar in the two age groups. With a median follow-up of 179 days for survivors among these 49 CAR T patients (range, 84-470), the 6month median progression-free and overall survival rates were 48% (95% CI: 33-63), and 71% (95% CI: 57-84), respectively. The 100-day complete response rate was 51%. Importantly, we did not observe evidence of a statistically significant difference in progression-free or overall survival between the older and younger groups of 255

Letters to the Editor

Table 1. Characteristics and toxicities of lymphoma patients treated with chimeric antigen receptor T-cell therapy.

Age in years, median (range) Female gender, n (%) CAR T, n (%) Axicabtagene ciloleucel Tisagenlecleucel Advanced stage at CAR T, n (%) Prior lines, median (range) Baseline LDH, median (range) Time to CAR T, median (range) DCI/CCI, median, (range) KPS <80, n (%) Functional limitation, n (%) Cognitive impairment, n (%) Prior fall, n (%) Weight loss, n (%) ICU admission, n (%) CRS, n (%) No CRS Grade 1-2 CRS Grade >2 CRS ICANS, n (%) No ICANS Grade 1-2 ICANS Grade >2 ICANS Infections, ≥G3, n (%) Prolonged cytopenia, n (%) Metabolic toxicities, ≥ grade 3, n (%) Other toxicities, ≥ grade 3, n (%)

Younger patients (<65 years, n=25)

Older patients (≥65 years, n=24)

56 (20 – 64) 2 (8)

72 (67 – 86) 13 (54)

21 (84) 4 (16) 14 (56) 3 (2 – 9) 298 (128 – 3722) 75 days (43 – 175) 2 (2 – 4) 7 (28) 5 (20) 8 (32) 7 (28) 8 (32) 9 (36)

15 (63) 9 (37) 14 (58) 3 (2 – 9) 240 (146 – 1409) 92 days (33 – 272) 3 (2 – 7) 9 (38) 8 (33) 11 (46) 7 (29) 5 (21) 6 (25)

7 (28) 15 (60) 3 (12)

4 (17) 18 (75) 2 (8)

16 (60) 6 (24) 4 (16) 15 (60) 16 (64) 3 (12) 9 (36)

11 (46) 7 (29) 6 (25) 10 (42) 10 (42) 8 (33) 12 (50)

<0.001 0.11

0.78 0.81 0.12 0.54 0.04 0.55 0.35 0.76 >0.99 0.52 0.54 0.61

0.60

0.26 0.16 0.10 0.39

CAR T: chimeric antigen receptor T-cell therapy; TimeCAR T: time (in days) from last relapse/disease progression to start of CAR T; LDH: lactate dehydrogenase; DCI/CCI: Deyo/Charlson Comorbidity Index; KPS: Karnofsky Performance Status; ICU: intensive care unit; CRS: cytokine release syndrome; ICANS, immune effector cell associated neurotoxicity syndrome.

patients by either chronological age (Figure 1A, B), functional limitation, cognitive impairment, or comorbidity burden (DCI/CCI >2, data not shown). At the time of last follow-up, only one treatment-related death had occurred within 100 days: a 69-year old woman had died as a result of prolonged cytopenia. An additional patient with a history of prior allogeneic hematopoietic cell transplantation died of influenza pneumonia 129 days after CAR T infusion. We present here for the first time, outcomes of older LBCL patients referred for CAR T in the context of geriatric vulnerabilities. Not surprisingly, we found that older patients who received CAR T had better post-relapse overall survival than those who were referred but not treated; it is likely that there was a selection bias for patients who did not have rapidly progressive disease, significant comorbidities, or suboptimal performance status. This could also explain why we did not have manufacture failures or deaths prior to infusion in this small cohort of patients. Interestingly, no excess toxicity was found in several real-world cohorts of CAR T patients including those with poor performance status in whom chemotherapy was historically associated with significant toxicity and mortality, adding to the complexity of 256

selection of patients for CAR T.8-10 Multicenter collaborative and registry studies are currently underway to prospectively identify geriatric impairments, to validate their prognostic impact, and to develop proper algorithms for the selection of patients. While prior studies compared outcomes of CAR T based on chronological age alone,11,12 we identified a significant burden of baseline geriatric vulnerabilities including functional limitations, multiple comorbid conditions, cognitive impairment, weight loss, decreased mobility, and polypharmacy in our patients. Importantly, our low treatment-related mortality and similar efficacy and toxicities across groups of patients of different chronological age and impairment suggest that older, vulnerable patients should be similarly considered for CAR T as younger patients. We acknowledge that while no major differences were seen between groups, we cannot rule out smaller differences due to small sample size and patient selection bias. Lastly, our findings also support the concept that reducing disease burden may help to improve function and performance status in older lymphoma patients.13 Interestingly, while geriatric assessment domains such as functional limitation and multimorbidity are prognostically important for lymphoma haematologica | 2021; 106(1)

Letters to the Editor

Figure 1. Survival outcomes. (A) Overall survival (OS) of lymphoma patients who were treated with chimeric antigen receptor T-cell therapy (CAR T) stratified by age. Green line: patients 65 years or older (n=24). Red line: patients younger than 65 years (n=25). Numbers at risk are tabulated. (B) Progression-free survival (PFS) of lymphoma patients who were treated with CAR T stratified by age. Green line: patients 65 years or older (n=24). Red line: patients younger than 65 years (n=25). Numbers at risk are tabulated.

patients undergoing chemoimmunotherapy or autologous transplantation,1,14 they may not be prognostic in the setting of CAR T. In summary, although limited by small sample size and likely patient selection bias, our results highlight potential benefits of CAR T without excessive cytokine release syndrome, ICANS, and other high-grade toxicities, in selected older patients. These findings extend beyond published results for older patients in the ZUMA-1 and JULIET trials, and provide novel insights and the entry point for large-scale investigation of geriatric vulnerabilities in patients considered for CAR T. We contend that older patients should not be automatically excluded from CAR T based solely on chronological age, multi-morbidity, functional limitation, or cognitive impairment; rather, their care should be individualized including greater attention to non-oncological geriatric issues. It is likely that a detailed geriatric assessment in combination with organ function evaluation will allow better selection of older patients who could benefit from this curative treatment.15 Richard J. Lin,1,2 Stephanie M. Lobaugh,3 Martina Pennisi,1 Hei Ton Chan,1,Â° Yakup Batlevi,1 Josel D. Ruiz,1 Theresa A. Elko,1 Molly A. Maloy,1 Connie L. Batlevi,2,4 Parastoo B. Dahi,1,2 Sergio A. Giralt,1,2 Paul A. Hamlin,2,4 Elena Mead,2,5 Ariela Noy,2,4 M. Lia Palomba,2,4 Bianca D. Santomasso,2,6 Craig S. Sauter,1,2 Michael Scordo,1,2 Gunjan L. Shah,1,2 Beatriz Korc-Grodzicki,2,7 Soo Jung Kim,7 Mari Lynne Silverberg,1 Chelsea A. Brooklyn,1 Sean M. Devlin3 and Miguel-Angel Perales1,2 1 Adult BMT Service, Memorial Sloan Kettering Cancer Center; haematologica | 2021; 106(1)

Department of Medicine, Weill Cornell Medical College; 3Department of Epidemiology & Biostatistics, Memorial Sloan Kettering Cancer Center; 4Lymphoma Service, Memorial Sloan Kettering Cancer Center; 5 Critical Care Service, Memorial Sloan Kettering Cancer Center; 6 Neurology Service, Memorial Sloan Kettering Cancer Center and 7 Geriatrics Service, Memorial Sloan Kettering Cancer Center, New York, NY, USA Â°Current address: Columbia University Medical Center, New York, NY, USA

Correspondence: MIGUEL-ANGEL PERALES - peralesm@mskcc.org doi:10.3324/haematol.2019.243246 Disclosures: SAG: advisory Board for Amgen, Actinuum, Celgene, Johnson & Johnson, Jazz pharmaceutical, Takeda, Novartis, Kite, Spectrum Pharma; research funding from Amgen, Actinuum, Celgene, Johnson & Johnson, Miltenyi, Takeda; CLB: research funding from Epizyme, Novartis, Janssen, BMS, Miragen, Mediimmune; consultancy from Defined Health; HTC: consultancy: Atara Biotherapeutics; GLS: research funding from Janssen and Amgen; MS: consultancy and research funding: Angiocrine Bioscience, Inc.; consultancy: McKinsey & Company; PAH: research support and consulting fees from Portola Pharmaceuticals, Inc. CSS: consultant on advisory boards for: Juno Therapeutics, Sanofi Genzyme, Spectrum Pharmaceuticals, Novartis, Precision Biosciences, Kite, a Gilead Company and GSK. Research funds for investigator-initiated trials from: Juno Therapeutics and Sanofi-Genzyme; AN: research funding from Pharmacyclics and Raphael. Honoraria from Prime Oncology, Medscape, and EUSA. Advisory Board for Janssen; MLP: advisory Board for Celgene, Consultant for Merck and Pharmacyclics; MAP: honoraria from Abbvie, Bellicum, Bristol-Myers Squibb, Incyte, Merck, Novartis, Nektar Therapeutics, and Takeda. He serves on DSMBs for Servier 257

Letters to the Editor

and Medigene, and the scientific advisory boards of MolMed and NexImmune. Research support for clinical trials from Incyte, Kite (Gilead) and Miltenyi Biotec. Contributions: RJL and MAP designed the study and wrote the manuscript. RJL, SML, SMD, and MAP analyzed the data. RJL, MSP, HTC, YB, JDR, TAE, MAM, MLS, and CAB contributed essential data management. BK and SJK performed geriatric assessment. All remaining authors enrolled patients and contributed data to the study. Every author critically appraised and approved the submitted manuscript. Funding: this research was supported in part by the National Institutes of Health/National Cancer Instituteâ&#x20AC;&#x2122;s Cancer Center Support Grant P30 CA008748 and the Program Project Grant P01 CA023766. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. RJL acknowledges research support from the Elsa U. Pardee Foundation for Cancer Research, New York State Empire Clinical Research Investigator Program (ECRIP), and Parker Institute for Cancer Immunotherapy (PICI) at MSK. This work was presented in part at the 61st American Society of Hematology Annual Meeting & Exposition from December 6-10, 2019, Orlando, FL, USA.

References 1. Soubeyran P-L, Cordoba R. Approaches for vulnerable and frail older patients with diffuse large B-cell lymphomas. Curr Opin Oncol. 2019;31(5):369-373. 2. Crump M, Neelapu SS, Farooq U, et al. Outcomes in refractory diffuse large B-cell lymphoma: results from the international SCHOLAR-1 study. Blood. 2017;130(16):1800-1808. 3. Hunter BD, Rogalski M, Jacobson CA. Chimeric antigen receptor Tcell therapy for the treatment of aggressive B-cell non-Hodgkin lymphomas: efficacy, toxicity, and comparative chimeric antigen receptor products. Expert Opin Biol Ther. 2019;19(11):1157-1164. 4. Lin RJ, Elko TA, Devlin SM, et al. Impact of geriatric vulnerabilities on allogeneic hematopoietic cell transplantation outcomes in older patients with hematologic malignancies. Bone Marrow Transplant. 2020;55(1):157-164.

258

5. Perales M-A, Bonafede M, Cai Q, et al. Real-world economic burden associated with transplantation-related complications. Biol Blood Marrow Transplant. 2017;23(10):1788-1794. 6. Lee DW, Santomasso BD, Locke FL, et al. ASBMT consensus grading for cytokine release syndrome and neurological toxicity associated with immune effector cells. Biol Blood Marrow Transplant. 2019; 25(4):625-638. 7. Pennisi M, Jain T, Santomasso BD, et al. Comparing CAR T-cells toxicity grading systems: application of ASTCT grading system and implications for management. Blood Adv. 2020;4(4):676-686. 8. Nastoupil LJ, Jain MD, Spiegel JY, et al. Axicabtagene ciloleucel (AxiCel) CD19 chimeric antigen receptor (CAR) T-cell therapy for relapsed/refractory large B-cell lymphoma: real world experience. Blood. 2018;132(Suppl 1):91. 9. Jacobson CA, Hunter B, Armand P, et al. Axicabtagene ciloleucel in the real world: outcomes and predictors of response, resistance, and toxicity. Blood. 2018;132(Suppl 1):92. 10. Smith SD, Reddy P, Sokolova A, et al. Eligibility for CAR T-cell therapy: an analysis of selection criteria and survival outcomes in chemorefractory DLBCL. Am J Hematol. 2019;94(4):E117-E116. 11. Neelapu SS, Locke FL, Bartlett NL, et al. Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N Engl J Med. 2017;377(26):2531-2544. 12. Schuster SJ, Bishop MR, Tam CS, et al. Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma. N Engl J Med. 2019;380(1):45-56. 13. Bowcock SJ, Fontana V, Patrick HE. Very poor performance status elderly patients with aggressive B cell lymphomas can benefit from intensive chemotherapy. Br J Haematol. 2012;157(3):391-393. 14. Spina M, Merli F, Puccini B, et al. The elderly project by the Fondazione Italiana Linfomi: a prospective comprehensive geriatric assessment (CGA) of 1353 elderly patients with diffuse large B-cell lymphoma. In: Hematologic Oncology, 2019:p S2. Published by John Wiley & Sons, Ltd. 15. Jain T, Bar M, Kansagra AJ, et al. Use of chimeric antigen receptor T cell therapy in clinical practice for relapsed/refractory aggressive B cell non-Hodgkin lymphoma: an expert panel opinion from the American Society for Transplantation and Cellular Therapy. Biol Blood Marrow Transplant. 2019;25(12):2305-2321.

haematologica | 2021; 106(1)

Letters to the Editor

Macrophage-HFE controls iron metabolism and immune responses in aged mice More than one in 200 individuals of northern European origin are affected by HFE-hemochromatosis, a highly prevalent genetic iron overload disorder.1 Hfe deficiency specifically in hepatocytes causes low expression of the iron regulatory hormone hepcidin, and iron accumulation in organs.2 Here we report a previously unrecognized role of HFE in macrophages as a regulator of iron and immune responses, which is independent of the hepatocytic role of HFE in maintaining systemic iron homeostasis. Our previous work solved a long-standing debate by establishing that HFE acted in hepatocytes to control hepcidin levels, thereby preventing systemic iron overload.2 These studies classified HFE-hereditary hemochromatosis as a liver disease. Low hepcidin levels in HFE-hereditary hemochromatosis increase the stability of the iron exporter ferroportin (FPN), allowing for excess efflux of iron from macrophages and duodenal enterocytes into the circulation.3 As a consequence, macrophages from HFE-patients and Hfe-mutant mice are relatively iron-depleted, despite systemic iron overload.4,5 Macrophages play versatile roles in maintaining tissue homeostasis. They are involved in the recycling of aged red blood cells, tissue repair and regeneration, and in activation of immune responses.6 Alternative macrophages in response to selected sets of pro- and antiinflammatory stimuli has been linked to changes in intracellular iron levels.7,8 On the other hand, systemic iron overload increased susceptibility of mice towards infections with Vibrio vulnificus9 and Yersinia pseudotuberculosis,10 while depletion of iron by deferoxamine improved survival of mice receiving lethal doses of lipopolysaccharides (LPS).11 Interestingly, lack of Hfe in mice was beneficial, protecting against invasive Salmonella enterica serovar typhimurium infection12 and macrophages derived from Hfe-/- mice displayed an attenuated inflammatory response to LPS and Salmonella challenge.13 One explanation for these findings may be that low iron levels in Hfe-deficient macrophages restrict iron as a nutrient and signal for Salmonella growth. Alternatively, Hfe may exert a previously unrecognized role in immune cells. Given that iron dyshomeostasis is associated with aging and that aging may undermine the immune responses of macrophages, we tested whether selective lack of Hfe in macrophages may affect iron and immune responses in aged mice. Analyses of iron-related parameters in 45-week old HfeLysMCre mutant mice, which carry a specific deletion of Hfe in myeloid cells,2 showed significant decreases of non-heme iron levels in the liver, spleen and duodenum (Figure 1A, B). The degree of iron reduction was moderate in comparison to that caused by nutritional iron deficiency (Figure 1A). Consistent with decreased iron levels in the liver and in freshly isolated primary hepatocytes from HfeLysMCre mice (Figure 1A, C), serum hepcidin levels were decreased, while circulating iron and erythropoietin levels were unchanged (Figure 1D). Hematologic indices and the numbers of macrophages, granulocytes, T cells and B cells in the spleen and bone marrow were not significantly different between HfeLysMCre mutant and control mice (Figure 1E, Online Supplementary Figure S1). In line with low serum hepcidin levels, hepatic expression of hepcidin (Hamp1) and genes known to be co-regulated with hepcidin in a BMP/SMAD-dependent manner (such as Bmp6, Smad7, Id1 and Pai1) were decreased (Figure 1F, G). By contrast, the levels of transferrin receptor 1 (TfR1), haematologica | 2021; 106(1)

which is responsible for transferrin-mediated iron uptake, failed to increase in the liver of HfeLysMCre mice (Figure 1G), likely due to a milder decrease in the liver iron stores in HfeLysMCre mice than in mice subjected to nutritional iron deficiency (Figure 1A, Online Supplementary Figure S3A). In the spleen of 45-week old HfeLysMCre mice, the decrease in iron levels correlated with increased protein levels of Fpn and TfR1, (Figure 1H), contrasting the effect of nutritional iron deficiency which was associated with low Fpn levels (Online Supplementary Figure S3B). In the duodenum of 45-week old HfeLysMCre mice, increased mRNA expression of iron importers such as Dcytb and Dmt1 was observed, whereas the levels of Fpn and TfR1 were not statistically different (Figure 1I). We conclude that 45-week old HfeLysMCre mice show significant alterations in iron metabolism. This is in contrast to the iron state of 12-week old HfeLysMCre mice, which showed no changes in any of the iron-related parameters investigated (Figure 1J-M).2 We next monitored systemic responses of 45-week old HfeLysMCre mice to iron overload and immune challenges in the absence of macrophage-Hfe. We subjected 45-week old HfeLysMCre mutant and control mice to parenteral iron overload. The mice were given a single intra-peritoneal injection of iron-dextran solution. The ability to accumulate iron in tissues and to increase hepcidin was comparable between HfeLysMCre mutant and control mice (Figure 2A, B), suggesting that macrophage HFE is dispensable for iron accumulation and hepcidin responses in the setting of the parenteral iron overload used in this study. By contrast, 45-week old HfeLysMCre mice had a better survival in response to LPS-induced endotoxin shock (10 mg/kg; LPS from E. coli 055:B5, L2630) (Figure 1H), compared to 12-week old HfeLysMCre mice which succumbed to the endotoxin shock at a similar rate as that in the controls (Figure 2C, D). Likewise, constitutive Hfe-/- and hepatocyte-specific Hfe-mutant mice (HfeAlfpCre), and mice maintained on an iron-deficient diet prior to LPS challenge, were not protected from endotoxin shock (Figure 2E, F). Based on these data we conclude that a specific lack of Hfe in macrophages in 45-week old mice may provide an adaptive mechanism aimed at protecting the host from inflammatory injury. Given that the release of serum hepcidin, various proand anti-inflammatory cytokines and chemokines is considered a central feature of the inflammatory response, we next measured the expression of a set of serum immune mediators in 45-week old, HfeLysMCre mutant mice during endotoxin shock. The overall expression pattern of hepcidin, pro- and anti-inflammatory cytokines and chemokines was similar between mutant HfeLysMCre mice and control animals (Online Supplementary Figure S2), implying that the endotoxin shock triggered common pathways. Interestingly, HfeLysMCre mutant mice showed a trend towards a higher degree of induction of granulocyte colony-stimulating factor, interferon γ, interleukin (IL)-1b, IL-17α (Online Supplementary Figure S2C) and reduced induction of eotaxin, IL2, IL4, IL5, IL6, IL10, IL13, macrophage inflammatory protein 1α, and tumor necrosis factor α (TNF-α) mediators compared to control mice (Online Supplementary Figure S2D), raising the question of whether differential expression of these immune mediators may explain the remarkably higher tolerance of aged HfeLysMCre mice to LPS-induced shock. To better understand the role of macrophage-Hfe at the cellular level, we next isolated primary macrophages from HfeLysMCre, HfeAlfpCre (deletion of Hfe in hepatocytes) and constitutive Hfe knock-out (Hfe-/-) mice and com259

Letters to the Editor

Figure 1. Lack of macrophage-Hfe in 45-week old mice contributes to moderate systemic iron deficiency in contrast to physiological iron homeostasis present in 12-week old HfeLysMCre mice. (A) Non-heme iron levels in the liver, spleen and duodenum from 45-week old HfeLysMCre mutant and Hfeflox control mice, and from mice kept on an iron-deficient diet (IDD). n indicates the number of livers (n=19, 15, 5), spleens (n=19, 15, 5) and duodenums (n=6, 6, 5) isolated from HfeLysMCre, Hfeflox and IDD mice, respectively. (B) Perls staining for iron deposits in the liver, spleen and duodenum of 45-week old HfeLysMCre mutant and control mice. Scale bar 50 µm (liver) and 100 µm (spleen and duodenum). Representative stainings of three sections are shown. (C) Intracellular total iron levels measured by total-reflection x-ray fluorescence in primary hepatocytes (HC), Kupffer cells (KC), liver sinusoidal endothelial cells (LSEC), and hepatic stellate cells (HSC) from 45-week old HfeLysMCre and Hfeflox mice (n=3-5). (D) Serum hepcidin, iron and erythropoietin (EPO) levels, and (E) hematologic parameters including red blood cell (RBC) count, hemoglobin (Hgb) concentration, hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), and the percentages of white blood cells (WBC) including lymphocytes (Lym), monocytes (Mo), and granulocytes (Gra) in the blood of 45-week old HfeLysMCre mutant (n=6) and Hfeflox control mice (n=6). (F) Relative mRNA expression of iron-related genes in the liver of 45-week old HfeLysMCre mutant (n=6) and Hfeflox control mice (n=9) determined by real-time polymerase chain reaction (PCR). (G, H) Representative immunoblot analysis of pSMAD1, transferrin receptor 1 (TfR1) and ferroportin (FPN), relative to bactin levels (shown in histograms on the right), in the livers and spleens of HfeLysMCre mutant (n=8) and control mice (n=8). (I) Relative mRNA expression of iron transporters in the duodenum of 45-week old HfeLysMCre mutant (n=4) and Hfeflox control mice (n=8) measured by real-time PCR. (J-L) Non-heme iron levels (J), serum iron and hepcidin levels (K), and relative mRNA expression of hepcidin (Hamp1) and Bmp6 (L) in the livers of 12-week old HfeLysMCre mutant (n=5) and Hfeflox control mice (n=6). (M) Representative immunoblot analysis of TfR1 and FPN, relative to b-actin levels (shown in histograms on the right), in the spleens of 12-week old HfeLysMCre mutant (n=5) and Hfeflox control mice (n=6). M: Prestained Protein Marker PageRuler. n indicates the number of mice used in the analysis. Data are shown as mean ± standard error of mean. Statistically significant differences are indicated as *P<0.05, **P<0.005, ***P<0.0005.

260

haematologica | 2021; 106(1)

Letters to the Editor

pared iron-related parameters. These mouse models enable the dissociation of the cell-autonomous actions of HFE in macrophages (present in HfeLysMCre and Hfe-/- mice) from those that are a consequence of altered systemic hepcidin expression (present in Hfe-/- and HfeAlfpCre mutant mice).2 We showed that lack of Hfe in macrophages derived from aged Hfe-/- and aged Hfe LysMCre mice correlated with reduced intracellular total iron levels whereas macrophages derived from HfeAlfpCre mice,

which carry a wild-type Hfe allele in all extra-hepatocytic cell types, including macrophages, showed no statistically significant difference in iron levels compared to their respective controls although there was a tendency towards less iron in HfeAlfpCre cells (Figure 3A). Consistent with our findings in the spleen of HfeLysMCre mice (Figure 1G), we showed that FPN protein and mRNA levels were significantly increased in Hfe-deficient macrophages while TfR1 expression was unaltered irre-

Figure 2. Lack of macrophage-Hfe in 45-week old mice is dispensable during parenteral iron overload but contributes to better survival of HfeLysMCre mutant mice during lipopolysaccharide-induced endotoxin shock. (A, B) Non-heme and plasma iron levels (A), and hepcidin mRNA expression (B) in 45-week old HfeLysMCre mutant (n=6) and control mice (n=6) following parenteral iron overload. Data are shown as the mean Âą standard error of mean. Statistically significant differences are indicated as *P<0.05, **P<0.005, ***P<0.0005, ****P<0.0001. n indicates the number of mice used in the analysis. (C-F) Survival of 45-week old HfeLysMCre mutant and control mice (n=11, 11) (C), 12-week old HfeLysMCre mutant and control mice (n=10, 10) (D), 45-week old Hfe-/- and control mice, and mice maintained on an iron-deficient diet for 4 weeks (IDD) (n=13, 13, 12) (E) and 12-week old HfeAlfpCre, Hfe-/- and control mice (n=9, 9, 9), which were injected with lipopolysaccharide (10 mg/kg) and monitored every 6 h for up to 4 days. The percentage survival rate was expressed by Kaplan-Meier curves. Statistical analyses were performed using the log-rank test and the results were considered statistically significant when *P<0.05. ns: not statistically significant; Fe-ip: intraperitoneal iron injections; LPS: lipopolysaccharide.

haematologica | 2021; 106(1)

261

Letters to the Editor

Figure 3. Hfe in macrophages controls cellular iron homeostasis via ferroportin. (A) Intracellular total iron levels in bone marrow-derived macrophages (BMDM) from 45-week old Hfe-/-, HfeLysMCre and HfeAlfpCre mutant mice (n=3, 4). (B) Representative immunoblot analysis of transferrin receptor 1 (TfR1) and ferroportin (FPN) relative to b-actin levels (shown in histograms below) in 45-week old Hfe-/-, HfeLysMCre and HfeAlfpCre mutant mice. (C) Relative mRNA quantification of iron-related genes, inflammatory cytokines and toll-like receptors genes by real-time polymerase chain reaction in BMDM derived from 45-week old HfeLysMCre mutant and control mice (n=6, 6). (D, E) Intracellular total iron levels (D) and immunoblot analysis (E) of TfR1 and FPN, relative to b-actin levels (shown in histograms on the right), in BMDM from 12-week old HfeLysMCre mutant and Hfeflox control mice (n=5, 4). M: Prestained Protein Marker PageRuler (Thermo Scientific). (F) Schematic illustration of the Hfe cDNA construct cloned into the pcDNA6.2 vector under the control of the cytomegalovirus (CMV) promoter, tagged with V5 epitope at the C-terminal end and a poly-A tail. (G) Relative mRNA expression of Hfe and Fpn, and (H) intracellular total iron levels in transiently transfected BMDM with vector carrying the Hfe cDNA construct (indicated as ‘o/e Hfe’) or empty vector (‘mock’) (n=4, 5). (I) Representative immunoblot analysis, from two independent experiments, of FPN, TfR1, ferritin and relative quantification to β-actin levels (shown in histograms on the right) in BMDM transiently transfected with a vector carrying Hfe cDNA or an empty vector (‘mock’) (n=4, 6). (J) Phosphorylation status of STAT3(Ser727) in the BMDM from HfeLysMCre mutant and Hfeflox control mice (n=7, 7) and in BMDM transiently transfected with a vector carrying Hfe cDNA or an empty vector (‘mock’) (n=9, 9) measured by Bio-Plex Pro Cell Signaling MAPK Panel. AU: arbitrary units. Data are shown as mean ± standard error of mean. Statistically significant differences are indicated as *P<0.05, **P<0.005. n indicates the number of mice used in the analysis.

262

haematologica | 2021; 106(1)

Letters to the Editor

spective of the genotypes (Figure 3B). The mRNA expression of hepcidin and other iron-related genes, pro- and anti-inflammatory cytokines, and toll-like receptors was not affected by the lack of macrophage-Hfe (Figure 3C). Importantly, intracellular total iron levels, FPN and TfR1 protein expression, were not changed in macrophages isolated from 12-week old HfeLysMCre mutant mice (Figure 3D, E), supporting our previous findings.2 Conversely, overexpression of Hfe in primary wild-type macrophages counteracted FPN induction and intracellular iron deficiency (Figure 3F-I). We observed an interesting correlation between HFE expression in macrophages and the STAT3 signaling pathway, whereby the latter’s activity was significantly decreased in Hfe-deficient macrophages and 3-fold increased upon Hfe overexpression (Figure 3J). Additional phospho-proteins studied, including pp38MAPK, pERK1/2, pMEK1, pJNK, pATF-2, pp90RSK, pHSP27, and p53, were not significantly different between Hfe-deficient and Hfe-overexpressing macrophages (data not shown). The contribution of STAT3 signaling to FPN expression in Hfe-deficient macrophages deserves further investigations. Collectively, the results from primary macrophages show that macrophage-Hfe expression controls iron accumulation by regulating expression of the iron exporter FPN. However, the LysM-Cre does not target all macrophage types with the same efficacy and its expression varies among tissues. This may explain the lack of an iron-poor phenotype in Küpffer cells in contrast to bone marrow-derived macropahges. Our findings are in line with an earlier study by Drakesmith et al., who showed that HFE expression in THP1 macrophages inversely correlated with FPN levels and iron release, whereas iron uptake by TfR1 was not affected.14 Other studies using mouse or human Hfe/HFE-deficient blood monocyte/macrophage cultures4,5 also demonstrated increased FPN expression in isolated cells. Similarly, reciprocal bone marrow15 and liver transplantation5 studies between wild-type and Hfe-/- mice suggested a diminished capacity of macrophages to store iron, being consistent with a function of HFE in inhibiting iron release.14 However, under all these conditions, the observed effects could not be dissociated from the systemic effects of hepcidin. In summary, our findings provide compelling new data on the long-standing question of the role of HFE in macrophages. Naveen Kumar Tangudu,1 Dilay Yilmaz,1 Katharina Wörle,2 Andreas Gruber,2 Silvia Colucci,3 Kerstin Leopold,2 Martina U. Muckenthaler3 and Maja Vujić Spasić 1 1 Institute of Comparative Molecular Endocrinology, Ulm University, Ulm; 2Institute of Analytical and Bioanalytical Chemistry, Ulm University, Ulm and 3Department of Pediatric Hematology, Oncology and Immunology, University of Heidelberg, Molecular Medicine Partnership Unit, Heidelberg, Germany Correspondence: MAJA VUJIĆ SPASIĆ - maja.vujic@uni-ulm.de

haematologica | 2021; 106(1)

doi:10.3324/haematol.2019.235630 Disclosures: no conflicts of interests to disclose. Contributions: conceptualization: MVS ; methodology: NKT, DY, KW, AG, SC and KL; investigation and formal analysis: NKT, KW, AG, SC, KL, MUM and MVS; original draft: MVS; review & editing: all authors contributed. Acknowledgments: we thank the staff of the Animal Facility at Ulm University and Prof. Gröne (DKFZ, Heidelberg, Germany) for their contribution to this work and Dr. S. Altamura for providing the samples from FpnKI mice. Funding: This work was supported by the German Research Foundation (DFG, VU75/2-1) and by Ulm University (support for MVS). MUM acknowledges funding from the DFG (SFB1036) and support from the late Dr. Andre Bachmann, who strongly believed in a beneficial role of the C282Y genotype.

References 1. Beutler E, Felitti VJ, Koziol JA, Ho NJ, Gelbart T. Penetrance of 845G--> A (C282Y) HFE hereditary haemochromatosis mutation in the USA. Lancet. 2002;359(9302):211-218. 2. Vujic Spasic M, Kiss J, Herrmann T, et al. Hfe acts in hepatocytes to prevent hemochromatosis. Cell Metab. 2008;7(2):173-178. 3. Nemeth E, Tuttle MS, Powelson J, et al. Hepcidin regulates cellular iron efflux by binding to ferroportin and inducing its internalization. Science. 2004;306(5704):2090-2093. 4. Jacolot S, Yang Y, Paitry P, Ferec C, Mura C. Iron metabolism in macrophages from HFE hemochromatosis patients. Mol Genet Metab. 2010;101(2-3):258-267. 5. Garuti C, Tian Y, Montosi G, et al. Hepcidin expression does not rescue the iron-poor phenotype of Kupffer cells in Hfe-null mice after liver transplantation. Gastroenterology. 2010;139(1):315-322.e1. 6. Wynn TA, Chawla A, Pollard JW. Macrophage biology in development, homeostasis and disease. Nature. 2013;496(7446):445-455. 7. Vinchi F, Costa da Silva M, Ingoglia G, et al. Hemopexin therapy reverts heme-induced proinflammatory phenotypic switching of macrophages in a mouse model of sickle cell disease. Blood. 2016;127(4):473-486. 8. Tangudu NK, Alan B, Vinchi F, et al. Scavenging Reactive Oxygen Species Production Normalizes Ferroportin Expression and Ameliorates Cellular and Systemic Iron Disbalances in Hemolytic Mouse Model. Antioxid Redox Signal. 2018;29(5):484-499. 9. Arezes J, Jung G, Gabayan V, et al. Hepcidin-induced hypoferremia is a critical host defense mechanism against the siderophilic bacterium Vibrio vulnificus. Cell Host Microbe. 2015;17(1):47-57. 10. Miller HK, Schwiesow L, Au-Yeung W, Auerbuch V. Hereditary Hemochromatosis Predisposes Mice to Yersinia pseudotuberculosis Infection Even in the Absence of the Type III Secretion System. Front Cell Infect Microbiol. 2016;6:69. 11. Wang S, Liu C, Pan S, et al. Deferoxamine attenuates lipopolysaccharideinduced inflammatory responses and protects against endotoxic shock in mice. Biochem Biophys Res Commun. 2015;465(2):305-311. 12. Nairz M, Theurl I, Schroll A, et al. Absence of functional Hfe protects mice from invasive Salmonella enterica serovar Typhimurium infection via induction of lipocalin-2. Blood. 2009;114(17):3642-3651. 13. Wang L, Johnson EE, Shi HN, et al. Attenuated inflammatory responses in hemochromatosis reveal a role for iron in the regulation of macrophage cytokine translation. J Immunol. 2008;181(4):2723-2731. 14. Drakesmith H, Sweetland E, Schimanski L, et al. The hemochromatosis protein HFE inhibits iron export from macrophages. Proc Natl Acad Sci U S A. 2002;99(24):15602-15607. 15. Makui H, Soares RJ, Jiang W, Constante M, Santos MM. Contribution of Hfe expression in macrophages to the regulation of hepatic hepcidin levels and iron loading. Blood. 2005;106(6):2189-2195.

263

Letters to the Editor

Factor IX alteration p.Arg338Gln (FIX Shanghai) potentiates FIX clotting activity and causes thrombosis Hemophilia B is almost exclusively caused by mutations of the F9 gene and inherited in an X-linked recessive manner.1 The amino acid residue arginine of factor IX (FIX) with CpG sequences2 is a hot spot for mutations,3 involving CG to TG or CA transitions and causing hemophilia B.4 Interestingly, for the coding sequence of residue Arg338 of FIX, only the transition CGA to TGA, which causes nonsense mutations, had been identified in a patient with severe hemophilia B,5 whereas the other type of CpG transition predicting the missense mutation p.Arg338Gln has been strangely absent from the list of F9 mutations reported in the hemophilia B database. It had been suspected that the particular type of amino acid replacement at this site may lead to conditions other than hemophilia.6,7 In a Chinese patient with deep vein thrombosis (DVT), we identified a F9 mutation at the same site as that in FIX Padua, but differently from this latter, the arginine is changed to a polar amino acid residue glutamine (p.Arg384Gln, HGVS numbering; or p.Arg338Gln, legacy numbering) due to the nucleotide transition c.1151G>A. A coagulation function assay with both the patient’s plasma and recombinant FIX p.Arg338Gln (p.R338Q) showed that the mutation also increased FIX activity. When the FIX with the p.R338Q modification was expressed in vivo after being transferred with self-complementary adeno-associated virus (scAAV) vectors, the deficiency of FIX in a mouse model of hemophilia B was restored, and the bleeding phenotype was significantly ameliorated. A 13-year old boy was referred to Ruijin Hospital for consultation regarding recurrent DVT. The patient was well and had no history of other major disorders. The two episodes of DVT had affected both femoral and popliteal veins of his left leg and the diagnosis was confirmed by ultrasound studies. He had no history of left leg trauma and no vascular abnormality was identified in the imaging study. There was no confirmed family history of venous thromboembolism. The patient was being treated with warfarin and this had not been discontinued during the study because of the potential risk of DVT recurrence. The study was approved by the ethics committee of Ruijin Hospital affiliated to Shanghai Jiaotong University School of Medicine. Written consent was acquired from family members and the patient’s mother on behalf of both parents. Under treatment with warfarin, the activity of three other K-dependent coagulation factors, factors II, VII and X, was significantly suppressed in the patient; however, his plasma maintained highly elevated FIX activity (FIX:C around 143.1% of normal). In addition, the plasma FIX antigen (FIX:Ag) level was much lower, at approximately 27.5% of normal. The

patient’s mother also had relatively higher FIX clotting activity (FIX:C 155.3%) and her FIX:Ag was 75.8% of normal (Table 1). Thus, the ratios between FIX:C and FIX:Ag were significantly higher for both the patient and his mother, being 5.20 and 2.06, respectively. The antithrombin III (AT III) level was within the normal range and lupus anticoagulant and antiphospholipid antibodies were negative in both the patient and his mother. Although we were unable to determine the patient’s protein C (PC) and protein S (PS) levels accurately, because of his treatment with warfarin, the levels of PC and PS were both within normal ranges in the mother (data not shown). The patient’s father was not available for analysis. The coding sequence and flanking regions of the F9 gene were amplified by polymerase chain reaction and directly sequenced. The genes coding for PS, PC, AT III and common genetic thrombotic risk factors, including the factor V Leiden mutation and prothrombin G20210G/A, were screened. Direct sequencing of the F9 gene revealed a nucleotide substitution c.1151G>A (Figure 1A), which predicated a missense mutation p.Arg338Gln (p.R338Q). The genetic variant was traced back to the patient’s mother (Figure 1B). This pedigree was negative for both factor V Leiden and the prothrombin G20210A mutation. Analysis of the genes for PS, PC and AT III failed to reveal any pathogenic changes (data not shown). The F9 cDNA containing the p.R338Q mutation were inserted into the mammalian expression vector pCDNA3.1 and transfected into HEK293T cells using lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) to produce FIX (the methods are described in the Online Supplementary File). The amount and coagulation activity of FIX were determined by enzyme-linked immunosorbent assay and an activated partial thromboplastin time-based clotting assay. Recombinant FIX was subjected to sodium dodecylsulfate polyacrylamide gel electrophoresis and detected by western blotting using a FIX-specific antibody (PAHFIX-S, Haematologic Technologies, Essex Junction, VT, USA). As shown in Figure 1C, the recombinant FIX Shanghai had a similar migration pattern as that of the wild-type FIX and the mutation increased the FIX activity by approximately 5fold as determined by the ratio of FIX:C to FIX:Ag (Figure 1D), which was similar to the change of FIX activity observed in the patient. We further investigated the expression and function of FIX p.R338Q in a murine model of hemophilia B using both high and low doses of recombinant adeno-associated virus (rAAV) vectors, with wild-type FIX and FIX p.R338L. The plasmid scAAV-TTR-hFIXco (AAV-hFIXco) was constructed by cloning codon-optimized human FIX (hFIXco) cDNA8 under the control of the human transthyretin (TTR) promoter in a scAAV vector plasmid, and with two inverted terminal repeat sequences of AAV2 flanking the human FIX gene expression cassettes.

Table 1. The coagulation profile of the patient.

Patient Patient’s mother

PT (s)

INR

APTT (s)

FIX:Ag

27.3 9.3

2.26 0.81

38.9 29.1

27.5 75.8

Coagulation factor antigen/activity (%) FIX:C FVIII:C FII:C FVII:C

FX:C

AT III (%)

143.1 155.3

10.2 96.5

91 97

126.8 155.9

22.5 93.6

29.6 118.5

PT: prothrombin time; INR: international normalized ratio; APTT: activated partial thromboplastin time; AT III: antithrombin III; F: factor; Ag: antigen; C: coagulation factor activity. Normal reference range for APTT: 25.1~39.5 s; PT: 9.9~15 s; coagulation factor activity: 50~150%; AT III activity: 84.6~120.2%.

264

haematologica | 2021; 106(1)

Letters to the Editor

The plasmids scAAV-TTR-hFIXcoR338Q (AAV-hFIXcoR338Q) and scAAV-TTR-hFIXcoR338L (AAV-hFIXcoR338L) were obtained by site-directed mutagenesis of the plasmid AAV-hFIXco at p.R338L and p.R338Q, respectively. Recombinant AAV type 8 (rAAV8) vectors were generated by triple plasmid transfection (methods provided in the Online Supplementary File). Two vector doses of approximately 5×1010 and 4×1011 rAAV8 viral genomes (vg), which were equivalent to 2×1012 or 1.6×1013 vg/kg in a clinical setting, were diluted in 200 mL of phosphatebuffered saline and injected into 4to 8-week old hemophilia B mice via the tail vein (n=5 mice per group). FIX:Ag and FIX:C were determined in the mice by enzyme-linked immunosorbent assay and an activated partial thromboplastin-based assay, using pooled normal human plasma as the standard. Twentyfour weeks after injection of AAV vectors packed with either of the FIX species, the FIX:C of FIX p.R338Q and p.R338L was three to five times higher than that of the wild-type FIX at most of the time points examined (P<0.05) (Figure 2A, D, G, J). The expression (FIX:Ag) of FIX p.R338Q was similar to that of FIX p.R338L, but lower than that of wild-type FIX, particularly at a low

dose (Figure 2B, E, H, K). The in vivo expression study validated that the FIX p.R338Q mutant had intrinsically higher clotting activity than wild-type FIX, with 4- to 5fold higher FIX-specific activity as determined in mice given either high or low doses of the vector (Figure 2C, F and I, L, respectively). The FIX clotting activity enhancement brought by p.R338Q was on average 20% less than that brought by p.R338L. The amino acid residue Arg338 has a profound effect on the procoagulant activity of FIX and laboratory studies on the artificial mutation FIX p.Arg338Ala showed that replacement of the positively charged arginine by a hydrophobic alanine could increase the FIX activity by 2~3-fold.9 In a pedigree with thrombophilia in Padua (Italy), replacement of the positively charged amino acid residue Arg338 with a nonpolar amino acid residue, leucine, led to an approximately 8-fold increase in FIX clotting activity.10 In our Chinese patient with thrombophilia, we found the long-sought CpG transition-related mutation c.1151G>A (p.Arg338Gln), which we designated FIX Shanghai. Although there have been contradictory reports on the mutation’s impact on clotting activity, pre-

Figure 1. F9 gene analysis and determination of FIX Shanghai clotting function. (A). A F9 gene variant c.1151G>A was identified in the patient, which predicted a missense mutation in factor IX (FIX) p.Arg338Gln (p.R338Q). (B) The mutation was inherited from the patient’s mother. (C) Subjected to electrophoresis, the FIX mutant p.R338Q expressed in HEK293 cells migrated in the same fashion as wild-type (wt) FIX on a sodium dodecylsulfate polyacrylamide gel. (D) Both recombinant FIX p.R338Q and FIX from the patient’s plasma showed more potent FIX activity, as represented by the increased ratio between FIX:C and FIX:Ag.

haematologica | 2021; 106(1)

265

Letters to the Editor

viously determined by using an artificially made mutant,11 the early onset and recurrences of DVT in our patient carrying FIX Shanghai evidently substantiate the effect of this mutation on upregulating coagulation activity. The mutation greatly increased the clotting activity of the patientâ&#x20AC;&#x2122;s plasma FIX, as did the recombinant mutant expressed in HEK293 cells, although on average 20% less than the FIX Padua mutation. Consistent with findings in

humans, hemophilia B mice into which the FIX Shanghai mutant was transferred by AAV vectors also presented with increased FIX activity in vivo. The dosage of AAV vectors is directly related to the safety and efficacy of a gene therapy for hemophilia B,12 and it would be ideal to achieve high FIX expression with lower doses of AAV vectors. Discovery of the FIX Padua mutation promoted the development of gene therapy.13

continued on the next page 266

haematologica | 2021; 106(1)

Letters to the Editor

continued from the previous page

Figure 2. In vivo evaluation of FIX Shanghai following delivery by adeno-associated virus vectors into hemophilic B mice. Evaluation of codon-optimized hFIXco, hFIXco-R338L and hFIXco-R338Q by recombinant adeno-associated virus (rAAV) delivery in hemophilia B (HB) mice. Each 4- to 8-week old HB mouse was injected with a total amount of 5x1010 vector genomes (low dose) or 4×1011 vector genomes (high dose) via the tail vein (n=5 in each group). Phosphate-buffered saline (PBS) was injected as a negative control (n=3). FIX clotting activity and protein quantity were measured by the activated partial thromboplastin time (APTT) (A, G) and enzyme-linked immunosorbent assay (ELISA) (B, H) using plasma samples collected at the indicated times after rAAV administration. A pooled human plasma was used as the normal standard. Specific activities (C, I) were calculated as APTT/ELISA at the corresponding points. Data are mean ± standard error of mean. Statistical analysis was done with analysis of variance testing using GraphPad Prism 6 (Graphpad Software, Inc. La Jolla, CA, USA). Representative results for A, B, C, G, H and I, at weeks 0, 2, 6 and 24 are shown in D, E, F, J, K and L, respectively. P values less than 0.05 were considered statistically significant (ns: >0.05; *P≤0.05; **P≤0.01; ***P≤0.001).

haematologica | 2021; 106(1)

267

Letters to the Editor

With high FIX clotting activity, the AAV vector carrying the F9 Padua variant achieved therapeutic levels of FIX in patients without toxic side effects.14 With a similar effect on FIX activity as FIX Padua, the AAV-delivered FIX Shanghai in a hemophilia B mouse model yielded sustained FIX expression, and the FIX clotting activity achieved in the mice by FIX Shanghai was significantly better than that of wild-type FIX and similar to that of FIX Padua. The adoption of a naturally occurring hypercoagulative FIX mutation, FIX Shanghai, is thus an attractive approach for the development of a gene therapy regimen for hemophilia B with high efficacy and fewer safety concerns. In conclusion, the new FIX variant identified in the current study, FIX Shanghai (FIX p.Arg338Gln), greatly increased FIX clotting activity and potentiated the risk of thrombosis. Before use in clinical trials, the potential toxicity of this variant should be studied further in animals. However, its high activity also makes it an ideal candidate for the development of novel gene therapy regimens for hemophilia B. Wenman Wu,1,2* Lin Xiao,3* Xi Wu,1 Xiaoling Xie,1 Ping Li,3 Changming Chen,1 Zhaoyue Zheng,3 Jiangang Ai,4 C. Alexander Valencia,5,6,7 Birong Dong,5 Qiulan Ding,1 Biao Dong3,5 and Xuefeng Wang1,2 1 Department of Laboratory Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; 2Faculty of Medical Laboratory Science, Ruijin Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, China; 3State Key Laboratory of Biotherapy, West China Hospital, Sichuan University and National Collaborative Innovation Center, Chengdu, Sichuan, China; 4Department of Laboratory Medicine, Jiangxi Provincial Peopleâ&#x20AC;&#x2122;s Hospital, Nanchang, China; 5National Clinical Research Center for Geriatrics, West China Hospital, Sichuan University, Chengdu, Sichuan, China; 6Lake Erie College of Osteopathic Medicine, Erie, PA, USA and 7Aperiomics, Inc., Sterling, VA, USA *WW and LX contributed equally as co-first authors.

Correspondence: XUEFENG WANG - wxf63@shsmu.edu.cn QIULAN DING - qiulan_ding@shsmu.edu.cn BIAO DONG - biaodong@scu.edu.cn doi:10.3324/haematol.2019.216713 Disclosures: no conflicts of interests to disclose. Contributions: WW designed experiments, analyzed data, and wrote the paper; LX execute animal work, analyzed data and wrote the paper; PL and ZZ prepared AAV vectors; XX, CC, JA expressed recombinant protein and performed function assay; AV wrote the paper; BD planed the experiments of AAV vectors; XW

268

performed the clotting function assays and FIX antigen determination of patient; XW and QD established the diagnosis of the patients, over sighted experiments and wrote the paper; BD planned and conducted the experiments of AAV vectors and wrote the paper. Funding: this work was supported by NSFC 81670130, WW; NSFC 81570115, XW; NSFC 81570114, QD; NSFC 81571792, BD.

References 1. Goodeve AC. Hemophilia B: molecular pathogenesis and mutation analysis. J Thromb Haemost. 2015;13(7):1184-1195. 2. Koeberl DD, Bottema CD, Buerstedde JM, Sommer SS. Functionally important regions of the factor IX gene have a low rate of polymorphism and a high rate of mutation in the dinucleotide CpG. Am J Hum Genet. 1989;45(3):448-457. 3. Vitkup D, Sander C, Church GM. The amino-acid mutational spectrum of human genetic disease. Genome Biol. 2003;4(11):R72. 4. Green PM, Montandon AJ, Bentley DR, Ljung R, Nilsson IM, Giannelli F. The incidence and distribution of CpG----TpG transitions in the coagulation factor IX gene. A fresh look at CpG mutational hotspots. Nucleic Acids Res. 1990;18(11):3227-3231. 5. Ludwig M, Schwaab R, Eigel A, et al. Identification of a single nucleotide C-to-T transition and five different deletions in patients with severe hemophilia B. Am J Hum Genet. 1989;45(1):115-122. 6. Gostout B, Vielhaber E, Ketterling RP, et al. Germline mutations in the factor IX gene: a comparison of the pattern in Caucasians and non-Caucasians. Hum Mol Genet. 1993;2(3):293-298. 7. Bottema CDK, Ketterling RP, Vielhaber E, et al. The pattern of spontaneous germ-line mutation: relative rates of mutation at or near CpG dinucleotides in the factor IX gene. Hum Genet. 1993; 91(5):496-503. 8. Nathwani AC, Gray JT, Ng CY, et al. Self-complementary adenoassociated virus vectors containing a novel liver-specific human factor IX expression cassette enable highly efficient transduction of murine and nonhuman primate liver. Blood. 2006;107(7):2653-2661. 9. Chang J, Jin J, Lollar P, et al. Changing residue 338 in human factor IX from arginine to alanine causes an increase in catalytic activity. J Biol Chem. 1998;273(20):12089-12094. 10. Simioni P, Tormene D, Tognin G, et al. X-linked thrombophilia with a mutant factor IX (factor IX Padua). N Engl J Med. 2009; 361(17):1671-1675. 11. Monahan PE, Sun J, Gui T, et al. Employing a gain-of-function factor IX variant R338L to advance the efficacy and safety of hemophilia B human gene therapy: preclinical evaluation supporting an ongoing adeno-associated virus clinical trial. Hum Gene Ther. 2015;26(2):6981. 12. Manno CS, Arruda VR, Pierce GF, et al. Successful transduction of liver in hemophilia by AAV-factor IX and limitations imposed by the host immune response. Nat Med. 2006;12(3):342-347. 13. Lozier JN. Factor IX Padua: them that have, give. Blood. 2012; 120(23):4452-4453. 14. George LA, Sullivan SK, Giermasz A, et al. Hemophilia B gene therapy with a high-specific-activity factor IX variant. N Engl J Med. 2017;377(23):2215-2227.

haematologica | 2021; 106(1)

Letters to the Editor

Impact of graft composition on outcomes of haploidentical bone marrow stem cell transplantation Haploidentical hematopoietic stem cell transplantation (haplo-HSCT) is a therapeutic option for patients without an HLA-matched donor and several studies recently demonstrated comparable outcomes following these or HLA-matched transplants.1-3 Multiple haploidentical donors are often available, including parents, children, and half-matched siblings. While influences of several donor characteristics on post-transplant outcomes have been extensively evaluated, the impact of bone marrow cellular graft composition on patients receiving post-transplantation cyclophosphamide-based graft-versus-host disease (GvHD) prophylaxis has not been studied. We hypothesized that graft composition independently affects the outcomes of haplo-HSCT and evaluated graft, donor, and recipient characteristics as predictors of outcomes in a homogeneous population of 147 patients ≥18 years old who received a fresh bone marrow graft between February 2009 and August 2015 at the MD Anderson Cancer Center (MDACC). Validation was done in a subsequent cohort of 111 patients treated between August 2015 and May 2019. All patients received the same melphalan-based conditioning regimen, post-transplantation cyclophosphamide-based GvHD prophylaxis and standard antimicrobial prophylaxis, as previously described.4 The MDACC Institutional Review Board approved this retrospective study. Univariable analysis using Cox proportional hazards

regression analysis and Fine and Gray competing-risks regression was performed to evaluate donor, recipient, disease, and transplant characteristics and graft cellular characteristics including CD34+, total nucleated cells (TNC), CD3+CD4+, CD3+CD8+, CD3+CD4+/CD3+CD8+, CD19+, and CD3-CD56+ cell populations for associations with outcomes. Predictors that were significant at the 0.1 level on univariable analysis were considered for multivariable analysis using classification and regression tree (CART) analysis to classify donor, recipient, and graft characteristics in order of their statistical impact and identify interaction effects among these three categories of predictors. For the internal validation, we performed bootstrapping analysis to estimate bias-corrected confidence intervals around the relative risk assessing the association between CD3+CD4+/CD3+CD8+ ratio and outcomes based on 1,000 resampled datasets. Details of graft cellular assessment and statistical methods are summarized in Online Supplementary Material 1. Most patients (54%) underwent haplo-HSCT for acute myeloid leukemia and myelodysplastic syndrome. Forty-three percent of patients had a high or very high disease risk index (DRI). Most patients (73%) had a hematopoietic cell transplantation comorbidity index score ≤3. The patients’ characteristics are summarized in Online Supplementary Material 2. The median follow-up of survivors was 37 (range, 780) months. Transplantation outcomes are summarized in Online Supplementary Material 3. Associations between outcomes and donor and recipient characteristics are presented in Online Supplementary Material 4. The median CD4+/CD8+ cell ratio (CD4/CD8) was 1.1

Table 1. Correlation between graft cellular subsets and donor characteristics evaluated in univariate analysis.

Characteristics

Median Median N. of patients CD34+ TNC dose, (n=147) cell dose, x106/kg (IQR) x106/kg (IQR)

Age, years >30

≤30

>50

≤50

120

Gender Female

Male

CMV serostatus 94 Reactive Nonreactive 53

% CD8+ % CD4+ cells cells Median P P Median (IQR) value (IQR) value

2.5 (1.7, 3.4) 2.5 (2, 3.4) 2.1 (1.5, 2.7) *2.6 (1.9, 3.5)

319 (264, 407) 329 (236, 400) 315 (229, 404) 321 (256, 404)

41 (37, 47) 38 (35, 44) 43 (38, 51) 39 (35, 44)

0.04

2.3 (1.9, 3.1) 2.6 (1.7, 3.7)

315 (264, 378) 332 (249, 410)

42 (37, 47) 39 (33, 44)

0.06

2.5 (1.8, 3.5) 2.4 (1.6, 3)

326 (268, 406) **295 (209, 375)

40 (36, 46) 41 (35, 46)

0.9

0.01

35 (30, 42) 35 (30, 42) 36 (30, 45) 35 (30, 42)

0.9

35 (31, 41) 35 (30, 44)

0.9

CD4/CD8 ≤0.8 P % value

0.4

0.7

0.003 37 (31, 44) 33 (28, 37)

0.04

0.1

0.04

18 19

8 0.4 8 (6, 13) (6, 13) 9 9 (6, 12) (6, 15) 9 <0.01 6 (7, 13) (3, 8) 9 10 (6, 12) (7, 14)

0.8

0.3

0.7

9 (6, 13) 8 (7, 13)

0.7

8 (6, 12) 10 (6, 13)

0.1

0.9

8 (6, 13) 9 (7, 14)

0.7

8 (6, 11) 10 (7, 13)

0.06

0.04

24 10

29 0.8

% CD56+ % CD19+ CD4/CD8 cells cells >1.5 Median P Median P P % value (IQR) value (IQR) value

IQR: interquartile range; TNC: total nucleated cells; CD4/CD8: CD4+/CD8+ cell ratio; CMV: cytomegalovirus. *P=0.04. **P=0.02 for comparison of 25th percentile

haematologica | 2021; 106(1)

269

Letters to the Editor

A Figure 1. Legend on following page.

excludes outside values

270

haematologica | 2021; 106(1)

Letters to the Editor

Figure 1. Distributions of CD4+ and CD8+ cell percentages and risk stratifications based on graft, donor and recipient-characteristics. (A) The distributions of %CD4+ and %CD8+ cells in unmanipulated bone marrow grafts vary according to CD4+/CD8+ cell ratio (CD4/CD8). Distributions of %CD4+ and %CD8+ cells are skewed in low (≤0.85) and high (>1.5) CD4/CD8 quartiles. CD8+ cells are predominant in grafts with a CD4/CD8 ≤0.85 and CD4+ cells are predominant in grafts with a CD4/CD8 >1.5. Data are illustrated in a box-and-whisker plot with the whiskers indicating the 25th and 75th percentiles. Lines within the box plots indicate the median %CD4+ and %CD8+ cells in each CD4/CD8 quartile. (B-F) Results of multivariate classification and regression tree (CART) analysis are shown for severe acute graft-versus-host disease (B), early non-relapse mortality (C), late non-relapse mortality (D), disease progression (E), and progression-free survival (F). Two figures are presented for each outcome depicting the risk classification algorithm (left figure) generated by CART analysis, and the corresponding cumulative incidence curves (right figure) for each subgroup represented in the risk stratification algorithm. N: number; aGVHD: acute graft-versus-host disease; %CI: percent cumulative incidence; NRM: non-relapse mortality; HCT-CI: Hematopoietic Cell Transplantation - Comorbidity Index; CMV: cytomegalovirus serostatus; DRI: Disease Risk Index; prg: progressive disease; TNC: total nucleated cells; PFS: progression-free survival.

(range, 0.4-3.1). As shown in Figure 1A, grafts with CD4/CD8 ≤0.85 had higher %CD8+ and grafts with a CD4/CD8 >1.5 had higher %CD4+. The %CD4+ and %CD8+ distributions were more balanced in grafts with a CD4/CD8 between 0.85 and 1.5. Correlations between donor characteristics (age, gender, and cytomegalovirus status) and graft composition are presented in Table 1. Donor age >30 years was correlated with higher %CD4+ (median 41% vs. 38%, P=0.04) and CD4/CD8 >1.5 (24% vs. 10%, P=0.04), whereas donor age >50 years was correlated with lower %CD19+ (median 6% vs. 10%, P=0.0003) and lower infused CD34+ cell dose (median 2.6 vs. 2.1, P=0.04). Grafts from female donors had higher %CD4+ (median 42% vs. 39%, P=0.06) and were more likely to have a CD4/CD8 >0.85 (83% vs. 68%, P=0.04). Overall, cytomegalovirus seropositivity correlated with higher %CD8+ (median 37% vs. 33%, P=0.003), a CD4/CD8 ≤0.85 (median 31% vs. 15%, P=0.04), and an infused TNC dose >250 x 106/kg (81% vs. 64%, P=0.02). Multivariable CART analyses have shown significant associations of recipient and donor characteristics and graft cellular composition with post-transplant outcomes, and generated risk groups with different posttransplant outcomes, as summarized in Table 2 and Figure 1B-F. haematologica | 2021; 106(1)

The incidence of severe acute GvHD ranged from 0% to 54% (Figure 1B). The highest risk of severe acute GvHD was associated with grafts from female donors, age >30 years with ≤6% CD56+ cells in the graft (n=11, cumulative incidence [CI] 54%). The incidence of early non-relapse mortality (NRM), defined as <60 days after transplantation, was 0-22% (Figure 1C). Grafts with a low CD4/CD8 (<0.85) and low infused CD19+ cell dose (≤2.5x106 CD19+ cells/kg) were associated with a higher early NRM rate. Notably, all cases of early NRM were due to infections. The incidence of late NRM, defined as >60 days after transplantation, ranged from 9% to 58% (Figure 1D). The high-risk group for late NRM included three subgroups: (i) recipients of grafts with a CD4/CD8 >1.5; (ii) patients aged >60 years who received grafts with a CD4/CD8 ≤1.5 from cytomegalovirus-seropositive donors; and (iii) patients with comorbidity scores >3 who received grafts with a CD4/CD8 ≤1.5 from cytomegalovirus-seropositive donors. Late NRM was primarily related to GvHD. The incidence of disease progression ranged from 0% to 73% (Figure 1E). Three risk groups for disease progression were defined based on DRI, TNC dose, and CD4/CD8. The group at highest risk of progression included patients with a high/very-high DRI whose 271

Letters to the Editor

grafts had either a low TNC dose (relapse 67%) or a low (≤0.85) CD4/CD8 (relapse 73%). The progression-free survival (PFS) rate ranged from 6% to 87% (Figure 1F). Three risk groups were defined based on DRI, TNC dose, CD4/CD8, and CD56+ cell dose. Patients with low/intermediate DRI and grafts with a balanced (>0.85-1.5) CD4/CD8 and >2.5x106 CD56+ cells/kg had the highest PFS rate (87%, hazard

ratio [HR]=0.2, P=0.007). Patients (n=39) with high/very-high DRI and either a graft with low TNC/kg or a graft with an unbalanced (≤0.85 or >1.5) CD4/CD8 had the lowest PFS rate (11%, HR=2.9, P<0.001). The remaining patients (n=84, reference group) had an intermediate PFS rate of 48%. To validate our findings, we tested the predictive value of CD4/CD8 in a subsequent cohort of 111

Table 2. Cellular subsets and donor and recipient characteristics as predictors of outcomes in multivariate classification and regression tree analysis.

Outcome Severe acute GvHD Female donor >30 y <6% CD56+ cells ≥6% CD56+ cells Male donor any age or female donor ≤30 y CD4/CD8 >1.5 CD4/CD8 ≤1.5, female donor CD4/CD8 ≤1.5, male donor Early non-relapse mortality ≤2.5x106 infused CD19+ cells/kg CD4/CD8 ≤0.85 CD4/CD8 >0.85 >2.5x106 infused CD19+ cells/kg Late non-relapse mortality CD4/CD8 >1.5 CD4/CD8 ≤1.5 Recipient >60 y, donor CMV R Recipient >60 y, donor CMV NR Recipient ≤60 y, HCT-CI >3, donor CMV R Recipient ≤60 y, HCT-CI >3, donor CMV NR Recipient ≤60 y, HCT-CI ≤3 Disease progression High vs. high DRI ≤245x106 TNC/kg >245x106 TNC/kg and CD4/CD8 ≤0.85 OR >1.5 >245x106 TNC/kg and CD4/CD8 >0.85-1.5 Intermediate or low DRI CD4/CD8 ≤0.85 CD4/CD8 >0.85, ≤8% CD56+ cells CD4/CD8 >0.85, >8% CD56+ cells Progression-free survival* High or very high DRI ≤245x106 TNC/kg >245x106 TNC/kg, CD4/CD8 ≤0.85 OR >1.5 >245x106 TNC/kg, CD4/CD8 >0.85-1.5 Low or intermediate DRI CD4/CD8 ≤0.85 OR >1.5 CD4/CD8 >0.85-1.5, ≤2.5x106 CD56+ cells/kg CD4/CD8 >0.85-1.5, >2.5x106 CD56 cells/kg

%CI

HR (95% CI)

11 29

54% 21%

Reference 0.3 (0.1-0.9)

0.03

High Intermediate

17 23 66 147

12% 9% 0%

0.2 (0.03-0.8) 0.1 (0.02-0.6) NA

0.03 0.01 <0.01

Intermediate Intermediate Low

18 53 76 139 25

22% 4% 0%

Reference 0.1 (0.03-0.8) NA

0.03 <0.01

High Low Low

49%

0.5 (0.2-1.5)

0.3

High

12 10 13 11 68 146

58% 10% 40% 9% 10%

Reference 0.1 (0.01-0.9) 0.5 (0.1-1.6) 0.1 (0.01-0.7) 0.1 (0.03-0.3)

0.04 0.2 0.02 <0.001

High Low High Low Low

18 12 31

73% 67% 37%

Reference 0.9 (0.4-2.1) 0.4 (0.2-0.9)

0.8 0.02

High High Intermediate

20 27 38 146

30% 11% 0%

0.3 (0.1-0.8) 0.1 (0.03-0.4) NA

0.02 <0.01 <0.01

Intermediate Low Low

18 21 22

6% 16% 33%

19 (5.5-66) 13 (3.7-43) 6.6 (1.9-23)

<0.01 <0.01 0.003

High High Intermediate

36 26 23

46% 61% 87%

5.4 (1.6-18) 3.5 (0.97-13) Reference

0.007 0.054

Intermediate Intermediate Low

N. of patients

Proposed Risk Stratification

146

%CI: percent cumulative incidence; HR: hazard ratio; 95% CI: 95% confidence interval; GvHD: graft-versus-host disease; y: years; CD4/CD8: CD4+/CD8+/ cell ratio, NRM: nonrelapse mortality; CMV: cytomegalovirus; R: reactive; NR: non-reactive; HCT-CI: Hematopoietic Cell Transplantation - Comorbidity Index; DRI: Disease Risk Index; TNC: total nucleated cells. *PFS proportion, rather than %CI, is provided.

272

haematologica | 2021; 106(1)

Letters to the Editor

patients treated in the same way as the training cohort. Demographic data are detailed in Online Supplementary Material 2. Results from the validation cohort were consistent with those from the training study: predominance of either CD4+ or CD8+ cells was associated with adverse outcomes, whereas a balanced distribution (0.91.1) was associated with superior PFS. CD4/CD8 ≤0.9 was significantly associated with higher relapse rate (HR=2.5, P=0.04), while CD4/CD8 >1.1 was associated with a significantly higher rate of NRM (HR=2.5, P=0.03) and lower rate of relapse (HR=0.4, P=0.03), in univariate analysis. These trends were independent of the DRI. Irrespective of the DRI, a CD4/CD8 between 0.9-1.1 was associated with superior PFS; however, the association did not reach statistical significance (HR=0.6, P=0.2). In this study, we found that the composition of unprocessed bone marrow grafts had a major independent impact on the outcomes of patients who underwent haplo-HSCT with post-transplantation cyclophosphamide-based GvHD prophylaxis, and identified an optimal balanced CD4/CD8 associated with the longest PFS. Grafts with low or high CD4/CD8 were associated with inferior survival and different patterns of failure. Previous studies of haplo-HSCT with granulocyte colony-stimulating factor-primed bone marrow and peripheral blood stem cells and antithymocyte globulinbased GvHD prophylaxis showed that a low CD4/CD8 is associated with adverse outcomes.5,6 Our findings suggested that the predominance of either a CD4+ or CD8+ cell population, which translated into high and low (inverted) CD4/CD8 ratios, respectively, was associated with inferior outcomes. The predominance of CD8+ cells was associated with a high incidence of early NRM and disease progression, whereas the predominance of CD4+ cells was associated with a high incidence of severe acute GvHD and late NRM. These findings were confirmed in a bootstrap analysis and in a separate cohort of patients subsequently treated at our institution with the same conditioning and a fresh bone marrow graft. A better PFS was again noted in the validation cohort for grafts with a “balanced” CD4/CD8 ratio (0.9-1.1) (although this association did not reach statistical significance in the confirmatory group, likely due to lower number of patients and change in practice over time to avoid older female donors). The adverse effects associated with a predominance of either a CD4+ or CD8+ cell population could be interpreted in the context of the required cooperation between these cell subsets in mounting an effective immune response.7-11 A direct anti-tumor effect for CD4+ cells, independently of CD8+ cells, has also been suggested.7-9,12-14 Our data also indicate that CD4/CD8 has significant synergistic effects with CD19+ cells in early NRM and a significant role for CD4/CD8 and natural killer cells in disease progression. Specifically, for early NRM, a low CD4/CD8 was associated with higher NRM only in recipients whose grafts had a low dose of infused CD19+ cells. Grafts with low doses of infused CD19+ cells and a CD4/CD8 ≤0.85 represented 12% of the graft pool and yet they accounted for 67% of the cases of early NRM. In the case of disease progression, for recipients with a low or intermediate DRI, the antitumor effect associated with a CD4/CD8 >0.85 was amplified by concurrent high levels of natural killer cells in the graft. This is in line with recent evidence suggesting a synergistic antitumor effect between CD4+ and natural killer cells.9 In conclusion, we found that donor graft composition haematologica | 2021; 106(1)

is a major independent determinant of transplant outcomes in patients receiving haploidentical bone marrow transplantation. A “balanced” graft with an optimal CD4/CD8 ratio close to 1 was associated with the best transplant outcomes. These findings have implications not only for donor selection but also for risk mitigation and suggest that customized grafts with optimal CD4/CD8 ratios, and CD19+ and CD56+ cell numbers should be explored in the future. Rima M. Saliba,1 Lauren Veltri,1,° Gabriela Rondon,1 Julianne Chen,1 Gheath Al-Atrash,1 Amin Alousi,1 Charles Martinez,1 LaJerald Augustine,2 Chitra M. Hosing,1 Betul Oran,1 Katayoun Rezvani,1,2 Elizabeth J. Shpall,1,2 Partow Kebriaei,1 Issa F. Khouri,1 Uday Popat,1 Richard E. Champlin1 and Stefan O. Ciurea1 1 Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center, Houston, TX and 2Cell Therapy Laboratory, The University of Texas MD Anderson Cancer Center, Houston, TX, USA ° Current affiliation: Department of Hematology/Oncology, West Virginia University, Morgantown, WV, USA Correspondence: STEFAN O. CIUREA - sciurea@mdanderson.org doi:10.3324/haematol.2019.227371 Disclosures: no conflicts of interests to disclose. Contributions: RMS conceived the study, performed statistical analysis, and wrote the manuscript. LV contributed with some data collection and manuscript writing. GR, JC, CM, LA contributed with data collection. GA, SA, AA, CMH, BO, KR, EJS, PK, IFK, UP, REC contributed with patient care, critical review of the manuscript. SOC contributed with study design, data collection, interpretation of results and manuscript writing. All authors reviewed and approved the submission. Acknowledgments: we acknowledge Joseph Munch from the Department of Scientific Publications at MD Anderson Cancer Center for his help with editing this manuscript.

References 1. Ciurea SO, Zhang MJ, Bacigalupo AA, et al. Haploidentical transplant with posttransplant cyclophosphamide vs matched unrelated donor transplant for acute myeloid leukemia. Blood. 2015; 126(8):1033-1040. 2. Kanate AS, Mussetti A, Kharfan-Dabaja MA, et al. Reducedintensity transplantation for lymphomas using haploidentical related donors vs HLA-matched unrelated donors. Blood. 2016;127(7):938-947. 3. Bashey A, Zhang X, Sizemore CA, et al. T-cell-replete HLA-haploidentical hematopoietic transplantation for hematologic malignancies using post-transplantation cyclophosphamide results in outcomes equivalent to those of contemporaneous HLAmatched related and unrelated donor transplantation. J Clin Oncol. 2013; 31(10):1310-1316. 4. Gaballa S, Ge I, El Fakih R, et al. Results of a 2-arm, phase 2 clinical trial using post-transplantation cyclophosphamide for the prevention of graft-versus-host disease in haploidentical donor and mismatched unrelated donor hematopoietic stem cell transplantation. Cancer. 2016;122(21):3316-3326. 5. Luo XH, Chang YJ, Xu LP, Liu DH, Liu KY, Huang XJ. The impact of graft composition on clinical outcomes in unmanipulated HLA-mismatched/haploidentical hematopoietic SCT. Bone Marrow Transplant. 2009;43(1):29-36. 6. Xu LP, Luo XH, Chang YJ, et al. High CD4/CD8 ratio in allografts predicts adverse outcomes in unmanipulated HLA-mismatched/haploidentical hematopoietic stem cell transplantation for chronic myeloid leukemia. Ann Hematol. 2009;88(10):10151024. 7. Mumberg D, Monach PA, Wanderling S, et al. CD4(+) T cells eliminate MHC class II-negative cancer cells in vivo by indirect

273

Letters to the Editor

effects of IFN-gamma. Proc Natl Acad Sci U S A. 1999;96 (15):8633-8638. 8. Muranski P, Restifo NP. Adoptive immunotherapy of cancer using CD4(+) T cells. Curr Opin Immunol. 2009;21(2):200-208. 9. Perez-Diez A, Joncker NT, Choi K, et al. CD4 cells can be more efficient at tumor rejection than CD8 cells. Blood. 2007;109(12):5346-5354. 10. Ferguson FG, Wikby A, Maxson P, Olsson J, Johansson B. Immune parameters in a longitudinal-study of a very old population of Swedish people - a comparison between survivors and nonsurvivors. J Gerontol A Biol Sci Med Sci. 1995;50(6):B378B382.

274

11. Pawelec G, Ferguson FG, Wikby A. The SENIEUR protocol after 16 years. Mech Ageing Dev. 2001;122(2):132-134. 12. Corthay A, Skovseth DK, Lundin KU, et al. Primary antitumor immune response mediated by CD4(+) T cells. Immunity. 2005; 22(3):371-383. 13. Dudley ME, Wunderlich JR, Robbins PF, et al. Cancer regression and autoimmunity in patients after clonal repopulation with antitumor lymphocytes. Science. 2002;298(5594):850-854. 14. Hunder NN, Wallen H, Cao JH, et al. Treatment of metastatic melanoma with autologous CD4+T cells against NY-ESO-1. New Engl J Med. 2008;358(25):2698-2703.

haematologica | 2021; 106(1)

Letters to the Editor

T-cell immune response controls the high incidence of adenovirus infection in adult allogenic hematopoietic transplantation recipients

Table 1. Characteristics of the human adenovirus (HAdV) viremia episodes depending on the level of the HAdV DNAemia.

Human adenovirus (HAdV) can produce disseminated disease with high mortality in pediatric allogenic hematopoietic stem cell transplant (allo-HSCT) recipients. In adult recipients, prospective studies addressing the incidence of HAdV infection and disease are lacking. The aim of this study was to characterize the kinetics of HAdV replication, the specific T-cell immune response, and the clinical impact of HAdV infection in adult alloHSCT recipients. We found a high incidence of HAdV viremia, which was controlled by the T-cell immune response and without association with disseminated HAdV disease. Therefore, these data argue against the pre-emptive therapy for HAdV infection in adult alloHSCT recipients. The high variability on the reported rates of HAdV infection in HSCT recipients reflects the gap of knowledge regarding the natural history of the HAdV infection in these patients.1-5 Although the overall mortality of HAdV infections is low in adults (1%), life-threatening disease has been associated with HAdV DNAemia.5-7 However, this statement derives from studies made in pediatric allo-HSCT77 or retrospective evaluations of pediatric and adult patients selected after disseminated HAdV disease diagnosis.6 In cases of disseminated disease, lethality rates reach up to 60-80% in children and adults.5,7,8 One hundred seventeen patients received an alloHSCT from February 2016 to May 2018. Ninety-five of them (>18 years old) fulfilled the study requirements and were included (see the Online Supplementary Materials and Methods). Demographics and clinical characteristics of the patients are detailed in the Online Supplementary Table S1. Fifty-eight (61.1%) patients showed HAdV viremia episodes, all cases belonging to the HAdV species C. Most episodes were transient viremia or blips, defined as HAdV detected in one isolated sample (59 of 80, 73.8%), with a median of one episode per patient. HAdV viremia occurred early (weeks -1 to +7) and blips earlier than persistent episodes (median 3 weeks [range: 1-5] vs. 7 weeks [range: 0-10], respectively). Viral loads showed wide variability, with lower levels observed in blips versus persistent episodes (1.5x102 [range: 8.1x101-5.4x102] versus 9.8x102 [range: 2.1x102 -8.3x104] copies/mL), and a median duration of 14 days in persistent episodes. Thirty-six (37.9%) patients had HAdV viremia <5x102 copies/mL and 22 (23.2%) ≥5x102 DNA copies/mL. The differences of viremia episodes between both groups are detailed in (Table 1). Six of 22 patients with ≥5x102 copies/mL had blips previously. The incidence of HAdV viremia was similar to that reported in pediatric patients5 but higher than those observed in adult allo-HSCT recipients (15-19.7%).1,2 In agreement with other reports,9,10 blips and persistent viremia episodes were detected early in the follow-up, suggesting that the HAdV viremia may arise from reactivation of the virus. This data could support the implementation of a HAdV pre-transplant screening, as has been already suggested.10 In adult allo-HSCT recipients, postulated risk factors for HAdV infection and disease include haploidentical or unrelated donor cord transplant, graft-versus-host disease (GvHD) of grade 3/4, and treatment with alemtuzumab or anti-timocytic globulin (ATG).11 We did not find any

Patients 36 (37.9) 22 (23.2) Episodes per patient (median, IQR) 1 (1-2) 1 (1-1) Blips episodes 59 (73.8) 18 (75.0) Week of follow-up (median, IQR) 3 (1-5) 5 (2-8) Maximum viral load 1.3x102 3.2x103 (copies/mL; median, IQR) (7.9x101-2.1x102) (1.1x103-2.1x104) Persistent episodes 15 (26.8) 6 (25.0) Week of follow-up (median, IQR) 6 (-1-11) 7 (1-8) Duration of viremia 14 (7-14) 14 (14-14) (days; median, IQR) Maximum viral load 1.9x102 2.0x104 (copies/mL; median, IQR) (8.1x101-2.6x102) (1.8x103-7.3x105)

haematologica | 2021; 106(1)

Low DNAemia* High DNAemia** n (%) n (%)

*DNAemia < 5x102 copies/mL; **DNAemia ≥ 5x102 copies/mL; IQR: interquartile range.

association between HAdV viremia and these risk factors (Online Supplementary Table S2). Regarding the ATG use, it is important to point out that this result may be due to the low number of patients treated (10,5%). The use of rapamycin for GvHD prophylaxis in patients with a reduced intensity conditioning (20% of the patients treated with ATG and 60% of the whole series) may be another explanation for this lack of association. Results from our laboratory not yet published show an antiHAdV activity of rapamycin, with an IC of 2.3 mM (selective index of 55), as reported for human cytomegalovirus (HCMV).12 The lack of association with any demographic data, transplantation or clinical characteristics was independent of the level of HAdV viremia (data not shown). In this study no patient developed symptomatic or disseminated HAdV disease nor received specific HAdV therapy during the follow-up. In addition, there was no chronologic association between HAdV viremia and the presence of other infections. Ten patients died due to different causes: five patients directly related to the transplantation; two patients of GvHD, two patients of thrombotic microangiopathy and one patient of venoocclusive disease). Three patients had other infections: two patients with septic shock from pneumonia and Clostridioides difficile infection and one patient with Aspegillus fumigatus pneumonia). One patient relapsed with hematologic disease and one patient with multifactorial pulmonary fibrosis. There was not difference in the mortality between patients with (n=5, 8.6%) or without (n=5, 13.5%) HAdV viremia (Online Supplementary Table S2). The data of the present study does not support, in adult allo-HSCT patients, the threshold of ≥5x102 HAdV DNA copies/mL in blood as a marker of invasive HAdV infection,9,13 since no patient developed disseminated HAdV disease, independently of the HAdV viremia level. The HAdV-specific T-cell immune response was evaluated (Online Supplementary Materials and Methods) in 90 out of the 95 patients (in five patients there was loss of follow-up in specific time points). The number of patients with specific HAdV-specific T-cell immune response increased during the follow-up: 30 (33.3%), 65 (72.2%) and 78 (86.7%) of 90 patients at days +21, +56 and +100 after transplantation, respectively. Twelve 50

275

Letters to the Editor

(13.3%) of the 90 patients did not develop HAdV-specific a T-cell immune response during the follow-up; three (25.0%) of them had one HAdV viremia blip episode. The number of patients with CD4+ T cells expressing IL-2 by day +56 (P<0.05) and CD8+ T cells expressing IL-2 (P<0.05) and TNF-α (P<0.05) by day +21, was higher in those with versus without HAdV viremia (Online Supplementary Table S3). When the HAdV-specific T-cell immune response was compared between both groups higher expressions of IL-2 by CD4+ T cells, and IL-2 and TNF-α by CD8+ T cells were observed by days +56 and +21, respectively (Online Supplemental Figure S4). In contrast, when we compared the HAdV-specific T-cell

Media of viral load (copies/mL)

Media of viral load (copies/mL) Media of viral load (copies/mL)

Media of viral load (copies/mL)

immune response in patients with HAdV viremia higher versus lower than 5x102 copies/mL the maximum cytokine expression was observed in CD4+ T cells expressing TNF-α by day +100 (0.2% vs. 0.13%, P<0.05) and higher expressions of IL-2 by CD4+ T cells and INFγ by CD8+ T-cells were observed by day +100 after the transplantation in patients with HAdV viremia ≥5x102 DNA copies/mL (Figure 1, Online Supplementary Table S5). Interestingly, patients with pre-transplantation (days -7 and 0) persistent HAdV viremia (n=4) had CD4+ expression of INF-γ (P=0.05) and CD8+ expression of IL-2, TNF-α and INF-γ (P<0.05) higher and/or earlier (at +21 and +56 days) than those with blips (n=12); this

Figure 1. Human adenovirus-specific CD4+ T cells expressing IL-2 and TNF-α and CD8+ T cells expressing IFN-γ, on days +21, +56 and +100 after transplantation. (A) Patients with viremia ≥5x102 copies/mL. (B) Patients with viremia <5x102 copies/mL.

276

haematologica | 2021; 106(1)

Letters to the Editor

immune response paralleled with almost complete disappearance of viremia episodes (Figure 2). Finally, the percentage of CD4+ and CD8+ T cells and the CD4+/CD8+ T-cell ratio were similar in patients with and without HAdV viremia and independent of the level of viremia (Online Supplementary Figure S1). In this study, we have characterized the HAdV-specific cellular immune response to identify the most appropriate immunologic marker to guide the management of HAdV infections in adult allo-HSCT recipients. Unlike Guérin-El Khourouj et al.,14 who reported that HAdV-specific CD4+ IFN-γ monitoring may be of help for the detection of high-risk allo-HSCT patients,14 we found that the expression of IL-2 and TNF-α by CD8+ T cells were the earlier markers in HAdV viremia patients. Moreover, specific CD4+ and CD8+ T-cell immune responses increased with a higher HAdV load in the blood. One interesting finding in the present study was the almost complete clearance of HAdV viremia, after 3 weeks of follow-up, in patients with persistent viremia before transplantation, which had higher expressions of INF-γ by CD4+ and of IL-2, TNF-α and INF-γ by CD8+ T cells, although the small number of these patients only permit to suggest this association. A limitation of our study is the lack of specific HAdV T-cell immune response evaluation in donors, which precludes to know if it has some impact on the kinetics and control of HAdV replication after the transplantation. Currently, no definitive recommendations for the monitoring of HAdV viral loads and specific T-cell

Copies/mL

immune reconstitution as a basis for prophylaxis and therapy of HAdV infections in adult patients excist. In the pediatrics setting, the recommendations are limited to the threshold of HAdV DNAemia ≥1x103 copies/mL for the initiation of pre-emptive therapy, which is derived in most cases from retrospective and observational stu-dies.3,15 A recent study suggests that the pretransplant detection of ≥1x106 HAdV DNA in stool is correlated with a higher risk of invasive HAdV infection.13 In a European survey on the management of HAdV infection in allo-HSCT, physicians initiate preemptive therapy in adult patients with cidofovir in 29% and 14% of high and low-risk patients, respectively.15 The high adverse effects observed with cidofovir represent a challenge for this decision in the absence of clinical data from prospective studies supporting the prediction value of viremia on disseminated HAdV disease in adult patients. In summary, our results show that HAdV infection is a frequent event after allo-HSCT in adult patients, in contrast with the data from previous studies. The HAdV viremia may be persistent and with high viral loads. However, we found neither an association between the presence of HAdV viremia and the subsequent development of disseminated HAdV disease nor mortality. The HAdV-specific T-cell mediated immune response seems to control HAdV infections in adult recipients and to promote the early viremia clearance in adult patients with persistent viremia before transplantation. Therefore, the results from the present study do not support pre-emp-

Figure 2 Human adenovirus-specific CD4+ and CD8+ T-cell immune responses in patients with blips and presistent viremia at days -7 and/or 0 pretransplantation. (A) and (C) CD4+ and CD8+ T cells, respectively, expressing IL-2, TNF-α and IFN-γ, on days +21, +56 and +100 after transplantation in patients with blips. (B) and (D) CD4+ and CD8+ T cells, respectively, expressing IL-2, TNF-α and IFN-γ, on days +21, +56 and +100 after transplantation in patients with persistent viremia.

haematologica | 2021; 106(1)

277

Letters to the Editor

tive therapy based only in the HAdV viremia in adult allo-HSCT recipients. Javier Sánchez-Céspedes,1* José Antonio Marrugal-Lorenzo,1* Cecilia Martín-Gandul,1 Nancy Rodríguez-Torres,2 Enrique Montero-Mateos,1 Ana Serna-Gallego,1 Virginia Escamilla-Gómez,2 Laura Merino,1 Ildefonso Espigado,2 Jerónimo Pachón,3,4 José Antonio Pérez Simón2,3 and Manuela Aguilar-Guisado1 1 Unit of Infectious Diseases, Microbiology and Preventive Medicine, Institute of Biomedicine of Seville (IBIS), University Hospital Virgen del Rocío (CSIC), University of Seville; 2Department of Hematology, University Hospital Virgen del Rocío, Institute of Biomedicine of Seville (IBIS/CSIC/CIBERONC), University of Seville; 3Department of Medicine, University of Seville and 4Institute of Biomedicine of Seville (IBIS), University Hospital Virgen del Rocío (CSIC), University of Seville, Seville, Spain *JSC and JAML contributed equally as co-first authors. Correspondence: JAVIER SANCHEZ-CESPEDES - janchez-ibis@us.es MANUELA AGUILAR-GUISADO - maguilarguisado@yahoo.es doi:10.3324/haematol.2019.240101 Disclosures:No conflicts of interests to disclose. Contributions: JSC and MAG: study design, data collection, analysis and interpretation, literature search, manuscript preparation; JAPS, IE and JP: study design and data analysis and interpretation; JAML: data collection and analysis, figure preparation and statistical analysis; CMG and ASG: data collection; NRT, EMM, VEG and LM: provision of samples, data collection. manuscript preparation. Funding: this work was supported by Plan Nacional de I+D+i 2013‐2016 and Instituto de Salud Carlos III, Subdirección General de Redes y Centros de Investigación Cooperativa, Ministerio de Ciencia, Innovación y Universidades, Spanish Network for Research in Infectious Diseases (REIPI RD16/0016/0009) - cofinanced by European Development Regional Fund “A way to achieve Europe”, Operative program Intelligent Growth 2014‐2020, the Instituto de Salud Carlos III, Subdirección General de Evaluación y Fomento de la Investigación (PI15/00489), and the Spanish Adenovirus Network (AdenoNet, BIO2015/68990-REDT). JSC is a researcher belonging to the program “Nicolás Monardes” (C-0059-2018), Servicio Andaluz de Salud, Junta de Andalucía, Spain.

References 1. Chakrabarti S, Mautner V, Osman H, et al. Adenovirus infections following allogeneic stem cell transplantation: incidence and outcome in relation to graft manipulation, immunosuppression, and

278

immune recovery. Blood. 2002;100(5):1619-1627. 2. Gustafson I, Lindblom A, Yun Z, et al. Quantification of adenovirus DNA in unrelated donor hematopoietic stem cell transplant recipients. J Clin Virol. 2008;43(1):79-85. 3. Hiwarkar P, Kosulin K, Cesaro S, et al. Management of adenovirus infection in patients after haematopoietic stem cell transplantation: state-of-the-art and real-life current approach: a position statement on behalf of the Infectious Diseases Working Party of the European Society of Blood and Marrow Transplantation. Rev Med Virol. 2018; 28(3):e1980. 4. Kang JM, Park KS, Kim JM, et al. Prospective monitoring of adenovirus infection and type analysis after allogeneic hematopoietic cell transplantation: a single-center study in Korea. Transpl Infect Dis. 2018;20(3):e12885. 5. Lion T. Adenovirus infections in immunocompetent and immunocompromised patients. Clin Microbiol Rev. 2014;27(3):441-462. 6. Erard V, Huang ML, Ferrenberg J, et al. Quantitative real-time polymerase chain reaction for detection of adenovirus after T cell-replete hematopoietic cell transplantation: viral load as a marker for invasive disease. Clin Infect Dis. 2007;45(8):958-965. 7. Lion T, Baumgartinger R, Watzinger F, et al. Molecular monitoring of adenovirus in peripheral blood after allogeneic bone marrow transplantation permits early diagnosis of disseminated disease. Blood. 2003;102(3):1114-1120. 8. Ganzenmueller T, Buchholz S, Harste G, Dammann E, Trenschel R, Heim A. High lethality of human adenovirus disease in adult allogeneic stem cell transplant recipients with high adenoviral blood load. J Clin Virol. 2011;52(1):55-59. 9. Kosulin K, Pichler H, Lawitschka A, Geyeregger R, Lion T. Diagnostic parameters of adenoviremia in pediatric stem cell transplant recipients. Front Microbiol. 2019;10:414. 10. Piatti G. Pre-transplant screening for latent adenovirus in donors and recipients. Open Microbiol J. 2016;10:4-11. 11. Matthes-Martin S, Feuchtinger T, Shaw PJ, et al. European guidelines for diagnosis and treatment of adenovirus infection in leukemia and stem cell transplantation: summary of ECIL-4 (2011). Transpl Infect Dis. 2012;14(6):555-563. 12. Chou S, Ercolani RJ, Derakhchan K. Antiviral activity of maribavir in combination with other drugs active against human cytomegalovirus. Antiviral Res. 2018;157:128-133. 13. Kosulin K, Berkowitsch B, Matthes S, Pichler H, et al. Intestinal adenovirus shedding before allogeneic stem cell transplantation is a risk factor for invasive infection post-transplant. EBioMedicine. 2018; 28:114-119. 14. Guerin-El Khourouj V, Dalle JH, Pedron B, et al. Quantitative and qualitative CD4 T cell immune responses related to adenovirus DNAemia in hematopoietic stem cell transplantation. Biol Blood Marrow Transplant. 2011;17(4):476-485. 15. Gonzalez-Vicent M, Verna M, Pochon C, et al. Current practices in the management of adenovirus infection in allogeneic hematopoietic stem cell transplant recipients in Europe: the AdVance study. Eur J Haematol. 2019;102(3):210-217.

haematologica | 2021; 106(1)

Letters to the Editor

SIRPαFc treatment targets human acute myeloid leukemia stem cells Acute myeloid leukemia (AML) is an aggressive hematologic malignancy in which therapy resistance and relapse are driven by leukemia stem cells (LSC) that exhibit genetic and functional heterogeneity.1 Molecularly-targeted therapies may fail if some LSC subclones are inherently resistant, underscoring the need to develop complementary therapeutic strategies that will broadly target all subclones. Immunotherapies that enhance host innate immunity and target malignant cell properties that are common to all LSC subclones represent one such approach. Hematologic and solid tumors can upregulate expression of CD47, a self-marker that binds to SIRPα, an immunoglobulin (Ig) superfamily receptor expressed on macrophages, dendritic cells and neurons that transmits an inhibitory “don’t eat me” signal and prevents phagocytosis, enabling tumor cells to evade innate immune surveillance.2-6 We previously showed that disruption of SIRPα-CD47 signaling enhances phagocytosis of hematopoietic tumor cells in a SLAMF7-dependent manner.7-9 Here, we show that a human SIRPαFc protein (TTI-621, Trillium Therapeutics Inc., Ontario, Canada) blocks the CD47 “don’t eat me” signal and potently reduces leukemic engraftment in xenograft models, with high response rates in a diverse cohort of primary AML samples. The majority of samples from patients with unfavorable prognostic features were responsive to SIRPαFc treatment, including cases with a high LSC17 score, which is associated with poor initial response and short survival following standard treatments. Secondary transplantation experiments demonstrated that SIRPαFc treatment reduced LSC frequency. These data support further development of this novel immunotherapy for treatment of high-risk AML patients. Based on the results of our prior proof-of-concept study demonstrating promising therapeutic potential of SIRPαFc in a small number of patient samples (n=4),7 we conducted a large efficacy study of TTI-621, a fully human recombinant SIRPαFc fusion protein composed of the IgV domain of human SIRPα fused to a human IgG1-Fc moiety, in AML xenograft models. TTI-621 is a dual function molecule that in addition to blocking CD47-SIRPα interaction, delivers an activating signal through binding of the IgG1-Fc moiety to Fc receptors. TTI-621 binds minimally to human erythrocytes,9 a property that differentiates it from CD47 blocking antibodies and may limit potential clinical toxicity related to hemolytic anemia. In order to understand overall response rates in AML, we evaluated a cohort of 30 patient samples with diverse cytogenetic and molecular abnormalities (Online Supplementary Table S1). For each sample, bulk peripheral blood mononuclear cells were transplanted into cohorts of sublethally-irradiated NSG mice (10 mice/group). Following a 2-week engraftment period, mice were trea-ted with human SIRPαFc (TTI621) or a control IgG for 4 weeks and leukemic engraftment was determined by flow cytometry (Figure 1A). For each sample, the definition of response was based on the relative reduction in human leukemic engraftment in SIRPαFc-treated versus control-treated mice, using previously reported cutoffs.10 In control-treated mice, mean engraftment levels at the end of the treatment period were 51% in the injected femur (range: 17-96%) and 29% in the non-injected haematologica | 2021; 106(1)

bones (range: 0-96%) (Table 1). For 23 of 30 samples classified as good responders, treatment with SIRPαFc resulted in a 91% mean reduction of leukemic engraftment in the injected femur relative to control-treated mice (range: 53-100%, P<0.0001, Table 1, Figure 1B-C). These samples also showed a significant relative reduction in leukemic engraftment in non-injected bones (mean 96%, range: 35-100%, P<0.0001). It is noteworthy that significant responses were seen even for samples that genera-ted high levels of leukemic engraftment in control-treated mice: in many of these cases, leukemic cells were almost completely eliminated by SIRPαFc treatment. Six samples demonstrated a lesser response that was largely observed in the non-injected bones, with a mean reduction of 69% compared to controls (range: 43-93%, P=0.04, Table 1, Figure 1D-E), and were classified as partial responders. Only one sample among the 30 tested did not demonstrate a reduction in leukemic engraftment in either the injected femur or non-injected bones following SIRPαFc treatment. There were no statistically significant differences in clinical parameters or CD47 expression by RNAsequencing between the partial-/non-responders and good responders. Importantly, the majority of samples from cases with unfavorable prognostic features such as age greater than 60, adverse cytogenetic risk category, and secondary AML, as well as samples obtained from relapsed/resistant patients, were responsive to SIRPαFc treatment (Table 2). Notably, 20 of 26 (77%) of samples with a high LSC17 score (Online Supplementary Table S1), which we have previously shown identifies patients who do not derive significant clinical benefit from standard therapies,11 responded well to single agent SIRPαFc treatment, with the remaining six cases showing a partial response. The high response rates observed following SIRPαFc treatment in this diverse cohort of primary AML samples suggest that this therapeutic approach will have broad efficacy in AML, with potent activity even in patients with high-risk features. Although many AML patients achieve remission following standard chemotherapy, treatment-resistant LSC drive disease recurrence. In order to determine whether SIRPαFc treatment eliminated LSC in addition to the bulk disease, we transplanted leukemia cells harvested from primary treated mice into non-treated secondary recipients at limiting dilution (Figure 1F-G and Online Supplementary Table S2). For this analysis, we tested AML samples with sufficient human leukemia cells remaining after SIRPαFc treatment, including three partial responders and the one primary non-responder. In all four cases, we observed a lower LSC frequency (3.9-fold to 10.3-fold, P=0.002-0.024) in mice transplanted with SIRPαFc-treated cells compared to controls. These results provide direct evidence that SIRPαFc treatment targeted LSC in primary mice, despite the observation of no or only a partial reduction of bulk disease. The increasing evidence of genetic and functional heterogeneity in AML underscores the importance of developing novel therapeutic approaches that do not depend on prior identification of survival vulnerabilities in all LSC clones. SIRPαFc targeting of widely expressed CD47 harnesses one component of the patient’s innate immune surveillance machinery to eliminate leukemic cells with aberrant or increased expression of prophagocytic mar-kers related to their aberrant genome and epigenome. Molecules such as calreticulin3 and SLAMF78 have been previously reported to trigger phagocytosis of hematopoietic and solid tumor cells, however their 279

Letters to the Editor

Table 1. Efficacy of SIRPαFc treatment in acute myeloid leukemia xenografts.

Patient ID AML3 AML4 AML5 AML6 AML7 AML8 AML9 AML10 AML11 AML12 AML13 AML14 AML15 AML16 AML17 AML18 AML19 AML20 AML21 AML22 AML23 AML24 AML25 AML26 AML27 AML28 AML29 AML30 AML31 AML32

Injected RF Mean engraftment (%) Control SIRPαFc 47.5 61.5 17.9 81.6 20.1 34.7 22.4 22.2 16.6 67.7 41.1 33.3 92.1 53.3 27.9 22.5 75.8 38.1 83.5 78.6 23.5 18.8 79.0 82.5 63.4 33.8 20.6 96.1 68.5 94.8

0.9 2.4 0.0 38.1 4.5 0.4 0.0 2.7 5.0 2.7 0.1 1.0 0.1 0.3 0.7 0.1 0.5 2.2 26.9 4.1 0.3 0.0 25.4 65.1 57.0 23.7 26.6 83.5 53.0 85.9

RR (%)

<0.0001 <0.0001 0.021 0.001 0.012 <0.0001 <0.0001 0.001 0.015 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 NS NS 0.042 NS NS NS NS

98 96 100 53 78 99 100 88 70 96 100 97 100 99 98 100 99 94 68 95 99 100 68 21 10 30 -29 13 23 9

Non-injected BM Mean engraftment (%) Control SIRPαFc 20.4 23.6 2.2 41.9 9.4 4.3 7.1 0.6 4.6 17.4 8.9 0.0 70.4 30.6 13.3 7.2 38.0 7.9 74.7 27.1 1.4 3.7 42.9 61.2 69.9 22.6 7.2 90.1 58.3 96.1

0.1 0.0 0.0 0.4 0.3 0.1 0.0 0.4 0.2 0.0 0.0 0.0 0.1 0.2 0.1 0.1 0.0 0.0 0.0 0.0 0.1 0.0 6.0 7.2 31.9 3.4 3.7 51.7 4.3 93.7

RR (%)

In vivo response*

0.002 <0.0001 0.012 <0.0001 0.006 0.0003 <0.0001 NS 0.001 <0.0001 <0.0001 NA <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0002 0.0003 <0.0001 <0.0001 0.0004 0.005 <0.0001 0.013 0.003 <0.0001 NS

99 100 99 99 97 98 100 35 96 100 100 NA 100 99 100 99 100 100 100 100 96 100 86 88 54 85 48 43 93 2

R R R R R R R R R R R R R R R R R R R R R R R PR PR PR PR PR PR NR

AML: acute myeloid leukemia; RR: relative reduction; R: responder; PR: partial responder; NR: non-responder; RF: injected right femur; BM: non-injected bone marrow; NA: not assessable; NS: not statistically significant. *In vivo response criteria: R:>50% RR in RF; PR: 20-50% RR in RF or ≥20% RR in BM only; NR: no significant difference between SIRP Fc- and control-treated mice (NS) or <20% RR in both RF and BM.

expression is inconsistent on AML cells. The heterogeneity of AML gene expression between patients and critically among subclones within individual patients is likely to confer variability in prophagocytic signaling ligands. Indeed, differences in prophagocytic signaling may have contributed in part to the observed variability in response to SIRPαFc treatment in our xenograft models. Nevertheless, most AML samples appear to rely heavily on SIRPαFc-CD47 interaction to evade macrophagemediated surveillance, with only 1 of 30 samples tested failing to exhibit any response to treatment. This high response rate, while supportive of potent clinical efficacy, precluded an in-depth comparative analysis of potential prophagocytic marker expression between responders and non-responders. SIRPαFc acts as a decoy receptor that “unmasks” tumor cells and triggers phagocytosis by multiple polarized macrophage subsets.12 The IgG1 Fc region of SIRPαFc likely also assists in the activation of macrophages by engaging Fc receptors, raising the possi280

bility of off-target effects against normal cells. However, we have previously shown that SIRPαFc treatment preferentially targets leukemic cells over normal hematopoietic cells in phagocytosis assays, consistent with the notion that AML cells provide stronger and/or distinct prophagocytic signals to macrophages.7 These observations suggest that there is a therapeutic window for SIRPαFc treatment that will enable elimination of leukemic cells while preserving normal hematopoiesis. The mice used as recipients in our AML xenograft experiments lack B and T cells; thus, contributions of host adaptive immunity to leukemia elimination could not be evaluated. However, SIRPαFc-triggered macrophage phagocytosis of target cells in vitro has been shown to augment tumor antigen presentation and lead to expansion of tumor-specific CD8+ T cells,13 which may contribute to therapeutic efficacy in patients. In the immediate post-chemotherapy setting, however, the contribution of adaptive immunity to the anti-leukemic activity of SIRPαFc may be limited as lymphocyte populations haematologica | 2021; 106(1)

Letters to the Editor

haematologica | 2021; 106(1)

Figure 1. SIRPαFc treatment eliminates leukemia stem cells in primary acute myeloid leukemia xenografts. (A) Schematic illustrating the experimental protocol used for in vivo drug testing. (B), (D) Representative flow cytometric analysis of human CD45+CD33+ acute myeloid leukemia (AML) engraftment in the injected femur of mice transplanted with a representative responder sample (AML19) (B) or a representative partial responder sample (AML26) (D) and treated with control IgG (top panels) or SIRPαFc (bottom panels). (C), (E) Summary of human AML engraftment in the injected femur and non-injected bones of mice treated with control IgG or SIRPαFc, as determined by flow cytometry. Three representative responder samples are shown in (C) and the non-responder (AML32) and two representative partial responder samples (AML26, AML27) are shown in (E), (F-G) Summary of human AML engraftment in the injected femurs of untreated secondary recipient mice 12 weeks after transplantation of indicated doses of AML cells pooled after harvesting from injected femurs (RF) or non-injected bone marrow (BM) of primary mice treated with control IgG or SIRPαFc, as determined by flow cytometry. A representative partial responder (AML28) is shown in (F) and the non-responder (AML32) is shown in (G). Each symbol represents one mouse. Bars indicate mean values. I.F.: intrafemoral; I.P.: intraperitoneal; *P≤ 0.05; **P≤ 0.01; ***P≤ 0.001; ****P≤ 0.0001.

281

Letters to the Editor

appear to be compromised for many months following treatment.14 The strong response to SIRPαFc monotherapy we observed in the AML xenograft model may be further enhanced by clinical combination with agents such as azacitidine, which has been shown to trigger immune-mediated clearance of cancer cells by inducing a viral mimicry state.15 Our data demonstrate the potent efficacy of SIRPαFc monotherapy in a preclinical AML model. Clinically, TTI-621 may be most effective in the remission setting following recovery of host macrophage function, employed as maintenance therapy to eliminate LSC in AML patients with detectable residual disease, preventing relapse. The optimal balance between efficacy and safety of novel anti-cancer therapies can only be assessed in a clinical setting. Currently, TTI-621 is being evaluated in an early phase clinical trial for treatment of patients with relapsed or refractory hematologic malignancies (NCT02663518), and appears to be well tolerated without causing significant anemia, in keeping with the observation that TTI-621, unlike anti-CD47 antibodies, does not bind appreciably to human red blood cells.9 TTI-621 has also demonstrated efficacy as an intratumoral treatment for relapsed or refractory cutaneous Tcell lymphomas (NCT02890368), with rapid local treatment effects observed in 91% of patients. Overall, these findings support the development of TTI-621 as a novel immunotherapy for AML and other malignancies. Oleksandr Galkin,1 Jessica McLeod,1 James A. Kennedy,1,2° Liqing Jin,1 Nathan Mbong,1 Mark Wong,3 Robert A. Uger,3 Mark D. Minden,1,2,4,5 Jayne S. Danska5,6,7 and Jean C.Y. Wang1,2,4 1 Princess Margaret Cancer Center, University Health Network; 2 Division of Medical Oncology and Hematology, Department of Medicine, University Health Network; 3Trillium Therapeutics Inc., Mississauga; 4Department of Medicine, University of Toronto; 5 Department of Medical Biophysics, University of Toronto; 6Hospital for Sick Children, Toronto and 7Department of Immunology, University of Toronto, Toronto, Ontario, Canada °Current affiliation: Odette Cancer Center, Sunnybrook Health Sciences Center, Toronto Correspondence: JEAN C.Y. WANG - jean.wang@uhnresearch.ca doi:10.3324/haematol.2019.245167 Disclosures: RAU and MW are employees of Trillium Therapeutics Inc. (TTI); TTI provided grant funding and drug to support this work. There is an existing license agreement between TTI and University Health Network/Hospital for Sick Children and LJ, JSD and JCYW may be entitled to receive financial benefits further to this license and in accordance with their respective institutions’ intellectual property policies. Contributions: OG, JM, LJ and NM performed experiments; OG, JAK, LJ performed data analysis; JSD and JCYW supervised the study; MDM provided patient samples and clinical data; OG and JCYW wrote the manuscript; MW, RAU, JSD and JCYW revised the manuscript. Funding: this work was funded by the Government of Canada through Genome Canada and the Ontario Genomics Institute (OGI-091) with funds from the Government of Ontario and Trillium Therapeutics Inc.

References 1. Klco JM, Spencer DH, Miller CA, et al. Functional heterogeneity of genetically defined subclones in acute myeloid leukemia. Cancer Cell. 2014;25(3):379-392. 2. Jaiswal S, Jamieson CH, Pang WW, et al. CD47 is upregulated on circulating hematopoietic stem cells and leukemia cells to avoid phago-

282

Table 2. Clinical characteristics of SIRPαFc responders and partial/non-responders.

Responders (n=23) Age at AML diagnosis (years) median (range) 62 (26 – 86) PB WBC count at diagnosis (×109/L) median (range) 99 (4 – 227) BM blasts at diagnosis (%) n=19 median (range) 62 (23 – 95) Age at AML diagnosis (n) < 60 11 ≥ 60 12 De novo vs. Secondary AML (n) De novo 16 Secondary* 7 Disease stage of sample (n) Diagnosis 17 Relapsed/Resistant 6 MRC cytogenetic risk at diagnosis (n) n=21 Intermediate 15 Adverse 6 NPM1 status (n) n=12 NPM1c 6 NPM1 wild-type 6 FLT3-ITD status (n) n=13 FLT3-ITD positive 10 FLT3-ITD negative 3 LSC17 score (n) High score 20 Low score 3

Partial and non-responders (n=7) 55 (33 – 69) 43 (5 – 235) n=7 85 (51 – 92) 5 2 4 3 5 2 n=7 5 2 n=4 3 1 n=4 2 2 6 1

*Secondary includes therapy-related acute myeloid leukemia (AML) and postmyelodysplastic/myeloproliferative neoplasms (MDS/MPN) AML. PB: peripheral blood; WBC: white blood cell; BM: bone marrow.

cytosis. Cell. 2009;138(2):271-285. 3. Chao MP, Jaiswal S, Weissman-Tsukamoto R, et al. Calreticulin is the dominant pro-phagocytic signal on multiple human cancers and is counterbalanced by CD47. Sci Transl Med. 2010;2(63):63ra94. 4. Chao MP, Alizadeh AA, Tang C, et al. Therapeutic antibody targeting of CD47 eliminates human acute lymphoblastic leukemia. Cancer Res. 2011;71(4):1374-1384. 5. Zhao XW, van Beek EM, Schornagel K, et al. CD47-signal regulatory protein-alpha (SIRPalpha) interactions form a barrier for antibodymediated tumor cell destruction. Proc Natl Acad Sci U S A. 2011; 108(45):18342-18347. 6. Willingham SB, Volkmer JP, Gentles AJ, et al. The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors. Proc Natl Acad Sci U S A. 2012;109(17):66626667. 7. Theocharides APA, Jin L, Cheng PY, et al. Disruption of SIRPα signaling in macrophages eliminates human acute myeloid leukemia stem cells in xenografts. J Exp Med. 2012;209(10):1883-1899. 8. Chen J, Zhong MC, Guo H, et al. SLAMF7 is critical for phagocytosis of haematopoietic tumour cells via Mac-1 integrin. Nature. 2017; 544(7651):493-497. 9. Petrova PS, Viller NN, Wong M, et al. TTI-621 (SIRPalphaFc): a CD47-blocking innate immune checkpoint inhibitor with broad antitumor activity and minimal erythrocyte binding. Clin Cancer Res. 2017;23(4):1068-1079. 10. Chen WC, Yuan JS, Xing Y, et al. An integrated analysis of heterogeneous drug responses in acute myeloid leukemia that enables the

haematologica | 2021; 106(1)

Letters to the Editor

discovery of predictive biomarkers. Cancer Res. 2016;76(5):12141225. 11. Ng SW, Mitchell A, Kennedy JA, et al. A 17-gene stemness score for rapid determination of risk in acute leukaemia. Nature. 2016; 540(7633):433-437. 12. Lin GHY, Chai V, Lee V, et al. TTI-621 (SIRPalphaFc), a CD47-blocking cancer immunotherapeutic, triggers phagocytosis of lymphoma cells by multiple polarized macrophage subsets. PLoS One. 2017; 12(10):e0187262. 13. Viller NN, Truong T, Linderoth E, et al. Abstract B028: blockade of

haematologica | 2021; 106(1)

the CD47 "do not eat" signal by TTI-621 (SIRPÎąFc) leads to enhanced antitumor CD8+ T cell responses in vitro. Cancer Immunol Res. 2016;4(Suppl 11):B028. 14. Verma R, Foster RE, Horgan K, et al. Lymphocyte depletion and repopulation after chemotherapy for primary breast cancer. Breast Cancer Res. 2016;18(1):10. 15. Roulois D, Loo Yau H, Singhania R, et al. DNA-demethylating agents target colorectal cancer cells by inducing viral mimicry by endogenous transcripts. Cell. 2015;162(5):961-9.

283

Letters to the Editor

Comparative analysis of targeted novel therapies in relapsed, refractory chronic lymphocytic leukemia Therapeutic options for chronic lymphocytic leukemia (CLL) patients have expanded over recent years as efficacious novel agents (NA) have received licensing approval, initially in relapsed, refractory (R/R) disease. Ibrutinib is a Bruton tyrosine kinase (BTK) inhibitor (BTKi) targeting Bcell receptor (BCR) pathway signaling. It has potent activity in relapsed disease as monotherapy1 with significant benefit compared to ofatumumab and combined with bendamustine-rituximab (BR) versus BR.2 Idelalisib also targets the BCR pathway by inhibiting the delta isoform of phosphoinositide 3-kinase (PI3Ki). It is licensed in combination with rituximab following improved progression-free survival (PFS) compared to rituximab in heavily pre-treated patients.3 However, toxicity concerns now limit its use in R/R CLL, and the randomized ASCEND trial4 has shown PFS superiority for the second generation BTKi acalabrutinib versus idelalisib-rituximab (hazard ratio [HR] 0.29; P<0.0001). Moreover, a large retrospective series has previously compared ibrutinib and idelalisib-rituximab as first NA (NA1)5 and found a similar, indirect PFS advantage for ibrutinib. Venetoclax is a potent, oral small-molecule BCL2 inhibitor (BCL2i) licensed as monotherapy ± rituximab in R/R CLL.6,7 Venetoclax is increasingly utilized at first relapse in combination with rituximab for a 2-year fixed duration.6 However, to date, no prospective trials have directly compared ibrutinib with venetoclax as NA1 in R/R CLL. It remains a key unanswered question as to which of these two NA optimizes the balance of safety and efficacy when used as the NA1 in R/R CLL. To address this, we report a large, international study to establish the efficacy of ibrutinib and venetoclax ± anti-CD20 as NA1 in R/R CLL. To our knowledge, this is the largest series comparing these two approaches as NA1. To address this question, a single NA database was created from prior international, collaborative, multicenter retrospective analyses examining each NA to describe and compare the characteristics and outcomes of CLL patients treated with ibrutinib or venetoclax ± anti-CD20 as the NA1 between 2015-2019. To be eligible for comparison, patients could not have been treated with a targeted agent in the front-line setting and had to receive ibrutinib or venetoclax as NA1 no later than line 5. Twenty centers participated in this study from academic and community sites. To define the study cohort for the primary analysis, medical chart reviews were performed to identify all con-

secutive patients with CLL at each institution who received ibrutinib or venetoclax as NA1 in the R/R setting including only patients in routine clinical practice. To verify that cohorts were similar, baseline characteristics were collected prior to NA exposure. Using a standardized case report form, investigators collected data on preNA demographics, disease/prognostic characteristics (including del(17)p, complex karyotype [CK] defined: ≥3 aberrations, IGHV status, and TP53 mutation), clinical and genetic characteristics, number and type of prior therapies. Toxicity data regarding dose interruptions, reductions, NA discontinuation, and reasons for discontinuation were collected. Other primary endpoints were overall response rate (ORR) and PFS for the NA1. Investigators were requested to classify responses by International Workshop on CLL (iwCLL) criteria as complete remission (CR), partial remission (PR) (including PR with lymphocytosis), stable disease (SD), and progressive disease (PD).8 PFS was defined as the time from NA1 to last follow-up, progression of CLL or death. Overall survival (OS) was defined as the time from NA1 to death from any cause. PFS and OS were estimated by the Kaplan-Meier method.9 All other comparisons were descriptive. Comparisons were made using Cox regression or log-rank tests. Statistical analyses were performed using STATA 10.1. Follow-up was censored at most recent visit or death. A total of 433 patients received a NA1: 385 ibrutinib; 48 venetoclax ± anti-CD20 (80% monotherapy). Data for PFS were available for 417 patients. Median follow-up was 14 months (ibrutinib) and 13.5 months (venetoclax), respectively. Each group had a median of two prior lines (range 1-4 for each). Median age of each cohort (ibrutinib: 69 years; venetoclax: 65 years) were similar, as were the proportion with adverse features such as del(17p) (Table 1). Overall response rate and CR were: ibrutinib (71% and 12%, respectively) and venetoclax (96% and 56%, respectively). Venetoclax-treated patients had a statistically significantly higher CR rate versus ibrutinib (P<0.001). Figure 1A shows PFS of all patients stratified according to NA1. Using ibrutinib as the comparator (HR, 1.0), PFS HR for venetoclax was 0.29 (95%Cl: 0.10-0.92; P=0.036) (Figure 1A). Using ibrutinib as the comparator for OS, HR was 1.2 (95%Cl: 0.47-3.1; P=0.70) for the venetoclax-treated cohort (Figure 1B). When considering del(17p), del(11q), CK and IGHV status, we could not identify a subgroup that benefited in terms of PFS when comparing each NA1 (data not shown). Considering the whole cohort, only those with a CK (HR, 1.99 [1.1-3.80]; P=0.04) had a statistically significant inferior PFS. Dose interruptions were recorded in 33% ibrutinib and 32% venetoclax-treated patients (Table 2). Dose reduc-

Table 1. Baseline characteristics of each novel therapy cohort.

Baseline characteristics First novel therapy Median age at treatment, years (range) Median n. of prior therapies, (range) Del(17p) Complex karyotype Elevated LDH

N 382

Ibrutinib 69 (27-95)

N 48

Venetoclax +/- anti-CD20 65 (39-87)

385

2 (1-4)

255 157 189

24% 32% 45%

47 17 20

34% 24% 45%

N/n: number; LDH: lactate dehydrogenase.

284

haematologica | 2021; 106(1)

Letters to the Editor

tions were recorded in 22% of ibrutinib and 26% venetoclax-treated patients. Overall discontinuation rates were 41% ibrutinib and 25% venetoclax (P=0.06). For ibrutinib, the most common discontinuation reasons were adverse events (AE) (22%), CLL progression (8%) and Richterâ&#x20AC;&#x2122;s transformation (RT) (2%). For venetoclax, the most common discontinuation reasons were allogeneic stem cell transplantation (10%), CLL progression (4%), and unrelated death event (4%). RT was uncommon and comparable across cohorts (Table 2). To our knowledge, this is the largest series comparing ibrutinib and venetoclax-based treatment in R/R CLL. In the first such analysis, we provide an indirect comparison of patients receiving ibrutinib and venetoclax as a NA1. We note that CR rates were higher with venetoclax treat-

ment, which appeared to translate to a PFS advantage over ibrutinib but not an OS advantage. In light of this, and in the absence of randomized data comparing these approaches, our data provide reassurance that either option remains a reasonable approach as NA1 in R/R CLL. The PFS difference seen within this series may in part relate to the tolerability of each agent within routine clinical practice. Higher rates of discontinuation due to AE and frequent dose reductions/interruptions have been reported to be most pronounced in ibrutinib-treated patients10,11 and AE are the most dominant cause for ibrutinib discontinuation within our data. Our study has several limitations. It is retrospective, and its conclusions remain hypothesis generating. However, in the absence of ongoing randomized trials

Figure 1. Survival outcome of patients with relapsed, refractory chronic lymphocytic leukemia CLL according to first novel agent received. (A) Progressionfree survival according to novel targeted agent. (B) Overall survival according to novel targeted agent. HR: hazard ratio; CI: confidence interval.

haematologica | 2021; 106(1)

285

Letters to the Editor

Table 2. Dosing patterns and discontinuations according to novel agent.

Dosing patterns First novel therapy Dose reductions Dose interruptions Discontinuation rate Most common discontinuation reason Richter’s transformation rate

Ibrutinib

Venetoclax +/anti-CD20

22% 33% 41% Adverse event 22% 2%

26% 32% 25% Allogeneic stem cell transplantation 10% 2%

comparing these two standard-of-care NA1 in CLL, our study may help to inform practice and aid future prospective trial design. We recognize the unavoidable biases associated with retrospective reporting, missing data, the lack of centralized response assessment and pathology review, and the potential for overestimating CR rates; however, any misclassification is likely to be non-differential. Information regarding dosing, interruptions, and toxicity were not prospectively collected leading to the risk of bias. To address this, we partnered with investigators with extensive experience in using NA and experienced in ‘real-world’ evidence collation. While the group has published similar datasets exploring NA sequencing,5,10-13 this analysis is the first published indirect comparison of ibrutinib and venetoclax in R/R CLL as first NA from our ‘real-world’ database and provides important new data in terms of NA sequencing and selection. Patients within our series had received a median of two prior lines pre-NA, with a relatively high rate of del(17p) (25% across all patients) and CK (31% across all patients) for whom data were available. This likely reflects the fact that, early on, these agents were prescribed to heavily pretreated, poor-risk patients. Use of NA is quickly being brought forward to first relapse4 and front-line treatment.14,15 It is possible that survival analyses would produce different results in less heavily pre-treated patients. However, given the lack of prospective comparative data, we believe our comparison provides valuable information for healthcare providers in the decision-making process regarding which NA to use first in R/R CLL. Given the nature of this population, making inferences regarding front-line therapy is challenging and was not our primary goal. We also did not aim to specifically answer questions regarding subsequent sequencing beyond NA1; this work is currently ongoing. Finally, the patient numbers in the venetoclax group are small compared to the ibrutinib group and we acknowledge analyses may be underpowered to detect any possible differences. Although the baseline characteristics are similar, smaller cohorts could provide greater heterogeneity. The difference in baseline characteristics reflects prescribing practice over recent years where the strongest evidence for NA1 within trials1 and non-trial10,11 was initially formed with ibrutinib. Finally, our data were collected prior to recent analyses suggesting CK with ≥5 aberrations selected the highest risk patients, and as such, our CK data should be interpreted in that light. In conclusion, venetoclax and ibrutinib-based therapy as NA1 provide comparable OS outcomes in R/R CLL patients treated outside of trials. Our data suggest a significant PFS advantage for venetoclax-treated patients as 286

NA1, a finding that requires independent validation and reassessment for patients treated at first relapse outside of clinical trials. The selection of either as the NA1 should therefore be based on individual patient factors, drug access, deliverability, and patient preference. Toby A. Eyre,1* Nicole Lamanna,2* Lindsey E. Roeker,3 Chaitra S Ujjani,4 Brian T. Hill,5 Paul M. Barr,6 Erick Lansigan,7 Bruce D. Cheson,8 Maryam Yazdy,8 John N. Allan,9 Joanna Rhodes,10 Stephen J. Schuster,10 Chadi Nabhan,11 Alan Skarbnik,12 Lorie Leslie,13 Prioty Islam,14 Katherine Whitaker,3 Catherine C. Coombs,15 Hande H. Tuncer,16 John M. Pagel,17 Ryan Jacobs,18 Allison M. Winter,5 Neil Bailey,17 Andrea Sitlinger,14 Anna H. Schuh,1 George Follows,19 Christopher P. Fox,20 Danielle M. Brander,14 Mazyar Shadman7 and Anthony R. Mato3 1 Department of Haematology, Churchill Hospital, Oxford University Hospitals NHS Foundation Trust, Oxford, UK; 2Herbert Irving Comprehensive Cancer Center (New York-Presbyterian Columbia University Medical Center), New York, NY, USA; 3 Memorial Sloan Kettering Cancer Center, New York, NY, USA;4 Seattle Cancer Care Alliance/Fred Hutchinson Cancer Research Center, Seattle, WA, USA; 5Department of Hematology and Medical Oncology, Taussig Cancer Institute, Cleveland Clinic Foundation, Cleveland, OH, USA; 6Division of Hematology/Oncology, Wilmot Cancer Institute, University of Rochester, Rochester, NY, USA; 7 Dartmouth-Hitchcock Medical Center, Lebanon, NH, USA; 8 Georgetown University Hospital, Lombardi Comprehensive Cancer Center, Washington DC, USA; 9Weill Cornell Medicine, Long Island City, NY, USA; 10Division of Hematology and Oncology, Abramson Cancer Center, University of Pennsylvania, Philadelphia, PA, USA; 11 Aptitude Health, Atlanta, GA, USA; 12Novant Health, Charlotte, NC, USA; 13John Theurer Cancer Center at Hackensack University Medical Center, Hackensack, NJ, USA; 14Hematologic Malignancies and Cellular Therapy, Duke University Medical Center, Durham, NC, USA; 15Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC, USA; 16Department of Medicine, Cancer Center, Tufts Medical Center, Boston, MA, USA; 17Center for Blood Disorders and Stem Cell Transplantation, Swedish Cancer Institute, Seattle, WA, USA; 18Department of Hematology, Lymphoma Division, Levine Cancer Institute, Charlotte, NC, USA; 19Department of Haematology, Addenbrookes, Cambridge, UK and 20Nottingham University Hospitals NHS Trust, Nottingham, UK *TAE and NL contributed equally as co-first authors.

Correspondence: TOBY A. EYRE - toby.eyre@ouh.nhs.uk doi:10.3324/haematol.2019.241539 Disclosures: ARM holds a consultancy role for TG Therapeutics (in addition DSMB), Abbvie, Pharamacyclics, Johnson & Johnson, Regeneron, Astra Zeneca, and Celgene and has received research funding from TG Therapeutics, Abbvie, Pharamacyclics, Johnson & Johnson, Regeneron, Portola, DTRM, and Acerta; LER has received a travel grant from AbbVie; RJ has received honoraria from Genentech and AbbVie; BTH has served on advisory boards for and received research funding from Genentech/AbbVie; NL participated in advisory boards for AbbVie, AstraZeneca, Genentech, Gilead, Janssen, Juno, Pharmacyclics, is a member of a scientific advisory board for Celgene, and had research funding to the institution from AbbVie, Acerta, AstraZeneca, Beigene, Genentech, Gilead, Juno, Oncternal, TG Therapeutics, and Verastem; DMB serves as consultant for, is a member of the scientific advisory board of, and institution is the site of a PI clinical trial (grant paid to the institution) from AbbVie and Genentech; CSU has received consulting fees from AstraZeneca, Genentech, Gilead, and research support from AbbVie and Pharmacyclics; MSY received honoraria from Bayer as a speaker; APS has received research funding from BMS, has received consulting fees from Pharmacyclics, haematologica | 2021; 106(1)

Letters to the Editor

Janssen, AbbVie, Genentech, AstraZeneca, and serves on the Speakers Bureau for Pharmacyclics, Janssen, AbbVie, Genentech, Jazz Pharmaceuticals, Gilead, Kite Seattle, Genetics, Verastem, and Stock ownership for COTA healthcare; JMP has received consulting fees from AbbVie, Pharmacyclics, Genentech, Jazz Pharmaceuticals and serves on advisory boards for Pharmacyclics, AbbVie, and Genentech, speakers bureau for AbbVie, Pharmacyclics, Genentech, Jazz Pharmaceuticals, Gilead Sciences, Verastem, Kite, and Seattle Genetics; CCC has received honoraria from Loxo, Octapharma, H3 Biomedicine, Pharmacyclics and AbbVie and has received consulting fees from AbbVie; PMB has received consulting fees from AbbVie/PCYC, Gilead, Verastem, Genentech; CN is employed by Aptitude Health; BDC has received consulting fees from AbbVie, TG Therapeutics, Roche-Genentech, Pharmacyclics, Gilead, Epizyme, Celgene, Morhphosys, Astellas, AstraZeneca, Bayer, Karyopharm, receives research funding to the institution from AbbVie, TG Therapeutics, Roche-Genentech, Pharmacyclics, Gilead, Epizyme, Celgene, Trillium, AstraZeneca; TAE received Honora ria from Roche, Gilead, Abbvie, and Janssen, and holds an advisory board role for Gilead and has received research funding from Gilead, and travelled to conferences for Takeda, Abbvie, and Janssen; AG is an employee of COTA, Acerta, Genentech, Pharmacyclics/Janssen, and Kite/Gilead, reports receiving other commercial research support and speakers bureau honoraria from, and is a consultant/advisory board member for Takeda, Kite/Gilead, Pharmacyclics/Janssen, Genentech, and Acerta, and holds ownership interest (including patents) in COTA; CF reports receiving commercial research grants from Gilead, Abbvie, and Roche, reports receiving other commercial research support from Adienne, reports receiving speakers bureau honoraria from Abbvie, Celgene, Roche, and Janssen, and is a consultant/advisory boardmember for Abbvie, Celgene, Roche, Janssen, Sunesis, Gilead, Takeda, and Atarbio; no potential conflicts of interest were disclosed by the other authors. Contributions: TE, NL, ARM, LR and MS designed the study; all authors performed data collection and analysis; TE wrote the manuscript and all authors reviewed it; ARM supervised the study. Funding: this study was supported by funding from the Oxford NIHR Biomedical Research Centre. Views expressed are those of the authors and not necessarily those of the NHS or the NIHR or the United Kingdomâ&#x20AC;&#x2122;s Department of Health.

9. 10.

11.

12.

13.

14.

References 1. Byrd JC, Furman RR, Coutre SE, et al. Targeting BTK with ibrutinib

haematologica | 2021; 106(1)

15.

in relapsed chronic lymphocytic leukemia. N Engl J Med. 2013; 369(1):32-42. Chanan-Khan A, Cramer P, Demirkan F, et al. Ibrutinib combined with bendamustine and rituximab compared with placebo, bendamustine, and rituximab for previously treated chronic lymphocytic leukaemia or small lymphocytic lymphoma (HELIOS): a randomised, double-blind, phase 3 study. Lancet Oncol. 2016;17(2):200211. Furman RR, Sharman JP, Coutre SE, et al. Idelalisib and rituximab in relapsed chronic lymphocytic leukemia. N Engl J Med. 2014;370(11):997-1007. Ghia P, Pluta A, Wach M, et al. Acalabrutinib vs rituximab plus idelalisib (IdR) or bendamustine (BR) by investigator choice in relapsed/refractory (RR) chronic lymphocytic leukemia: phase 3 ASCEND Study. Hematol Oncol. 2019;37(S2):86-87. Mato AR, Hill BT, Lamanna N, et al. Optimal sequencing of ibrutinib, idelalisib, and venetoclax in chronic lymphocytic leukemia: results from a multi-center study of 683 patients. Ann Oncol. 2017; 28(5):1050-1056. Seymour JF, Kipps TJ, Eichhorst B, et al. Venetoclax-rituximab in relapsed or refractory chronic lymphocytic leukemia. N Engl J Med. 2018;378(12):1107-1120. Stilgenbauer S, Eichhorst B, Schetelig J, et al. Venetoclax in relapsed or refractory chronic lymphocytic leukaemia with 17p deletion: a multicentre, open-label, phase 2 study. Lancet Oncol. 2016;17(6):768778. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760. Bland JM, Altman DG. Statistics notes: survival probabilities (the Kaplan-Meier method). BMJ. 1998;317(7172):1572. UK CLL Forum. Ibrutinib for relapsed/refractory chronic lymphocytic leukemia: a UK and Ireland analysis of outcomes in 315 patients. Haematologica. 2016;101(12):1563-1572. Mato AR, Nabhan C, Thompson MC, et al. Toxicities and outcomes of 616 ibrutinib-treated patients in the United States: a real-world analysis. Haematologica. 2018;103(5):874-879. Mato AR, Thompson M, Allan JN, et al. Real-world outcomes and management strategies for venetoclax-treated chronic lymphocytic leukemia patients in the United States. Haematologica. 2018; 103(9):1511-1517. Eyre TA, Kirkwood AA, Gohill S, et al. Efficacy of venetoclax monotherapy in patients with relapsed chronic lymphocytic leukaemia in the post-BCR inhibitor setting: a UK wide analysis. Br J Haematol. 2019;185(4):656-669. Fischer K, Al-Sawaf O, Bahlo J, et al. Venetoclax and obinutuzumab in patients with CLL and coexisting conditions. N Engl J Med. 2019; 380(23):2225-2236. Woyach JA, Ruppert AS, Heerema NA, et al. Ibrutinib regimens versus chemoimmunotherapy in older patients with untreated CLL. N Engl J Med. 2018;379(26):2517-2528.

287

Letters to the Editor

Platelets from patients with myocardial infarction can activate T cells Platelets are not just the cells that mediate coagulation. Over the last years, there has been increasing and compelling evidence that these cells may act as key regulators in immune responses and participate in the pathogenesis of various immune-mediated diseases.1-3 It has been shown that platelets can function as antigen presenting cells and activate T cells through MHC-I.4 Patients with myocardial infarction (MI) have activated platelets but whether these platelets interact with the immune system is not completely clear. It has been well established that immediate treatment with factors that prevent platelet aggregation is crucial for these patients; the sooner these agents are given, the better the outcome.5 T cells have been implicated in the pathogenesis of MI.6 Autoreactive T cells may have a role in the destabilization of the atherosclerosis plaque and plaque rupture. Regulatory T cells (CD4+CD25+high, Foxp3+; Treg) represent an important subset of T cells essential for immune homeostasis that controls activation of T cells.7 Up to now, data on Treg have shown that these cells have a protective role in atherosclerosis. Treg alterations in patients with MI with ST elevation (STEMI) have not been fully elucidated and8 there are no available data on the possible interaction of activated platelets with the T cells in these patients.9-12 The micro RNAs (miRNAs) represent small non-coding sequences with the ability to suppress different genes. In mice, the deficient expression of the miR155 was associated with enhanced atherosclerosis, decreased plaque stability and decreased Treg.13 However, the role of miR155 has not been established in patients with MI. We aimed to investigate whether T cells can be activated by the circulating activated platelets from patients with STEMI and explore how the Treg cell population is affected. After written informed consent had been obtained, peripheral blood mononuclear cells (PBMC) were isolated from 33 patients with STEMI (29 men, 4 women, mean age 62.5 years) at the time of hospital admission at diagnosis, before any treatment, as well as 5 days and 30 days later. Ten healthy subjects and five patients with unstable angina served as the control and disease-control group, respectively. We also isolated platelet rich plasma or plasma alone from the patients and healthy subjects and used it in mixed cultures with the isolated PBMC. The membrane expression of CD69 was used as a marker for T-cell activation. Three additional patients who received aspirin or clopidogrel before admission were also analyzed as an internal control group (Online Supplementary Appendix). We analyzed the percentages of CD4+CD25+highFOXP3+ cells representing the regulatory subset (Tregs) at all three time points, as described above (Day 0, day 5, and at one month of follow-up).14 The miR155 levels were evaluated using real-time polymerase chain reaction in samples from patients and healthy individuals. In STEMI patients, the expression of miR155 was evaluated at admission and when the patients were discharged (Day 0 and Day 5, respectively). This study was approved by the Ethics Committee of the Patras University Hospital. We first examined the activation of T cells in mixed cultures as described above (Online Supplementary Appendix). The expression of CD69 on the surface of T cells was used as a marker for T-cell activation. T cells that were incubated with platelets from patients with STEMI showed a statistically significant increase in 288

CD69 expression (2.163% vs. 0.575%; P=0.013) compared to the T cells that were incubated with platelets from healthy individuals (Figure 1Î&#x2018;). We did not record any activation in T cells that were incubated with plasma alone either from patients with STEMI or from healthy individuals. These results show that the activation of T cells incubated with platelets is specific only for platelets from patients with STEMI. It is also of note that T cells that were incubated with platelets from patients previously treated with aspirin or clopidogrel did not show any activation (data not shown) and the T cell-CD69 surface expression was comparable to that seen in T cells cocultured with platelets obtained from healthy individuals. It has been proposed that platelets may also have a role in the activation of the immune system and may represent key players in the immune homeostasis; our data are in agreement with this. Tregs can recognize self from non-self antigens and can down-regulate activation of T cells.7 Based on our results that activated platelets can activate T cells, we next examined the percentages of Tregs using flow cytometry to explore whether Tregs are altered in order to control the platelet-initiated activation. Upon presentation, patients with STEMI did not show any statistically significant different numbers in Tregs compared to healthy individuals. However, 5 days later, patients with STEMI displayed a statistically significant increase in Treg numbers compared with the two control groups. One month later, Treg numbers returned to the initial presentation levels (Figure 1B). To our knowledge, this is the first time that Tregs have been studied serially at different time points following a STEMI. Tregs increase during the first days after the STEMI and possibly represent the T-cell subset that is trying to eliminate the activation of the immune system and the inflammatory response. Recently, it was shown that miR155 is up-regulated in patients with viral myocarditis.15 Also, differential expression of the Tregs was shown to correlate with the expression of miR155.13 To explore the mechanism of Treg upregulation, the levels of miR155 were evaluated. Patients with STEMI displayed comparable levels of miR155 at presentation to healthy individuals. However, five days later, patients with STEMI had a statistically significant decrease in miR155 levels (P<0.001) that was inversely correlated with the increased Treg numbers observed at the same time point (Figure 1C). Alterations in different miRNAs expression have been associated with rheumatoid arthritis and psoriasis, but also with solid tumors. Results from mouse models implicate deficient expression of the miR155 with decreased plaque stability and decreased Tregs.13 Up to now, the role of miR155 has not been established in patients with MI. Here we show for the first time that patients with STEMI have decreased miR155 levels that inversely correlate with Treg numbers. Activated T cells may have a dual role in plaque rupture. T cells secrete IFN-Îł which inhibits the formation of new collagen, essential for the stability of the plaque in coronary arteries. Additionally, such T cells interact with macrophages leading to increased collagen degradation. Decreased formation combined with increased degradation of collagen contributes to plaque rapture.6 Our study shows that platelets from patients with STEMI can activate T cells ex vivo. This activation of circulating platelets may lead to an increase of the regulatory T-cell subset, possibly via the down-regulated expression of miR155. The increase in Treg observed in parallel with the decreased miR155 expression perhaps represent an effort of the immune system to control those auto-reactive haematologica | 2021; 106(1)

Letters to the Editor

Figure 1. Platelets from patients with STEMI can activate T cells and increase regulatory T cells. (A) Platelets from patients with STEMI activate T cells ex vivo. T cells incubated with platelets obtained from patients with STEMI displayed a statistically significant increased expression of the surface activation marker CD69 (P=0.013) compared to T cells incubated with platelets from healthy individuals. There was no activation in T cells when incubated with plasma from patients with STEMI. PLT control: activation in T cells when treated with platelets from healthy control subjects; PLT STEMI: activation in T cells when treated with platelets from patients with STEMI; Plasma STEMI: activation of T cells in cultures when treated only with plasma from patients with STEMI. (B) Patients with STEMI display a statistically significant increase in Treg numbers. Patients with STEMI at admission (Day 0) show comparable levels of Tregs with the healthy control group and with patients with unstable angina. Five days after admission, patients with STEMI displayed statistically significant increase in Tregs compared to controls, and one month later, Treg numbers in STEMI patients return to the admission levels. Treg at presentation versus 5 days after admission: 2.875% versus 4.521%; P=0.0002. Treg 5 days after admission versus 1 month follow-up (fu) after admission: 4.521% versus 2.745%; P=0.0004. Control: healthy control subjects; Myocardial infarction day 0: STEMI patients at admission; Myocardial infarction day 5: STEMI patients 5 days after admission; Myocardial infarction one month fu: STEMI patients one month after initial admission at fu; Unstable angina: Control group with unstable angina. (C) Decreased levels of miR155 in patients with STEMI. Patients with STEMI at admission (MI Day 0) show comparable mRNA levels of miR155 with the healthy control group (P=0.79). Five days after admission, patients with STEMI displayed a statistically significant decrease in miR155 mRNA levels compared to controls (P=<0.001) and this decrease was associated with the increase in Tregs observed in the same patients at the same time point. Control miR155: healthy control group; MI Day 0 miR155: patients with STEMI at admission; MI Day 5-Exit miR155: patients with STEMI 5 days after admission.

T cells that participate in plaque rapture. From the current study, we cannot confirm whether platelets are only activated by necrosis alone or whether necrosis directly activates the immune response. Our observations are preliminary and further studies are needed to establish a causal link between activated platelets and T cells. Moreover, the precise characterization of the mechanisms implicated in activated platelet:T-cell interaction might help prevent myocardial infarction. 1

Elena E. Solomou, Katerina Katsanaki, Elena Kalyvioti,1 Vasilios Gizas,1 Angelos Perperis,1 Dora Babali,1 Evgenia Verigou,1 Haralambos Gogos,1 haematologica | 2021; 106(1)

George Hahalis,1 Periklis Davlouros1 and Dimitrios Alexopoulos2 1

Department of Internal Medicine, University of Patras Medical School, Rion, Patras and 2Bâ&#x20AC;&#x2122; Department of Cardiology, University of Athens Medical School, Athens, Greece Correspondence: ELENA E. SOLOMOU - elenasolomou@hotmail.com doi:10.3324/haematol.2019.243402 Disclosures: no conflicts of interest to disclose. Contributions: EES designed research, performed experiments, analyzed data and wrote the paper; KK and EK analyzed data and per289

Letters to the Editor

formed experiments; VG, AP and DB performed experiments; EV analyzed data; HG, GH and PD analyzed data; DA analyzed data and wrote the paper.

8. Pastrana JL, Sha X, Virtue A, et al. Regulatory T cells and atheroscle-

rosis. J Clin Exp Cardiolog. 2012;2012(Suppl 12):2. 9. Murphy AJ, Tall AR. Disordered haematopoiesis and athero-throm-

bosis. Eur Heart J. 2016;37(14):1113-2111. 10. Hofmann U, Beyersdorf N, Weirather J, et al. Activation of CD4+

References 1. Davis RP, Miller-Dorey S, Jenne CN. Platelets and coagulation in

infection. Clin Transl Immunology. 2016;5(7):e89. 2. Middleton EA, Weyrich AS, Zimmerman GA. Platelets in pulmonary

immune responses and inflammatory lung diseases. Physiol Rev. 2016;96(4):1211-1259. 3. Carestia A, Kaufman T, Schattner M. Platelets: new bricks in the building of neutrophil extracellular traps. Front Immunol. 2016; 7:271. 4. Lesley M. Chapman, et al Platelets present antigen in the context of MHC class I. J Immunol. 2012;189(2):916-923. 5. Briasoulis A, Telila T, Palla M, Siasos G, Tousoulis D. P2Y12 receptor antagonists: which one to choose? a systematic review and metaanalysis. Curr Pharm Des. 2016;22(29):4568-4576. 6. Libby P. Mechanisms of acute coronary syndromes and their implications for therapy. N Engl J Med. 2013;368(21):2004-2013. 7. Sakaguchi S. Naturally arising Foxp3-expressing CD25+CD4+ regulatory T cells in immunological tolerance to self and non-self. Nat Immunol. 2005;6(4):345-352.

290

T lymphocytes improves wound healing and survival after experimental myocardial infarction in mice. Circulation. 2012; 125(13):1652-1663. 11. Sharir R, Semo J, Shimoni S, et al. Experimental myocardial infarction induces altered regulatory T cell hemostasis, and adoptive transfer attenuates subsequent remodeling. PLoS One. 2014; 9(12):e113653. 12. del Rosario Espinoza Mora M, Böhm M, Link A. The Th17/Treg imbalance in patients with cardiogenic shock. Clin Res Cardiol. 2014;103(4):301-313. 13. Donners MM, Wolfs IM, Stöger LJ, et al. Hematopoietic miR155 deficiency enhances atherosclerosis and decreases plaque stability in hyperlipidemic mice. PLoS One. 2012;7(4):e35877. 14. Solomou EE, Rezvani K, Mielke S, et al. Deficient CD4+ CD25+ FOXP3+ T regulatory cells in acquired aplastic anemia. Blood. 2007; 110(5):1603-1606. 15. O’Connell RM, Kahn D, Gibson WS, et al. MicroRNA-155 promotes autoimmune inflammation by enhancing inflammatory T cell development. Immunity. 2010;33(4):607-619.

haematologica | 2021; 106(1)

Letters to the Editor

Elotuzumab, lenalidomide, and dexamethasone as salvage therapy for patients with multiple myeloma: Italian, multicenter, retrospective clinical experience with 300 cases outside of controlled clinical trials Recently, monoclonal antibodies (mAb) directed to antigens expressed by plasma cells demonstrated major clinical activity in multiple myeloma (MM), thus gaining a relevant role in the treatment of MM patients.1 Two mAb targeting signaling lymphocytic activation, molecule F7 (SLAMF7/CS1) and CD38, have been increasingly included in relapse/refractory (RR) MM (RRMM) therapeutic regimens.2 Among them, elotuzumab (Elo), targets the glycoprotein receptor SLAMF7 causing, on the one side, activation of natural killer cells, and on the other, myeloma cell death through antibody-dependent cellular cytotoxicity (ADCC).3,4 Surprisingly, Elo failed to demonstrate significant anti-tumor activity as single-agent,5 ultimately refining its anti-myeloma action in combination with immunomodulatory drugs (IMiD) such as lenalidomide (R). A randomized phase III clinical trial comparing the clinical benefits of R plus dexamethasone (Rd) versus Rd plus Elo (EloRd) resulted in a longer progression-free survival (PFS) in patients allocated to the experimental arm.6 Moreover, extended assessments at 3-year,7 4-year,8 and 5-year9 follow-ups showed a significantly higher overall response rate (ORR) for EloRd as compared to the Rd regimen. Accordingly, the addition of Elo to Rd significantly reduced the risk of death by 27%.8 Here we report data of an Italian 'real-life' experience on EloRd as therapy for RRMM patients treated outside of controlled clinical trials, after receiving marketing approval in Italy in April 2017. Overall, 300 RRMM patients treated with EloRd according to the marketing approved schedule between April 2017 and April 2019 at 40 Italian centers entered this study (Online Supplementary Methods). Baseline characteristics are shown in Table 1. Approximately one quarter of patients (24.3%) had resistance to their most recent line of therapy, symptomatic relapse was observed in 171 patients (57%), and a biochemical relapse in 56 (18.7%). Over one-third (38.3%) received autologous stem cell transplant (ASCT), while 26% of patients had received prior lenalidomide treatment. Fluorescence in situ hybridization (FISH) data were available in only 64 patients; ten patients showed high-risk abnormalities. Thirty patients with mild renal impairment received R at the starting dose of 10 mg, while 24 with severe renal impairment received the starting dose of 15 mg every other day; 123 elderly patients (>75 years) received a weekly dose of 20 mg. At the time of last database update, the median number of courses administered was 12 (range, 1-43). A total of 188 (62.7%) patients stopped EloRd treatment at the time of the cut-off date (median follow-up: 19 months [range 1–36 months]) mainly owing to disease progression (151 cases), 13 patients for toxicity (9 infections, 2 lenalidomide-related severe skin rash, one dexamethasone-related psychosis, and one lenalidomide-related hepatotoxicity), 22 patients due to therapy-unrelated deaths, while two patients underwent ASCT. Infusion reactions occurred in 19 patients (6.3%, all grade 1-2) and were promptly resolved in all patients (no discontinuation reported). Major adverse events (AE) are shown in Table 2 and included grade 3/4 neutropenia (19%), anemia (15.7%), lymphocytopenia (12.7%), and thrombocytopenia (10%) while infection rates and pneumonia were approximately 34% and 16%, respectively. Notably, haematologica | 2021; 106(1)

Table 1. Main characteristics of patients at baseline.

N. of patients (%) Age, (years) <75 ≥75 Sex Male Female Paraproteins (isotype) Immunoglobulin G Immunoglobulin A Immunoglobulin D Light chain only Non-secretory Creatinine clearance (mL/min) ≥60 <60 Stage ISS, (n=238) I II III Number of previous lines of therapy 1 2 3 3 4 Previous ASCT No Yes Previous therapies Bortezomib Lenalidomide Cytogenetic profile Standard risk High risk* Not evaluated Disease status Biochemical relapse Symptomatic relapse Refractory to last treatment Time from diagnosis to EloRd treatment (years) ≥3.5 <3.5

177 (59) 123 (41) 157 (52.3) 143 (47.7) 188 (62.7) 55 (18.2) 2 (0.7) 53 (17.7) 2 (0.7) 214 (71.3) 86 (28.7) 91 (38.3) 95 (39.9) 52 (21.8) 186 (62) 70 (23.3) 20 (6.7) 24(8) 185 (61.7) 115 (38.3) 282 (94) 78 (26) 49 (16.3) 10 (3.3) 241 (80.4) 56 (18.7) 171 (57) 73 (24.3) 154 (51.3) 146 (48.7)

N: number; ISS: International Staging System; ASCT: autologous stem cell transplant; EloRd: lenalidomide plus dexamethasone (Rd) plus elotuzumab. *Patients harboring a t(4;14), t(14;16), or del(17p) were classified as having high-risk disease and all other cases as being at standard risk.

there was no significant difference in incidence of AE between younger (≤75 years) and elderly patients (data not shown). Age is an important factor in the treatment decisionmaking process for MM patients because of its association with frailty, increased comorbidities, poor tolerability, and higher risk of complications.10 In our series, approximately 41% of patients were aged ≥75 years, and 291

Letters to the Editor

of these about 29% presented with renal impairment. In addition, the Eloquent-2 phase III trial6 and subsequent updates7-9 have suggested that this triplet drug regimen is safe. Although these results should be treated with some caution given the retrospective nature of the present study and the different clinical features of patients included in the two series (i.e., the median number of previous lines of therapies, 2 in the Eloquent-2 trial and one in our retrospective series), our real-world cohort documented similar AE profiles, except lymphopenia incidence, possibly due to the above mentioned median number of previous lines of therapies and the reduced dexamethasone dose for patients >75 years. Of note, no significant differ-

ences in terms of incidence of AE were documented between younger (<75 years) and elderly patients. Thus, the choice of adding elotuzumab to Rd seems to have been rewarded by the clinical benefit observed in our 'real-world' cohort, and as previously described, in the Eloquent-2 phase III trial6-9 across key subgroups including elderly patients, as well as in individuals with a reduced renal function, thus offering a paradigm for case selection.11 The ORR was 77%, with 23 complete remissions (CR) (7.6%) and 88 very good partial remissions (VGPR) (29.3%). The median time to first response was 1.7 months, while the median time to best response was 3.5

Figure 1. Progression-free survival (PFS) of the retrospective cohort of relapse/refractory (RR) multiple myeloma patients treated with lenalidomide plus dexamethasone (Rd) plus elotuzumab (EloRd). Kaplan-Meier curve of PFS of the entire cohort (A). Forest plot of Cox univariate analysis for PFS according to clinical laboratory variables (B). b2M: beta-2-microglobulin; ASCT: autologous stem cell transplant; CrCl: creatinine clearance; CRAB: hypercalcemia, renal failure, anemia, and bone disease; N: number.

292

haematologica | 2021; 106(1)

Letters to the Editor

Table 2. Incidence of serious adverse events.

EloRd (n=300) Grade 3/4 adverse events Hematologic toxicities Lymphocytopenia Anemia Thrombocytopenia Neutropenia Non-hematologic toxicities Infections Pneumonia Fatigue Diarrhea

N. of cases (%) 38 (12.7) 47 (15.7) 30 (10) 57 (19) 103 (34.3) 50 (16.7) 62 (20.7) 22 (7.3)

months (Online Supplementary Figure S1) with approximately 73% and 91% of patients reaching the best response at 6 and 12 months, respectively. The ORR of our 'real-world' cohort was comparable with that of the Eloquent-2 trial (77% vs. 79%),6 with a similar number of patients reaching good quality responses. Moreover, the median time to achieve the best response was 3.5 months as compared to 2.8 months according to the Eloquent-2 independent review and 3.8 months based on the investigators’ assessment.6 Among the variables analyzed, a significantly worse treatment response was observed among patients previously exposed to lenalidomide and those with refractory disease status at EloRd start (Online Supplementary Table S1). However, the association between response and disease status at EloRd was no longer significant after Bonferroni correction. Nevertheless, in the light of the etiological nature of our study, multivariate ordinal regression analysis was still performed, showing prior lenalidomide exposure as the unique variable adversely and independently associated with the best response (odds ratio: 2.04, 95%CI: 1.2-3.3; P=0.005). Thus, prior lenalidomide exposure should be considered to be an additional concern when choosing EloRd treatment. During the follow-up period, 173 patients out of 300 experienced disease progression or died; 94 patients died. Median PFS was 17.6 months (95%CI: 14.1-21.0), and the 1-year PFS was 59.1% (Figure 1A), both results were very similar to the 19.4 months and 68% determined in the Eloquent-2 trial.6 Univariate Cox analyses showed that refractory disease status at EloRd initiation (HR:1.572, 95%CI: 1.054-2.345; P=0.027), a shorter time (i.e., <3.5 years) from diagnosis to EloRd start (HR:1.505, 95%CI: 1.036-2.187; P=0.032), more than two previous lines of therapy (HR:1.777, 95%CI: 1.227-2.572; P=0.002), and serum albumin level below 35 g/L (HR:1.849, 95%CI: 1.241-2.757; P=0.003) were associated with a significantly higher risk to progress or of death (Figure 1B). Notably, in the multivariate Cox model, albumin <35 g/L (HR:1.721, 95%CI: 1.147-2.581; P=0.009), time from diagnosis to EloRd start <3.5 years (HR:1.811, 95%CI: 1.179-2.781; P=0.007), and >2 lines of previous therapy (HR:2.116, 95%CI: 1.39-3.22; P<0.0001) maintained an independent prognostic impact on PFS. This multivariate model confirmed the Eloquent-2 trial results,6 demonstrating that those cases with a short disease history could be more prone to progress. Notably, a precocious EloRd treatment in MM patients with asymptomatic biochemical relapse failed to improve PFS in our haematologica | 2021; 106(1)

series, strengthening the concept that only patients with recurrence of the most typical clinical manifestations of MM (hypercalcemia, renal failure, anemia, and bone disease; CRAB) should start treatment. Moreover, ten cases with high-risk cytogenetic abnormalities showed a significantly shorter PFS than 54 cases with standard risk (Online Supplementary Figure S2). Nevertheless, the low number of evaluated cases (19.6%) does not allow any conclusions to be drawn regarding efficacy of EloRd in high-risk patients. Another objective of this analysis was to determine the relationship between the quality of response and PFS. EloRd-treated patients who achieved ≥VGPR were associated with higher PFS rates compared with EloRd-treated patients who achieved a partial response (PR) or less (Online Supplementary Figure S3). The results of Cox regression are detailed in the inbox in Online Supplementary Figure S3. Specifically, PFS rates at one year were approximately 80.1% (HR:1, reference group), 59% (HR: 2.15, 95%CI: 1.46-3.16; P<0.0001), and 23.8% (HR: 6.74, 95%CI: 4.47-10.15; P<0.0001) for cases achieving ≥VGPR, partial response (PR), and <PR, respectively (Online Supplementary Figure S3). Interestingly, no difference was demonstrated between complete remission (CR)/near CR (nCR) cases as compared with those achieving VGPR. The median time-to-next-treatment (TTNT) (25.3 months ([95%CI: 22.1-28.3]) (Online Supplementary Figure S4) of our 'real-world' cohort was comparable to that of the Eloquent-2 trial (25.3 vs. 33.4 months).7 Finally, median overall survival (OS) was not reached and 1-year OS was 65.5% (Online Supplementary Figure S5). To our knowledge, our survey is one of the largest EloRd series in terms of number of patients evaluated. This regimen represents one example of five triple schedules that have received high-level evidence and uniform consensus as a preferred treatment regimen for patients with RRMM.12,13 In conclusion, our 'real-world' cohort clearly showed that EloRd was a safe and effective regimen for RRMM patients, confirming the results obtained in the Eloquent-2 controlled clinical trial.6-9 Based on both studies, we suggest incorporating EloRd as a first salvage regimen in lenalidomide naïve patients and in patients with relatively longer disease duration. New ongoing trials will assess the efficacy of elotuzumab either in combination or with other IMiD14 in the context of an intensified treatment algorithm.15 Massimo Gentile,1,2 Giorgina Specchia,3 Daniele Derudas,4 Monica Galli,5 Cirino Botta,1,2 Stefano Rocco,6 Concetta Conticello,7 Catello Califano,8 Nicola Giuliani,9 Silvia Mangiacavalli,10 Enrico Attingenti,11 Alessandra Lombardo,12 Marino Brunori,13 Elena Rossi,14 Elisabetta Antonioli,15 Roberto Ria,16 Renato Zambello,17 Nicola Di Renzo,18 Giuseppe Mele,19 Gianpaolo Marcacci,20 Pellegrino Musto,21 Silvana Capalbo,22 Nicola Cascavilla,23 Claudio Cerchione,24 Angelo Belotti,25 Clelia Criscuolo,26 Giuseppina Uccello,27 Paola Curci,3 Ernesto Vigna,1,2 Vincenzo Fraticelli,1,2 Donatella Vincelli,28 Angela Bonalumi,29 Agostina Siniscalchi,30 Raffaella Stocchi,31 Massimo Martino,32 Stelvio Ballanti,33 Dominella Gangemi,34 Alfredo Gagliardi,35 Barbara Gamberi,36 Alessandra Pompa,37 Anna Grazia Recchia,2 Giovanni Tripepi,38 Annalisa Pitino,39 Ferdinando Frigeri,11 Ugo Consoli,27 Sara Bringhen,40 Elena Zamagni,41 Francesca Patriarca,31 Valerio De Stefano,14 Francesco Di Raimondo,7 Salvatore Palmieri,6 Maria Teresa Petrucci,42 Massimo Offidani,43 Mario Boccadoro,40 Michele Cavo41 and Fortunato Morabito2,44 293

Letters to the Editor

1 Hematology Unit, Azienda Ospedaliera of Cosenza, Cosenza, Italy; 2Biotechnology Research Unit, AO di Cosenza, Cosenza, Italy; 3 Hematology Section, University of Bari, Bari, Italy; 4Department of Hematology, Businco Hospital, Cagliari, Italy; 5Hematology and Bone Marrow Transplant Unit, Azienda Socio-Sanitaria Territoriale-Papa Giovanni XXIII, Bergamo, Italy; 6Hematology Unit, Ospedale Cardarelli, Napoli, Italy; 7Department of Medical and Surgical Specialties, Hematology Section, University of Catania, Catania, Italy; 8 Onco-Hematology Unit, Ospedale Umberto I, Nocera Inferiore, Italy; 9 Hematology Unit, Parma University Hospital, Parma, Italy; 10 Hematology Division, Department of Hematology-Oncology, IRCCS Fondazione Policlinico San Matteo, Pavia, Italy; 11UOC Ematologia a Indirizzo Oncologico, AORN “Sant’Anna e San Sebastiano”, Caserta, Italy; 12Onco-Hematology Unit Azienda Ospedaliera Santa Maria of Terni, Terni, Italy; 13Internal Medicine, Ospedale S. Croce, Azienda Ospedaliera Ospedali Riuniti Marche Nord, Presidio of Fano, Fano, Italy; 14Istituto di Ematologia, Università Cattolica, Fondazione Policlinico Gemelli IRCCS, Roma, Italy; 15Hematology Unit, Careggi Hospital, Florence, Italy; 16Department of Biomedical Science, University of Bari "Aldo Moro" Medical School, Internal Medicine "G. Baccelli", Policlinico, Bari, Italy; 17Department of Medicine, Hematology and Clinical Immunology Section, Padua University School of Medicine, Padua, Italy; 18UOC di Ematologia e Trapianto di Cellule Staminali, PO "Vito Fazzi", Lecce, Italy; 19Ospedale A. Perrino, Haematology, Brindisi , Italy; 20Hematology and Bone Marrow Transplant Unit, Istituto Nazionale Tumori, Fondazione “G. Pascale”, IRCCS, Napoli, Italy; 21Unit of Hematology and Stem Cell Transplantation, IRCCS-CROB, Referral Cancer Center of Basilicata, Rionero in Vulture, Italy; 22Department of Hematology, Hospital University Riuniti, Foggia, Italy; 23Department of Hematology and Bone Marrow Transplant, IRCCS Casa Sollievo della Sofferenza, San Giovanni Rotondo, Italy; 24Hematology Unit, IRCCS - Istituto Scientifico Romagnolo per lo Studio e la Cura dei Tumori, Meldola, Italy; 25Hematology Unit, A.O. Spedali Civili, Brescia, Italy; 26 Hematology Unit, AO S.G. Moscati, Aversa, Italy; 27Hematology Department, G. Garibaldi Hospital, Catania, Italy; 28Hematology Unit, Department of Hemato-Oncology and Radiotherapy, Great Metropolitan Hospital "Bianchi-Melacrino-Morelli", Reggio Calabria, Italy; 29Section of Hematology, Department of Medicine, University of Verona, Verona, Italy; 30Hematology Unit and Pathology Department, S. Eugenio Hospital Rome, Rome, Italy; 31Department of Hematology, DISM, University of Udine, Udine, Italy; 32Stem Cell Transplantation Unit, Department of Hemato-Oncology and Radiotherapy, Great Metropolitan Hospital "Bianchi-Melacrino-Morelli", Reggio Calabria, Italy; 33Institute of Haematology and Stem Cell Transplantation, Ospedale Santa Maria della Misericordia, University of Perugia, Perugia, Italy; 34Department of Hematology, DISM, University of Udine, Udine, Italy; 35Complex Operative Unit of Hematology, S. Maria di Loreto Nuovo Hospital, Napoli, Italy; 36Hematology Unit, AUSL-IRCCS, Reggio Emilia, Italy; 37Hematology Unit, Fondazione IRCCS Ca' Granda, Ospedale Maggiore Policlinico, Milano, Italy; 38 Nephrology Center of National Research Institute of Biomedicine and Molecular Immunology, Reggio Calabria, Italy; 39CNR-IFC, Roma, Italy; 40Division of Hematology, University of Torino, AOU Città della Salute e della Scienza di Torino, Torino, Italy; 41Seràgnoli Institute of Hematology and Medical Oncology, University of Bologna, Bologna, Italy; 42Department of Cellular Biotechnologies and Hematology, Sapienza University of Rome, Rome, Italy; 43Hematology Unit, AOU Ospedali Riuniti di Ancona, Ancona, Italy and 44Hemato-Oncology Department, Augusta Victoria Hospital of Jerusalem, Jerusalem, Israel

294

Correspondence: MASSIMO GENTILE - massim.gentile@tiscali.it doi:10.3324/haematol.2019.241513 Disclosures: No conflicts of interests to disclose. Contributions: MG, GS, VDS, FDR, MTP, MO, MB, MC and FM designed the study, analyzed and interpreted data, and wrote the manuscript; MG, GT, AP and FM performed statistical analysis; DD, MG, CB, SR, CC, CCa, NG, SG, EA, AL, MB, ER, EA, RR, RZ, NDR, GM, GMa, PM, SC, NC, CC, AB, CCr, GU, PC, EV, VF, DV, ABo, AS, RS, MM, SB, DG, AG, BG, AP, FF, UC, SB, EZ, FP, SP and AGR provided the patients and collected clinical data; all authors gave their final approval for the manuscript.

References 1. Touzeau C, Moreau P, Dumontet C. Monoclonal antibody therapy

in multiple myeloma. Leukemia. 2017;31(5):1039-1047. 2. Durer C, Durer S, Lee S, et al. Treatment of relapsed multiple myelo-

ma: evidence-based recommendations. Blood Rev. 2020;39:100616. 3. Collins SM, Bakan CE, Swartzel GD, et al. Elotuzumab directly

enhances NK cell cytotoxicity against myeloma via CS1 ligation: evidence for augmented NK cell function complementing ADCC. Cancer Immunol Immunother. 2013;62(12):1841-1849. 4. Hsi ED, Steinle R, Balasa B, et al. CS1, a potential new therapeutic antibody target for the treatment of multiple myeloma. Clin Cancer Res. 2008;14(9):2775-2784. 5. Zonder JA, Mohrbacher AF, Singhal S, et al. A phase 1, multicenter, open-label, dose escalation study of elotuzumab in patients with advanced multiple myeloma. Blood. 2012;120(3):552-559. 6. Lonial S, Dimopoulos M, Palumbo A, et al. Elotuzumab therapy for relapsed or refractory multiple myeloma. N Engl J Med. 2015;373(7):621-631. 7. Dimopoulos MA, Lonial S, White D, et al. Elotuzumab plus lenalidomide/dexamethasone for relapsed or refractory multiple myeloma: ELOQUENT-2 follow-up and post-hoc analyses on progression-free survival and tumour growth. Br J Haematol. 2017;178(6):896-905. 8. Dimopoulos MA, Lonial S, Betts KA, et al. Elotuzumab plus lenalidomide and dexamethasone in relapsed/refractory multiple myeloma: extended 4-year follow-up and analysis of relative progression-free survival from the randomized ELOQUENT-2 trial. Cancer. 2018;124(20):4032-4043. 9. Lonial S, Dimopoulos MA, Weisel K, et al. Extended 5-y follow-up (FU) of phase 3 ELOQUENT-2 study of elotuzumab + lenalidomide/dexamethasone (ELd) vs Ld in relapsed/refractory multiple myeloma (RRMM). J Clin Oncol. 2018;36(Suppl 15):8040. 10. Willan J, Eyre TA, Sharpley F, Watson C, King AJ, Ramasamy K. Multiple myeloma in the very elderly patient: challenges and solutions. Clin Interv Aging. 2016;11:423-435. 11. Offidani M, Corvatta L. A review discussing elotuzumab and its use in the second-line plus treatment of multiple myeloma. Future Oncol. 2018;14(4):319-329. 12. Moreau P, San Miguel J, Sonneveld P, et al. Multiple myeloma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28(Suppl 4):iv52-iv61. 13. Network NCC. Multiple myeloma (version 4.2018). Available from: http://www.nccn.org/ [Last access 9 Aug 2018] 14. Dimopoulos MA, Dytfeld D, Grosicki S, et al. Elotuzumab plus pomalidomide and dexamethasone for multiple myeloma. N Engl J Med. 2018;379(19):1811-1822. 15. Thomas SK, Shah JJ, Morphey AN, et al. Updated results of a phase II study of lenalidomide-elotuzumab as maintenance therapy postautologous stem cell transplant (AuSCT) in patients (Pts) with multiple myeloma (MM). Blood. 2018;132(Suppl 1):1982.

haematologica | 2021; 106(1)

Letters to the Editor

state. We compared the bias (eGFR-mGFR) in patients with repeated measures and categorized patients as having low intrapatient variability if the absolute difference in the bias on repeated measures was <5 mL/min/1.73m2. Summary statistics including mean and standard deviation (SD) for continuous variables and counts and percentages for categorical variables were reported and compared using statistical tests as appropriate. For the agreement analyses between mGFR and eGFR, mean bias (95% limits of agreement) and the SD of bias were calculated using Bland-Altman methods assuming constant variance. A t-test and F-test were used to compare bias and variance between the CKiD equation and four other pediatric eGFR equations. The Pearson correlation (r) or Spearman rank correlation (r), and Lin concordance correlation (CCC) with 95% confidence intervals are presented. The percentage difference between eGFR and mGFR and the percentage of values within ±10% (P10[%]) and ±30% (P30[%]) are presented. A linear regression model was used to assess the effects of age, treatments and/or their interaction on bias between eGFR and mGFR. Next, we analyzed patients with repeated measurements. We performed generalized least squares models to assess the effects of age, treatments and/or their interaction on bias. We modeled the correlations of repeated measurements on the same subjects using a compound symmetry correlation structure. All P-values were two-sided and considered statistically significant when <0.05. Analyses were performed using SAS v9.4 (Cary, NC, USA) and R-3.5.2 (Vienna, Austria). Three-hundred sixty-four 99mTc-DTPA mGFR examinations were performed in 198 subjects. Among the 198 individuals with an initial mGFR, 196 also had SCr and 124 had SCr and CyC measured within 4 weeks of mGFR. The median age of the 198 participants at the time of their initial mGFR was 8.2 years (range, 2.1-18.0). The mean (± SD) mGFR was 141± 26 mL/min/1.73 m2. Eighty-nine (45%) participants were female. No difference was observed in the mean age of female and male participants (8.6 vs. 8.5 years, P=0.83). However, the mean mGFR (± SD) was significantly higher in males than females (145±26 vs. 137±26 mL/min/1.73 m2, P=0.024). At the time of mGFR, participants were receiv-

High bias and low precision for estimated versus measured glomerular filtration rate in pediatric sickle cell anemia Individuals with sickle cell anemia (SCA) develop glomerular injury that progresses to chronic kidney disease.1 Measured GFR (mGFR) is the gold standard test for monitoring glomerular function. Pediatric SCA research has used 99mTc-DTPA to determine the mGFR but this test is not feasible for annual monitoring because of its high cost and time commitment.2 A more convenient approach to track glomerular function is to measure serum creatinine (SCr) and/or cystatin C (CyC) and calculate the estimated glomerular filtration rate (eGFR). Several pediatric eGFR equations were validated in nonSCA populations; however, each of these equations has limitations and none has been validated in SCA.3-7 It is imperative to identify the eGFR equation with the least bias of eGFR relative to mGFR and highest precision for SCA clinical care and research. Bias defines the accuracy (eGFR minus mGFR) and standard deviation of this bias defines the precision of eGFR. The Institutional Review Board-approved Sickle Cell Clinical Research and Intervention Program (SCCRIP) requires annual eGFR and mGFR using 99mTc-DTPA every 3 to 6 years (NCT:020988635).8 To determine the accuracy and precision of five pediatric eGFR equations, we tested the hypotheses that: (i) in comparison to mGFR, estimates using the Chronic Kidney Disease in Children (CKiD) eGFR equation would have lowest bias and highest precision; (ii) bias would be similar by therapy and age; and (iii) the intrapatient variability for the bias would be low among patients with repeated measures. Among patients with eGFR and mGFR obtained within a 4-week period enrolled in the SCCRIP study, we performed an agreement analysis between five standard pediatric eGFR equations derived from either SCr, SCr and blood urea nitrogen (BUN), CyC, or SCr and CyC and mGFR by 99mTc-DTPA clearance.5-7 We excluded five participants with either chronic kidney disease or severe hyperfiltration (mGFR: <60 or >240, eGFR: >350 mL/min/1.73m2). We determined eGFR and mGFR at outpatient appointments during clinician-determined steady

Table 1. Bias, precision, agreement and accuracy of the estimated compared to glomerular filtration rate (GFR) compared to 99mTc-DTPA measurements of GFR.

eGFR equation

Pediatric creatinine-based equations Schwartz creatinine-based equation (2009) Schwartz creatinine BUN- based equation (2009) Pediatric cystatin-based equations Schwartz cystatin C-based equation (2012) Filler equation (2003) Pediatric creatinine-cystatin equation Pediatric CKiD creatinine-cystatin equation (2012)

Bias (95% limits of agreement) mGFR-eGFR

SD of bias

CCC

196 196

-21.4 (-80, 37.2)* 10.7 (-36.4, 57.8)*

29.9# 24.02#

0.38 (0.28, 0.47) 0.44 (0.33, 0.53)

126 126

47.7 (-1.8, 97.2)* 12.5 (-43.1, 68.1)

25.3# 28.4#

124

17.0 (-22.0, 55.9)

19.86

Correlation coefficient (r or r)

P10 (%)

P30 (%)

r=0.49a r=0.50

34.7% 41.3%

69.9% 92.9%

0.08 (0.04, 0.12) 0.3 (0.15, 0.44)

r=0.36a r=0.36a

4.8% 26.6%

36.3% 87.1%

0.44 (0.34, 0.53)

r=0.66

41.9%

95.2%

Schwartz creatinine-based equation: eGFR= 41.3 x (height (m)/SCr); Schwartz creatinine/BUN-based equation: eGFR= 40.7 x (height (m)/SCr)0.64x(30/BUN)0.202; Schwartz cystatin C-based equation: eGFR=70.69 x (SCysC)-0.931; Filler cystatin C equation: log (GFR) = 1.962 + [1.123 × log (1/CysC)]; CKiD equation: GFR=39.8 x [height/SCr]0.456 x [1.8/CysC]0.418 x [30/BUN]0.079 x [1.076male] [1.00 female] x [height (m)/1.4]0.179 *P-value <0.05 for the comparison of bias in the other equations to that in the Pediatric CKiD creatinine-cystatin equation based on a two-sided t-test. #P-value <0.05 for the comparison of variance in bias in the other equations to that in the Pediatric CKiD creatininecystatin equation based on a two-sided F-test. a their Pearson correlation coefficients (r) are almost identical to r. N: number; mGFR: measured glomerular filtration rate; eGFR: estimated glomerular filtration rate; SD; standard deviation; SD: standard deviation, CCC: Lin concordance correlation coefficient, r: Spearman rank correlation, r: Pearson correlation, P10%: percentage of eGFR values within ±10% of GFR; P30%: percentage of eGFR values within ±30% of GFR; SCr: serum creatinine; CysC: cystatin C; BUN: blood urea nitrogen; CKiD; Chronic Kidney Disease In Children. 10

haematologica | 2021; 106(1)

295

Letters to the Editor

ing either hydroxyurea alone (n=73), hydroxyurea and chronic transfusion therapy (n=17), chronic transfusion therapy alone (n=12), or no SCA-modifying therapy (n=96). We present the mean bias, precision (SD of bias), and agreement (Lin concordance correlation, Pearson/Spearman rank correlation, P10 and P30) of the fit for mGFR as compared to five eGFR equations. (Table 1, Figure 1). The Filler CyC or the Schwartz SCr/BUN eGFR equations had the smallest mean bias. The CKID

eGFR equation had the lowest SD, highest correlation (r=0.66), concordance correlation (0.44), and P30. We compared the bias in four eGFR equations to the that of the CKiD equation; the CKiD equation had significantly lower bias than the Schwartz SCr equation (P=2Ă&#x2014;10-22) but did not have statistically lower bias than the other three eGFR equations. (Table 1, Figure 1). The CKiD equation had a statistically significant lower SD than all other equations (P<0.05).

Figure 1. Correlation between the results of five equations to estimate glomerular filtration rate and glomerular filtration rate measured with DTPA. GFR: glomerular filtration rate; CCC: Lin concordance correlation coefficient; 95% CI: 95% confidence interval; BUN: blood urea nitrogen; CKiD: Chronic Kidney Disease in Children; CyC: cystatin C; SCr: serum creatinine.

296

haematologica | 2021; 106(1)

Letters to the Editor

Table 2. Change in measured glomerular filtration rate (mGFR) on repeat evaluation versus change in estimated glomerular filtration rate (eGFR) equations on repeat evaluation and number (%) of participants with high accuracy of repeat change. (mGFR – mGFR ) – (eGFR – eGFR ). repeat

baseline

Δ mGFR vs. Δ Schwartz Cr

Number 103 Difference in the change in mGFR vs. eGFR -13.8 (20.4) on repeat evaluation (mL/min/1.73 m2), Mean (SD)* Number and % of patients with intrasubject Δ 6 (5.8%) bias† <5 mL/min/1.73 m2 Number and % of patients with intrasubject Δ 18 (17.5%) bias† <10 mL/min/1.73 m2

repeat

Δ mGFR vs. Δ Schwartz BUN Cr

Δ mGFR vs. Δ Schwartz CyC

Δ mGFR vs. Δ mGFR vs. Δ Filler CyC Δ CKiD CyC & Cr

103 15.3 (18.5)

44 45.3 (13.8)

44 10.5 (16.1)

42 18.0 (13.0)

6 (5.8%)

7 (15.9%)

6 (13.6%)

7 (16.7%)

21 (20.4%)

17 (40.5%)

12 (27.3%)

17 (40.5%)

*Mean (SD) of intrasubject bias estimates. †intrasubject Δ bias calculated as range of bias estimates. Δ mGFR; mGFR Cr: creatinine; BUN; blood urea nitrogen; CysC: cystatin C; CKiD: Chronic Kidney Disease in Children; SD: standard deviation.

repeat

We sought to investigate the association of age and therapy on the bias. Age was not significantly associated with bias except when using the Schwartz SCr equation (Schwartz SCr vs. mGFR P=0.005; other P-values: 0.30.9). Therapy was not significantly associated with bias between mGFR or any of the five eGFR equations (range of P-values: 0.2-0.6). Finally, we did not identify an association of age with bias that was modified by hydroxyurea therapy for any of the five eGFR equations. We analyzed the intrapatient variability in the bias among participants who had initial and at least one additional mGFR and eGFR measurements. Using all five eGFR equations, the mean difference in the bias between the initial mGFR and eGFR and the bias on the repeat evaluation for individuals ranged from 10.5 mL/min/1.73m2 (using the Filler CyC equation) to 45.3 mL/min/1.73m2 (using the Schwartz CyC equation) (Table 2). No eGFR equation identified more than 20% of patients with a change in bias from baseline that was within 5 mL/min/1.73m2 of the bias on the repeated measures. Our findings show that the Schwartz SCr/BUN (10.7 mL/min/1.73m2) and the Filler CyC (12.5 mL/min/1.73m2) equations had the lowest bias while the CKiD equation had the highest correlation (0.66). In comparison, the CKiD equation was developed and validated with a bias of -0.2 mL/min/1.73m2 and a correlation of 0.92.5 This comparison highlights the limitations of using the current eGFR equations as validated in pediatric SCA research and clinical care. Despite the overall limitation in the bias and precision, our data suggest that clinicians monitoring annual eGFR should obtain SCr and CyC to decrease the bias and imprecision of SCr-alone eGFR equations. Our data were limited to children; additional research is needed in adult SCA.9 In one recent study in SCA adults (n=12) comparing mGFR (determined using iohexol) and eGFR equations, the eGFR calculated by CyC alone again had the lowest bias (0.2 mL/min/1.73 m2) but high imprecision (SD 26.3 mL/min/1.73 m2).10 A second important finding of our study is that high intrapatient variability in the bias was observed using repeated evaluations. It is well established that systematic bias exists between mGFR and eGFR.9,11 Researchers may accept this bias if it is consistent on repeated measures. The CKiD study demonstrated this low intrapatient variability on repeated measures; the annual change in the bias from baseline (mGFR-eGFR) to annual followup (mGFR-eGFR) was approximately 1 mL/min/1.73 haematologica | 2021; 106(1)

baseline

– mGFR

baseline

). Δ eGFR (eGFR

repeat

– eGFR

baseline

m2.12 Our data using the CKiD equation identified a mean difference in the bias on repeat evaluations of 18 mL/min/1.73 m2. This high intrapatient variability in the bias on repeat evaluations should preclude assumptions in a pediatric SCA clinical trial using renoprotective agents that eGFR changes from baseline to exit confirm the magnitude or direction of changes in mGFR. Next, this study identifies sex differences in mGFR in pediatric SCA. The mGFR in males was significantly higher than in females, which replicates SCA murine data that male mice develop a higher mGFR than females.13 A GFR difference by sex has not been well studied in SCA; however, some adult SCA studies have found that male patients with chronic kidney disease have a greater annual decline in eGFR, increased risk of acute kidney injury, and a higher mortality rate as compared to females.14,15 Next, age did not influence the bias between mGFR in four of the five eGFR equations and hydroxyurea therapy did not affect the bias between mGFR and five eGFR equations. Several novel findings relevant to clinicians and researchers emerge from this study; however, some limitations are worth noting. First, CyC and SCr levels were accepted if performed within 4 weeks of mGFR. Second, adult mGFR data were not available. Therefore, future prospective research should perform all tests on the same day and include high-risk adult participants with eGFR <60 mL/min/1.73 m2. In conclusion, we demonstrate that for pediatric clinical care, annual eGFR equations should include SCr and CyC although recognizing the limitations of this approach. For natural history studies, we suggest using mean eGFR over several time points to minimize imprecision. Pediatric trials of novel renoprotective therapies should use mGFR. There is an urgent need to develop either a more precise eGFR equation validated for SCA or an efficient, economical mGFR method. Validated tests of glomerular function are essential to reduce the morbidity and mortality associated with SCA kidney disease. Jeffrey D. Lebensburger,1 Jeffrey Gossett,2 Rima Zahr,3 Winfred C. Wang,4 Kenneth I. Ataga,5 Jeremie H. Estepp,4 Guolian Kang2 and Jane S. Hankins4 1 Department of Pediatrics, University of Alabama, Birmingham, AL; 2 Department of Biostatistics, St. Jude Children’s Research Hospital, University of Alabama at Birmingham, Birmingham, AL; 3Division of Pediatric Nephrology and Hypertension, University of Tennessee Health Science Center, Memphis, TN; 4Department of Hematology, St. Jude Children’s Research Hospital, Memphis TN and 5Center for 297

Letters to the Editor

Sickle Cell Disease, University of Tennessee Health Science Center, Memphis, TN, USA Correspondence: JEFFREY D. LEBENSBURGER - jlebensburger@peds.uab.edu doi:10.3324/haematol.2019.242156 Disclosures: JDL is a consultant for Novartis and has received funding from Pfizer ASPIRE. JHE receives research support from Pfizer and Eli Lilly and Co.and serves as a consultant for Daiichi Sankyo and Global BloodTherapeutics; WCW receives research support from Novartis and Agios; JSH receives research support from Global Blood Therapeutics; KIA is a member of an advisory board for Novartis, Global Blood Therapeutics, Novo Nordisk, Editas Medicine, and Bioverativ. Contributions: JDL and JSH developed the concept and wrote the manuscript. JG and GK performed the data analysis. JHE, KIA, WCW, RZ assisted in the study design and manuscript preparation.

6. 7. 8.

9. 10.

11.

References 12. 1. Sharpe CC, Thein SL. How I treat renal complications in sickle cell disease. Blood. 2014;123(24):3720-3726. 2. Wang WC, Ware RE, Miller ST, et al. Hydroxycarbamide in very young children with sickle-cell anaemia: a multicentre, randomised, controlled trial (BABY HUG). Lancet. 2011;377(9778):1663-1672. 3. Aygun B, Mortier NA, Smeltzer MP, Hankins JS, Ware RE. Glomerular hyperfiltration and albuminuria in children with sickle cell anemia. Pediatr Nephrol. 2011;26(8):1285-1290. 4. McPherson Yee M, Jabbar SF, Osunkwo I, et al. Chronic kidney disease and albuminuria in children with sickle cell disease. Clin J Am Soc Nephrol. 2011;6(11):2628-2633. 5. Schwartz GJ, Schneider MF, Maier PS, et al. Improved equations esti-

298

13. 14. 15.

mating GFR in children with chronic kidney disease using an immunonephelometric determination of cystatin C. Kidney Int. 2012;82(4):445-453. Filler G, Lepage N. Should the Schwartz formula for estimation of GFR be replaced by cystatin C formula? Pediatr Nephrol. 2003;18(10):981-985. Schwartz GJ, Munoz A, Schneider MF, et al. New equations to estimate GFR in children with CKD. J Am Soc Nephrol. 2009;20(3):629637. Hankins JS, Estepp JH, Hodges JR, et al. Sickle Cell Clinical Research and Intervention Program (SCCRIP): a lifespan cohort study for sickle cell disease progression from the pediatric stage into adulthood. Pediatr Blood Cancer. 2018;65(9):e27228. Asnani M, Reid M. Cystatin C: a useful marker of glomerulopathy in sickle cell disease? Blood Cells Mol Dis. 2015;54(1):65-70. Yee ME, Lane PA, Archer DR, Joiner CH, Eckman JR, Guasch A. Losartan therapy decreases albuminuria with stable glomerular filtration and permselectivity in sickle cell anemia. Blood Cells Mol Dis. 2018;69:65-70. Yee MEM, Lane PA, Archer DR, Joiner CH, Eckman JR, Guasch A. Estimation of glomerular filtration rate using serum cystatin C and creatinine in adults with sickle cell anemia. Am J Hematol. 2017;92(10):E598-E599. Ng DK, Schwartz GJ, Warady BA, Furth SL, Munoz A. Relationships of measured iohexol GFR and estimated GFR with CKD-related biomarkers in children and adolescents. Am J Kidney Dis. 2017;70(3):397-405. Kasztan M, Fox BM, Lebensburger JD, et al. Hyperfiltration predicts long-term renal outcomes in humanized sickle cell mice. Blood Adv. 2019;3(9):1460-1475. Stallworth JR, Tripathi A, Jerrell JM. Prevalence, treatment, and outcomes of renal conditions in pediatric sickle cell disease. Southern Med J. 2011;104(11):752-756. McClellan AC, Luthi JC, Lynch JR, et al. High one year mortality in adults with sickle cell disease and end-stage renal disease. Br J Haematol. 2012;159(3):360-367.

haematologica | 2021; 106(1)

Letters to the Editor

Low incidence of EPOR mutations in idiopathic erythrocytosis Since the discovery of JAK2 mutations, the etiological diagnosis of erythrocytosis has been greatly facilitated. When positive, the diagnosis of polycythemia vera (PV) is highly probable; when negative, it is necessary to focus on secondary causes, particularly hypoxia due to cardiorespiratory diseases or tumor-related causes. However, the cause of erythrocytosis often remains unknown even after extensive investigation, which leads to the diagnosis of idiopathic erythrocytosis, or primary familial and congenital polycythemia. We report here the low incidence of mutation (1.1%) in the erythropoietin receptor (EPOR) gene in a large series of 270 consecutive unrelated cases with primary familial and congenital polycythemia. Three mutations in the EPOR gene were observed, two of which have already been reported as pathogenic (Pro381Glnfs*2 and Ser407*). The third mutation is a new variant (Ser432Alafs*21) that was discovered in one patient from New Caledonia. Interestingly, a different French laboratory found the same variant in three other patients from New Caledonia without known family ties. Erythrocytosis and polycythemia are common terms referring to the increase of hematocrit (Ht) and hemoglobin (Hb) concentrations. However, to ensure a diagnosis of true erythrocytosis, certain criteria must be met. Testing must reveal a red cell mass (RCM) >125%1 or Ht >56% in women or >60% in men. Primary erythrocytosis is caused by an intrinsic abnormal proliferation of the erythroid compartment. The most common is PV, the diagnosis of which is well standardized according to the 2016 World Health Organization (WHO) criteria,2 which includes the JAK2 mutations, found in 98% of cases. Erythrocytosis can also result from an acquired abnormality that stimulates erythropoietin (EPO) secretion, resulting in secondary polycythemias. Increased secretion of EPO can arise from hypoxia pathway activation, often observed in heart or lung diseases, and also from ectopic secretion of EPO by tumors. Congenital secondary erythrocytosis results, among others, from hemoglobin with high oxygen affinity due to Îą, b or Îł globin mutations, or from a very rare mutation in the bisphosphoglycerate mutase (BPGM) coding gene that causes 2,3-BPG levels to decrease.3 Yet many cases of polycythemia remain idiopathic because the most common causes are not identified. Since the advent of molecular biology and next-generation sequencing (NGS), the genes involved in the regulation of the hypoxia sensing pathway (PHD2/EGLN1, HIF2A/EPAS1, VHL, EPO) or in proliferation and differentiation of erythroid progenitors (JAK2 or the EPO recep-

tor gene [EPOR]) are more easily identified and their pathological pathways can be clarified. As recently reviewed, new mutations in the HIF/EPO pathway have been reported in the last few years, underscoring the rapid evolution of current knowledge on the causes of erythrocytosis.4 For example, we recently demonstrated that exon skipping and cryptic exon retention were new VHL alterations and revealed a novel complex splicing regulation of the VHL gene.5 Over the past 4 years, 270 patients (229 males, 41 females) with idiopathic erythrocytosis have been tested using a dedicated NGS panel of genes involved: (i) in the regulation of the hypoxia pathway (PHD1 [EGLN2], PHD2 [EGLN1], PHD3 [EGLN3], HIF-1A, HIF-2A [EPAS1], HIF-3A, VHL, VHLL); (ii) in proliferation and differentiation of erythroid progenitors (EPO, EPOR, JAK2, LNK [SH2B3], CBL); or (iii) in mature cell function (bisphosphoglyceratemutase (BPGM]). In order to restrict our analysis to truly idiopathic polycythemia, patients were selected according to clinical and biological criteria: increased RCM >125% (measured using a chromium 51 method) or Ht value >56% in women or >60% in men were necessary for the confirmation of absolute erythrocytosis. Each patient underwent testing for blood electrolytes, serum EPO, JAK2 mutations (on exon 14 and exon 12), functional respiratory tests, abdominal ultrasound scan, arterial and venous blood gas (in search of a P50<24 mmHg that could be indicative of Hb with high oxygen affinity) in order to eliminate classical secondary causes. In particular, the measurement of venous P50 is a rapid, low-cost test that can orientate further screening towards an inexpensive sequencing technique for the HBB, HBA1 and HBA2 genes.6,7 Finally, because the detection of variants in the genes with high similarity of sequences such as HBA1 and HBA2 is challenging using NGS, we decided not to include HBB, HBA1 and HBA2 genes in our NGS panel.8 Patients were included in the NGS program if their results showed no abnormalities in the criteria cited above. Using the NGS panel, variants considered as pathogen were observed in 62 (23%) patients, mainly in the genes involved in the regulation of the hypoxia pathway. Concerning the EPOR gene, only three different mutations, all in exon 8, were identified in three of the 270 patients that were tested (Table 1). The first anomaly, a frameshift mutation (c.1142_1143delCC, p.Pro381Glnfs*2), was found in patient #1: a 52-year old woman. A family history of probable erythrocytosis in her father and brother was reported, but further information was not available. The second EPOR mutation, a nonsense mutation (c.1220C>A, p.Ser407*), was found in patient #2, a 14year old girl, who had been diagnosed with idiopathic

Table 1. Hematologic and genetic data of the probands with EPOR mutations.

Case Age

Sex

RBC Hemoglobin Hematocrit (x1012/L) (Hb, g/L) (Hct, %)

5.23

172

7.1

200

3 4

31 8

M M

6.44 7,3

216 197

MCV (fL)

EPO (UI/L)

Mutation cDNA

Mutation protein

Family History

c.1142_1143delCC

p.Pro381Glnfs*2

<1.0

c.1220C>A

p.Ser407*

<10.5 <0,6

c.1293del c.1293del

p.Ser432Alafs*21 p.Ser432Alafs*21

Yes (father and brother) Yes (mother and sister) Yes (unknown) Yes

50%

61.3 57,5

RBC: red blood cells; MCV: mean corpuscular volume; F: female; M: male.

haematologica | 2021; 106(1)

299

Letters to the Editor

Table 2. Complications and treatment of the probands with EPOR mutations.

Case

Vascular complication

Phlebotomy

1 2 3 4

Yes (myocardial infarction and transient ischemic attack) No No No

Yes (twice/year) Yes No No

erythrocytosis at the age of 13 years. Her Hb was 172g/L when the NGS was performed. She also had a family history of idiopathic erythrocytosis (mother and sister), but the cause was unknown. These two EPOR mutations have already been described9,10 and are characterized as pathogenic. The third mutation is a new frameshift mutation (c.1293del, p.Ser432Alafs*21) that was found in patient #3: a 31-year old male native of New Caledonia now living in continental France. This result was confirmed by Sanger sequencing performed both on DNA extracted from an additional blood sample and from hair follicles in order to ensure the mutation was constitutional. He had been diagnosed with isolated idiopathic erythrocytosis at the age of 20 years. At the time of analysis, he was not showing any sign of splenomegaly nor any other medical condition. Due to his relocation in Europe and estrangement from his family, no data were available regarding a possible family history of erythrocytosis. Since this mutation had not been previously reported in the literature or in databases, we enquired as to whether colleagues involved in EPOR sequencing had found a similar variant. Interestingly, this EPOR variant had indeed been discovered by another French laboratory. Three patients were identified. An 8-year old boy (patient #4) had been referred for isolated erythrocytosis in a context of a family history of idiopathic erythrocytosis in his mother, maternal aunt and uncle, and maternal grandfather who all presented isolated erythrocytosis that had not been previously investigated (Figure 1). The same laboratory had also found this mutation in a 48-year old female with isolated erythrocytosis, and a 54- year old female, also presenting with isolated erythrocytosis. For the second case (48-year old female), investigation of her family history revealed that her niece also carried the mutation. Both cases originated from New Caledonia, but further investigation found no relation between their families. However, the fact that the same, rare variant was identified independently in four patients from New Caledonia strongly supports a founder effect. Since the discovery of the link between mutations in the gene encoding EPOR and erythrocytosis,11 many variants have been described. All are localized in exon 8, which encodes the distal region of the EPO receptor.12 Most of the mutations cause a truncation in the intracellular domain of the receptor. It has now been established that this shortening leads to a loss of internalization and downregulation of the receptor, mostly due to the impairment of negative regulators binding to EPOR. Indeed, SHP1, which is a major inhibitor, binds to the residue p.Tyr454, and another inhibitor, SOCS3, to p.Tyr426, p.Tyr454, and p.Tyr456, all of which are lost in the various truncations described thus far. This results in an excessive activation of the receptor, showing hypersensitivity to EPO, combined with an increased number of receptors at the membrane, eventually leading to erythrocytosis.13-15 Nonetheless, recent studies have shown that 300

Figure 1. Pedigree of patient #4. The propositus is indicated by the arrow. The black circles and squares represent people in the family with erythrocytosis.

the precise mechanisms leading to primary erythrocytosis vary depending on the type of the mutation,16 explaining the wide heterogeneity and complexity of this condition. Indeed, the mutant generates a new C-terminal cytoplasmic part that could induce the EPOR signaling pathway, as demonstrated in a recent study.16 The new variant described here (c.1293del, p.Ser432Alafs*21) is localized in a hot spot area where many pathogenic mutations have previously been reported.13,16-18 Due to its localization and the resulting truncation, it is expected to cause loss of the major binding sites of negative regulators, thus explaining the observed phenotype. The discovery of the same variant in three other families, for a total of four different families all coming from New Caledonia, with a history of erythrocytosis strengthens the hypothesis of a functional impact of the variant. Regarding complications, only patient #1 had a vascular complication as myocardial infarction and transient ischemic attack. On the contrary, neither hemorrhage nor progression to hematologic malignancies was noted in patients with EPOR mutations (Table 2). In the whole cohort of patients with erythrocytosis, EPO values ranged from 0.6 to 140 IU/L, with the vast majority of cases (>90%) having a serum EPO in the normal range (4.3-29 UI/L). Off note, patients with an EPOR mutation all had low serum EPO, which is a classic feature of EPOR mutations. In conclusion, our study confirms that mutations in EPO-R are relatively rare among patients with idiopathic erythrocytosis. However, patients with low serum EPO levels and/or a family history of erythrocytosis should be checked for this anomaly. We also describe a novel mutation with a founder effect in New Caledonia. haematologica | 2021; 106(1)

Letters to the Editor

Mathilde Filser,1 Bernard Aral,2 Fabrice Airaud,3 Aurélie Chauveau,4 Aisha Bruce,5 Yann Polfrit,6 Anne Thiebaut,7 Martin Gauthier,8 Cédric Le Maréchal,9,10 Eric Lippert,4,9 Stéphane Béziau,3,11 Céline Garrec,3 Betty Gardie11,12,13* and François Girodon1,13,14* 1 Service d’Hématologie Biologique, Pôle Biologie, CHU de Dijon, France; 2Laboratoire de Génétique Chromosomique et Moléculaire, Pôle Biologie, CHU de Dijon, France; 3Service de Génétique Médicale, CHU de Nantes, Nantes, France; 4Hématologie Biologique, CHU de Brest, Brest, France; 5Department of Pediatrics, University of Alberta, Edmonton and Stollery Children’s Hospital, Alberta Health Services, Alberta, Canada; 6Service de Pédiatrie, CH NOUMEA, Nouméa, Nouvelle-Calédonie; 7Service d’Hématologie, CHU Grenoble Alpes, Grenoble, France; 8Service d’Hématologie, Toulouse-Oncopole University Cancer Institute (IUCT-O), Toulouse, France; 9Université de Brest, Inserm, EFS, UMR 1078, GGB, Brest, France; 10Laboratoire de Génétique Moléculaire et d’Histocompatibilité, CHU de Brest, Brest, France; 11L’Institut du Thorax, INSERM, CNRS, UNIV Nantes, Nantes, France; 12Ecole Pratique des Hautes, EPHE, PSL Research University, Paris, France; 13Laboratory of Excellence GR-Ex, France and 14Inserm U1231, Université de Bourgogne, Dijon, France *BG and FG contributed equally to this work as co-senior authors. Correspondence: FRANÇOIS GIRODON - francois.girodon@chu-dijon.fr doi:10.3324/haematol.2019.244160 Disclosures: no conflicts of interests to disclose. Contributions: MF and FG wrote the manuscript; AB, YP, AT, MG provided samples; BA, AC, FA, EL, CLM and CG performed analyses; SB and BG revised the manuscript. Funding: this study was supported by grants from the ANR (PRTS 2015 "GenRED"); the labex GR-Ex, reference ANR-11-LABX-0051.

References 1. Pearson TC, Messinezy M, Westwood N, et al. A polycythemia

vera update: diagnosis, pathobiology, and treatment. Hematology Am Soc Hematol Educ Program. 2000:51-68. 2. Barbui T, Thiele J, Gisslinger H, et al. The 2016 WHO classification and diagnostic criteria for myeloproliferative neoplasms: document summary and in-depth discussion. Blood Cancer J. 2018;8(2):15. 3. McMullin MF. Idiopathic erythrocytosis: a disappearing entity. Hematology Am Soc Hematol Educ Program. 2009:629-635.

haematologica | 2021; 106(1)

4. Lappin TR, Lee FS. Update on mutations in the HIF: EPO pathway

and their role in erythrocytosis. Blood Rev. 2019;37:100590. 5. Lenglet M, Robriquet F, Schwarz K, et al. Identification of a new

VHL exon and complex splicing alterations in familial erythrocytosis or von Hippel-Lindau disease. Blood 2018;132(5):469-483. 6. Rumi E, Passamonti F, Pagano L, et al. Blood p50 evaluation enhances diagnostic definition of isolated erythrocytosis. J Intern Med. 2009;265(2):266-274. 7. Oliveira JL, Coon LM, Frederick LA, et al. Genotype-phenotype correlation of hereditary erythrocytosis mutations, a single center experience. Am J Hematol. 2018;93(8):1029-1041. 8. Girodon F, Airaud F, Garrec C, Bézieau S, Gardie B. Gene panel sequencing in idiopathic erythrocytosis. Haematologica. 2017; 102(1):e30. 9. Al-Sheikh M, Mazurier E, Gardie B, et al. A study of 36 unrelated cases with pure erythrocytosis revealed three new mutations in the erythropoietin receptor gene. Haematologica. 2008;93(7):10721075. 10. Toriumi N, Kaneda M, Hatakeyama N, et al. A case of primary familial congenital polycythemia with a novel EPOR mutation: possible spontaneous remission/alleviation by menstrual bleeding. Int J Hematol. 2018;108(3):339-343. 11. de la Chapelle A, Traskelin AL, Juvonen E. Truncated erythropoietin receptor causes dominantly inherited benign human erythrocytosis. Proc Natl Acad Sci U S A. 1993;90(10):4495-4499. 12. Chauveau A, Luque Paz D, Lecucq L, et al. A new point mutation in EPOR inducing a short deletion in congenital erythrocytosis. Br J Haematol. 2016;172(3):475-477. 13. Sokol L, Luhovy M, Guan Y, Prchal JF, Semenza GL, Prchal JT. Primary familial polycythemia: a frameshift mutation in the erythropoietin receptor gene and increased sensitivity of erythroid progenitors in erythropoietin. Blood. 1995;86(1):15-22. 14. Perrotta S, Cucciolla V, Ferraro M, et al. EPO receptor gain-of-function causes hereditary polycythemia, alters CD34+ cell differentiation and increases circulating endothelial precursors. PLoS One. 2010;5(8):e12015. 15. Watowich SS, Xie X, Klingmuller U, et al. Erythropoietin receptor mutations associated with familial erythrocytosis cause hypersensitivity to erythropoietin in the heterozygous state. Blood. 1999;94(7):2530-2532. 16. Pasquier F, Marty C, Balligand T, et al. New pathogenic mechanisms induced by germline erythropoietin receptor mutations in primary erythrocytosis. Haematologica. 2018;103(4):575-586. 17. Furukawa T, Narita M, Sakaue M, et al. Primary familial polycythaemia associated with a novel point mutation in the erythropoietin receptor. Br J Haematol. 1997;99(1):222-227. 18. Bento C, Percy MJ, Gardie B, et al. Genetic basis of congenital erythrocytosis: mutation update and online databases. Hum Mutat. 2014;35(1):15-26.

301

Letters to the Editor

Chemoimmunotherapy with rituximab, cyclophosphamide and prednisolone in IgM paraproteinemic neuropathy: evidence of sustained improvement in electrophysiological, serological and functional outcomes There is ongoing debate as to the optimum treatment for IgM paraproteinemic neuropathy (PPN), with no treatment yet shown objectively to alter the long-term natural history of slow neurological decline. We present a case series of 25 patients with IgM PPN treated between August 2010 and October 2016 with standard R-CP (rituximab, cyclophosphamide, prednisolone) chemoimmunotherapy. This is the first report of detailed 2-year serological, neurological and neurophysiological outcome data in patients treated prospectively with this regimen. The treatment was well tolerated and produced significant improvement for at least 2 years in several neurological, electrophysiological and serological outcomes. Peripheral neuropathy is a well-recognized complication of IgM paraproteinemia, and occurs both in those with a monoclonal gammopathy of undetermined significance (MGUS) and in patients with an underlying lymphoproliferative disorder (LPD). Recent guidelines suggest that patients with renal or neurological complica-

Table 1. Patients' characteristics (n=25). Mean age at presentation, years (range) Gender, n. (%) Duration of neuropathy symptoms prior to treatment Hematologic diagnosis, n. (%)

Paraprotein concentration (g/L) Anti-MAG antibody status Clinical features, n. (%)

Neurophysiology, n. (%) Prior treatment for paraproteinemic neuropathy, n. (%)

tions of paraproteinemia should be reclassified as having MGCS (monoclonal gammopathy of clinical significance)1 to reflect the fact that they may merit treatment. In 50-60% of cases, the M protein shows reactivity to a neural antigen known as myelin associated glycoprotein (MAG).2 There is accumulating evidence that the antiCD20 monoclonal antibody, rituximab, benefits a proportion of patients with IgM paraproteinemic neuropathy.3-6 However, there is insufficient evidence to assess the efficacy of rituximab alone compared with its use as part of a combined chemoimmunotherapy (CIT) regimen. International guidelines recommend that in patients with progressive disability due to anti-MAG neuropathy, immunosuppressive treatment should be considered as an alternative to rituximab monotherapy.7 A retrospective analysis of 45 PPN patients treated with a variety of treatment protocols also suggests a role for CIT.8 Finally, a recent demonstration of a high prevalence of mutated myeloid differentiation factor 88 (MYD88), which is strongly associated with Waldenstrom’s macroglobulinemia (WM), in patients with anti-MAG neuropathy supports the use of treatment regimens effective in WM.9 Here we report a large prospective case series of patients with progressive disability due to confirmed IgM paraproteinamic neuropathy treated with standard R-CP chemoimmunotherapy. R-CP is standard R-CVP, minus

68 (50-81) Male: 20 (80) Female: 5 (20) Median 5 years (range 1-20) IgM MGCS: 16 (64) WM: 8 (32) CLL : 1 (4) Median 4.7 (range 1.6-17.9) Anti-MAG positive: 18 (72) (>1,000 Bühlmann titer units) Anti-MAG negative: 7 (28) (a) Defined neurological rating scales:* Value of median baseline ONLS: 3 Motor weakness (MRC sum score <80): 18 (72) Sensory sum score >10: 20 (80) (b) Clinical opinion based on patient-reported symptoms and/or examination findings: Functional impairment affecting quality of life: 25 (100) Distal limb numbness or paresthesiae: 25 (100) Sensory ataxia: 18 (72) Demyelinating neuropathy: 19 (76) Demyelinating / axonal neuropathy: 6 (24) None: 20 (80) Intravenous immunoglobulin (IVIG): 2 (8) Steroids: 2 (8) IVIG, cyclophosphamide and plasma exchange: 1 (4)

MCGS: monoclonal gammopathy of clinical significance; WM: Waldenstrom’s macroglobulinemia; CLL: chronic lymphocytic leukemia; MAG: myelin-associated glycoprotein. *Explanation of the neurological rating scales: ONLS: Overall Neuropathy Limitation Score. This reflects patients’ functional ability in the arms (range 0-5) and legs (range 0-7) (higher scores represent increasing levels of disability). A change of 1 point in the ONLS could for example reflect a change from walking with or without an aid. MRC sum score: this reflects muscle strength throughout the limbs. Note that in IgM paraproteinemic neuropathy, the muscles involved are predominantly the distal leg muscles, and the MRC sum score in this condition would rarely be lower than 70. A drop of just 1 point to a score of 79/80 could for example represent the development of focal weakness in the foot, and so is of clinical importance. Sensory sum score: normal sensation is equivalent to a score of 0, and worsening sensory impairment in 4 modalities (light touch, pinprick, vibration, and joint position sense) is represented by an increasing score (range 0-64). A change of 4 points could for example represent numbness receding from the knees to the ankles in both legs, leading to improved balance due to better proprioception.

302

haematologica | 2021; 106(1)

Letters to the Editor

vincristine to avoid neurological toxicity. It is very similar to the DRC (dexamethasone, rituximab, cyclophosphamide) protocol10 used frequently in the treatment of WM, and similar to immunosuppressive protocols used in rheumatological disorders. All patients had IgM paraproteinemia, fulfilling criteria either for WM or for MGCS, together with electrophysiological evidence of neuropathy. Comprehensive assessment was undertaken to establish a likely causal relationship between the IgM paraprotein and the neuropathy, and to exclude other conditions such as diabetic or vasculitic neuropathy. In total, 33 patients have started treatment. We report results from 25 patients at one year and from 18 patients at 2 years (i.e., those who have reached these time points) (Table 1). Baseline investigations included bone marrow biopsy and computed tomography (CT) body scan, as well as full standard hematology and biochemistry screen, serum protein electrophoresis, and anti-MAG antibody titers using an enzyme-linked immunosorbent assay (ELISA). MYD88 testing was not carried out as the patients presented here pre-dated the routine clinical use of this assay. Following fully informed written consent, treatment comprised six cycles (every 21 days) of the following regimen: rituximab 375 mg/m2 intravenously and cyclophosphamide 750 mg/m2 intravenously on day 1; and prednisolone 50 mg/m2 orally on days 1-5. All patients completed the full treatment protocol and all received prophylaxis for at least 6 months post treatment to reduce treatment-related complications. The treatment protocol was generally well tolerated.

Complications were rare: one flu pneumonitis, one dental abscess, and seven other minor complications, all of which resolved fully. This tolerance compares favorably with the larger rituximab monotherapy randomized controlled trial in which eight serious adverse events occurred in 26 patients.4 It has recently emerged that, out of 33 patients who have been treated in total, three patients have since developed IDH1 (isocitrate dehydrogenase) negative glioblastoma. Two patients were anti-MAG positive with MGCS, one patient was anti-MAG negative with underlying WM. Given that the incidence of glioblastoma multiforme (GBM) is 3 in 100,000, this is concerning. At this stage it is unclear whether the GBM cases: (i) are a random occurrence; (ii) relate to the known association between WM and GBM;11 (iii) reflect an unknown shared risk factor between IgM PPN and GBM; or (iv) relate to the use of CIT in IgM PPN. The last point seems unlikely as, firstly, epidemiological studies of GBM show no link between immunosuppression and GBM incidence,12 and, secondly, this CIT regimen has been used widely in hematology and rheumatology for over 20 years, with no suggestion of a link with GBM. The serology results in individual patients are shown in Figure 1. At 2 years post treatment completion, paraprotein concentration fell compared to baseline in 17 of 18 (94%) patients (Figure 1A). Median paraprotein concentration fell from 4.7 g/L at baseline to 2.0 g/L at 2 years (P=0.000*). Twelve of 13 anti-MAG positive patients showed a reduction in anti-MAG titer compared to baseline (Figure 1B). Median anti-MAG titer at baseline was

titer

Figure 1. Response of IgM paraprotein and anti-myelin associated glycoprotein (MAG) titers to R-CP (rituximab, cyclophosphamide, prednisolone) chemoimmunotherapy treatment. (A) IgM titer at baseline and 2 years post treatment. (B) Anti-MAG titer at baseline and 2 years post treatment. Patients 1-13 in Figure 1A are the anti-MAG positive patients, and these correspond to patients 1-13 in Figure 1B

haematologica | 2021; 106(1)

303

Letters to the Editor

38,956 BĂźhlmann titer units (BTU), falling to 14,783 BTU at 2 years (P=0.002*). Neurological symptoms and disability were measured using the Overall Neuropathy Limitation Score (ONLS).13 A 1 point change in ONLS reflects a meaningful change in patient function, e.g., the difference between walking with or without an aid. Median ONLS showed a significant improvement from 3 at baseline to 2 at 1 year (P=0.006*) and to 1 at 2 years (P=0.053). Motor function was quantified using the Medical Research Council (MRC) Sum Score of muscle power (range 0-80),14 in which, e.g., a loss of just 1 point equates to the development of a partial footdrop. Eighteen patients (72%) had motor weakness at baseline, and 13 have 2-year followup data: MRC sum score in these patients showed a significant improvement in the median score (79) compared to baseline (76) (P=0.031*), and 9 of 13 patients (69%) improved by at least 1 point. Sensory Sum Score reflects the extent of sensory loss in limbs:14 e.g., a change of 4 points reflects numbness receding from the knees to the ankles in both legs. Sensory Sum Score improved by at least 4 points in 9 of 18 (50%) patients compared to baseline. On direct questioning about their symptoms 2 years post treatment, 65% of patients reported either improvement (8 of 17 patients, 47%) or stabilization (3 of 17 patients, 18%). Electrophysiology provides an objective, quantifiable measure of nerve function, which allows us to demon-

strate whether the neuropathy is progressing, stable or improving. Our electrophysiological studies (see Figure 2 for definitions) showed that the mean distal motor latency (DML) sum score improved from 10.7 milliseconds (ms) at baseline to 9.1 ms at 2 years (P=0.005*) (Figure 2A). Sensory nerve action potential (SNAP) sum score also showed a trend to improvement, increasing from 11.7 microvolts (uv) to 15.7 uv at 2 years, (P=0.26) (Figure 2B). Additional neurophysiology is presented in eight off-protocol patients, as there is little in the literature regarding long-term changes in neurophysiology in patients with IgM PPN. In these patients, both motor and sensory nerve conduction studies deteriorated over a 54month period. The mean SNAP sum score significantly worsened from a baseline of 17.4 uv to 8.2 uv at 54 months (P=0.028*) (Figure 2D). The mean DML sum score also deteriorated from a baseline mean of 12.9 ms to 15.1 ms over the same period (P=0.069) (Figure 2C). The clinical importance of these findings is notable: we demonstrate that untreated patients show a progressive deterioration in neurophysiology which, despite not being directly comparable due to the differing follow-up times, is in contrast to those treated with this CIT regimen, who showed a significant improvement in motor response and a trend towards sensory improvement 2 years after treatment. Our case series is not sufficiently large to study the factors which may predict individual patient responses by

Figure 2. Motor and sensory electrophysiological changes 2 years post R-CP (rituximab, cyclophosphamide, prednisolone) chemoimmunotherapy treatment compared with historical off-treatment group. (A) Distal motor latency: mean sum scores at baseline, 12 and 24 months in protocol group: n=23 at 12 months and n=15 at 24 months. (B) Sensory nerve action potentials: mean SNAP sum score at baseline, 12 and 24 months in protocol group: n=23 at 12 months and n=15 at 24 months. (C) Distal motor latency: mean sum score trend in historical off-protocol group over a 54-month period (n=8). (D) Sensory nerve action potentials: mean SNAP sum score trend in historical off-protocol group over a 54-month period (n=8). Error bars represent the standard deviation. DML mean sum score: mean of right median and ulnar nerve distal motor latencies; Mean SNAP sum score: mean of right median, ulnar and radial nerve sensory nerve action potentials; DML: distal motor latency; SNAP: sensory nerve action potential; ms: milliseconds; mv: microvolts.

304

haematologica | 2021; 106(1)

Letters to the Editor

sub-group analysis. We plan to undertake a further analysis at a later time point with a larger cohort of patients in order to correlate neurological response with individual patient factors. This is the first report in IgM paraproteinemic neuropathy of the prospective use of a standardized CIT protocol with detailed outcome measures at 2 years post treatment. Although not a formal randomized comparison, our outcome data compare favorably with the rituximab monotherapy data from the literature. Based on more than 200 rituximab-treated patients, the average number of patients responding to treatment is 30-50% at time points of up to one year,6 whereas our data illustrate improvements in 10 of 18 (55%) patients for ONLS, 9 of 13 patients (69%) for MRC sum score, 9 of 18 patients (50%) for sensory sum score and 11 of 17 patients (65%) for PROMS, at a longer time point of 2 years. However, despite the fact that CIT has been used safely in hematooncology and rheumatology for decades, this case series raises the possibility of an increased incidence of malignant primary central nervous system tumor in patients with IgM PPN treated with such protocols. It is important that other centers treating IgM PPN patients with CIT be vigilant for the occurrence of malignancy, particularly GBM. Further epidemiological study is needed to look at the incidence of GBM in other case series of IgM PPN, with or without treatment, and in patients treated with CIT for other indications. Nancy T. H. Colchester,1 David Allen,2 Haider A. Katifi,1 Tracy Burt,3 Robert N. Lown,3 Ashwin A. Pinto1 and Andrew S. Duncombe3 1 Department of Neurology, Wessex Neurological Centre, University Hospital Southampton; 2Department of Neurophysiology, Wessex Neurological Center, University Hospital Southampton and 3 Department of Haematology, University Hospital Southampton, Southampton, UK Correspondence: NANCY T. H. COLCHESTER - nancy.colchester@uhs.nhs.uk doi:10.3324/haematol.2019.243139 Disclosures: no conflicts of interests to disclose. Contributions: NC analyzed the results, wrote sections of the manuscript and conducted neurological assessments; DA performed, analyzed and wrote up all the neurophysiology; HK oversaw patient selection and reviewed from a neurology perspective; TB designed the database and collated the serological results; RL helped supervise treatment. AP contributed patients for assessment; AD wrote the treatment protocol and guideline, assessed patient suitability, supervised treatment and wrote sections of the manuscript. All authors contributed to reviewing and editing the manuscript.

haematologica | 2021; 106(1)

Acknowledgments: the authors would like to thank the patients for their engagement in the treatment and outcome monitoring. We also thank Rutaba Eshita, Mazen Sabah, Chinar Osman and Annamaria Kiss-Csenki, Consultant Neurologists at University Hospital Southampton (RE and CO) and at Hampshire Hospitals NHS Foundation Trust (MS and AK) for their work in performing the neurological assessments. We also thank Ian Galea, Associate Professor in Neurology at the University of Southampton and Consultant Neurologist, for his advice regarding the statistical analysis.

References 1. Fermand JP, Bridoux F, Dispenzieri A, et al. Monoclonal gammopa-

thy of clinical significance: a novel concept with therapeutic implications. Blood. 2018;132(14):1478-1485. 2. Nobile-Orazio E, Manfredini E, Carpo M, et al. Frequency and clinical correlates of anti-neural IgM antibodies in neuropathy associated with IgM monoclonal gammopathy. Ann Neurol. 1994;36(3):416424. 3. Dalakas MC, Rakocevic G, Salajegheh M, et al. Placebo-controlled trial of rituximab in IgM anti-myelin-associated glycoprotein antibody demyelinating neuropathy. Ann Neurol. 2009;65(3):286-293. 4. Leger JM, Viala K, Nicolas G, et al. Placebo-controlled trial of rituximab in IgM anti-myelin-associated glycoprotein neuropathy. Neurology. 2013;80(24):2217-2225. 5. Lunn MP, Nobile-Orazio E. Immunotherapy for IgM anti-myelin associated glycoprotein paraprotein-associated peripheral neuropathies. Cochrane Database Syst Rev. 2016;10:CD002827. 6. Dalakas MC. Rituximab in anti-MAG neuropathy: more evidence for efficacy and more predictive factors. J Neurol Sci. 2017;377:224226. 7. D’Sa S, Kersten MJ, Castillo JJ, et al. Investigation and management of IgM and Waldenstrom-associated peripheral neuropathies: recommendations from the IWWM-8 consensus panel. Br J Haematol. 2017;176(5):728-742. 8. Hospital M, Viala K, Dragomir S, et al. Immunotherapy-based regimen in anti-MAG neuropathy: results in 45 patients. Haematologica. 2013;98(12):e155-157. 9. Vos JM, Notermans NC, D’Sa A, et al. High prevalence of the MYD88 L265P mutation in IgM anti-MAG paraprotein-associated peripheral neuropathy. J Neurol Neurosurg Psychiatry. 2018; 89(9):1007-1009. 10. Dimopoulos MA, Agagnostopulos A, Kyrtsonis MC, et al. Primary treatment of Waldenstrom’s macroglobulinaemia with dexamethasone, rituximab and cyclophosphamide. J Clin Oncol. 2007; 25(22):3344-3349. 11. Lamaida E, Caputi F, Rapana A, Bracale C, Graziussi G. Waldenstrom’s macroglobulinaemia associated with glioblastoma. A case report. Rev Neurol (Paris). 1996;152(10):637-639. 12. Thakkar JP, Dolecek TA, Horbinski C, et al. Epidemiologic and molecular prognostic review of glioblastoma. Cancer Epidemiol Biomarkers Prev. 2014;23(10):1985-1996. 13. Van Nes SI, Faber CG, Merkies IS. Outcome measures in immunemediated neuropathies: the need to standardise their use and to understand the clinimetric essentials. J Peripher Nerv Syst. 2008; 13(2):136-147. 14. RMC Trial Group. Randomised controlled trial of methotrexate for chronic inflammatory demyelinating polyradiculoneuropathy (RMC trial): a pilot, multicentre study. Lancet Neurol. 2009;8(2):158-164.

305

Letters to the Editor

FcγRIIb-BCR coligation inhibits B-cell receptor signaling in chronic lymphocytic leukemia B-cell receptor (BCR) signaling plays a key role in the pathogenesis of chronic lymphocytic leukemia (CLL) cells, which exhibit higher intrinsic BCR activity than normal B cells.1 In normal B cells, BCR signaling can be inhibited by the low affinity receptor FcγRIIb (CD32b). This receptor exerts its negative effect when coligated with the BCR by cognate immune complexes (IC) and contributes to maintain peripheral tolerance and to shape the B-cell repertoire. FcγRIIb-BCR coligation counteracts BCR signaling by inducing phosphorylation of the intracellular immunoreceptor tyrosine-based inhibition motif (ITIM) of FcγRIIb, and the subsequent recruitment of (SH2)-containing inositol phosphatases (SHIP), which inhibits Ca++ flux and cell proliferation.2,3 Considering the major role of BCR in the pathogenesis of CLL and the activated nature of the leukemic cells,1 we aimed at investigating whether FcγRIIb could regulate BCR signaling in this malignancy. For that purpose, we compared molecular and functional effects of BCR ligation and BCR-FcγRIIb cross-linking on purified B-CLL cells and normal B cells after exposure to equimolar concentrations of F(ab’)2 anti-human immunoglobulin M (IgM) (10 mg/mL) or whole anti-human IgM (15 mg/mL) for 5 minutes. In agreement with previous reports,4,5 BCR ligation induced highly heterogeneous responses on CLL cells. Twelve of 22 (55%) CLL samples responded to BCR

binding with an increase of phospho-AKT (p-AKT) and/or phospho-ERK (p-ERK) levels, as well as an increase of intracellular Ca++ influx (Online Supplementary Figure S1). BCR responsiveness was neither significantly associated with FcγRIIb expression (P=0.053) nor with IgM expression (P=0.307). As a subset of CLL patients show high levels of circulating IC and express poly-reactive BCR, it is reasonable to expect CLL cell interactions with a variety of IC, which could activate FcγRIIb inhibitory signaling. In murine B cells, recruitment of SHIP through FcγRIIb-BCR coligation leads to inactivation of anti-apoptotic kinase AKT7,8 and/or inhibition of ERK.9,10 Nevertheless, the inhibitory effect of FcγRIIb on BCR signaling has been only cursorily investigated in CLL.11,12 Gamberale et al.11 suggested that FcγRIIb may inhibit BCR signaling in CLL cells by reducing ERK phosphorylation. Here, similarly to what we observed in normal B cells (Online Supplementary Figure S2A-C), the crosslinking of FcγRIIb and BCR in leukemic B cells from BCR-responsive patients induced a significant increase in tyrosine phosphorylation of the ITIM domain of FcγRIIb as well as SHIP phosphorylation (pSHIP) and translocation to the membrane (Figure 1). In addition, as observed in normal B cells, FcγRIIb-BCR coligation blocked the BCR-induced increase of p-AKT and p-ERK levels (Figure 1C). These inhibitory effects of FcγRIIb were reverted when coligation with BCR was abrogated by adding the specific anti-FcγRIIb monoclonal antibody 2B6 to the whole anti-IgM pAb. Interestingly, ligation of BCR alone also induced a significant increase

Figure 1. Molecular effects of B-cell receptor (BCR) ligation or FcγRIIb-BCR co-ligation in chronic lymphocytic leukemia cells with responsive BCR. (A) Immunoblot for two representative chronic lymphocytic leukemia (CLL) patient’s samples that responded to BCR ligation. Purified B cells from CLL patients were exposed to F(ab’) anti-human immunoglubin M (IgM) (10 mg/mL) and whole anti-human IgM (15 mg/mL) polyclonal antibodies (pAb) or specific anti-human FcγRIIb monoclonal antibody (2B6) (1 mg/mL) for 5 minutes before being lysed and immunoblotted for p-ITIM domain of FcγRIIb, p-SHIP, p-AKT and p-ERK. Equal loading was checked by immunoblotting with GAPDH (n=9). (B) Immunoblot analyses of SHIP in membrane fraction extracts of CLL cells from two representative patient’s samples. The remaining extract containing cytosolic fraction of non-treated cells was used as control. The absence of GAPDH expression indicated no cytosolic contamination of the membrane fraction. Cadherin was used as a membrane fraction marker, as well as a control for protein loading. (C) Bar charts show mean values and standard deviation for the relative expression of p-ITIM, p-SHIP, p-AKT and p-ERK in CLL patients with responsive BCR (n=9). P-values were calculated using two tail paired t-test (*P<0.05, **P<0.01, ***P<0.001). F: F(ab’) anti-human IgM pAb; W: whole anti-human IgM pAb; 2B6: specific antihuman FcγRIIb monoclonal antibody. 2

306

haematologica | 2021; 106(1)

Letters to the Editor

Figure 2. Involvement of SHIP in the inhibitory effect of FcγRIIb-B-cell receptor coligation on chronic lymphocytic leukemia cells. (A) Relative SHIP expression in chronic lymphocytic leukemia (CLL) B cells transfected with short intefering RNA (siRNA) SHIP-1 (n=3). (B) Relative AKT and ERK phosphorylation in SHIP-1 siRNA transfected (n=3) and non-transfected (n=9) CLL B cells. (C) Immunoblot for a representative CLL patient’s sample transfected or not transfected with siRNA SHIP-1. (D) Immunoblot for a representative CLL patient’s sample exposed to 10 mM of 3AC or EtOH. F(ab’) fragments of anti-human immunoglobulin M (IgM) (10 mg/mL) or whole anti-human IgM (15 mg/mL) polyclonal antibodies (pAb) were used to ligate the BCR alone or to coligate it with FcγRIIb, respectively. Cells were lysed and immunoblotted for SHIP, p-SHIP, p-AKT and p-ERK. Equal loading was checked by immunoblotting with GAPDH. Mean values, standard deviation and statistical significance between data were determined by two-tailed paired t-test (*P<0.05, **P<0.01). F: F(ab’) anti-human IgM pAb; W: whole antihuman IgM pAb; 2B6: specific anti-human FcγRIIb monoclonal antibody. 2

of p-SHIP levels, as reported before13,14 indicating that SHIP can also function independently of FcγRIIb. We also studied the potential involvement of SHIP in the inhibition of p-AKT and/or p-ERK induced by FcγRIIb-BCR cocrosslinking. For that purpose, we inhibited SHIP using two different approaches; cell transfection with a specific SHIP-1 short interfering RNA (siRNA) and cell exposure to 3AC (a molecule that inhibits the enzymatic activity of SHIP). In both, SHIP-1 siRNA transfected (Figure 2A-C) or 3AC-exposed CLL cells (Figure 2D); we observed that BCR ligation was incapable of inducing an increase of p-AKT levels, suggesting that SHIP is responsible for the activation of AKT upon BCR ligation in CLL cells. Since AKT could not be activated by BCR ligation in these cells, the involvement of SHIP in the reduction of AKT activation after BCR-FcγRIIb coligation could not be assessed. In contrast, in both, SHIP-1 siRNA transfected or 3AC-exposed CLL cells, activation of ERK (successfully achieved upon BCR ligation), could not be inhibited by BCR-FcγRIIb coligation, demonstrating that SHIP is involved in the inhibitory effect of FcγRIIb through inhibition of ERK signaling in CLL cells. In normal B cells, the ability of FcγRIIb-BCR coligation to inhibit p-ERK was not altered by 3AC exposure or SHIP-1 siRNA transfection (Online Supplementary Figure S3). We further explored whether BCR-FcγRIIb coligation could inhibit the BCR-induced activation of CLL cells. Similarly to normal B cells (Online Supplementary Figure S4A-B), the crosslinking of FcγRIIb and BCR with whole anti-IgM pAb significantly inhibited the increase in Ca++ haematologica | 2021; 106(1)

flux (Figure 3A) and the increase in the proportion of CD69+ cells (Figure 3B) induced by BCR. We also observed that FcγRIIb-BCR coligation reduced the percentage of apoptotic cells induced by BCR activation after 48 hours (h) with or without IL4/CD40L, with no inhibitory effects being observed in proliferation ratios after 72h (Figure 3C-D, Online Supplementary Figure S5). In normal B cells, the increase of Ca++ flux, CD69+ cells and proliferation induced by BCR ligation was reduced by FcγRIIb-BCR coligation, but no changes were detected in apoptosis (Online Supplementary Figure S4). Collectively, our study shows that FcγRIIb-BCR coligation can reduce BCR signaling in CLL cells responsive to BCR by reducing the downstream activation of AKT and ERK pathways. Moreover, our results indicate that SHIP may be involved in the FcγRIIb inhibitory effect by blocking the ERK pathway. This inhibitory effect compromises BCR-induced leukemic cell activation, since molecular and cellular inhibitory effects were reverted when coligation was abrogating by using a specific anti-human FcγRIIb monoclonal antibody. These data are in agreement with a recent report published by Lemm et al.12 which suggests that SHIP is a potential regulator of BCR signaling in CLL cells and may explain our previous findings suggesting a correlation between high expression of FcγRIIb and better prognosis in early stage CLL.15 Rosa Bosch,1,2,3* Alba Mora,1,2,3,4* Carolina Cuellar,1,2,3 Gerardo Ferrer,5 Sergey Gorlatov,5 Josep Nomdedéu,7 Emili Montserrat,8 Jorge Sierra,1,2,3 Kanti R. Rai,5 Nicholas Chiorazzi5 and Carol Moreno1,2,3,4 307

Letters to the Editor

Figure 3. Activation, apoptosis and proliferation of chronic lymphocytic leukemia cells upon B-cell receptor (BCR) ligation or FcγRIIb-BCR coligation. (A) Calcium kinetics for a representative chronic lymphocytic leukemia (CLL) patient’s sample in which purified leukemic cells were loaded with Indo1 ratio and acquired for 60 seconds prior to the addition of F(ab’) anti-immunoglobulin M (anti-IgM), whole anti-IgM or anti-FcγRIIb. (B) Box plots represents the median mean fluorescence intensity ratio of Indo1 (4 mM) for each stimuli in CLL patient’s samples (n=10). Purified B cells from CLL patients with responsive BCR were exposed to the indicated stimuli for 48 hours (h) or 72h and then analyzed by flow cytometry for CD69 expression on viable cells (B), Annexin V and TO-PRO®-3 (Annexin V+ cells were considered apoptotic cells) (C) or EdU incorporation using the Click-iTTM EdU Alexa FluorTM 488 Flow Cytometry Assay Kit (D) (n=10). Box plots encompass the data points within 1.5-times the interquartile range from the first or third quartile. P-values were calculated using two tail paired t-test (*P<0.05, **P<0.01). F: F(ab’) anti-human IgM ployclonal antibodies (pAb); W: whole anti-human IgM pAb; 2B6: specific anti-human FcγRIIb monoclonal antibody. 2

1 Laboratory of Oncology/Hematology and Transplantation, Institute of Biomedical Research, IIB Sant Pau, Barcelona, Spain; 2Department of Hematology, Hospital de la Santa Creu i Sant Pau, Autonomous University of Barcelona, Barcelona, Spain; 3Biomedical Research Institute (IIB-Sant Pau) and José Carreras Leukemia Research Institute, Barcelona, Spain; 4Department of Medicine, Autonomous University of Barcelona, Barcelona, Spain; 5Karches Center for Oncology Research, The Feinstein Institute for Medical Research, Manhasset, NY, USA; 6MacroGenics, Inc., Rockville, MD, USA; 7 Laboratory of Hematology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain and 8Department of Hematology, Institute of Hematology and Oncology, Hospital Clínic, IDIBAPS, Barcelona, Spain. *RB and AM contributed equally as co-first authors. Correspondence: CAROL MORENO - cmorenoa@santpau.cat doi:10.3324/haematol.2019.245795 Disclosures: Sergey Gorlatov is an employee of MacroGenics, the manufacturer of the anti-FcγRIIb mAb used in this study, and receives compensation and stock options as part of his employment. All other authors declare no competing financial interests. Contributions: RB and AM performed the laboratory work; RB, AM, GF, EM, NC and CM analyzed the results and wrote the

308

paper; CC, SG, JN, JS and KR supplied essential data, samples or reagent; RB, AM, NC and CM designed the study. All authors have reviewed and approved the final version of the manuscript Funding: this work was supported in part by the Spanish Ministry of Science through grants from Red Temática de Investigación Cooperativa en Cáncer (RD12/0036/0071 to JS) Instituto de Salud Carlos III (FIS PI11/01740 and FIS PI19/00753 to CM), grants from Catalan Government AGAUR (2014 SGR 1281), Cellex Foundation Barcelona, Asociación Española contra el Cáncer (AECC), José Carreras Foundation and from the NIH, National Cancer Institute (RO1 CA0 81554 to NC).

References 1. Stevenson FK, Krysov S, Davies AJ, Steele AJ, Packham G. B-cell receptor signaling in chronic lymphocytic leukemia. Blood. 2011; 118(16):4313-4320. 2. Muta T, Kurosaki T, Misulovin Z, Sanchez M, Nussenzweig MC, Ravetch JV. A 13-amino-acid motif in the cytoplasmic domain of Fc gamma RIIB modulates B-cell receptor signalling. Nature. 1994;368(6466):70-73. 3. Phillips NE, Parker DC. Cross-linking of B lymphocyte Fc gamma receptors and membrane immunoglobulin inhibits antiimmunoglobulin-induced blastogenesis. J Immunol. 1984; 132(2):627-632. 4. Guarini A, Chiaretti S, Tavolaro S, et al. BCR ligation induced by

haematologica | 2021; 106(1)

Letters to the Editor

5. 6. 7. 8. 9. 10.

IgM stimulation results in gene expression and functional changes only in IgV H unmutated chronic lymphocytic leukemia (CLL) cells. Blood. 2008;112(3):782-792. Petlickovski A, Laurenti L, Li X, et al. Sustained signaling through the B-cell receptor induces Mcl-1 and promotes survival of chronic lymphocytic leukemia B cells. Blood. 2005;105(12):4820-4827. Gamberale R, Geffner JR, Giordano M. Immune complexes and apoptosis in B-cell chronic lymphocytic leukemia. Leuk Lymphoma. 2002;43(2):251-255. Carver DJ, Aman MJ, Ravichandran KS. SHIP inhibits Akt activation in B cells through regulation of Akt membrane localization. Blood. 2000;96(4):1449-1456. Aman MJ, Lamkin TD, Okada H, Kurosaki T, Ravichandran KS. The inositol phosphatase SHIP inhibits Akt/PKB activation in B cells. J Biol Chem. 1998;273(51):33922-33928. Tamir I, Stolpa JC, Helgason CD, et al. The RasGAP-binding protein p62dok is a mediator of inhibitory FcgammaRIIB signals in B cells. Immunity. 2000;12(3):347-358. Liu Q, Oliveira-Dos-Santos AJ, Mariathasan S, et al. The inositol

haematologica | 2021; 106(1)

11.

12.

13.

14.

15.

polyphosphate 5-phosphatase ship is a crucial negative regulator of B cell antigen receptor signaling. J Exp Med. 1998;188(7):1333-1342. Gamberale R, Fernandez-Calotti P, Sanjurjo J, et al. Signaling capacity of FcgammaRII isoforms in B-CLL cells. Leuk Res. 2005;29(11):1277-1284. Lemm EA, Valle-Argos B, Smith LD, Richter J. Preclinical evaluation of a novel SHIP1 phosphatase activator for inhibition of PI3K signaling in malignant B-cells. Clin Cancer Res. 2020;26(7):1700-1711. Pauls SD, Marshall AJ. Regulation of immune cell signaling by SHIP1: A phosphatase, scaffold protein, and potential therapeutic target. Eur J Immunol. 2017;47(6):932-945. Leung WH, Tarasenko T, Bolland S. Differential roles for the inositol phosphatase SHIP in the regulation of macrophages and lymphocytes. Immunol Res. 2009;43(1-3):243-251. Bosch R, Mora A, Vicente EP, et al. FcgammaRIIb expression in early stage chronic lymphocytic leukemia. Leuk Lymphoma. 2017; 58(11):2642-2648.

309

CASE REPORTS Severe delayed hemolytic transfusion reaction due to anti-Fy3 in a patient with sickle cell disease undergoing red cell exchange prior to hematopoietic progenitor cell collection for gene therapy To our knowledge, most patients with sickle cell disease (SCD) who have undergone hematopoietic progenitor cell (HPC) collection for gene therapy have been conditioned with a series of monthly simple or exchange transfusions. The rationale for this practice is to improve safety and decrease bone marrow inflammation, potentially leading to higher CD34+ mobilization and decreased risk of adverse events during HPC collection. Whether transfusion actually improves CD34+ mobilization is not clear however, and we previously showed that patients can be safely and effectively mobilized without prior transfusion.1 The risk of delayed hemolytic transfusion reactions (DHTR) is well known and notably appears to be higher in patients who are episodically, rather than chronically, transfused.2 Here we report the case of a 34-year-old male with SCD (genotype SS) who had a severe DHTR after a single red cell exchange (RCE) performed prior to plerixafor mobilization and autologous HPC collection for a study protocol. The patient has a history of multiple SCD-related complications, including aplastic crisis, pneumonia, recurrent acute chest syndrome, recurrent priapism, and silent cerebral infarcts. He has never needed regular transfusions but had occasional transfusions for acute chest syndrome and pain crises. His symptoms were controlled on hydroxyurea (HU), achieving a hemoglobin (Hb) F of ~20%. As per our protocol, HU was discontinued for 3 weeks prior to mobilization; 1 week prior to mobilization his HbF of 19% remained at his HU-treated baseline. The patient is group A RhD-positive and K-negative. His primary medical care facility for the last 10 years provided red blood cells (RBC) for transfusion matched for CEK and antigens to which he had made antibodies, but he had received transfusions at other hospitals with unknown antigen matching standards. His last transfusion was 4 years prior to this RCE. At the time of RCE, his direct antiglobulin test (DAT) was negative and his antibody (Ab) screen showed anti-E, anti-Fya, and antiLea; the anti-Fya and anti-Lea had been detected 5 and 6 years prior, respectively. Although not currently detected, he had a history of anti-Jka from his primary medical facility but the anti-Jka was not detected when samples were tested by the reference laboratory. RBC genotyping (see Table 1) agreed with his RBC Ab history except his RBC were predicted to be Jk(a+) by human erythrocyte antigen (HEA) test. Subsequent sequencing confirmed the predicted JK*A/B genotype and a mutation that does not encode an amino acid change; RNA testing showed the presence of JK*A and JK*B transcripts, consistent with the patient’s RBC expressing Jka and Jkb antigens. He was also found to have a common mutation in the GATA-1 binding motif of the erythroid FY*B promoter, which silences Fyb expression on erythrocytes but not on other cells of his body.3,4 Despite his red cell Ab history, he had no history of symptomatic hemolytic transfusion reactions, hyperhemolysis, or warm autoimmune hemolytic anemia. As per study protocol, on the day prior to stem cell mobilization and collection, the patient underwent RCE without adverse events with nine crossmatch-compatible red cell 310

units that were group A, D+, E−, K−, and Fy(a−). Although anti-Lea was microscopically positive at polyethylene glycol (PEG) IgG, anti-Lea is not routinely matched for at the study institution nor had been matched for with transfusions at his primary medical facility. After RCE, his Hb rose from 9.3 to 11.2 g/dL, and his HbS% decreased from 78% to 22% (HbA% of 72%). The following day, the patient was given a single 240 µg/kg dose of plerixafor for HPC mobilization and started leukapheresis about 3 hours afterward. He had no adverse events and was discharged home after his leukapheresis. Five days after RCE, the patient developed throbbing low back pain, subjective fevers, dark urine, and scleral icterus, symptoms differing from his typical vaso-occlusive crises usually manifesting as right-sided flank/upper back discomfort and priapism. He contacted the on-call research study provider the day after symptom onset and then his sickle cell physician 2 days after symptom onset, who instructed the patient to go to the Emergency Department (ED) for work-up. When he presented to the ED closest to him on the third day after symptom onset, he was found to have a fever of 38.8°C, tachycardia to 131 bpm, blood pressure of 133/78 mm Hg, and an oxygen saturation of 93% by pulse oximeter. His labs were significant for a hemoglobin of 6.1 g/dL, hematocrit (Hct) of 17.0%, reticulocyte count of 18%, total bilirubin of 4.7 mg/dL with indirect bilirubin of 3.8 mg/dL (baseline total bilirubin of 2.3 mg/dL, baseline indirect bilirubin of 1.5), lactate dehydrogenase of 875 U/L (baseline lactate dehydrogenase 408 U/L), haptoglobin of <20 mg/dL, and hemoglobin S% of 77%. His urinalysis was positive for bilirubin, blood, and protein. His Ab screen was positive, and his RBC reacted in the DAT (1+W) with anti-IgG and anti-C3. Ab investigation was performed with low ionic strength salt solution, papain, and PEG indirect antiglobulin tests (IAT), and the patient’s serum demonstrated reactivity with all cells positive for Fya and Fyb, even with papain treatment, which abolishes Fya and Fyb expression on RBC but does not affect Fy3. These results pointed to a new alloantibody, anti-Fy3. The patient was diagnosed with a DHTR. His hospital course was complicated by worsening hypoxemia with fevers, and he was found to have segmental pulmonary emboli in the bilateral lower lobes and possible pneumonia. He was treated with supplemental oxygen, enoxaparin, antibiotics, fluids and opioids for pain control; further transfusion was avoided given hemodynamic stability and fear of inducing hyperhemolysis. In addition to the newly detected antibody that was directed against the Fy3-antigen of the Duffy blood group system,5 the previously detected anti-E, anti-Fya, and anti-Lea were also present. Furthermore, the patient’s plasma reacted weakly with all E−, Fy(a−b−) and Le(a−) panel cells by papain and PEG IAT. An acid eluate prepared from the patient’s red cells reacted weakly with all cells tested by PEG IAT, consistent with the presence of a warm autoantibody. Lookback on the red cell units transfused during the exchange showed that 8 of the 9 units were Fy(a-b+); only 1 of the 9 units was Fy(a−b−). All units were Jk(a+); three units had Lea typing available and were Le(a-). Therefore, if hemolysis was from the newly detected anti-Fy3, he would be expected to lose 7.2 grams/dL of Hb due to hemolysis of the transfused Fy(a−b+), resulting in a patient hemoglobin of 4 g/dL (see Figure 1). Indeed, his Hb nadired at 4.4 g/dL, suggesting hemolysis of Fy(a−b+) cells. haematologica | 2021; 106(1)

Case Report Table 1. Human erythrocyte antigen (HEA) phenotype by DNA analysis report.

Rh c +*

C +

e +

Kell E 0

V +

VS +

K 0

k +

Kpa 0

Kidd Kpb +

Jsa 0

Jsb +

Jka +

Jkb +

Duffy Fya 0

Fyb (0)#

MNS M +

N +

S +

s +

U +

+* = c+ (partial). (0)# = GATA silencing mutation for Fyb.

He was discharged home 14 days after his initial ED presentation on anticoagulation. Approximately 6 weeks post-discharge, his Hb and Hct had improved to 8.4 g/dL and 22.4%, respectively, and he was restarted on hydroxyurea. He has not required further transfusion since his DHTR event. A follow-up Ab screen at his primary medical care facility 67 days after his RCE was positive and demonstrated anti-E, anti-Fya, and anti-Lea. The anti-Fy3 was no longer demonstrable. There are only five prior case reports in the literature describing acute and delayed hemolytic transfusion reactions due to anti-Fy3,6-10 an antigen in the Duffy blood group system. This system is comprised of five antigens, but antibodies against only three antigens, Fya, Fyb and Fy3, are known to cause hemolytic transfusion reactions. Fya and Fyb antigens differ by a single point mutation, while Fy3 refers to a common epitope on the FY protein separate from the Fya/Fyb polymorphism (Figure 2). The four major phenotypes are Fy(a+b+), Fy(a+b−), Fy(a−b+), or Fy(a−b−). Approximately 68% of Africans and people of African descent have lost expression of the FY protein on RBC and phenotype as Fy(a−b−); FY loss protects against malarial invasion because FY is a receptor for the malarial parasites P. vivax and P. knowlesi.3,11 The absence of FY protein on the RBC is primarily due to a point mutation in the GATA promotor region on a FYB allele such that the erythroid transcription factor GATA-1 fails to bind;12 thus, RBC lack expression of Fyb.3 Patients with the promoter mutation are not at risk for anti-Fyb, despite serologically typing as Fy(b-), because they have a structurally normal FY*B allele and Fyb is expressed on other tissues. These patients remain at risk for anti-Fy3, which can be clinically significant, and patients with an existing anti-Fya are at increased risk.9-13 The reasons for this have long been studied but are yet to be determined.

We speculate that DHTR due to anti-Fy3 may be rare for several reasons: Fy3 may be of low immunogenicity; anti-Fy3 has been under-recognized as a cause of DHTR; and C−E− RBC are likely to also be Fy(a−b−). We speculate that patients with an existing anti-Fya may develop anti-Fy3 if their Fy3 tissue promoter is defective and thus also become immunized to Fy3 antigen at the time of stimulation to Fya. This case illustrates several important points. First, this patient’s symptoms were initially not recognized as a potential DHTR. As gene therapy treatments for SCD become more common, transplant providers with less experience in treating patients with SCD should be educated regarding SCD-specific complications that may be encountered during treatment. Secondly, we suggest extended antigen matching for additional clinically significant antigens for episodically transfused immune red cell responders, i.e, patients with multiple alloantibodies. Importantly, in those who have made anti-Fya and are Fy(a−b−) due to the GATA mutation, Fy(a−b−) units should be considered, if possible, especially with high volume RCE. Fy(a−b−) units are primarily found in blood donors of African ancestry; matching may be difficult in patients with multiple alloantibodies. Thirdly, this case illustrates the importance of optimal transfusion management of patients with SCD, which includes red cell genotyping, performing antibody screens approximately 1 month post-transfusion to detect newly formed alloantibodies that may evanesce, and establishing a nationwide patient database of prior red cell antigen profiles and red cell antibody history that is available across hospital systems. The time frame of the appearance of anti-Fy3 is consistent with an anamnestic response to an antibody that, to our knowledge, was not previously detected. Finally, although it is standard practice to perform simple

Figure 1. Estimations of amounts of Fy3+ hemoglobin post exchange and of total hemoglobin post Fy3+-specific hemolysis. In order to estimate the amount of Fy3+ blood transfused in grams/dL (g/dL), the percentage of Fy3+ units (8 of 9 units) was multiplied by the percentage of hemoglobin A (HbA) post-exchange and multipled by the total hemoglobin (Hb) post-exchange in g/dL. In order to estimate the total Hb after hemolysis of the Fy3+ units, the amount of Hb from the Fy3+ units was subtracted from the total Hb post-exchange.

Fy6

Fya/Fyb

Gly/Asp

Fy3

COOH haematologica | 2021; 106(1)

Figure 2. Diagram representing the predicted 7-transmembrane domain structure of the Duffy glycoprotein. Indicated are the amino and carboxy terminals and location of the Fya/Fyb polymorphism at amino acid 42. Also indicated, on the third extracellular loop, is the predicted location of the sequences necessary for the binding of monoclonal anti-Fy3. The actual sequence required for binding of human polyclonal anti-Fy3 is not known. Also shown is the predicted region for binding of monoclonal anti-Fy6. N-glycosylation sites are indicated by .

311

Case Reports

or exchange transfusions prior to stem cell mobilization with plerixafor in patients with SCD, studies are needed to determine whether treatment modalities other than transfusion, such as hydroxyurea1 or other newly approved drugs, can be effective alternatives prior to plerixafor mobilization. In addition to the risk of serious DHTR, transfusion does not necessarily lead to good mobilization14 and also may not be an option for highly alloimmunized or Jehovahâ&#x20AC;&#x2122;s Witness patients. Elizabeth F. Stone,1 Scott T. Avecilla,2 David L. Wuest,2 Christine Lomas-Francis,1 Connie M. Westhoff,1 David L. Diuguid,3 Michel Sadelain,2 Farid Boulad2 and Patricia A. Shi1,4 1 New York Blood Center, New York; 2Memorial Sloan Kettering Cancer Center, New York; 3Division of Hematology, Columbia University Medical Center, New York and 4SickleCell Program, Division of Hematology, Albert Einstein College of Medicine, Bronx, NY, USA Correspondence: PATRICIA A. SHI/ELIZABETH F. STONE pshi@nybc.org/elizabeth.stone@mountsinai.org doi:10.3324/haematol.2020.253229 Disclosures: no conflicts of interest to disclose. Contributions: EFS and PAS wrote the manuscript; STA, DLW, DLD, FB, and PAS cared for the patient, CLF and CMW performed immunohematology studies, and all authors edited or reviewed the manuscript. Acknowledgments: the authors would like to thank our patient for his study participation; the Doris Duke Charitable Foundation for a 2011 Innovation in Clinical Research Award for trial support (to PAS and MS); and Sanofi-Genzyme for provision of plerixafor. Funding: this work was also supported by the Memorial Sloan Kettering Core Grant (P30 CA008748) funded by the National Cancer Institute.

312

References 1. Boulad F, Shore T, van Besien K, et al. Safety and efficacy of plerixafor dose escalation for the mobilization of CD34(+) hematopoietic progenitor cells in patients with sickle cell disease: interim results. Haematologica. 2018;103(5):770-777. 2. Narbey D, Habibi A, Chadebech P, et al. Incidence and predictive score for delayed hemolytic transfusion reaction in adult patients with sickle cell disease. Am J Hematol. 2017;92(12):1340-1348. 3. Meny GM. The Duffy blood group system: a review. Immunohematology. 2010;26(2):51-56. 4. Karczewski KJ, Francioli LC, Tiao G, et al. Variation across 141,456 human exomes and genomes reveals the spectrum of loss-of-function intolerance across human protein-coding genes. bioRxiv. 2019: 531210. 5. Albrey JA, Vincent EE, Hutchinson J, et al. A new antibody, anti-Fy3, in the Duffy blood group system. Vox Sang. 1971;20(1):29-35. 6. Olteanu H, Gerber D, Partridge K, Sarode R. Acute hemolytic transfusion reaction secondary to anti-Fy3. Immunohematology. 2005; 21(2):48-52. 7. Reyes MA, Illoh OC. Hyperhemolytic transfusion reaction attributable to anti-Fy3 in a patient with sickle cell disease. Immunohematology. 2008;24(2):45-51. 8. Went R, Wright J, Webster R, Stamps R. Anti-Fy3 in sickle cell disease: a difficult transfusion problem. Br J Haematol. 2009; 144(4):621-622. 9. Vengelen-Tyler V. Anti-Fya preceding anti-Fy3 or -Fy5: a study of five cases. Transfusion. 1985;25:482 (abstract). 10. Win N, Lee E, Needs M, Chia LW, Stasi R. Measurement of macrophage marker in hyperhaemolytic transfusion reaction: a case report. Transfus Med. 2012;22(2):137-141. 11. Weppelmann TA, Carter TE, Chen Z, et al. High frequency of the erythroid silent Duffy antigen genotype and lack of Plasmodium vivax infections in Haiti. Malar J. 2013;12:30. 12. Tournamille C, Colin Y, Cartron JP, Le Van Kim C. Disruption of a GATA motif in the Duffy gene promoter abolishes erythroid gene expression in Duffy-negative individuals. Nat Genet. 1995; 10(2):224-228. 13. Kosinski K, Molthan L, White L. Three examples of anti-Fy3 produced in Negroes. Rev Fr Transfus Immunohematol. 1984;27:619624. 14. Esrick EB, Manis JP, Daley H, et al. Successful hematopoietic stem cell mobilization and apheresis collection using plerixafor alone in sickle cell patients. Blood Adv. 2018;2(19):2505-2512.

haematologica | 2021; 106(1)

Case Reports

Clonal independence of JAK2 and CALR or MPL mutations in comutated myeloproliferative neoplasms demonstrated by single cell DNA sequencing JAK2, CALR and MPL driver mutations activate JAK/STAT signalling in BCR-ABL1 translocation negative classic myeloproliferative neoplasms (MPN). Detection of these mutations in patients is included in major criteria for establishing diagnoses of polycythemia vera (PV), essential thrombocythemia (ET) and primary myelofibrosis (PMF).1,2 JAK2 Val617Phe and exon 12 mutations are detected in approximately 98% of PV, while JAK2 Val617Phe, CALR exon 9 frameshift mutations and MPL exon 10 mutations collectively are detectable in 83% of ET and 92% of PMF.2 Mutations in these genes are typically mutually exclusive, however, JAK2 Val617Phe/JAK2 exon 12,3,4 JAK2/CALR,5,6 JAK2/MPL4,6,7 and very rarely CALR/MPL8 comutations have been reported. Further, the presence of multiple independent JAK2 Val617Phe mutations has been demonstrated in the majority of ET patients in one case series, and is one of several lines of evidence that suggest a predisposition in some patients to the development of JAK2 (and other MPN driver) mutations, the nature of which has not been elucidated but may include, genetic predisposition, a preceding clonal mutation or a permissive microenvironment.9 Herein we describe two unusual cases of patients with an MPN, each harboring two driver mutations (JAK2/CALR and JAK2/MPL) and the use of single cell DNA sequencing analysis to further investigate the clonal architecture of these neoplasms. In both cases we demonstrate that the driver mutations represent independent clones and give insight into the potential clonal complexity of this rare phenomenon. The first patient studied was a 70-year-old male with a long (approximately 30-year) history of PV, which transformed to myelofibrosis (MF) 14 years prior to analysis. Approximately 10 years after fibrotic transformation (i.e., 4 years prior to analysis) the patient commenced treatment with azacitidine and ruxolitinib for leukemic transformation. At this timepoint, trisomy of 1q was observed in his peripheral blood by G-banded karyotyping and confirmed by fluorescence in situ hybridization (FISH) analysis of 1q21 (CKS1B) in 62% of analyzed cells. His blast count normalized and he remains stable on azacitidine and ruxolitinib with post-polycythemic MF without blast excess. Next generation sequencing (NGS)-based gene panel analysis of 26 genes involved in myeloid malignancy (previously described10) was performed on a stored bone marrow aspirate DNA sample from 14 years prior at the point of myelofibrotic transformation (the earliest timepoint available). This testing detected a JAK2 p.(Val617Phe) (NM_004972.3:c.1849G>T) at a variant allele frequency (VAF) of 62% as well as a CALR p.(Lys385Asnfs*47) (NM_004343.3:c.1154_1155ins TTGTC) mutation at a VAF of 17%. Testing of a peripheral blood sample stored at the time of leukemic transformation again demonstrated the JAK2 and CALR mutations but in addition a TP53 p.(His178Pro) (NM_000546.5:c.533A>C) mutation was detected at a VAF of approximately 5%. We sought to understand the clonal relationship of the JAK2 and CALR mutations in this patientâ&#x20AC;&#x2122;s disease by performing single cell mutation analysis of mononuclear cells collected from a recent peripheral blood sample using a custom Tapestri gene panel to sequence 10,564 haematologica | 2021; 106(1)

single cells (Mission Bio) (see Online Supplementary Materials and Methods). The previously identified JAK2, CALR and TP53 mutations were genotyped in 96%, 52% and 94% of all filtered cells, and detected in 64%, 26% and 29% of genotyped cells with an aggregate single cell VAF (by read count) of 62%, 16% and 16%, respectively (Online Supplementary Table S1). Considering only the 38% of cells with genotypes determined for all three mutations, subclone analysis identified the presence of three major subclones (Figure 1A and Online Supplementary Table S2). The JAK2 mutation was detected as a homozygote in 72.48% of the cells comprising major clones (i.e., clone size [CS] 72.48%) in the absence of CALR and TP53 mutations. Heterozygous CALR and TP53 mutations were identified together within the second clone (CS 14.05%) and a wild-type (i.e., JAK2/CALR/TP53 mutation negative) clone was also detected (CS 13.47%). Amplicon-based copy number analysis of the single cell data also detected the trisomy of 1q (observed as copy number gain of panel gene MCL1 located at 1q21.3) and specifically identified this amplification within the JAK2 mutant clone (mean copy number 3.20) (Figure 1B). These data are consistent with the existence of two distinct clones, both with additional lesions suggestive of advancing forms of MPN.11 The second patient studied was a 74-year-old male who presented with persistent thrombocytosis (platelets 560-590x109/L) in association with symptomatic splenomegaly. A JAK2 p.(Val617Phe) (c.1849G>T) mutation was detected in his peripheral blood sample by an allele-specific assay and a subsequent bone marrow biopsy showed a mildly hypercellular bone marrow with fibrosis (grade MF-1 to MF-2/3), mild granulocytic hyperplasia, increased megakaryopoiesis and megakaryocytic atypia, which in conjunction with the JAK2 mutation was considered most consistent with PMF. He was commenced on ruxolitinib and responded clinically with decreased abdominal pain. A 26 gene myeloid NGS panel performed on his bone marrow aspirate (as above) detected a MPL p.(Trp515Ser) (NM_005373.2:c.1544G>C) mutation at a VAF of 19% and an established pathogenic SRSF2 p.(Pro95Arg) (NM_003016.4:c.284C>G) mutation at a VAF of 53% in addition to the previously known JAK2 mutation at a VAF of 15%. Given that spliceosome mutations, including those in SRSF2, occur exclusively in a heterozygote state in myeloid malignancies due to their dependency on the wild-type allele,12,13 this finding is consistent with the entire hematopoietic compartment sampled being comprised of a single heterozygous SRSF2 mutant clone with subclonal JAK2 and MPL mutations. In order to further investigate the clonal relationship between the JAK2 and MPL mutations within this SRSF2 mutant parental clone, single cell analysis was performed on 8,181 total nucleated cells from a peripheral blood sample (see Online Supplementary Materials and Methods). The previously identified JAK2 and MPL mutations were genotyped in 91% and 44% of all cells, and detected in 21% and 23% of genotyped cells with an aggregate single cell VAF (by read count) of 11% and 13%, respectively (Online Supplementary Table S1). Mutually exclusive heterozygous JAK2 (CS 21.06%) and MPL (CS 16.76%) subclones were detected in the 36% of cells with genotypes determined for both mutations. A JAK2/MPL mutation negative population was also detected (CS 62.17%) (Online Supplementary Table S2). This analysis demonstrates the independent origin of the JAK2 and MPL mutations within the context of an SRSF2 mutant parental clone in this patient (Figure 1C). 313

Case Reports

C Figure 1. Single cell clone analysis of two myeloproliferative neoplasm patients. (A and C) Major clone analysis of patients 1 and 2, respectively, demonstrating allele frequency distribution of single cells within each of the major clones identified by the presence of JAK2 Val617Phe, CALR Lys385Asnfs*47 and TP53 His178Pro mutations in patient 1, and JAK2 Val617Phe and MPL Trp515Ser mutations in patient 2. Zygosity (wild-type [wt], heterozygous [het], homozygous [hom]) is inferred from the median allele frequency (indicated). Variant data were filtered and major clones identified as indicated in the supplementary methods. (B) MCL1 copy number in patient 1 is shown for the corresponding clones, indicating a copy number gain of chromosome 1q21.3 within the JAK2hom;CALRwt;TP53wt clone.

The analysis of both of these rare and illustrative patients at the single cell level provides interesting and important insights into the clonal architecture of both comutated MPN and MPN in general. Firstly, we have unambiguously established the existence of these driver mutations in different subclones. Molecular characterization of MPN at a subclonal level has previously been performed using colony assays grown on medium from single cells in the presence of cytokines and growth factor to generate sufficient material for further characterization. Whilst this technique has given important insights to date,3,4 it is technically limited by the number of cells that can be assessed, the potential for mixed colonies as well as the preferential growth of particular clones with exogenous stimulation. Although different biases in the form of cell doublets and allelic drop-out may exist within single cell sequencing technologies, direct sequencing of >8,000 individual primary cells conclusively demonstrates the occurrence of JAK2/CALR/MPL driver mutations in these patients in distinct clones within each patient. The majority of comutated cases of MPN described in the literature to date have been noted to have JAK2 mutations present at low VAF (typically <4%)4,6 compared to the other driver mutation (CALR/MPL). We reviewed the data from 2,664 consecutive cases of clinical NGS MPN sequencing performed at our own institution (Peter MacCallum Cancer Center, Melbourne, Australia) over 5 years (2014-2019) and identified 16 comutated cases (six JAK2/CALR, seven JAK2/MPL, three 314

double-mutated JAK2) among 1,082 JAK2 (n=740), CALR (n=282) or MPL (n=73) mutated patients and noted similar findings (Figure 2). The biological relevance of the low VAF JAK2 mutation in these types of cases is unclear and the possibility that it may represent a â&#x20AC;&#x153;bystanderâ&#x20AC;? mutation outside the true disease compartment cannot be excluded. Indeed it is noted that the JAK2 Val617Phe is one of the most common mutations observed in age related clonal hematopoiesis.14 In contrast, single cell analysis of our cases suggest that they truly represent biclonally driven MPN given (i) the high fraction of clones containing each driver mutation (ii) the acquisition of further genomic abnormalities (1q21 gain and TP53 mutation) providing evidence of clonal evolution within both compartments. Unfortunately, no historical sample was present in case 1 from the polycythemic phase of illness to understand the time course of the acquisition of JAK2/CALR mutations. We note however, the extreme rarity of CALR mutations in PV.15 The second case demonstrates that biclonal MPN driver mutations can arise from a common progenitor cell, a feature that has previously been observed in the context of single driver mutations.16 This possibility is not excluded for the first case given the relatively focussed mutation analysis of 26 genes only. The finding of biclonality in these two cases does not exclude the possibility that co-mutation of driver genes may occur within a single clone in other cases, especially considering the background incidence of JAK2 mutations in age related clonal hematopoiesis. haematologica | 2021; 106(1)

Allele frequency

Case Reports

Figure 2. Allele frequencies in six JAK2/CALR and seven JAK2/MPL comutated patients detected among a series of 2,664 consecutive diagnostic samples assessed for myeloproliferative neoplasms. The median allele frequency for JAK2 is significantly lower than for either comutated CALR (JAK2 vs. CALR, median 0.32% [range: 0.02-2.83%] versus 24.90% [range: 10.3043.98%]) or comutated MPL (JAK2 vs. MPL, median 1.96% [range 0.50%20.12%] versus 19.88% [range: 8.51-37.61%]).

In summary we have described two cases of biclonal MPN characterized by single cell sequencing confirming the presence of driver mutations in different clones. These data provide insights into the potential molecular complexity of MPN including intralineage clonal evolution. Moreover, these data highlight the importance of routine assessment of all three canonical driver mutations (i.e., JAK2/CALR/MPL) in patients with MPN in order to accurately characterize disease biology. Ella R. Thompson,1,2* Tamia Nguyen,1* Yamuna Kankanige,1,2 Paul Yeh,1,2,3,4 Michael Ingbritsen,1 Michelle McBean,1 Timothy Semple,3 Gisela Mir Arnau,2,3 Kate Burbury,4 Nora Lee,4 Amit Khot,4 David Westerman1,2,4 and Piers Blombery1,2,4 1 Pathology Department, Peter MacCallum Cancer Center; 2Sir Peter MacCallum Department of Oncology, University of Melbourne; 3 Research Division, Peter MacCallum Cancer Center and 4Clinical Haematology, Peter MacCallum Cancer Centre/Royal Melbourne Hospital, Melbourne, Victoria, Australia *ERT and TN contributed equally as co-first authors.

Correspondence: ELLA R. THOMPSON - ella.thompson@petermac.org. doi:10.3324/haematol.2020.260448 Disclosures: no conflicts of interests to disclose. Contributions: TN, MI, TS and GMA performed laboratory stud-

haematologica | 2021; 106(1)

ies; ET, TN, YK, PY and PB performed data analysis; KB, NL, AK provided clinical data; ET, TN, YK and PB wrote the manuscript; MM, DW and PB supervised the study. Acknowledgments: the authors would like to thank the Snowdome Foundation and the Wilson Center for Lymphoma Genomics for their funding support.

References 1. Swerdlow SH, Campo E, Harris NL, et al. WHO classification of

Tumours of Haematopoietic and Lymphoid Tissues. 4th ed. Lyon, France: International Agency for Research on Cancer, 2017. 2. Tefferi A, Pardanani A. Myeloproliferative Neoplasms: a contemporary review. JAMA Oncol. 2015;1(1):97-105. 3. Li S, Kralovics R, De Libero G, Theocharides A, Gisslinger H, Skoda RC. Clonal heterogeneity in polycythemia vera patients with JAK2 exon12 and JAK2-V617F mutations. Blood. 2008;111(7):3863-3866. 4. Beer PA, Jones AV, Bench AJ, et al. Clonal diversity in the myeloproliferative neoplasms: independent origins of genetically distinct clones. Br J Haematol. 2009;144(6):904-908. 5. Nomani L, Bodo J, Zhao X, Durkin L, Loghavi S, Hsi ED. CAL2 immunohistochemical staining accurately identifies CALR mutations in myeloproliferative neoplasms. Am J Clin Pathol. 2016; 146(4):431-438. 6. Usseglio F, Beaufils N, Calleja A, Raynaud S, Gabert J. Detection of CALR and MPL mutations in low allelic burden JAK2 V617F essential thrombocythemia. J Mol Diagn. 2017;19(1):92-98. 7. Lasho TL, Pardanani A, McClure RF, et al. Concurrent MPL515 and JAK2V617F mutations in myelofibrosis: chronology of clonal emergence and changes in mutant allele burden over time. Br J Haematol. 2006;135(5):683-687. 8. Bernal M, Jimenez P, Puerta J, Ruiz-Cabello F, Jurado M. Co-mutated CALR and MPL driver genes in a patient with myeloproliferative neoplasm. Ann Hematol. 2017;96(8):1399-1401. 9. Lambert JR, Everington T, Linch DC, Gale RE. In essential thrombocythemia, multiple JAK2-V617F clones are present in most mutantpositive patients: a new disease paradigm. Blood. 2009;1 14(14):3018-3023. 10. Yannakou CK, Jones K, McBean M, et al. ASXL1 c.1934dup;p.Gly646Trpfs*12-a true somatic alteration requiring a new approach. Blood Cancer J. 2017;7(12):656. 11. Marcellino BK, Hoffman R, Tripodi J, et al. Advanced forms of MPNs are accompanied by chromosomal abnormalities that lead to dysregulation of TP53. Blood Adv. 2018;2(24):3581-3589. 12. Kim E, Ilagan JO, Liang Y, et al. SRSF2 Mutations contribute to myelodysplasia by mutant-specific effects on exon recognition. Cancer Cell. 2015;27(5):617-630. 13. Lee SC, Dvinge H, Kim E, et al. Modulation of splicing catalysis for therapeutic targeting of leukemia with mutations in genes encoding spliceosomal proteins. Nat Med. 2016;22(6):672-678. 14. Genovese G, Kahler AK, Handsaker RE, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med. 2014;371(26):2477-2487. 15. Chauveau A, Nibourel O, Tondeur S, et al. Absence of CALR mutations in JAK2-negative polycythemia. Haematologica. 2017; 102(1):e15-e16. 16. Ortmann CA, Kent DG, Nangalia J, et al. Effect of mutation order on myeloproliferative neoplasms. N Engl J Med. 2015;372(7):601612.

315

Case Reports

Monoclonal gammopathy of clinical significance: in vivo demonstration of the anti-thrombotic effect of an acquired anti-thrombin antibody Monoclonal gammopathy of clinical significance (MGCS) is a novel concept that has been suggested to describe an abnormal plasma cell clone producing a monoclonal immunoglobulin (MIg) associated with pathological manifestations.1 One of the MGCS-related complications includes autoantibody activity. To date, MIg-associated bleeding disorders have been essentially reported with auto-antibodies directed against von Willebrand factor (VWF) or factor VIII (FVIII).2,3 To our knowledge, no MGCS associated with an auto-antithrombin antibody (ATA) has yet been described. ATA have been reported in several cases of bleeding4 or thrombosis5 but can also be asymptomatic.6 This is a rare disorder that may develop after bovine thrombin exposure7 or in association with anti-phospholipid syndrome,8 liver cirrhosis9 or viral infection.10 Due to their versatile clinical presentation and the absence of specific diagnostic tests, ATA constitute a heterogeneous group of disorders that are difficult to characterize. Here we present the case of a patient with a MIg

Routine coagulation tests

(IgG k) and an ATA. The use of a diverse array of in vitro and in vivo models were applied to establish the causal relationship between MIg, ATA and bleeding tendency, sustaining the concept of MGCS-related bleeding disorder. The case concerns a 40-year-old man with no previous personal or family history of bleeding until he first presented with a psoas hematoma and hematuria. A prolonged activated partial thromboplastin time (aPTT) associated with an IgG k monoclonal gammopathy (MIgG, monoclonal component evaluated at 12 g/L) was discovered on this occasion. Bone marrow plasma cell percentage was below 5%, β2-microglobulin and light chains were found normal and the patient did not present any CRAB symptoms as assessed by imaging (data not shown) and blood tests (Online Supplementary Data). All these elements allowed to exclude the diagnosis of myeloma. The prolonged aPTT was not corrected by the addition of normal pooled plasma (NPP) and prothrombin time (PT), fibrinogen as well as FVIII/VWF were within the normal range. No further hemostatic investigations were performed, and the bleeding tendency was related to an unspecified acquired disorder. Over the next 6 years, bleeding episodes (rectal bleeding, retroperitoneal hemorrhage, muscular hematomas) were managed by

Thrombin generation test

Thrombin time mixing studies (sec)

Time (min) Figure 1. In vitro patient’s laboratory evaluation. (A) Routine coagulation tests were performed on platelet poor plasma (PPP) obtained from citrated blood after double centrifugation using APTT SP HemosIL (IL), Innovin (Siemens), Dade Thrombin (Siemens) and using a STAR®-MAX (Stago) according to manufacturer’s instructions. The detection of lupus anticoagulant was performed using a diluted Russel’s Viper Venom Time (dRVVT, LAC screen and LAC confirm, Siemens) and using CS 5100 (Sysmex) according to manufacturer’s instructions. Thrombin time (TT) mixing studies were performed on patient’s PPP and a normal pooled plasma (NPP) obtained from 30 healthy volunteers. Mixing studies (vol/vol) were performed by measuring TT of serial dilutions from 1:1 to 1:64 of patient’s PPP in NPP, and in NPP diluted in Owren Koller (OK) buffer as control (NPP+OK, vol/vol). (B) Thrombin generation measurement using calibrated automated thrombogram (CAT). Thrombin generation test (TGT) was studied on PPP obtained from citrated blood. Coagulation was triggered by low concentration of tissue factor 1 (TF1) (1 pM) in the presence of phospholipid (4 μm) and using CAT according to manufacturer instructions (PPP Reagent Low, Stago). TGT was performed in patient’s PPP and normal pooled plasma (NPP) obtained from 30 healthy volunteers. Mixing studies (vol/vol) were performed by measuring TGT of serial dilutions from 1:1 to 1:32 of patient’s PPP in NPP, and in NPP diluted in Owren Koller buffer as control (NPP+OK, vol/vol). Insert: TGT was performed in patient’s and control platelet rich plasma (PRP) obtained from citrated blood. Coagulation was initiated using low concentration of tissue factor (TF) (1 pM) and using CAT according to manufacturer’s instructions (PRP Reagent, Stago). aPTT: activated partial thrombin time; PT: prothrombin time.

316

haematologica | 2021; 106(1)

Case Reports

recombinant activated factor VII (rFVIIa) and several courses of dexamethasone, that allowed to temporarily normalize aPTT. Both azathioprine and rituximab were unsuccessful in reducing the bleedings and normalizing aPTT. He was referred to us for precise diagnosis of the hemostatic disorder. Blood count was normal and monoclonal component was stable. A thrombin time (TT) was performed along with aPTT, both were markedly prolonged (>120 and 76 seconds [sec], respectively vs. 16 and 35 sec for control). The prolonged TT was observed using bovine and human thrombin. PT was slightly modified (11.9 vs. 9.5 sec for control, international normalized ratio [INR]: 1.36) (Figure 1A). Coagulation factors, including fibrinogen, FII, FV, FVII, FVIII, FIX, FX, FXI and reptilase time were normal. Any presence of anticoagulant treatment was excluded (data not shown). Mixing studies (vol/vol) were performed by measuring TT of serial dilutions of patient’s platelet poor plasma (PPP) in NPP, to quantitate the inhibitor level (modified Bethesda assay). The results showed prolonged TT from dilutions 1:1 to 1:16 (from >120-28 sec), and normalization from 1:32 (21 sec) suggesting the presence of an ATA estimated at 32 Bethesda Units. A lupus anticoagulant was also detected without any other autoantibody or dysimmune abnormality, except the patient’s MIgG (no anticardiolipin or anti-b2GPI antibodies). In order to highlight the acquired plasmatic ATAinduced hemorrhagic tendency ex vivo, a thrombin generation test (TGT) was performed (CAT, Stago®) (Figure 1B). Compared to control, TGT in patient’s platelet rich plasma (PRP), showed a prolonged lag time (LT) of 15.5 versus 4.7 minutes (min) associated with a decreased endogenous thrombin potential (ETP) of 834 versus 1,730 nM/min corresponding to a 300% increase of LT and a 50% decrease of ETP, respectively. In patient’s PPP, no peak of thrombin generation could be detected within 1

hour (h). Similarly to TT, TGT was performed using sseveral dilutions of the patient’s PPP in NPP. With each dilution, the LT was shortened, ranging from more than 60-6 minutes (min) versus 5.15 min in NPP. ETP never exceeded 575 nM/min for the 1:32 dilution compared to 1,182 nM/min in NPP. These results equal to an increased LT ranging from 660-20% and a decreased ETP ranging from 75-51% for dilutions from 1:1 to 1:32 respectively. In order to confirm that the ATA belong to the same clone as the MIgG, the electrophoresis pattern of the antibodies was studied using isoelectrofocusing on agarose gel followed by immunofixation with an anti-IgG antiserum (Hydragel CSF Isofocusing®, Sebia). The patient's ATA was isolated from patient’s plasma by affinity chromatography using activated agarose beads (Aminolink®, ThermoScientific) covalently coupled to human α-thrombin (2 mg). The patient's total IgG was purified using protein A agarose (ThermoScientific). The immunofixation of plasmatic IgG, purified total IgG and ATA showed a similar monoclonal banding, suggesting that the ATA was indeed part of the MIgG, and not a coincidental autoantibody (Figure 2). The diagnosis of MGCS-related bleeding disorder due to a MIgG and an ATA was retained. In order to eradicate the ATA, a combination of bortezomib and dexamethasone was initiated (at doses used in myeloma patients), but four cycles of this combination did not decrease the MIgG. Continuous thalidomide and dexamethasone treatment was also unsuccessful. The patient's TT was only temporarily shortened 24-48 h after intravenous infusion of IgG. Over the years, the patient continued to present bleeding episodes (hematospermia, psoas hematoma). Given the failure of the previous therapies and the efficacy of “on demand” hemostatic treatment, the patient was finally managed by close follow-up, self-initiated combination of steroids and rFVIIa in case in case of hemorrhage. In order to illustrate the anti-coagulant effect of the

Figure 2. Immunoglobulin G (IgG) immunofixation pattern of patient’s plasma, purified total IgG and anti-thrombin antibody. The immunofixation was performed using Hydragel CSF Isofocusing on a Hydrasis system according to manufacturer instructions (Sebia). The procedure included isoelectric focusing on agarose gel followed by immunofixation with an enzyme (peroxidase) labelled anti-IgG antiserum on patient’s samples in which IgG concentration was adjusted to 20 mg/L before migration. C1, C2: normal sera; QC: quality control (Sebia); IgG: patient’s purified total IgG; ATA: patient’s anti-thrombin antibody; plasma: citrated plasma.

haematologica | 2021; 106(1)

317

Case Reports

Thrombus formation at the site of injury A B

C Platelet accumulation at the site of injury

D Fibrin generation at the site of injury

Figure 3. In vivo effect of the infusion of patient’s purified immunoglobulin G on thrombus formation and fibrin generation using intravital microscopy and a mouse model of laser-induced arteriolar injury. Intravital videomicroscopy of the cremaster muscle microcirculation was performed as previously described.11,12 Wild-type mice were pre-anesthetized with intraperitoneal ketamine (Fort Dodge) and xylazine (Llyod). A canulus was inserted in the jugular vein to maintain anesthesia with pentobarbital (Sigma-Aldrich) and infuse platelet and fibrin labeling as well as patient’s purified Immunoglobulin G (IgG). Digital images were captured with a C9300 Digital Camera (Hamamatsu) connected to a VS4-185 Image Intensifier Gen III (VideoScope International). Laser injury: vessel wall injury was induced with a micropoint laser system (Photonics Instruments) focused through the microscope objective. Multiple thrombi were studied in a single mouse, with new thrombi formed upstream of earlier thrombi. Image analysis: fluorescence data were captured digitally over 3 minutes following vessel injury. Image analysis was performed with Slidebook Version 5.5 (Intelligent Imaging Innovations) as previously described.9 In order to account for the variability of thrombus formation in any given set of experimental conditions, the data from 10-15 thrombi in two separate mice were used to determine the median value of the integrated fluorescence intensity before and 10 minutes after intravenous infusion of 690 mg of patient’s purified IgG. Platelet and fibrin labeling were performed using anti-CD42 antibody conjugated with DyLight488 (0,1 mg/g mouse, green) (Emfret) and with anti-fibrin antibody conjugated with AlexaFluor647 (0,5 mg/g mouse, red) (Sekisui). The kinetics of platelet accumulation and fibrin formation at the site of injury were determined by calculating median fluorescence values at 649 and 488 nm over time. (A-B) Thrombus formation at the site of injury. Representative images of fluorescence signal associated with platelets (green) and fibrin (red) over 120 seconds (T0-T120) following vessel injury (white arrow) are shown within the context of brightfield histology in the absence (A, left panel) or 10 minutes after infusion of patient’s IgG (690 mg) (B, right panel). (C-D) Platelet accumulation (C) and fibrin generation (D) at the site of injury. The median integrated platelet fluorescence (C, F platelet) or fibrin fluorescence (D, F fibrin) as a function of time in the absence (blue line = 10 thrombi, two mice) or 10 minutes after infusion of patient’s IgG (black line = 15 thrombi, two mice).

patient’s ATA in vitro, aPTT and TT were performed after addition of the patient’s purified ATA IgG (2.45 mg/mL). Compared to control IgG (2.5 mg/mL), addition of the patient’s ATA IgG to NPP led to a prolonged aPTT (59 versus 33 sec) and a prolonged TT (>120 vs. 19 sec). In order to demonstrate the anti-prothrombotic effect of the patient’s ATA in vivo, infusion of the patient’s ATA IgG was carried out in a mouse model of laser-induced arteriolar thrombus formation using intravital microscopy as previously described.11 In this model, infusion of control IgG (up to 1 mg) induced no modification of platelet accumulation and fibrin generation at the site of injury in cremaster arterioles in wild-type mice.12 By contrast, infusion of the patient’s ATA IgG (690 mg) induced a dramatic inhibition of platelet accumulation and fibrin generation of more than 70% at the site of injury (Figure 3). In essence, the presence of ATA is a rare disorder that can be either completely asymptomatic or lead to a severe bleeding tendency. ATA produce biological abnormalities in clotting tests such as dramatically prolonged 318

aPTT and TT, while coagulation factors are normal or subnormal. It can sometimes be associated with an antiphospholipid activity,8,13 as in our case. The level of the inhibitor can be measured using a modified Bethesda assay such as the TT mixing studies we performed. Still, there is no correlation between biological results and clinical manifestation, making it impossible to anticipate the individual hemorrhagic risk in a patient. On the other hand, global coagulation tests such as TGT have been proposed to better reflect and evaluate the overall hemostatic capacity and to correlate more adequately with bleeding phenotypes in acquired hemorrhagic disorders such as acquired hemophilia A.14,15 In our study, the results of theTGT 1 revealed the acquired origin of the disorder by inducing a dramatic increase in LT and decrease in ETP when adding serial dilutions of patient’s PPP in NPP; and TGT 2 reflected the patient’s severe hemorrhagic phenotype as no thrombin generation could be detected during 1 hour in the patient’s PPP. Conversely, there was a measurable peak of thrombin generation in patient’s PRP. We hypothesize that the ATA haematologica | 2021; 106(1)

Case Reports

induces an inhibition of the initial (low concentration) TF-triggered thrombin formation as well as further amplification of thrombin generation in PPP. In PRP, the ATAinduced inhibition might not be sufficient to prevent thrombin-induced-platelet activation. Therefore, the release of platelet-stored coagulation factors, and the residual traces of thrombin, might be sufficient to partially overcome the action of ATA at the surface of activated platelets, leading to a delayed and decreased but detectable thrombin generation. We further demonstrate the anti-thrombotic potency of the patient’s purified ATA in vitro and in vivo. Indeed, purified the patient’s ATA IgG retained the capacity to reproduce the biological abnormalities observed in the patient’s PPP, such as prolongation of aPTT and TT. Concomitantly, using a murine model of laser-induced arteriolar injury, we showed that infusion of the purified patient’s ATA IgG was specifically responsible for a significant inhibition of thrombus formation at the site of injury, illustrating the ATA-induced impairment on thrombus formation in vivo. In conclusion, we report a severe acquired hemorrhagic disorder secondary to an ATA in a patient presenting with a MIgG. We demonstrated the anti-thrombotic property of the MIgG through in vitro, ex vivo and in vivo experiments that allowed establishing a link between the biological and clinical presentation. This is the first description of the direct responsibility of a MGCS-associated ATA in the occurrence of a patient’s hemorrhagic tendency. Berenice Schell,1 Celine Desconclois,1 Xavier Mariette,2,3 Cecile Goujard,3,4 Peter J. Lenting,3,5 Cecile V. Denis3,5 and Valerie Proulle1,3,5 1 Service Hematologie Biologique, Hopital Bicetre, AP-HP, Le Kremlin-Bicetre; 2Service Rhumatologie, Hopital Bicetre, AP-HP, INSERM UMR_S U1184, Le Kremlin-Bicetre; 3Universite ParisSaclay, Paris; 4Service Medecine Interne, Hopital Bicetre, AP-HP, Le Kremlin-Bicetre and 5INSERM UMR_S 1176, Le Kremlin-Bicetre, France Correspondence: VALERIE PROULLE - valerie.proulle@aphp.fr doi:10.3324/haematol.2019.242370 Disclosures:no conflicts of interests to disclose. Contributions: VP designed the study, made the diagnosis, performed experiments, wrote the manuscript and is taking primary responsibility for the article. BS analyzed the data and wrote the manuscript. CD performed experiments. XM and CG managed the patient. CD, XM, CG, PL and CD revised the manuscript.

haematologica | 2021; 106(1)

References 1. Fermand JP, Bridoux F, Dispenzieri A, et al. Monoclonal gammopa-

thy of clinical significance: a novel concept with therapeutic implications. Blood. 2018;132(14):1478-1485. 2. Tiede A. Diagnosis and treatment of acquired von Willebrand syndrome. Thromb Res. 2012;130(Suppl 2):S2-6. 3. Kruse-Jarres R, Kempton CL, Baudo F, et al. Acquired hemophilia A: updated review of evidence and treatment guidance. Am J Hematol. 2017;92(7):695-705. 4. La Spada AR, Skalhegg BS, Henderson R, Schmer G, Pierce R, Chandler W. Brief report: fatal hemorrhage in a patient with an acquired inhibitor of human thrombin. N Engl J Med. 1995; 333(8):494-497. 5. Costa JM, Fiessinger JN, Capron L, Aiach M. Partial characterization of an autoantibody recognizing the secondary binding site(s) of thrombin in a patient with recurrent spontaneous arterial thrombosis. Thromb Haemost. 1992;67(2):193-199. 6. Baglin TP, Langdown J, Frasson R, Huntington JA. Discovery and characterization of an antibody directed against exosite I of thrombin. J Thromb Haemost. 2016;14(1):137-142. 7. Lawson JH, Pennell BJ, Olson JD, Mann KG. Isolation and characterization of an acquired antithrombin antibody. Blood. 1990;76(11):2249-2257. 8. Hwang KK, Grossman JM, Visvanathan S, et al. Identification of anti-thrombin antibodies in the antiphospholipid syndrome that interfere with the inactivation of thrombin by antithrombin. J Immunol. 2001;167(12):7192-7198. 9. Barthels M, Heimburger N. Acquired thrombin inhibitor in a patient with liver cirrhosis. Haemostasis. 1985;15(6):395-401. 10. Chuang YC, Lin YS, Liu HS, Wang JR, Yeh TM. Antibodies against thrombin in dengue patients contain both anti-thrombotic and profibrinolytic activities. Thromb Haemost. 2013;110(2):358-365. 11. Proulle V, Furie RA, Merrill-Skoloff G, Furie BC, Furie B. Platelets are required for enhanced activation of the endothelium and fibrinogen in a mouse thrombosis model of APS. Blood. 2014;124(4):611-622. 12. Arad A, Proulle V, Furie RA, Furie BC, Furie B. Beta(2)-glycoprotein1 autoantibodies from patients with antiphospholipid syndrome are sufficient to potentiate arterial thrombus formation in a mouse model. Blood. 2011;117(12):3453-3459. 13. Zhang S, Wu Z, Li J, et al. Clinical performance of antibodies to prothrombin and thrombin in Chinese patients with antiphospholipid syndrome: potential interest in discriminating patients with thrombotic events and non-thrombotic events. Rheumatol Int. 2017; 37(4):579-584. 14. Matsumoto T, Nogami K, Shima M. A combined approach using global coagulation assays quickly differentiates coagulation disorders with prolonged aPTT and low levels of FVIII activity. Int J Hematol. 2017;105(2):174-183. 15. Millet A, Graveleau J, Gueret P, et al. Thrombin generation in patients with acquired haemophilia and clinical bleeding risk. Br J Haematol. 2011;153(1):136-139.

319

ERRATA CORRIGE Low-dose X-rays leave scars on human hematopoietic stem and progenitor cells: the role of reactive oxygen species Masayuki Yamashita1 and Toshio Suda2,3 1

Division of Stem Cell and Molecular Medicine, Center for Stem Cell Biology and Regenerative Medicine, The Institute of Medical Science, The University of Tokyo, Tokyo, Japan; 2International Research Center for Medical Sciences, Kumamoto University, Kumamoto, Japan and 3 Cancer Science Institute, National University of Singapore, Singapore

In the article by Yamashita M and Suda T entitled “Low-dose X-rays leave scars on human hematopoietic stem and progenitor cells: the role of reactive oxygen species” published in Haematologica 2020;105(8):1986-1988:

X-ray has been replaced with ionizing radiation as highlighted in the text below. doi:10.3324/haematol.2020.272898

Low-dose ionizing radiations leave scars on human hematopoietic stem and progenitor cells: the role of reactive oxygen species Masayuki Yamashita1 and Toshio Suda2,3 1

fter Röntgen’s discovery in 1895, an X-ray became a game changer in medicine.1 It was discovered as an invisible ray of light that passes through many objects, including human bodies, and visualizes the internal organs and structures as silhouettes. As now seen in medical radiography, such as chest X-rays and computed tomography (CT) scans, ionizing radiation (IR) has enabled investigation of deep tissues in humans that had been otherwise impossible without surgical intervention, contributing to the early detection and treatment of many diseases. However, as is often the case with new medicine, were shown to have a biohazard effect.2 They are identified as a type of IR: a stream of high energy photons that are strong enough to ionize atoms and disrupt molecular bonds in biomolecules, including DNA. As DNA encodes an essential blueprint of a cell, the DNA-damaging property of IR can be toxic. This effect, although used for killing cancer cells in radiotherapy, has raised concerns about the effect of IR on normal tissues and whether the benefits exceed the risks. Modern medicine relies heavily on radiography to assess human health. The annual doses of IR people receive are increasing. A recent study estimated that around 2% or 4,000,000 of the non-elderly adults in the US receive 20 milligray (mGy) or more per year due to medical requirements.3 Historically, risks associated with low-dose IR are considered to be almost negligible as it does not cause any acute toxicity, nor does it increase the risk of carcinogenesis, based on empirical linear fits of existing human data determined at high doses, such as those of Japanese atomic bomb survivors.4 Indeed, low-dose IR rarely induces DNA double strand breaks (DSB), which often cause mutations and are considered to be the most relevant lesion for the deleterious effects of IR.5 However, even though low-dose IR rarely cause DSB, they are reportedly less easy to repair than those induced by high-dose IR .6 Importantly, recent evidence suggests that cumulative doses of 50 mGy IR (doses equivalent to 5-10 brain CT scans when given in childhood) have long320

term detrimental effects on human health, including a more than 3-fold increase in the risks of acute lymphoblastic leukemia and myelodysplastic syndrome.7 Furthermore, mouse studies demonstrate that low-dose IR affect function of long-lived tissue-specific stem cells, including hematopoietic stem cells (HSC).8,9 Thus, understanding the persistent effect of low-dose IR on human tissue-specific stem cells is of particular importance in precisely evaluating the risks posed by radiography on public health. In this issue of Haematologica, Henry et al. compared the effects of low and high doses of IR on hematopoietic stem and progenitor cells (HSPC) obtained from human umbilical cord blood (CB) (Figure 1).10 HSPC sustain themselves via self-renewing ability, and give rise to all of the blood lineage cells, such as innate and acquired immune cells, erythrocytes and platelets, through multi-lineage differentiation. They found that a single dose of 20 mGy IR is sufficient to impair the self-renewing capacity of CB HSPC. Intriguingly, this effect is independent of canonical DNA damage response (DDR), as a 20 mGy dose fails to induce DSB markers γH2AX and 53BP1 foci, or DDR hallmarks phospho-ATM and -p53, all of which are induced by a 2.5 Gy dose. Instead, the authors demonstrate that it is mediated by reactive oxygen species (ROS), a highly reactive oxygen byproduct mainly generated via the cell respiratory process of oxidative phosphorylation (OXPHOS) in mitochondria, and p38/MAPK14, a key enzyme that, upon elevation of ROS, sends a signal to HSPC to inhibit their self-renewing potential.11 Thus, the results of Henry et al. indicate that low-dose IR impair human CB HSPC function through ROS and p38/MAPK14, but not via canonical DDR via ATM or p53. The high sensitivity of HSC to elevated levels of ROS is well established, first in ATM deficiency and later in the contexts of other stress conditions.11-13 Similarly, p38/MAPK14 activation in response to ROS elevation is identified as a common downstream pathway responsible for impairment of self-renewal in HSC.11,12 In contrast, what haematologica | 2021; 106(1)

Errata Corrige

Figure 1. Response of human cord blood (CB) hematopoietic stem and progenitor cells (HSPC) to low- and high-dose ionizing radiation (IR) irradiation demonstrated by Henri et al.10 A low dose of 0.02 Gy (20 mGy) IR induces reactive oxygen species (ROS) elevation coupled with decrease in mitochondrial membrane potential (ΔΨ), which leads to increase in oxidative stress represented by formation of 8-oxo-deoxyguanosine (8-oxo-dG) in DNA, nuclear expression of NRF2, and activation of p38/MAPK14. The p38/MAPK14 activation mediates a decline in self-renewing capacity of HSPC without affecting their differentiating potential. The low-dose IR do not induce γ-H2AX and 53BP1 foci that represents nuclear DNA double strand breaks (DSB), or canonical DNA damage response via phosphorylation of ATM and p53. In contrast, a high dose of 2.5 Gy IR irradiation causes both ROS elevation and nuclear DSB. As a result, ROS inhibition either by N-acetylcysteine (NAC) or catalase, or p38 inhibition by SB203580, can reverse the detrimental effect by low dose, but not high dose, of IR on self-renewal capacity of HSPC.

is often unclear is the upstream mediator that causes ROS elevation. In the context of low-dose IR, mouse studies have uncovered the hypersensitivity of HSC and esophageal stem cells to low-dose IR that is mediated by ROS elevation, although the molecular link between low-dose IR and elevated ROS has not yet been investigated.8,9 It is estimated that approximately 90% of ROS can be generated during OXPHOS in mitochondria,14 mainly through functions of complexes I and III.15 Interestingly, the results shown by Henry et al. indicate that ROS elevation in human CB HSPC upon exposure to 20 mGy IR is closely associated with loss of mitochondrial membrane potential, which reflects a decrease in proton gradient across the cristae and often correlates with mitochondrial dysfunction.10 Apart from nucleus, mitochondria are the only organelle in mammalian cells that contain DNA, which can also be damaged by low-dose IR.16 Mitochondrial DNA (mtDNA) encodes proteins that consist of complexes I and ATP synthase, both of which are essential for proper electron transport and OXPHOS. Of note, these components are located in the so-called “common deletion” region of mtDNA that is commonly deleted upon exposure to low-dose IR. mtDNA is not protected by histones, and is thus potentially more susceptible to IR-induced damage compared to nuclear DNA. Moreover, mtDNA is located in matrix inside haematologica | 2021; 106(1)

inner membranes where ROS is generated, and is thus more greatly affected by IR-induced oxidative stress than nuclear DNA. Damage in mtDNA is not so simple as that in nuclear DNA, as a cell can contain more than 1,000 copies of mtDNA. Furthermore, numbers of mitochondria are dynamically changed by fusion and fission, which play critical roles in maintaining functional mitochondria via inter-mitochondrial complementation and quality control.17 In addition, damaged mitochondria can be removed by autophagy, which contributes to maintenance of self-renewal capacity of HSC.18,19 Henry et al. show that mitochondrial mass in HSPC does not seem to change after irradiation of 20 mGy IR.10 Although this observation should be validated by other methods,20 it supports the idea that changes in mitochondrial mass are unlikely to be the cause of ROS elevation. Rather, it is tempting to speculate that damage in mtDNA induced by low-dose IR causes persistent changes in mitochondrial function that lead to initial elevation of ROS and longterm impairment of HSC function. This would be consistent with the results reported by Rodrigues-Moreira et al., which demonstrate that low-dose IR cause biphasic elevations of ROS and the second wave of ROS elevation causes persistent reduction in self-renewing capacity of mouse bone marrow HSC.9 Mitochondrial dysfunction, but not constant elevation of ROS, is implicated in age321

Errata Corrige

associated decline in HSC function.18 Since involvement of mtDNA remains unknown, investigating whether aged HSC have mtDNA damage would be of particular interest. Collectively, identifying molecular ‘scars’ left by low-dose IR on HSC would help provide a precise evaluation of the long-term detrimental effects by medical radiographic examination and also find common mechanisms that underlie hematopoietic aging and disease.

References 1. Röntgen WC. On a New Kind of Rays. Science. 1896;3(59):227-231. 2. Daniel J. The X-Rays. Science. 1896;3(67):562-563. 3. Fazel R, Krumholz HM, Wang Y, et al. Exposure to low-dose ionizing radiation from medical imaging procedures. N Engl J Med. 2009;361(9):849-857. 4. Preston DL, Kusumi S, Tomonaga M, et al. Cancer incidence in atomic bomb survivors. Part III. Leukemia, lymphoma and multiple myeloma, 1950-1987. Radiat Res. 1994;137(2 Suppl):S68-S97. 5. Hoeijmakers JH. Genome maintenance mechanisms for preventing cancer. Nature. 2001;411(6835):366-374. 6. Rothkamm K, Lobrich M. Evidence for a lack of DNA double-strand break repair in human cells exposed to very low x-ray doses. Proc Natl Acad Sci U S A. 2003;100(9):5057-5062. 7. Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet. 2012;380(9840):499505. 8. Fernandez-Antoran D, Piedrafita G, Murai K, et al. Outcompeting p53-mutant cells in the normal esophagus by redox manipulation. Cell Stem Cell. 2019;25(3):329-341.e6. 9. Rodrigues-Moreira S, Moreno SG, Ghinatti G, et al. Low-dose irradiation

322

10.

11. 12. 13. 14. 15. 16. 17. 18. 19. 20.

promotes persistent oxidative stress and decreases self-renewal in hematopoietic stem cells. Cell Rep. 2017;20(13):3199-3211. Henry E, Souissi-Sahraoui I, Deynoux M, et al. Human hematopoietic stem/progenitor cells display ROS-dependent long-term hematopoietic defects after exposure to low dose of ionizing radiations. Haematologica. 2020;105(8):2044-2055. Ito K, Hirao A, Arai F, et al. Reactive oxygen species act through p38 MAPK to limit the lifespan of hematopoietic stem cells. Nat Med. 2006;12(4):446-451. Miyamoto K, Araki KY, Naka K, et al. Foxo3a is essential for maintenance of the hematopoietic stem cell pool. Cell Stem Cell. 2007;1(1):101-112. Tothova Z, Kollipara R, Huntly BJ, et al. FoxOs are critical mediators of hematopoietic stem cell resistance to physiologic oxidative stress. Cell. 2007;128(2):325-339. Balaban RS, Nemoto S, Finkel T. Mitochondria, oxidants, and aging. Cell. 2005;120(4):483-495. Nissanka N, Moraes CT. Mitochondrial DNA damage and reactive oxygen species in neurodegenerative disease. FEBS Lett. 2018;592(5):728-742. Kawamura K, Qi F, Kobayashi J. Potential relationship between the biological effects of low-dose irradiation and mitochondrial ROS production. J Radiat Res. 2018;59(Suppl 2):ii91-ii97. Youle RJ, van der Bliek AM. Mitochondrial fission, fusion, and stress. Science. 2012;337(6098):1062-1065. Ho TT, Warr MR, Adelman ER, et al. Autophagy maintains the metabolism and function of young and old stem cells. Nature. 2017;543 (7644):205-210. Ito K, Turcotte R, Cui J, et al. Self-renewal of a purified Tie2+ hematopoietic stem cell population relies on mitochondrial clearance. Science. 2016;354(6316):1156-1160. de Almeida MJ, Luchsinger LL, Corrigan DJ, Williams LJ, Snoeck HW. Dye-Independent Methods Reveal Elevated Mitochondrial Mass in Hematopoietic Stem Cells. Cell Stem Cell. 2017;21(6):725729.e4.

haematologica | 2021; 106(1)