Haematologica, Volume 106, Issue 11 by Haematologica

haematologica Journal of the Ferrata Storti Foundation

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

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

Associate Editors Hélène Cavé (Paris), Monika Engelhardt (Freiburg), Steve Lane (Brisbane), Pier Mannuccio Mannucci (Milan), Pavan Reddy (Ann Arbor), David C. Rees (London), Francesco Rodeghiero (Vicenza), Gilles Salles (New York), Kerry Savage (Vancouver), Aaron Schimmer (Toronto), Richard F. Schlenk (Heidelberg), Sonali Smith (Chicago)

Statistical Consultant Catherine Klersy (Pavia)

Editorial Board Walter Ageno (Varese), Sarit Assouline (Montreal), Andrea Bacigalupo (Roma), Taman Bakchoul (Tübingen), Pablo Bartolucci (Créteil), Katherine Borden (Montreal), Marco Cattaneo (Milan), Corey Cutler (Boston), Kate Cwynarski (London), Mary Eapen (Milwaukee), Francesca Gay (Torino), Ajay Gopal (Seattle), Alex Herrera (Duarte), Shai Izraeli (Ramat Gan), Martin Kaiser (London), Marina Konopleva (Houston), Johanna A. Kremer Hovinga (Bern), Nicolaus Kröger (Hamburg), Austin Kulasekararaj (London), Shaji Kumar (Rochester), Ann LaCasce (Boston), Anthony R. Mato (New York), Neha Mehta-Shah (St. Louis), Alison Moskowitz (New York), Yishai Ofran (Haifa), Farhad Ravandi (Houston), John W. Semple (Lund), Liran Shlush (Toronto), Sara Tasian (Philadelphia), Pieter van Vlieberghe (Ghent), Ofir Wolach (Haifa), Loic Ysebaert (Toulouse)

Managing Director Antonio Majocchi (Pavia)

Editorial Office Lorella Ripari (Office & Peer Review Manager), Simona Giri (Production & Marketing Manager), Paola Cariati (Graphic Designer), Giulia Carlini (Graphic Designer), Igor Poletti (Graphic Designer), Marta Fossati (Peer Review), Diana Serena Ravera (Peer Review) , Laura Sterza (Account Administrator)

Assistant Editors Britta Dorst (English Editor), Rachel Stenner (English Editor), Bertie Vitry (English Editor), Massimo Senna (Information technology), Idoya Lahortiga (Graphic artist)

haematologica Journal of the Ferrata Storti Foundation

Brief information on Haematologica 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, Original articles, Review articles, Perspective articles, Editorials, Guideline articles, Letters to the Editor, Case reports & Case series and Comments. 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 at 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. Subscription. Detailed information about subscriptions is available at www.haematologica.org. Haematologica is an open access journal and access to the online journal 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 printed edition for the year 2021 are as following: 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 11: November 2021 About the Cover 2791

Images from the Haematologica Atlas of Hematologic Cytology: pure erythroid leukemia Rosangela Invernizzi https://doi.org/10.3324/haematol.2021.279628

Editorials 2792

Band 3, an essential red blood cell hub of activity Timothy J. Satchwell and Ashley M. Toye https://doi.org/10.3324/haematol.2021.278643

2794

Improving the treatment of childhood acute lymphoblastic leukemia by optimizing the use of 70-year-old drugs William E. Evans https://doi.org/10.3324/haematol.2021.278967

2796

Time to reconsider CD33 single nucleotide polymorphism in the response to gemtuzumab ozogamicin Jatinder K. Lamba and Soheil Meshinchi https://doi.org/10.3324/haematol.2021.279043

Review Articles 2799

2021 European Myeloma Network review and consensus statement on smoldering multiple myeloma: how to distinguish (and manage) Dr. Jekyll and Mr. Hyde Pellegrino Musto et al. https://doi.org/10.3324/haematol.2021.278519

2813

The possible role of mutated endothelial cells in myeloproliferative neoplasms Mirko Farina, Domenico Russo and Ronald Hoffman https://doi.org/10.3324/haematol. 2021.278499

Articles Acute Lymphoblastic Leukemia 2824 Increments in DNA-thioguanine level during thiopurine-enhanced maintenance therapy of acute lymphoblastic leukemia Rikke Hebo Larsen et al. https://doi.org/10.3324/haematol.2020.278166

Chronic Lymphocytic Leukemia 2834 Venetoclax plus bendamustine-rituximab or bendamustine-obinutuzumab in chronic lymphocytic leukemia: final results of a phase Ib study (GO28440) Stephan Stilgenbauer et al. https://doi.org/10.3324/haematol.2020.261107

2845

Hodgkin lymphoma arising in patients with chronic lymphocytic leukemia: outcomes from a large multi-center collaboration Deborah M. Stephens et al. https://doi.org/10.3324/haematol.2020.256388

Chronic Myeloid Leukemia 2853 The effect of eltrombopag in managing thrombocytopenia associated with tyrosine kinase therapy in patients with chronic myeloid leukemia and myelofibrosis Mahran Shoukier et al. https://doi.org/10.3324/haematol.2020.260125

Haematologica 2021; vol. 106 no. 11 - November 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Hematopoiesis 2859 Characterization and evolutionary origin of novel C2H2 zinc finger protein (ZNF648) required for both erythroid and megakaryocyte differentiation in humans Daniel C. J. Ferguson et al. https://doi.org/10.3324/haematol.2020.256347

Hemostasis 2874 Targeting shear gradient activated von Willebrand factor by the novel single-chain antibody A1 reduces occlusive thrombus formation in vitro Thomas Hoefer et al. https://doi.org/10.3324/haematol.2020.250761

Iron Metabolism & its Disorders 2885 Intravenous iron preparations transiently generate non-transferrin-bound iron from two proposed pathways Maciej W. Garbowski et al. https://doi.org/10.3324/haematol.2020.250803

2897

The TMPRSS6 variant (SNP rs855791) affects iron metabolism and oral iron absorption – a stable iron isotope study in Taiwanese women Simone Buerkli et al. https://doi.org/10.3324/haematol.2020.264556

Myelodysplastic Syndromes 2906 Replication stress signaling is a therapeutic target in myelodysplastic syndromes with splicing factor mutations Johanna Flach et al. https://doi.org/10.3324/haematol.2020.254193

Non-Hodgkin Lymphoma 2918 CDKN2A deletion is a frequent event associated with poor outcome in patients with peripheral T-cell lymphoma not otherwise specified (PTCL-NOS) Francesco Maura et al. https://doi.org/10.3324/haematol.2020.262659

2927

Transducin β-like protein 1 controls multiple oncogenic networks in diffuse large B-cell lymphoma Youssef Youssef et al. https://doi.org/10.3324/haematol.2020.268235

2940

A prognostic index predicting survival in transformed Waldenström macroglobulinemia Eric Durot et al. https://doi.org/10.3324/haematol.2020.262899

Platelet Biology & its Disorders 2947 Platelet proteome and function in X−linked thrombocytopenia with thalassemia and in silico comparisons with gray platelet syndrome Daniel Bergemalm et al. https://doi.org/10.3324/haematol.2020.249805

Red Cell Biology & its Disorders 2960 Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I (CDA-I) identifies defects in differentiation and nucleolar abnormalities Caroline Scott et al. https://doi.org/10.3324/haematol.2020.260158

2971

The interactome of the N-terminus of band 3 regulates red blood cell metabolism and storage quality Aaron Issaian et al. https://doi.org/10.3324/haematol.2020.278252

Haematologica 2021; vol. 106 no. 11 - November 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Letters to the Editor 2986

Cluster of differentiation 33 single nucleotide polymorphism rs12459419 is a predictive factor in patients with nucleophosmin1-mutated acute myeloid leukemia receiving gemtuzumab ozogamicin Katrin Teich et al. https://doi.org/10.3324/haematol.2021.278894

2990

Aurora A kinase as a target for therapy in TCF3-HLF rearranged acute lymphoblastic leukemia Jessica Leonard et al. https://doi.org/10.3324/haematol.2021.278692

2995

IL4-STAT6 signaling induces CD20 in chronic lymphocytic leukemia and this axis is repressed by PI3K inhibitor idelalisib Veronika Sandova et al. https://doi.org/10.3324/haematol.2021.278644

3000

A novel classification of hematologic conditions in patients with Fanconi anemia Yvonne Lisa Behrens et al. https://doi.org/10.3324/haematol.2021.279332

3004

3008

mTOR inhibitors sensitize multiple myeloma cells to venetoclax via IKZF3- and Blimp-1-mediated BCL-2 upregulation Naoki Osada et al. https://doi.org/10.3324/haematol.2021.278506

3014

3017

Allogeneic hematopoietic stem cell transplantation from non-sibling 10/10 HLA-matched related donors: a single-center experience Yaoyao Shen et al. https://doi.org/10.3324/haematol.2021.278933

Case Reports 3021

Massive cerebral venous thrombosis due to vaccine-induced immune thrombotic thrombocytopenia Sara Bonato et al. https://doi.org/10.3324/haematol.2021.279246

3025

Intracerebral hemorrhage associated with vaccine-induced thrombotic thrombocytopenia following ChAdOx1 nCOVID-19 vaccine in a pregnant woman Daniela P. Mendes-de-Almeida et al. https://doi.org/10.3324/haematol.2021.279407

Haematologica 2021; vol. 106 no. 11 - November 2021 http://www.haematologica.org/

ABOUT THE COVER Images from the Haematologica Atlas of Hematologic Cytology: pure erythroid leukemia Rosangela Invernizzi University of Pavia, Pavia, Italy E-mail: ROSANGELA INVERNIZZI - rosangela.invernizzi@unipv.it doi:10.3324/haematol.2021.279628

ure erythroid leukemia (French-American-British classification M6) is characterized by >80% of erythroid precursors in the bone marrow with ≥30% proerythroblasts and a non-significant myeloblastic component. Morphological features of this extremely rare entity are shown in the Figure. Remarkable anisopoikilocytosis of red cells and numerous morphologically abnormal erythroblasts are commonly observed in the peripheral blood (A). Bone marrow is infiltrated by very immature and atypical erythroid precursors. These large cells have round nuclei with reticular chromatin and prominent nucleolus, and hyperbasophilic, agranular, often vacuolated cytoplasm (B) or megaloblastic features and a voluminous paranuclear clear area (C). Leukemic erythroblasts are negative for myeloperoxidase and Sudan black B staining, while Perls’ reaction may show several ring sideroblasts (D) and periodic-acid Schiff (PAS) stain is strongly positive with a granular pattern in the early erythroid precursors and a diffuse pattern in late erythroblasts (E). Atypical erythroid precursors show localized paranuclear reactivity for acid phosphatase (F).1 Disclosures No conflicts of interest to disclose.

Reference 1. Invernizzi R. Acute myeloid leukemia and related precursor neoplasms. Haematologica. 2020;105(Suppl1):98-119.

haematologica | 2021; 106(11)

2791

EDITORIALS Band 3, an essential red blood cell hub of activity Timothy J. Satchwell1,2,3 and Ashley M. Toye1,2,3 1

School of Biochemistry, University of Bristol; 2National Institute for Health Research (NIHR) Blood and Transplant Research Unit in Red Blood Cell Products, University of Bristol and 3Bristol Institute of Transfusion Sciences, NHSBT Filton, Bristol, UK. E-mail: ASHLEY TOYE - Ash.M.Toye@bristol.ac.uk doi:10.3324/haematol.2021.278643

he red blood cell (RBC) is a marvel of cellular evolutionary specialization. Whilst often considered simplistic owing to its absence of nuclei and other cellular organelles, this inability to respond transcriptionally and to replenish components through new protein synthesis necessitates complex post-translational mechanisms through which the cell is able to control, adapt and regulate its key functions in the different environments it experiences traversing the circulation. In this context, the membrane of the erythrocyte, perhaps more than any other cell, plays a crucial role. Erythrocyte membranes have a unique structure, comprising an array of integral membrane proteins with varied antigenic, transport and mechanical functions arranged in diverse multiprotein complexes.1 Underlying the membrane and connected to multicomplexes via adaptor proteins is the spectrin-based cytoskeletal meshwork. These protein interactions together with the membrane lipids themselves impart the characteristic properties of the biconcave RBC. Central to the stability and functional regulation of the RBC membrane is band 3 (Anion exchanger 1, AE1), its most abundant protein at approximately 1.2x106 copies per cell. Its unassuming name (a relic of its original identification as the

third band on a Coomassie-stained sodium dodecyl sulfate polyacrylamide gel electrophoresis separation of RBC membrane ‘ghosts’2) belies the importance this protein plays not just through its direct contribution to gas exchange through the electroneutral exchange of chloride ions with the bicarbonate product (HCO -) of carbon dioxide metabolism, but as the major site of membrane cytoskeletal connectivity and a hub for hemoglobin and glycolytic enzyme binding. A substantial body of evidence supports the binding of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) to the cytosolic N-terminus of band 33,4 with competitive binding with the deoxygenated form of hemoglobin, first proposed by the Low group, to direct the metabolism of glucose between glycolysis and the pentose phosphate pathway in response to cellular needs and the requirement for antioxidative measures. Put simplistically, band 3 and its interactions act as the RBC molecular switch to sense the oxygenated/deoxygenated state, imparting an appropriate metabolic response (see summary schematic in Figure 1A, B). Crammed into a single paper published in this issue of Haematologica, Issaian and colleagues have completed a heroic effort to deepen our understanding of the role that glycolytic enzyme binding to band 3 plays in erythrocyte 3

Figure 1. The influence of band 3 on red blood cell metabolism in different conditions. The schematic, adapted from Issaian et al.5 shows different conditions and the influence of band 3 on red blood cell metabolism in three different scenarios. (A) In deoxygenation/low oxidant stress conditions, deoxyhemoglobin can bind to the N-terminus of band 3, while glycolytic enzymes become displaced from the same region and are then activated, resulting in increased fluxes through glycolysis and decreased fluxes through the pentose phosphate pathway. (B) In oxygenated or high oxidative stress conditions, glycolytic enzymes can bind to the N-terminus of band 3, resulting in their partial inhibition, decreasing the metabolic flux through glycolysis and increasing flux through the pentose phosphate pathway to generate the reducing equivalent NADPH necessary to counteract oxidant stress. (C) In the absence of the extreme N-terminus of band 3 during storage, under high oxidative stress conditions, glycolytic enzymes cannot bind band 3 N-termini, so glyceraldehyde 3-phosphate dehydrogenase flux continues, and the pentose phosphate pathway is suppressed, leading to enhanced oxidative stress. G6P: glucose-6-phosphate; PFK: phospho-fructokinase; FBP: fructose 1,6, biphosphatase; ALDO: aldolase; G3P: glucose-3-phosphate; GAPDH: glyceraldehyde 3-phosphate dehydrogenase; DPG: diphosphoglycerate; ATP: adenosine triphosphate; LAC: lactate; Hb: hemoglobin; PPP: pentose phosphate pathway: NADPH: nicotinamide adenine dinucleotide phosphate. Image adapted from Issaian et al.5

2792

haematologica | 2021; 106(11)

Editorials

metabolism.5 This was achieved using erythrocytes from well-characterized mouse models and a rare natural human variant (Band 3 Neapolis) with truncations at the band 3 N-terminus, in combination with detailed metabolic analyses. The authors demonstrate that loss of band 3 or mutation of the N-terminal GAPDH binding site causes a failure of pentose phosphate pathway activation, with concurrent defects in glutathione recycling and evidence of increasing oxidation products arising during storage in the absence of the extreme N-terminal 11 amino acids of band 3 (summarized in Figure 1C). Interestingly, the authors also show that reintroduction of the band 3 Nterminus using cell-penetrating peptides can rescue the observed metabolic defects. This metabolomics work is striking in the new level of detail it provides about the metabolism of normal and variant RBC. Another impressive feature of this manuscript is the depth of characterization of the band 3 N-terminal interactome provided by the authors. Varied proteomic approaches, including chemical crosslinking, characterized site-specific interaction interfaces between GAPDH and band 3, with additional prospective interacting proteins reported including enzymes involved in glutathione synthesis, recycling and lipid peroxidation pathways. However, dissecting the functional relevance of the band 3 interactome is beset with substantial challenges, with its own abundance, existence within at least three (and potentially many more sub-) populations each with genuine direct and additional indirect interactants,6 all complicating interpretation. It is unlikely and indeed in some cases impossible (where binding sites for multiple proteins overlap) that all of the proteins reported bind to all of the copies of band 3, or even every band 3-containing complex. Efforts such as this, and others to uncover not just the identity but the plasticity and relevance of such interactions under physiologically relevant conditions, be they altered oxygenation,7 tonicity,8 mechanical deformation9 or storage such as the work described here, are key to our evolving understanding of RBC structure-function relationships. Finally, given the essential role band 3 plays in RBC metabolism, the manuscript also touches on the potential of band 3 polymorphisms to influence storage characteristics, using data from the REDSIII storage study.10 Here a note of caution needs to be applied as, besides a handful of already known polymorphisms in the band 3 N-terminus (previously shown to be associated with hereditary spherocytosis), it remains unclear what precise impact these mutations have. Any alteration in band 3 abundance has knock-on effects on RBC membrane stability

haematologica | 2021; 106(11)

which will impact on the expression of band 3-dependent and -independent proteins alike. Nevertheless, more widely, further determination of the mechanistic basis by which polymorphisms (both pathogenic and non-pathogenic) can influence properties of the RBC beyond those most obviously apparent represents an important continuing endeavor. Disclosures No conflicts of interest to disclose. Contributions TJS and AMT wrote the review together. The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR or the Department of Health and Social Care. Funding TJS and the work in AMT’s laboratory is funded in part by a National Institute for Health Research (NIHR) Blood and Transplant Research Unit in Red Blood Cell Products award (IS-BTU-1214-10032) (University of Bristol) and also National Health Service Blood and Transplant (NHSBT) R&D grants (WP15-04; WP15-05).

References 1. Mankelow TJ, Satchwell TJ, Burton NM. Refined views of multiprotein complexes in the erythrocyte membrane. Blood Cells Mol Dis. 2012;49(1):1-10. 2. Fairbanks G, Steck TL, Wallach DF. Electrophoretic analysis of the major polypeptides of the human erythrocyte membrane. Biochemistry. 1971;10(13):2606-2617. 3. Chu H, Low PS. Mapping of glycolytic enzyme-binding sites on human erythrocyte band 3. Biochem J. 2006;400(1):143-151. 4. Lewis IA, Campanella ME, Markley JL, Low PS. Role of band 3 in regulating metabolic flux of red blood cells. Proc Natl Acad Sci U S A. 2009;106(44):18515-18520. 5. Issaian A, Hay A, Dzieciatkowska M, et al. The interactome of the N-terminus of band 3 regulates red blood cell metabolism and storage quality. Haematologica. 2021;106(11):2971-2985. 6. van den Akker E, Satchwell TJ, Williamson RC, Toye AM. Band 3 multiprotein complexes in the red cell membrane; of mice and men. Blood Cells Mol Dis. 2010;45(1):1-8. 7. Chu H, McKenna MM, Krump NA, et al. Reversible binding of hemoglobin to band 3 constitutes the molecular switch that mediates O2 regulation of erythrocyte properties. Blood. 2016;128(23): 2708-2716. 8. Hsu K, Lee TY, Periasamy A, et al. Adaptable interaction between aquaporin-1 and band 3 reveals a potential role of water channel in blood CO2 transport. FASEB J. 2017;31(10):4256-4264. 9. Anong WA, Weis TL, Low PS. Rate of rupture and reattachment of the band 3-ankyrin bridge on the human erythrocyte membrane. J Biol Chem. 2006;281(31):22360-22366. 10. Kanias T, Lanteri MC, Page GP, et al. Ethnicity, sex, and age are determinants of red blood cell storage and stress hemolysis: results of the REDS-III RBC-Omics study. Blood Adv. 2017;1(15):1132-1141.

2793

Editorials

Improving the treatment of childhood acute lymphoblastic leukemia by optimizing the use of 70-year-old drugs William E. Evans Emeritus Faculty, Past-President and CEO, St. Jude Children’s Research Hospital, Memphis, TN, USA E-mail: WILLIAM E. EVANS - william.evans@stjude.org doi:10.3324/haematol.2021.278967

n the current issue of Haematologica, Larsen, Schmiegelow and colleagues1 have provided clear and convincing evidence that the addition of a relatively low dose of thioguanine (≤12.5 mg/m2/day) can significantly increase the amount of thioguanine nucleotides incorporated into DNA (DNA-TG) of normal leukocytes in patients with ALL, when compared to treatment with only mercaptopurine in historical controls or in the same patients prior to the addition of low-dose thioguanine. As depicted in Figure 1, mercaptopurine requires intracellular metabolism by multiple enzymes to produce thioguanine nucleotides, whereas thioguanine is converted directly to thioguanine nucleotides. The deoxy thioguanine triphosphates are then available for incorporation into DNA, which is thought to be the principal mechanism of mercaptopurine’s antileukemic effects. Mercaptopurine is a mainstay of combination chemotherapy for the treatment of acute lymphoblastic leukemia (ALL), which is curative for over 90% of children and ~70% of adults, whereas thioguanine is not widely used to treat ALL. Mercaptopurine was the first antileukemic agent for which pharmacogenomics was shown to be an important determinant of the optimal dosage, with those inheriting non-functional variants of thiopurine methyltransferase (TPMT) more likely to develop dose-limiting hematologic toxicity if treated with conventional doses of mercaptopurine (75 mg/m2/day).2,3 In the 1990s, mercaptopurine became one of the first medications for which preemptive genotyping for common variants (in TPMT) were used to determine the optimal dosage,2,3 and 20 years later this strategy was expanded to include testing for inactivating variants in NUDT15 (nucleotide diphosphatase nudix hydrolase 15).4,5 Nonfunctional alleles of TPMT are the primary determinants of mercaptopurine toxicity in people of European and African ancestry, whereas NUDT15 variants are the primary determinants in people of Asian and Native American ancestry.6 Patients who inherit two non-functional alleles for either of these enzymes must be treated with only 5-10% of the conventional dose of mercaptopurine to avoid toxicity, whereas for heterozygous patients it is recommended reducing the starting dose by about 50%.4 Even with these dose reductions, these enzymedeficient patients maintain higher average erythrocyte thioguanine nucleotide levels than homozygous wildtype patients treated with full doses of mercaptopurine and have comparable cure rates. It is unclear whether TPMT-deficient or heterozygous patients require supplemental doses of thioguanine to achieve DNA-TG in the target range and, if so, what dosage of supplemental thioguanine should be given. As a complement to preemptive genotyping, monitoring the concentration of thioguanine nucleotides in erythrocytes is commonly used to identify patients who

2794

accumulate excessive levels of thioguanine nucleotides or patients who have low thioguanine nucleotide levels due to non-compliance with daily oral mercaptopurine therapy. Although measuring thioguanine nucleotides in erythrocytes is clinically useful, this is not measuring the active drug in the target tissue (leukemia cells) nor active drug at the presumed site of action (thioguanine incorporated into the DNA). It will be important to determine in a large prospective clinical trial whether measuring thioguanine incorporated into DNA is indeed a better metric of mercaptopurine treatment than measuring thioguanine nucleotides in erythrocytes, because measuring thioguanine in DNA requires a more complex assay, which may not be widely available. Larsen et al. report that such a clinical trial (ALLTogether-1) is ongoing. It is interesting that in the current report, Larsen et al. did not find any correlation between median erythrocyte thioguanine nucleotides and median thioguaninine incorporated into leukocyte DNA (Online Supplementary Figure S6C in the article by Larsen et al.1). Although it is not known how closely DNA incorporated into DNA of normal leukocytes reflects thioguanine incorporated into DNA of primary ALL cells in patients, it is reasonable to assume this is a better surrogate than measuring thioguanine nucleotides in the cytosol of erythrocytes, in part because only the trinucleotide is incorporated into DNA, whereas inactive mono- and di-phosphate nucleotides are measured in erythrocytes. Measuring thioguanine incorporated into the DNA of primary leukemia cells in patients would be the ideal metric, but this is not feasible because patients are generally in complete remission before mercaptopurine therapy is initiated and thus there are no leukemia cells to assess. It is also not known whether the incorporation of thioguanine into DNA of normal leukocytes has a uniform relation to thioguanine incorporated into DNA of leukemia cells of different molecular and lineage subtypes of ALL. These limitations notwithstanding, measuring thioguanine incorporated into DNA (DNA-TG) of normal leukocytes offers a potential advance for optimizing mercaptopurine treatment of ALL. A major unknown is what level of thioguanine incorporation into DNA is indicative of optimal treatment with mercaptopurine, which will require assessment of the relation between thioguanine in leukocyte DNA and event-free survival in a large enough cohort of uniformly treated patients so that all relevant covariates can be included in a multivariate analysis. It is also not known whether the level of increase in DNA-TG documented by Larsen et al. translates into an improvement in event-free survival, although Larsen et al. speculate that this could reduce the relapse hazard rate by as much as 59%, based on their prior research reporting a relapse hazard ratio of 0.81 per 100 fmol/mg DNA increase (95% confidence interval: 0.67-0.98; P=0.029).7 A

haematologica | 2021; 106(11)

Editorials

Figure 1. Schematic representation of the metabolism of mercaptopurine and thioguanine. (A, B) Metabolism of mercaptopurine (A) and thioguanine (B) and the enzymes involved hypoxanthine phosphoribosyltransferase (HPRT), inosine monophosphate dehydrogenase (IMPD), guanosine monophosphate synthase (GMPS), kinase, nucleoside kinases, thiopurine S-methyltransferase (TPMT), nucleotide diphosphatase nudix hydrolase 15 (NUDT15), 5'-nucleotidase, cytosolic II (NT5C2), xanthine oxidase (XO), aldehyde oxidase (AO) and phosphoribosyl pyrophosphate (PRPP). PRPP is a substrate in the reaction catalyzed by HPRT to form thiopurine nucleotides. The monophosphate of either the deoxy or ribonucleotide is formed by NUDT15, depending on the substrate. Enzymes shown in red catalyze inactivation of these thiopurine medications whereas those depicted in green are involved in activation to thioguanine nucleotides (TGN), which can be incorporated into DNA and RNA.

final issue that will need further study in a larger cohort of patients is whether the addition of this small dose of thioguanine is associated with additional toxicity during ALL therapy, as thioguanine has been associated with veno-occlusive disease in about 20% of children with ALL,8 and the risk with low-dose thioguanine when given concomitantly with mercaptopurine is unknown. Nowadays, we are faced with the challenge of pushing the cure rate of childhood ALL beyond 90% while also improving the quality of life for those we cure. Improving the cure rate from 20% to 90% over the past six decades has been largely achieved by optimizing the use of conventional chemotherapy, such as mercaptopurine, not by the development of new antileukemic agents. Most of these improvements have been incremental in nature, but their cumulative effects have produced remarkable progress.9,10 The work of Schmiegelow and colleagues1 may offer yet another small step toward maximizing the effects of medications that we have been using for many decades. Much hope and hype have been raised around

haematologica | 2021; 106(11)

the development of more targeted therapy for cancer, yet when available these targeted agents, such as tyrosine kinase inhibitors, are being added to and not replacing conventional chemotherapy in treating ALL. We would be wise not to abandon efforts to further improve the use of these older anticancer agents and avoid placing all our hope on so-called “targeted chemotherapy”. And it should not go unnoticed that we continue to expand our knowledge of how best to use anticancer agents developed 70 years ago, suggesting that in the coming decades we may still be optimizing the use of both targeted and conventional chemotherapy as we work to push the ALL cure rate closer to 100%. Disclosures No conflicts of interest to disclose. Acknowledgments I thank Mr. Josh Stokes for his many contributions to the preparation of the figure.

2795

Editorials

References 1. Larsen RH Rank CU, Grell K, et al. Increments in DNA-thioguanine level during thiopurine-enhanced maintenance therapy of acute lymphoblastic leukemia. Haematologica. 2021;106(11)2824-2833. 2. Yates CR, Krynetski EY, Loennechen T, et al. Molecular diagnosis of thiopurine S-methyltransferase deficiency: genetic basis for azathioprine and mercaptopurine intolerance. Ann Intern Med. 1997;126(8):608-614. 3. Relling MV, Hancock ML, Rivera GK, et al. Mercaptopurine therapy intolerance and heterozygosity at the thiopurine S-methyltransferase gene locus. J Natl Cancer Inst. 1999;91(23):2001-2008. 4. Relling MV, Schwab M, Whirl-Carrillo M, et al. Clinical Pharmacogenetics Implementation Consortium guideline for thiopurine dosing based on TPMT and NUDT15 genotypes: 2018 update. Clin Pharmacol Ther. 2019;105(5):1095-1105. 5. Moriyama T, Nishii R, Perez-Andreu V, et al. NUDT15 polymorphisms alter thiopurine metabolism and hematopoietic toxicity. Nat Genet. 2016;48(4):367-373.

6. Yang JJ, Landier W, Yang W, et al. Inherited NUDT15 variant is a genetic determinant of mercaptopurine intolerance in children with acute lymphoblastic leukemia. J Clin Oncol. 2015;33(11):12351242. 7. Nielsen SN, Grell K, Nersting J, et al. DNA-thioguanine nucleotide concentration and relapse-free survival during maintenance therapy of childhood acute lymphoblastic leukaemia (NOPHO ALL2008): a prospective substudy of a phase 3 trial. Lancet Oncol. 2017;18(4): 515-524. 8. Stork LC, Matloub Y, Broxson E, et al. Oral 6-mercaptopurine versus oral 6-thioguanine and veno-occlusive disease in children with standard-risk acute lymphoblastic leukemia: report of the Children's Oncology Group CCG-1952 clinical trial. Blood. 2010;115(14):27402748. 9. Pui CH, Yang JJ, Hunger SP, et al. Childhood acute lymphoblastic leukemia: progress through collaboration. J Clin Oncol. 2015;33(27): 2938-2948. 10. Pui CH, Evans WE. A 50-year journey to cure childhood acute lymphoblastic leukemia. Semin Hematol. 2013;50(3):185-196.

Time to reconsider CD33 single nucleotide polymorphism in the response to gemtuzumab ozogamicin Jatinder K. Lamba1 and Soheil Meshinchi2 1

Department of Pharmacotherapy and Translational Research, Center for Pharmacogenomics, College of Pharmacy, UF Health Cancer Center, University of Florida, Gainesville, FL and 2Division of Pediatric Hematology/Oncology, Fred Hutchinson Cancer Research Center, Seattle, WA, USA E-mail: JATINDER K. LAMBA - jlamba@cop.ufl.edu doi:10.3324/haematol.2021.279043

D33 is a highly sought-after target in acute myeloid leukemia (AML), with gemtuzumab ozogamicin (GO), a CD33-antibody conjugated to a DNA-damaging cytotoxin currently approved for the treatment of CD33+ adult and pediatric AML.1,2 The levels of expression of CD33 on the cell surface vary significantly between patients (up to 2 log-fold) and have been shown to be associated with disease characteristics as well as response to GO.3-5 However, as a biological threshold of CD33 expression that correlates with response to GO is lacking for incorporation into prospective trials to guide CD33-directed therapeutics, there is an urgent and unmet need to better define genomic variants that might predict response to GO. It has recently been reported that there is a splicing single nucleotide polymorphism (SNP) in CD33, rs12459419 (C>T, Ala14Val), which results in skipping of exon 2 and thus loss of the most immunogenic domain of CD33 – IgV. Given that the IgV domain is recognized by GO, this SNP holds great potential for predicting response to GO. Results from one of the largest studies to date (COGAAML05316) in children and young adults randomized to receive standard therapy with or without GO (GO arm, n=408; no-GO arm, n =408) indicated that there was a CD33 splicing SNP genotype-dependent clinical benefit from GO.7 This study showed that the rs12459419C>T change was significantly associated with CD33 cell surface levels (P<0.001). With respect to clinical outcomes, in patients with the CC genotype (~50% of patients) who expressed high levels of full-length CD33, addition of GO resulted in a significant reduction in relapse risk (by ~50%, P<0.001), an improved disease-free survival and a trend to a better event-free survival in the whole

2796

cohort (disease-free survival, P=0.004; event-free survival, P=0.055). In contrast, patients with the CT/TT genotype had no benefit from the addition of GO to the standard no-GO therapy.7 GO has been shown to improve outcomes in patients with favorable cytogenetics.8 Among low-risk patients in the COG-AAML0531 trial, a significant improvement in outcome with GO was observed in those with the CC genotype (relapse risk, P<0.001; disease-free survival, P=0.001; event-free survival, P=0.001; and overall survival, P=0.014),7 but not in patients with the CT/TT genotype. These results were consistent with the first report on the rs12459419 SNP which showed, albeit in a very small group of patients given GO after failing induction 1, an increase in minimal residual disease after GO treatment in patients with the TT genotype.9 These results raised hope for potentially personalizing GO treatment guided by germline SNP. However, initial attempts to validate these results in adult AML patients were apparently not successful, in two studies.10,11 Some of the factors that might explain the inconsistencies from these studies are summarized in Table 1 along with key points from all the studies discussed in this article. There are a few points worth mentioning from the first study in 536 adult AML patients enrolled on MRC 15 and MRC 17.10 First, in contrast to the COG-AAML0531 study, in which there was a single randomization, these studies included multiple randomizations with patients receiving varying numbers of courses (0, 1, or 2) and doses of GO (3 mg/m2 or 6 mg/m2) and different induction and consolidation therapies of varying intensity, with the outcome analysis based only on GO exposure at initial induction. This randomization complexity also led to a lack of

haematologica | 2021; 106(11)

Editorials

Table 1. Summary of major studies focused on the CD33 single nucleotide polymorphism and gemtuzumab response.

Study

Study cohort

Major findings

Other points/limitations

Lamba et al., 2017

Pediatric de novo AML COG-AAML0531 trial N= 916, (GO arm=408, No-GO arm =408) -Single randomization ADE vs. ADE+GO (3 mg/m2 in induction 1 and consolidation 2)

- Study was limited to only pediatric AML patients.

Gale et al., 2018

Adult AML patients: N=536, from different GO trials (MRC-AML15 and MRC-AML17) with varying GO and chemotherapy randomizations (induction and/or consolidation) and doses (3 or 6 mg/m2)

CD33 Splicing SNP rs12459419: - SNP genotype has strong association with CD33 expression. - CC genotype-benefits from GO addition: lower risk of relapse and higher DFS and EFS in GO vs. No-GO arm - CT/TT genotype-No benefit from GO addition: no difference in RR or DFS between GO vs. No-GO arms - Patients within low-risk group, CC genotype had significantly better outcome (RR, DFS, EFS and OS) with GO but no benefit of GO observed in CT/TT -No difference in outcome (RFS, OS) by rs12459419 genotypes in whole cohort -Within favorable risk group no impact of CD33 splicing SNP on RFS or OS. - CD33 expression not associated with GO response - CD33 SNP demonstrated association with its expression in a subset (n=249) of patients.

Short et al., 2020

Adult patients: N=104 (frontline =36; refractory/ relapsed =55) -20 mg/m2 decitabine + 3 mg/m2 GO on day 5 as induction; 5 consolidation cycles of decitabine + GO Adult AML patients: AMLSG-0909 trial: NPM1 mutation-positive, intermediate risk N= 545 (GO arm=273; No-GO arm = 272)

Teich et al., 2021

-Trend to higher CR in patients with high 6 CD33-SNPscore ≥0 which includes the splicing SNP rs12459419. -3’UTR CD33-SNP rs1803254 associated with worse CIR and RFS

-CD33 rs12459419 CC genotype improved RFS and CIR in GO vs. No-GO arm -CT/TT genotype no difference in RFS or CIR by arm

- Patients received different GO randomizations (induction and/or consolidation) and doses (3 mg/m2 or 6 mg/m2) - Patients from GO in consolidation were combined with the No-GO group; - RFS and OS not different in randomized cohort although favorable cytogenetics showed a trend for better outcome by GO -Study population very poor risk disease with heterogeneous patients (AML/MDS/ CMML/PMF; frontline/ relapse). -Decitabine+ GO showed low global response rate, limiting the number of patients for evaluation - Study only limited to NPM1 mutation positive AML within intermediate cytogenetic risk adult AML patients

DFS: disease free survival; RR: risk of relapse; RFS: relapse free survival; OS: overall survival; CR: complete remission; CIR: cumulative incidence of relapse.

demonstrated benefit of GO based on CD33 expression levels; Secondly, the fact that, in contrast to other studies, MRC studies did not show an association of CD33 expression and response to GO, lack of CD33 SNP-associated efficacy would not be surprising given that CD33 SNP are highly associated with CD33 expression. Thirdly, a trend to a difference in overall survival between patients included or not included in the study (P=0.06) implies a possible selection bias. Finally, relapse-free survival and overall survival were not significantly different in the randomized cohort although trends were observed for overall survival within patients with favorable cytogenetics (relapse-free survival, P=0.1; overall survival, P=0.05). The study by Short et al. included a very heterogeneous group of patients with very poor risk spanning from frontline to relapsed/refractory AML, myelodysplastic syndromes, chronic myelomonocytic leukemia, and primary myelofibrosis.11 Meta-analysis of results from MRC trials demonstrated a significant benefit of GO within favorable- and intermediate-risk groups but not in adverse cytogenetic risk groups,8 thus only including poor-risk AML might have had an impact on the results of Short et al. Additionally, the treatment regimen includ-

haematologica | 2021; 106(11)

ed a combination of decitabine for 5 days followed by the addition of GO (3 mg/m2) for induction and five consolidation cycles. However, it is worth mentioning that within this cohort, patients with a higher CD33 SNP score (≥0, which represents the CD33 rs12459419 CC genotype) tended to have a higher possibility of achieving complete remission.11 Thus, although the CD33 splicing SNP was observed to have an impact in AML in children and young adults, studies in adult AML were not able to demonstrate similar results, in part due to the complexities of the trial designs. In the study reported by Teich et al., in this issue of Haematologica,12 the impact of the CD33 splicing SNP in a homogeneous group of adult patients with NPM mutation-positive AML and primarily intermediate-risk cytogenetics randomized to receive standard therapy with or without the addition of GO (3 mg/m2) in two cycles of induction was investigated. Of note, this regimen was closer to that used in the COG-AAML0531 trial. Consistent with the results from COG AAML0531, in Teich’s study, the CC genotype of the CD33-rs12459419 SNP was associated with improved relapse-free survival and lower cumulative incidence of relapse in the GO arm than in the no-GO arm. Such a benefit of GO was not

2797

Editorials

observed in the rs12459419 CT/TT genotype group.12 This study with a single randomization and uniformly treated patients confirms the COG AAML0531 findings, providing additional evidence that CD33 genotype may constitute a valuable tool for predicting GO-responsive patients in cases of adult and pediatric AML. It also provides a caveat regarding the interpretation of studies with multiple GO randomizations that may obscure the clinical relevance of CD33 SNP in the context of GO treatment. The results published by Teich et al. show that the impact of CD33 SNP on GO response is not age-dependent and are of potential clinical utility in adult AML, especially given the abundance of the allele frequency (0.3) in patients with Caucasian ancestry. These exciting results warrant prospective and in-depth investigation and validation of the CD33 splicing SNP for its impact on response to GO in both pediatric and adult AML. Given that germline SNP genotyping is very quick and simple, and can be performed in a variety of specimens (blood sample, buccal swap, etc.), CD33 genotyping can be quickly translated to clinical testing. Disclosures No conflicts of interest to disclose. Contributions JKL and SM both contributed to writing this editorial. Funding JKL and SM are supported by the Leukemia Lymphoma Society (grant ID: 6610-20).

References 1. Norsworthy KJ, Ko CW, Lee JE, et al. FDA approval summary: Mylotarg for treatment of patients with relapsed or refractory CD33positive acute myeloid leukemia. Oncologist. 2018;23(9):1103-1108.

2798

2. Jen EY, Ko CW, Lee JE, et al. FDA approval: gemtuzumab ozogamicin for the treatment of adults with newly diagnosed CD33-positive acute myeloid leukemia. Clin Cancer Res. 2018;24(14):3242-3246. 3. Olombel G, Guerin E, Guy J, et al. The level of blast CD33 expression positively impacts the effect of gemtuzumab ozogamicin in patients with acute myeloid leukemia. Blood. 2016;127(17):21572160. 4. Pollard JA, Alonzo TA, Loken M, et al. Correlation of CD33 expression level with disease characteristics and response to gemtuzumab ozogamicin containing chemotherapy in childhood AML. Blood. 2012;119(16):3705-3711. 5. Pollard JA, Loken M, Gerbing RB, et al. CD33 expression and its association with gemtuzumab ozogamicin response: results from the randomized phase III Children's Oncology Group trial AAML0531. J Clin Oncol. 2016;34(7):747-755. 6. Gamis AS, Alonzo TA, Meshinchi S, et al. Gemtuzumab ozogamicin in children and adolescents with de novo acute myeloid leukemia improves event-free survival by reducing relapse risk: results from the randomized phase III Children's Oncology Group trial AAML0531. J Clin Oncol. 2014;32(27):3021-3032. 7. Lamba JK, Chauhan L, Shin M, et al. CD33 Splicing polymorphism determines gemtuzumab ozogamicin response in de novo acute myeloid leukemia: report from randomized phase III Children's Oncology Group trial AAML0531. J Clin Oncol. 2017;35(23):26742682. 8. Hills RK, Castaigne S, Appelbaum FR, et al. Addition of gemtuzumab ozogamicin to induction chemotherapy in adult patients with acute myeloid leukaemia: a meta-analysis of individual patient data from randomised controlled trials. Lancet Oncol. 2014;15(9):986996. 9. Lamba JK, Pounds S, Cao X, et al. Coding polymorphisms in CD33 and response to gemtuzumab ozogamicin in pediatric patients with AML: a pilot study. Leukemia. 2009;23(2):402-404. 10. Gale RE, Popa T, Wright M, et al. No evidence that CD33 splicing SNP impacts the response to GO in younger adults with AML treated on UK MRC/NCRI trials. Blood. 2018;131(4):468-471. 11. Short NJ, Richard-Carpentier G, Kanagal-Shamanna R, Impact of CD33 and ABCB1 single nucleotide polymorphisms in patients with acute myeloid leukemia and advanced myeloid malignancies treated with decitabine plus gemtuzumab ozogamicin. Am J Hematol. 2020;95(9):E225-E228. 12. Teich K, Krzykalla J, Kapp-Schwoerer S, et al. Cluster of differentiation 33 single nucleotide polymorphism rs12459419 is a predictive factor in patients with nucleophosmin1-mutated acute myeloid leukemia receiving gemtuzumab ozogamicin. Hematologica. 2021;106(11):2986-2989.

haematologica | 2021; 106(11)

REVIEW ARTICLE

2021 European Myeloma Network review and consensus statement on smoldering multiple myeloma: how to distinguish (and manage) Dr. Jekyll and Mr. Hyde Pellegrino Musto,1 Monika Engelhardt,2 Jo Caers,3,4 Niccolo’ Bolli,5,6 Martin Kaiser,7,8 Niels van de Donk,9 Evangelos Terpos,10 Annemiek Broijl,11 Carlos Fernández de Larrea,12 Francesca Gay,13 Hartmut Goldschmidt,14 Roman Hajek,15 Annette Juul Vangsted,16 Elena Zamagni,17 Sonja Zweegman,9 Michele Cavo,17 Meletios Dimopoulos,18 Hermann Einsele,19 Heinz Ludwig,20 Giovanni Barosi,21 Mario Boccadoro,13 Maria-Victoria Mateos,22 Pieter Sonneveld11 and Jesus San Miguel23 "Aldo Moro" University School of Medicine, Unit of Hematology and Stem Cell Transplantation, AOUC Policlinico, Bari, Italy; 2Department of Medicine I, Medical Center - University of Freiburg, Freiburg, Faculty of Medicine, University of Freiburg, Germany; 3 Department of Clinical Hematology, CHU of Liège, Liège, Belgium; 4Laboratory of Hematology, GIGA-I3, University of Liège, Liège, Belgium; 5Division of Hematology, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milano, Italy; 6Department of Oncology and Onco-Hematology, University of Milan, Milano, Italy; 7The Institute of Cancer Research, Division of Molecular Pathology, London, UK; 8The Royal Marsden Hospital, Department of Haematology, London, UK; 9Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Hematology, Cancer Center Amsterdam, Amsterdam, the Netherlands; 10Stem Cell Transplantation Unit, Plasma Cell Dyscrasias Unit, Department of Clinical Therapeutics, School of Medicine, National and Kapodistrian University of Athens, Alexandra General Hospital, Athens, Greece; 11Erasmus MC Cancer Institute & Erasmus University of Rotterdam, Rotterdam, the Netherlands; 12Amyloidosis and Myeloma Unit, Department of Hematology, Hospital Clínic, IDIBAPS, University of Barcelona, Barcelona, Spain; 13Myeloma Unit, Division of Hematology, University of Torino, Azienda Ospedaliero Universitaria Città della Salute e della Scienza, Torino, Italy; 14 University Hospital Heidelberg Internal Medicine V and National Center for Tumor Diseases (NCT), Heidelberg, Germany; 15Department of Hemato-oncology, University Hospital Ostrava and Faculty of Medicine, University of Ostrava, Ostrava , Czech Republic; 16Department of Hematology, Rigshospitalet, Copenhagen, Denmark; 17 Seràgnoli Institute of Hematology, Bologna University School of Medicine, Bologna, Italy; 18National and Kapodistrian University of Athens, School of Medicine, Department of Clinical Therapeutics, Athens, Greece; 19University Hospital Würzburg, Internal Medicine II, Würzburg, Germany; 20Wilhelminen Cancer Research Institute, 1st Department of Medicine, Center for Oncology, Hematology and Palliative Care, Wilhelminenspital, Vienna, Austria; 21Fondazione IRCCS Policlinico S. Matteo, Pavia, Italy; 22University Hospital of Salamanca/IBSAL/CIC-IBMCC, Salamanca, Spain and 23 Universidad de Navarra, CIMA, IDISNA, CIBERONC, Pamplona, Spain

Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2799-2812

ABSTRACT

ccording to the updated International Myeloma Working Group criteria, smoldering multiple myeloma (SMM) is an asymptomatic plasma cell disorder characterized by an M-component >3 g/dL, bone marrow plasma cell infiltration >10% and <60%, and absence of any myeloma-defining event. Active multiple myeloma is preceded by SMM, with a median time to progression of approximately 5 years. Cases of SMM range from the extremes of “monoclonal gammopathy of undetermined significance-like”, in which patients never progress during their lifetimes, to “early multiple myeloma”, in which transformation into symptomatic disease, based on genomic evolution, may be rapid and devastating. Such a “split personality” makes the prognosis and management of individual patients challenging, particularly with regard to the identification and possible early treatment of high-risk SMM. Outside of clinical trials, the conventional approach to SMM generally remains close observation until progression to active multiple myeloma. However, two prospective, randomized trials have recently demonstrated a significant clinical benefit in terms of time to progression, and of haematologica | 2021; 106(11)

Correspondence: PELLEGRINO MUSTO p.musto@tin.it Received: March 12, 2021. Accepted: June 15, 2021 Pre-published: July 15, 2021. https://doi.org/10.3324/haematol.2021.278519

©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.

2799

P. Musto et al.

overall survival in one of the two studies, for some patients with higher-risk SMM treated with lenalidomide ± dexamethasone, raising the question of whether such an approach should be considered a new standard of care. In this paper, experts from the European Myeloma Network describe current biological and clinical knowledge on SMM, focusing on novel insights into its molecular pathogenesis, new prognostic scoring systems proposed to identify SMM patients at higher risk of early transformation, and updated results of completed or ongoing clinical trials. Finally, some practical recommendations for the real-life management of these patients, based on Delphi consensus methodology, are provided.

Introduction “Asymptomatic”, smoldering multiple myeloma (SMM) is in the middle of the continuum of monoclonal gammopathies. SMM is more advanced and carries a higher disease burden than monoclonal gammopathy of undetermined significance (MGUS), but does not show the clinical features of end-organ damage, nor any of the other myeloma-defining events of active multiple myeloma (MM).1-5 Of all patients with MM, 8-14% have SMM: these patients have a median age of onset of 67 years, an annual incidence of 0.4 cases per 100,000 persons and a higher prevalence in Americans of African descent.6,7 The median time to progression to active MM is around 5 years, with a variable rate of approximately 10%/year during the first 5 years, 3%/year for the next 5 years, and 1%/year thereafter.7,8 Thus, SMM is a heterogeneous entity, ranging from ‘MGUS-like’ to ‘early MM’, in which malignant plasma cells can rapidly expand and lead to active MM. Yet, nearly one third of cases of SMM will never progress. Such a “split personality” of SMM remains intriguing and challenging.9 For decades, the conventional approach to SMM has been close observation, delaying treatment to the time of progression to active MM. Significant advances in the understanding of disease biology, improved risk stratification, and newer therapies with better efficacy and lower toxicity contributed to deeper responses and longer survival for patients with active MM. These advances have also challenged the management of SMM, raising the question of whether earlier treatment could: (i) avoid or delay the progression to MM; (ii) prevent the severe complications of end-organ damage; and (iii) potentially cure at least a proportion of patients with SMM. Thus, “to treat or not to treat” SMM, or even better, “are there patients with SMM who would benefit more from early treatment?” remains a controversial issue.4,10-15

Methodology Here, a European Myeloma Network (EMN) Expert Panel updates 2016 European perspectives on SMM,4 addressing current biological knowledge, new prognostic scoring systems and recent results of clinical trials, as well as providing practical recommendations. During two EMN Trialist meetings in 2019 and 2020, the areas of major concern in the management of SMM were selected by generating and rank-ordering key questions using the criterion of clinical relevance. Multistep procedures were utilized to achieve a consensus on recommendations. One panelist drafted the statements addressing the identified key questions. Subsequently, each panelist expressed his or her agreement with those statements 2800

and provided suggestions. The Delphi questionnaire method was used and a consensus of at least 80% was reached for all six final statements (median 93%; range, 83-100%).

New insights into the molecular pathogenesis of smoldering multiple myeloma All cases of MM evolve through MGUS and SMM stages, although these are often not clinically evident.16 The disease pathogenesis starts from intrinsic genomic defects in plasma cells, mainly translocations of oncogenes under the control of regulatory elements of the immunoglobulin heavy chain (IGH) locus or multiple trisomies of odd-numbered chromosomes.17 However, these are not sufficient to cause the progression from MGUS to SMM or MM, as in most cases the clone will not evolve in a patient’s lifetime. Furthermore, high-risk translocations such as t(4;14) and t(14;16) are less frequent in asymptomatic cases, while t(11;14) is more frequent in SMM.18 These different frequencies reflect different intrinsic propensities of the oncogenic translocation to drive symptomatic progression. Hyperdiploidy seems to slightly increase the risk of transformation to active MM as well.19 Interestingly, the number of chromosomal trisomies is lower in hyperdiploid MGUS than in hyperdiploid SMM,20,21 suggesting that ongoing acquisition of genomic lesions may be the cause of progression. Consistently, secondary copy-number abnormalities, such as del(1p), amp(1q), del(16q), and del(17p)18,22 are also more frequent in MM than in SMM. Next-generation sequencing studies evidenced a globally lower number of mutations in MGUS and SMM than in MM.23,24 This was particularly true for mutations in KRAS, NRAS, FAM46C, but also for genes in the NF-κB pathway and DNA repair pathway genes.18 Conversely, high-risk SMM showed a landscape of mutations and chromosomal abnormalities more similar to that of MM.25 Indeed, single-cell studies have highlighted, within a clonal plasma cell population, the presence of subclones with distinct phenotypes that could be linked to malignant progression and prove the step-wise evolution of MM.26 The analyses of paired SMM-MM genomes highlighted two patterns of progression: (i) evolution from minor or entirely new subclones, and (ii) no association with genomic changes.18,27-29 The former includes true asymptomatic cases that need to acquire new lesions to shift their clinical behavior towards an aggressive phenotype. The latter are aggressive cases just about to meet clinical criteria for progression, with generally a shorter time to evolution.25 Indeed, accumulating evidence suggests that changes in clonal substructure can be used to monitor SMM before end-organ damage develops.18,30 Genomic events associated with SMM progression include translohaematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

cations between the IGH locus and the MYC oncogene25,30,31 and accumulation of complex rearrangements.25,30,32 Last, the activity of mutational processes is different in MGUS/SMM and MM. Early mutations, acquired at the time of initiation, are caused by the DNA deaminase AID or from processes associated with cell aging. Late mutations, developing at the time of progression, arise from aberrant activity of the APOBEC family of cytidine deaminases, whose activity is absent in normal plasma cells.33-34 In clinical practice, recurrent translocations or hyperdiploidy can be assessed by fluorescence in situ hybridization or multiplex ligation-dependent probe amplification.35 Next-generation sequencing offers a more comprehensive evaluation of genomic lesions,36 but this approach must still be considered investigational and not a current standard. The progression of MM also depends on the tumor microenvironment.37,38 In particular, clonal plasma cells feed on proliferative and anti-apoptotic signals from stromal cells, including interleukin-6, insulin-like growth factor 1 and vascular endothelial growth factor. Osteoblasts and osteoclasts play opposing roles, with reduced osteoprotegerin secretion by osteoblasts and increased RANKL secretion by stromal cells promoting the activity of osteoclasts, which in turn secrete interleukin-6. In addition, progressive SMM is associated with increased neo-angiogenesis, in which MM-induced endothelial cells carry specific gene expression signatures associated with disease evolution. Furthermore, immune cells in the microenvironment are actively involved in MM progression. Innate and acquired immunity may prevent clonal plasma cell growth in early asymptomatic stages,39 as indirectly shown by the unrestrained growth of plasma cells from asymptomatic patients xenografted into immunocompromised mice.40 In addition, single-cell studies have shown how clonal plasma cells may shape a permissive immune environment already from the MGUS stage. In a continuous pattern, progression from SMM to MM is also associated with reduced MHC class II expression in CD14+ monocytes, an increase of regulatory T cells, loss of memory cytotoxic cells with skewing towards effector cells with suppressed or anergic phenotype, and upregulated interferon signaling promoting immunosuppression and MM growth.41 While these find-

ings are less likely to be translated soon into clinical applications, the possibility of harnessing the immune microenvironment to generate diagnostic tests to predict progression, or even therapeutic approaches to halt progression, is intriguing.

Current diagnosis and monitoring SMM is currently diagnosed according to International Myeloma Working Group (IMWG) criteria, based on amount of M-component, percentage of clonal bone marrow plasma cells (BMPC) infiltration and no evidence of end-organ damage (CRAB) or amyloidosis (Table 1).3 Updated IMWG criteria re-classified 10-15% of patients previously diagnosed as having SMM,3 on the basis of new myeloma-defining events (biomarkers of malignancy: >60% BMPC, free light chains [FLC] ratio >10042 and >1 focal lesion with magnetic resonance imaging [MRI]), the so-called “SLIM CRAB” criteria, as “ultra-high-risk” SMM. These patients are considered eligible for full MM treatments, having a risk of progression of about 80% at 2 years. The new IMWG criteria emphasize the role of imaging to risk-stratify SMM, with patients having two or more focal lesions >5 mm detected by MRI qualifying for active treatments.43,44 In particular, the IMWG recommends sensitive low-dose whole-body CT for the staging of monoclonal gammopathies.45 This imaging technique was recently validated in a prospective study of 100 patients with SMM at different timepoints to identify early bone lesions related to MM evolution.46 If low-dose wholebody CT is negative, whole-body or spine and pelvis MRI, where possible, should be pursued,43 although the latter MRI protocol can miss around 10% of non-axial lesions. Given the spatial distribution of focal bone lesions, it has been recently suggested that the cutoff for the number of focal lesions should be adapted according to the MRI protocol used.47 18F-fluorodeoxyglucose positron emission tomography (PET)-CT may be considered an appropriate alternative to low-dose whole-body CT. It is highly recommended to distinguish between SMM and active MM, if low-dose whole-body CT is negative and whole-body MRI is unavailable.48 In the case of uncertain or borderline lesions, radiological studies

Table 1. Current diagnostic criteria for monoclonal gammopathy of undetermined significance, smoldering multiple myeloma and multiple myeloma, according to International Myeloma Working Group criteria.

Disease

Criteria3

MGUS

Serum MC (non IgM type) <3 g/dL and clonal BMPC <10% Absence of myeloma-defining events, such as end-organ damage (CRAB)*, or other biomarkers of malignancy (SLiM)**, or amyloidosis, which can be attributed to the plasma cell proliferative disorder Serum MC (IgG or IgA) >3 g/dL or urinary MC >500 mg per 24 hours and clonal BMPC ≥10% and <60% Absence of myeloma-defining events, such as end-organ damage (CRAB)*, or other biomarkers of malignancy (SLiM)*, or amyloidosis, that can be attributed to the plasma cell proliferative disorder Clonal BMPC cells >10% or biopsy proven plasmacytoma and any one or more of myeloma-defining events, such as end-organ damage (CRAB)*, or other biomarkers of malignancy (SLIM)*, or amyloidosis, which can be attributed to the plasma cell proliferative disorder

SMM

MGUS: monoclonal gammopathy of undetermined significance; MC: M-component; BMPC: bone marrow plasma cells; SMM: smoldering multiple myeloma; MM: multiple myeloma. * CRAB: serum Calcium > 1 mg/dL above the upper limit of normal value or > 11 mg/dL; Renal insufficiency: serum creatinine > 2 mg/dL or creatinine clearance < 40 mL/min; Anemia: hemoglobin >2 g/dL below the lower limit of normal value or 10 g/dL; Bone lesions: one or more lytic lesions on skeletal radiography or computed tomography (including positron emission tomography). **SLiM: clonal bone marrow plasma cell percentage ≥ Sixty percent; serum involved/uninvolved free LIght chains ratio >100; Magnetic resonance imaging: more than one focal lesion ≥5 mm.

haematologica | 2021; 106(11)

2801

P. Musto et al.

should be repeated, preferably alternating different techniques, every 3-6 months or whenever clinically indicated, until a definitive diagnosis of symptomatic MM is reasonably made or excluded, (i.e., the suspicious lesion remains stable, is not accompanied by other signs of progression, or is disappearing). Clinical and laboratory monitoring of SMM should initially be performed every 2-3 months after diagnosis for 6-12 months.49 If test results are stable, patients may be followed every 4-6 months for another year and every 612 months thereafter. However, follow-up should be individualized based on risk of progression. Despite the absence of prospective data, according to the Expert Panel, imaging evaluation might preferably be repeated annually with MRI (because of the higher sensitivity for early damage) for the first 5 years (then stopped or the frequency reduced) or at clinical suspicion/pain or progressive increase of M-component (low-dose whole-body CT or MRI). A detailed imaging algorithm for patients with SMM is reported in the recent IMWG consensus on imaging.43 Importantly, combined, longitudinal evaluation and review of all relevant disease parameters may be required to interpret dynamics of the disease correctly.4

New prognostic scores During the last years, several risk scores for SMM progression, combining routinely used laboratory parameters, have emerged.4,8,9,49 These include the BMPC infiltration rate, an aberrant plasma cell phenotype (>95% clonal BMPC within the BMPC compartment), immunoparesis, the amount of serum M-component, an altered FLC ratio, and albumin levels.50-54 Further studies have highlighted the prognostic value of additional factors: (i) a progressive increase of the Mcomponent and Bence-Jones proteinuria over time;55,56 (ii) an evolving decrease in hemoglobin;57 (iii) bone marrow biopsy characteristics;58 (iv) presence of circulating plasma cells and their proliferative activity in the bone marrow;59,60 (v) bone involvement detected by MRI or PET-CT imaging;61,62 (vi) cytogenetic and molecular features of the clonal population;22,63-65 and (vii) serum levels of B-cell maturation antigen.66 In particular, a Southwest Oncology Group model, incorporating serum FLC values, serum Mcomponent and the University of Arkansas Medical Sciences 70-gene expression profile signature (GEP-70), predicted 2-year progression rates of 66.7%, 21.9% and 3.4% in patients with two or three, one, and no risk factors, respectively, ranking statistically first among other investigated clinical risk scores available at the time of analysis (Table 2).67 Since ultra-high-risk SMM patients are now considered to have MM,3 previously defined risk stratification models need to be re-assessed. Researchers at Mayo Clinic reexamined their initial cohort of 421 patients with SMM who met the 2014 IMWG criteria to re-classify risk factors for progression.68 The median time to progression to symptomatic MM was 57 months. Based on multivariate analysis data, a new Mayo risk model was proposed utilizing the same three parameters previously identified in 2008, but with different cut-offs: involved to uninvolved serum FLC ratio >20, serum M-component >2 g/dL, and BMPC infiltration >20% (20/2/20 SMM score, Table 2). 2802

On this basis, three risk groups were identified: low-risk: no risk factors; intermediate risk: one risk factor; highrisk: two or more risk factors. The estimated median times to progression were 109.8 months, 67.8 months and 29.2 months, respectively. Criticisms of this model, however, could be that the cut-offs were developed due to best subgroup separations, the entire cohort and subgroups were limited, and a validation analysis was not available at the time of publication. A panel of IMWG experts recently conducted a larger, multicenter, retrospective study of 1,996 SMM patients diagnosed according to the 2014 IMWG criteria, to develop a robust risk stratification model69 (Table 2). The follow-up from diagnosis was 3 years. The median time to progression of the entire cohort was 6.4 years, while the 2-, 5-, and 10-year risks of progression were 22%, 42% and 64%, respectively. Stepwise selection and multivariable analysis confirmed the value of 20/2/20 parameters and 1,363 patients with all three factors available were stratified in the same three categories, whose 2year progression rates were 6%, 18% and 44%, respectively. Additional analyses were conducted in 689 patients with a complete dataset by adding the presence of at least one of the following recurrent cytogenetic abnormalities: t(4;14), t(14;16), 1q gain, and del13q. Pooled data identified four risk categories, corresponding with 2-year risks of progression of 6%, 23%, 37%, and 63% (Table 2). To define a scoring tool providing a more individualized risk assessment, the three original risk factors (involved to uninvolved serum FLC ratio, serum Mcomponent, and BMPC infiltration), together with cytogenetic abnormalities, were included in a new logistic regression model based on the entire range of values instead of using single cut points (Table 2). Using this approach, the 2-year risks of progression were 3.8% in patients with a total score of 0-4, 26% in those with a score 5-8, 51% in those with a score 9-12, and 73% when the score was >12. The Czech Myeloma Group recently developed a simplified alternative model for SMM based exclusively on different serum parameters70 (Table 2). Data were collected from a training group of 287 patients and validated in an independent cohort of 240 patients. With a median follow-up of 2.4/2.5 years in the two groups, progression to MM occurred in 52% and 39% of patients, respectively. The median risks of progression per year were 11% and 10%, during the first 5 years after diagnosis, respectively. A serum FLC ratio of >30, immunoparesis (at least one uninvolved immunoglobulin below reference levels), and serum M-component ≥2.3 g/dL emerged as predictors of 2-year progression rate in a combined multivariate analysis. Based on these parameters, a new risk model was proposed with four groups of patients defined by the presence of none, one, two or three of these risk factors. Notably, the 2-year risks of progression in the two cohorts were 79% and 80% for those patients with three risk factors. In a recently proposed “genomic model”, next-generation sequencing was applied in a retrospective cohort of 214 patients with SMM, with whole exome sequencing performed on 166 tumors, and deep targeted sequencing on 48 tumors71 (Table 2). The model incorporated information on DNA repair/MAPK pathway gene alterations and MYC aberrations. All these genomic abnormalities independently predicted progression after accounting for haematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

Table 2. Most recent prognostic models for smoldering multiple myeloma.

Model

Risk factors

Risk Group

2-year PD rate (%)

SWOG 201467

MC >3 g/dL, sFLC >25 mg/dL, GEP-70 >0.26

Mayo 201868

sFLCr >20, MC >2 g/dL, BMPC >20%

IMWG 202069

MC >2 g/dL, sFLCr >20, BMPC >20%

Low (0 factors), n=60 Intermediate (1 factor), n=39 High (≥2 factors), n=18 Low (0 factors), n=143 Intermediate (1 factor), n=121 High (≥2 factors), n=153 Low (0 factors), n=522 Intermediate (1 factor), n=445 High (≥2 factors, n=396 Low (0 factors/score 0-4) *, n=241 Low-intermediate (1 factor/score 5-8), n=264 Intermediate (2 factors/score 9-12), n=233 High (3-4 factors/score >12), n=51 Low (0 factors), n=48/26 ** Intermediate (1 factor), n=44/34 High (2 factors), n=32/41 Very high (3 factors), n=15/12 0 factors, n=58 At least 1 factor, n=24

3.4 21.9 66.7 9.7 26.3 47.4 6 18 44 6/3.8 22/26.2 45.5/51.1 63.1/72.5 8.5 / 5.3 ** 20.9 / 7.5 41.9 / 44.8 78.7 / 81.3 86.4 14.4

+ high risk cytogenetics: [t(4;14), t(14;16), +1q, del(13q)]

CMG 202070

Immunoparesis (at least one uninvolved immunoglobulin below reference levels), sFLCr >30, MC ≥2.3 g/dL

Dana Farber 202071

DNA repair pathway gene alterations [mutations in TP53 and ATM, del(17p)], MAPK pathway gene mutations (KRAS, NRAS), MYC aberrations (translocations or copy-number variations).

Median TTP (months)

109.8 67.8 29.2

PD: progressive disease; TTP: time to progression; SWOG; Southwest Oncology Group;.MC: monoclonal component; sFLC: serum free light chains; sFLCr: serum free light chain ratio; BMPC: bone marrow plasma cells; GEP-70: University of Arkansas Medical Sciences 70-gene expression profile signature; CMG: Czech Myeloma Group. *sFCLr score: <10: 0; 10-25: 2; >25-40: 3; >40: 5. MC score: 0-1.5 g/dL: 0; >1.5-3 g/dL: 3; >3 g/dL: 4. BMPC score: 0-15%: 0; >15-20%: 2; >20-30%: 3; >30-40%: 5; >40%: 6. ** Validation cohort.

clinical risk staging. Patients without such alterations had a median time to progression of 7.2 years versus 1.2 in those with one or more alterations. In addition, the risk was cumulative, and patients with two or more alterations progressed the fastest. These results were validated in an external cohort of 72 patients with SMM previously sequenced. Importantly, this model outperformed the Mayo Clinic 2008 and 2018 prediction models.

Early treatment Possible treatments of SMM vary considerably, with aims ranging from disease control, to delaying progression, and ideally cure (Table 3). Eight randomized, controlled trials covering 885 patients were evaluated in a recent meta-analysis that compared early versus deferred treatment in SMM.72 These studies included patients treated with melphalan + prednisone,73-75 bisphosphonates ± thalidomide,76-78 siltuximab,79 and lenalidomide + dexamethasone.80,81 Overall, early treatment significantly decreased progression of SMM (risk ratio [RR] = 0.53, 95% confidence interval [95% CI]: 0.33-0.87, P=0.01), particularly in patients receiving melphalan + prednisone (RR=0.22, 95% CI: 0.08-0.64, P=0.005) or immunomodulatory drugs (RR=0.43, 95% CI: 0.31-0.59, P<0.00001), and in the high-risk SMM subgroup (RR=0.51, 95% CI: 0.37-0.70, P=0.0001). In the latter patients, treatment of SMM also significantly decreased mortality (RR=0.53, 95% CI: 0.29-0.96, P=0.04). Regarding the most relevant adverse events, the risk of secondary primary malignancies was significantly increased with early treatment (RR=4.13, 95% CI: 1.07-15.97, P=0.04). Major criticisms to this meta-analysis were the wide heterogeneity of therapeutic approaches used and the lack of risk stratification in all but one trial. haematologica | 2021; 106(11)

Treatment options with novel agents Lenalidomide The results of two phase III, prospective studies have been fully published, providing support to the use of lenalidomide (± dexamethasone) in patients with highrisk SMM. The Spanish QuiRedex phase III trial (NCT00480363) was a pivotal study conducted in 119 patients with highrisk SMM, defined by the presence of both BMPC >10% and M-component >3 g/dL or, if only one criterion was present, patients had to have <95% aberrant BMPC by immunophenotyping plus immunoparesis. Patients were randomized to receive nine 4-week induction cycles with lenalidomide + dexamethasone followed by maintenance with lenalidomide alone for 2 years or to undergo observation.80,81 The primary endpoint was time to progression. Updated results after a median follow-up of 10.8 years revealed a 46% reduction in the risk of death (hazard ratio [HR]=0.54; 95% CI: 0.3-0.9; P=0.034) and 73% in that of progression (HR=0.27; 95% CI: 0.16-0.42; P<0.0001) for early treatment versus observation.82 The median overall survival had not been reached in the treatment arm, while it was 7.8 years in the control arm. The updated median time to progression was 9.0 versus 2.1 years in patients receiving treatment and in the control group, respectively. No differences were observed between arms when overall survival was compared from the start of subsequent therapy in patients who progressed to active MM, suggesting that, once the patient has progressed, lenalidomide treatment would not induce the appearance of resistant clones. The frequency of secondary primary malignancies was higher in the treatment group than in the observation group (10% vs. 2%), but the cumulative risk did not differ significantly (P=0.07). Extensive phenotypic studies suggested that 2803

P. Musto et al. Table 3. Update of clinical trials in smoldering multiple myeloma with currently available results. Other ongoing studies without published results are reported in the text of the paper.

QuiRedex NCT00480363 [80-82]

Phase, n, of patients

Treatments

Risk stratification

Primary endpoint, FU

ORR, CR, MRD

PFS, TTP, OS

III, n=119

Induction: 28-day cycle [C1-9] Lenalidomide 25 mg p.o. days 1–21 + Dex 20 mg p.o. days 1–4, 12–15; Maintenance: 28-day cycle [C1-24] Lenalidomide 10 mg p.o days 1–21 (n=57) Observation (n=62) Continuous therapy: 28 day-cycle [C1 – PD]: Lenalidomide 25 mg p.o. days 1–21 (n=90) Observation (n=92) Induction: 28-day cycle [C1-8]: Carfilzomib 20/36 mg/m2 i.v. days 1,2, 8, 9, 15, 16 + Lenalidomide 25 mg p.o. days 1–21 + Dex 20 mg (C1 – 4) and 10 (C5– 8) p.o. or i.v. days 1, 2, 8, 9, 15, 16 Maintenance: 28-day cycle [C1 – 24] Lenalidomide 25 mg days 1–21 Induction: 28-day-cycle [C1-6]: Carfilzomib 20/36 mg/m2 i.v. days 1,2, 8, 9, 15, 16 + Lenalidomide 25 mg p.o. days 1–21 + Dex 40 mg days 1, 8, 15, 22 AuSCT: Melphalan 200 mg/m2 Consolidation: KRd as induction for 2 cycles Maintenance: 28 day-cycle Lenalidomide 10 mg days 1–21 + Dex 20 mg days 1, 8,15, 22 for 2 years Induction: 28-day cycle [C1-12] Ixazomib 4 mg p.o. days 1, 8, 15 + Dex 40 mg p.o. days 1, 8, 15, 22 Maintenance: 28-day cycle [C1-24] Ixazomib 4 mg p.o. days 1, 8, 15 Induction: 28-day-cycle [C1-9]: Ixazomib 4 mg p.o. days 1, 8, 15 + Lenalidomide 25 mg p.o. days 1–21 + Dex 40 mg p.o. days 1, 8, 15, 22 Maintenance: 28-day cycle [C10-24]: Ixazomib 4 mg p.o. days 1, 8, 15 + Lenalidomide 15 mg p.o. days 1–21 (n= 45) Daratumumab 16 mg/kg i.v. in 8-week cycles: Extended intense (n=41): [C1] every 1 week; [C2-3]

High-risk by BMPC >10% and MC >3 g/dL or, if only one criterion present, BMPC with aberrant phenotype >95% plus immunoparesis

TTP

After induction: ORR 79%, CR: 14%

Median TTP: 9.0 vs. 2.1 years (P=0.034)

SWOG E3A06 NCT01169337 [84]

III, n=182

NCT01572480 [23,85]

II, n=18

GEM-CESAR NCT02415413 [87,88]

III, n=90

NCT02697383 [89]

Pilot study, n=14

NCT02916771 [90]

II, n=26 (56 planned)

CENTAURUS NCT02316106 [92]

II, n=123

Intermediate or high-risk by BMPC >10%, or sheets and FLC ratio <0.26 or >1.65

High-risk by PETHEMA and MAYO 2008 criteria

High-risk by PETHEMA criteria (ultra-high risk patients were included, ref. 3)

Median FU: 10.8 years

PFS

After maintenance: ORR 90%, CR 26%

ORR/CR: 0% ORR 50%, CR 0%

Median FU: 35 months

ORR, MRD Median FU: 43.3 months

MRD Median FU: 32 months

ORR 0% ORR 100%, MRD negativity: 92% by MFC, 75% by NGS,

Median OS: NR vs. 7.8 years (P<0.0001)

3-year PFS: 91% vs. 66% (P=0.002)

Estimated 4-year PFS: 71% Estimated 4-year OS: 100%

ORR: 98% PFS 93% post-induction, (5 biochemical 98% post-AuSCT, 100% progressions) post-consolidation and maintenance; ≥CR: 38.4%, 61.5% and 68.6% at the same time-points MRD negativity 23%, 44% and 55% at the same time points

High-risk by PETHEMA or Mayo 2008 criteria

High risk, (ref. 8)

ORR

ORR 64% (no CR)

No progression to MM

ORR 89%, CR 19.2%

No progression to MM

CR and PD/DR ratio

ORR 56%, CR 4.9%,

Median FU 26 months

PD/DR 0.059

2-year PFS: 89.9% extended intense, vs.82.0% intermediate

Median FU 17 months

PFS Median number of cycles: 8

Intermediate or high-risk by BMPC ≥10% and <60% and at least one of the following: MC

Continued on the following page

2804

haematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

Continued from the previous page

NCT02960555 [96]

II, n=24 (planned n. 61)

NT01441973 [97]

II, n=31

NCT02279394 [98]

II, n=50

NCT02603887 [99]

Pilot study, n=13

NCT01484275 [79]

III, n=85

every other week; [C4-7] ≥3 g/dL (IgA ≥2 g/dl), every 4 weeks; urine MC >500 mg/24 h, [C8-20] every 8 weeks FLC ratio <0.126 or >8 Intermediate intense (n=41): [C1] every 1 week and [C2-20] every 8 weeks Short dosing (n=41): [C1] every 1 week Isatuximab 20 mg/kg i.v. High-risk (criteria NA) in 4 week cycle [C1] every 1 week; [C2-6] every other week; [C7-30] every 4 weeks Elotuzumab 20 mg/kg i.v. High risk by MC ≥3 g/dL [C1] days 1, 8, then with BMPC ≥10%; or MC [C2-progressive disease] 1–3 g/dL (alternatively monthly every 4 weeks urine MC >200 mg/24 h), BMPC ≥10% and FLC ratio Elotuzumab 10 mg/kg i.v. <0.125 or >8.0 [C1] days 1, 8, 15, 22, then [C2- progressive disease] monthly every 2 weeks Induction: 28-day cycle High-risk, (ref. 8) [C1-2] Elotuzumab 10 mg/kg i.v. days 1, 8, 15, 22 + Lenalidomide 25 mg p.o. days 1–21 + Dex 40 mg p.o. days 1, 8, 15, 22 [C3-8]: Stem cell collection; Elotuzumab 10 mg/kg i.v. days 1, 15 + Lenalidomide as 25 mg p.o. days 1–21 + Dex 40 mg p.o. days 1, 8, 15 Maintenance: 28-day cycle [C9-24] Elotuzumab 10 mg/kg i.v. days 1 + Lenalidomide 25 mg p.o. days 1–21 Pembrolizumab 200 mg i.v. Intermediate to high risk every 3 weeks × 8 cycles; by either PETHEMA, Mayo with option to continue up 2008 or SWOG criteria to 24 cycles if continued benefit Siltuximab 15 mg/kg i.v. in High risk by BMPC >10% 2 h every 4 weeks (n=43) and either MC >3 g/dL, or [C1-progressive disease] FLC ratio <0.126 / >8 (n=43) and MC >1 / <3 g/dL (32% ultra-high risk; ref 3)

intense vs. 75.3% short dosing ORR 54%, CR 9.8%, PD/DR 0.107

ORR Median number of cycles: 11.5 Relationship CD 56dim NK cells and MC protein reduction (not found)

ORR,38%, CR 0%, PD/DR 0.150 ORR 64%, CR 5%, with MRD negativity

ORR 10% (cumulative)

2-year PFS 69% (cumulative)

ORR 84%, CR 6%

No progression to MM

ORR 8%, CR 8%, MRD negativity 8%

15% of patients progressed to MM

1-year PFS: 84.5% siltuximab vs. 74.4% observation (P<0.06)

FU at least 28 months

PFS FU NA

ORR Median number of cycles: 8 PFS Median FU 29.2 months

Observation (n=2)

NCT01718899 [100]

I-IIa, n=20

PVX-410 vaccine cohort (n=3+6) 0.4-0.8 mg s.c.; every 2 weeks × 6 doses PVX-410 combination cohort (n=10): PVX-410 vaccine 0.8 mg s.c. every 2 weeks × 6 doses + Lenalidomide 25 mg p.o. days 1-21 every 28 days × 3 cycles

Moderate to high risk by MC ≥3 g/dL, BMPC >10%, abnormal FLC ratio (0.26-1.65) Moderate (2 risk factors) or high risk (3 risk factors)

Safety and immune response

Immune response 95%; (10/11 PVX-410 monotherapy, 9/9 PVX-410 combination)

Median PFS: NR siltuximab vs.: 23.5 months observation OS not reached in both arms PVX-410-alone: 3 progressions to MM (median TTP 36 weeks) Combination cohort: 1 progression to MM, median TTP NR

FU: follow-up, ORR: overall response rate; CR: complete remission; MRD: minimal residual disease, PFS: progression-free survival; TTP: time to progression; OS: overall survival; C: cycle; p.o.: per os; i.v. intravenous; s.c.: subcutaneous; BMPC: bone marrow plasma cells; MC: M-component; KRd: carfilzomib, lenalidomide, dexamethasone; NR: not reached; FLC: free light chains; Dex: dexamethasone; MFC: multicolor flow cytometry; NGS: next generation sequencing; AuSCT: autologous stem cell transplantation; NA, not available; NK: natural killer; MM: multiple myeloma; PD/DR: progressive disease/death rate per patient-year ratio; Mayo 2008: ref. 51; PETHEMA: ref. 50; SWOG: ref. 66.

haematologica | 2021; 106(11)

2805

P. Musto et al.

the impaired immune system of high-risk SMM patients could be reactivated by the immunomodulatory activity of lenalidomide.83 The Eastern Cooperative Oncology Group E3A06 phase III trial (NCT01169337) assessed the efficacy of lenalidomide monotherapy compared with observation in patients with intermediate- or high-risk SMM.80 Lenalidomide was administered orally as a single agent until disease progression, toxicity, or withdrawal for other reasons. Response to therapy was observed in 50% of patients in the lenalidomide arm, with no responses in the observation arm. With a median follow-up of 35 months, progression-free survival (the primary endpoint of the study) was significantly longer with lenalidomide than with observation (HR=0.28; 95% CI: 0.12-0.62; P=0.002), indicating a 72% decrease in the risk of progression. The most pronounced and significant benefit was seen in patients with high-risk SMM, as defined by both 2008 and 2018 Mayo Clinic risk models. Six deaths occurred, two in the lenalidomide arm and four in the observation arm. Grade 3 or 4 non-hematologic adverse events were observed in 25 patients (28%) on lenalidomide and in none of the control group. The treatment discontinuation rate was 50%, (in 40% of cases due to adverse events). Secondary primary malignancies occurred in 5.2% and 3.5% of treated and untreated patients, respectively. At the time of publication, data were not mature for evaluation of overall survival. Several different, more intensive treatments are being or are close to being tested in SMM. Such approaches, briefly described below, generally comprise triplets including a proteasome inhibitor with an immunomodulatory drug + dexamethasone, monoclonal antibodies or even autologous stem cell transplantation. These new strategies, however, have so far generally been explored only in pilot studies with a limited number of patients or in phase II trials, with still preliminary results.

Carfilzomib, lenalidomide and dexamethasone In a US pilot study (NCT01572480) in high-risk SMM, the carfilzomib, lenalidomide + dexamethasone (KRd) regimen resulted in an overall response rate of 100%.23,85 Very high rates of minimal-residual disease (MRD) negativity, as determined by multiparametric flow cytometry (92%) or next-generation sequencing (75%), were observed. After a median potential follow-up of 43.3 months, 63% of patients remained MRD-negative, with estimated 4-year progression-free and overall survival rates of 71% and 100%, respectively. The safety profile was consistent with previous reports for this regimen. A subsequent phase II study in high-risk SMM (Mayo Clinic or PETHEMA models) assessed KRd (8 cycles) + lenalidomide maintenance (KRd-R).86 After a median potential follow-up of 27.3 months, the overall response rate was 100% and 78% of patients achieved a best response of stringent complete remission. The primary objective of MRD-negative complete remission was achieved by 70.2%, with a median duration of 5.5 years. At the 5-year landmark, only 10% of patients had developed MM. No patient died, while grade 3-4 treatmentrelated adverse events occurred in 33% of patients, the most frequent being cytopenias, thromboembolism, rash, and lung infection. Another ongoing multicenter, open-label phase II trial by the HOVON group is randomizing patients with high2806

risk SMM, defined according to both the Mayo Clinic52 and Spanish51 criteria, and with ultra-high-risk SMM, as defined by the IMWG,3 to treatment with a KRd combination versus lenalidomide + dexamethasone alone for nine cycles, followed by lenalidomide maintenance for 2 years (NCT03673826). Among studies examining potentially curative strategies, the GEM-CESAR trial is a phase II, single-arm trial focusing on SMM patients at high risk of progression to active MM (>50% at 2 years), according to Spanish criteria (NCT02415413), and patients with ultra-high-risk SMM.87,88 The induction therapy in this trial is six 4-week cycles of KRd, high-dose melphalan and autologous stem cell transplantation as intensification therapy, followed by two KRd consolidation cycles and maintenance with lenalidomide + dexamethasone for up to 2 years. The primary endpoint is a sustained MRD-negativity rate of at least 50%, as determined by next-generation flow cytometry after induction and transplantation. At a first analysis of 90 patients younger than 70 years,87 main grade 3-4 toxicities during induction and maintenance treatments included neutropenia (6%), thrombocytopenia (11%), infections (18%) and skin rash (9%). With a median follow-up of 32 months, 98% of patients remained alive and 93% progression-free, with only five patients having had biochemical relapses. Updated results88 indicate that the overall response rate was 98% after induction, 98% after autologous stem cell transplantation, and 100% after consolidation; 68.6% of patients reached complete remission or better after consolidation, with 55% of them achieving MRD negativity.

Ixazomib-based regimens A combination of ixazomib + dexamethasone has been evaluated in a small pilot study in high-risk SMM89 (NCT02697383). At the time of the first analysis, with a median follow-up of 17 months, ten patients were continuing treatment: nine of them had achieved at least very good partial remission and no patient had progressed to symptomatic MM. Grade 3 non-hematologic adverse events included one intestinal complication and two grade 3 lung infections. The oral triplet ixazomib, lenalidomide + dexamethasone for nine cycles followed by 15 cycles of ixazomib + lenalidomide was investigated in patients with high-risk SMM in a phase II trial.90 Based on preliminary results in the first 26 patients, grade 3 hypophosphatemia, leukopenia, and neutropenia occurred in about 10% of patients, while 8% developed grade 4 neutropenia and hyperglycemia. The overall response rate in patients who received at least three cycles of treatment was 89% (23/26), including five complete responses (19.2%). None of the patients had shown progression to overt MM at the time of the analysis.

Monoclonal antibodies As for other hematological malignancies, monoclonal antibodies are currently also being tested in SMM.91

Daratumumab (anti-CD38) The randomized phase II CENTAURUS study (NCT02316106) evaluated daratumumab as a single agent in three different treatment schedules (extended intense, haematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

extended intermediate, or short dosing) in 123 patients with intermediate/high-risk SMM.92 The main endpoints were a complete remission rate >15% and the progressive disease/death rate per patient-year ratio. After a median follow-up of 26 months, the complete remission rates were 4.9%, 9.8%, and 0% for patients treated with the intense, intermediate, and short dosing schedules, respectively. Favorable progressive disease/death rate ratios and a median progression-free survival longer than 2 years were observed in all arms. Two-year progressionfree survival rates for intense, intermediate, and short dosing were 89.9%, 82%, and 75.3%, respectively. Target-saturating trough concentrations were generally preserved by intense dosing. No new safety issues occurred. Based on these data, the long dosing schedule is being investigated in the ongoing, randomized phase III AQUILA study (NCT03301220) comparing subcutaneous daratumumab for up to 39 cycles versus active observation.93 The study will randomize about 360 patients with high-risk SMM according to one or more of the following features: serum M-component ≥3 g/dL, IgA SMM, immunoparesis, serum FLC ratio of ≥8 but <100, BMPC percentage in the range of 51% to 60%. Another phase III, randomized trial is comparing lenalidomide + dexamethasone versus daratumumab, lenalidomide + dexamethasone in high-risk SMM (NCT03937635, DETER-SMM) patients, with overall survival as the primary endpoint. A target of 288 patients diagnosed within 1 year and having an abnormal serum FLC ratio (≤0.125 or ≥ 8.0) and involved chain <100 mg/L), and/or serum M-component >3 g/dL and/or presence of t(4;14) or del 17p or 1q will be enrolled. The ongoing, phase II ASCENT trial (NCT03289299) is evaluating daratumumab, carfilzomib, lenalidomide and dexamethasone (DKRd) without autologous stem cell transplantation in high-risk SMM.94 Treatment consists of 12 DKRd cycles, followed by maintenance with lenalidomide and daratumumab for 12 cycles. Preliminary safety data on 46 patients who reported grade 3-4 adverse events documented cytopenia, thromboembolic events, infections, hypertension, diarrhea, and allergic reactions, which all occurred in less than 10% of subjects. The relative median dose intensity was ≥80% for all drugs, demonstrating the feasibility of such an approach. Single-agent daratumumab given intravenously for up to 20 cycles is also being evaluated in a phase II study in lower-risk SMM (NCT03236428).95 Feasibility was demonstrated in 28 patients. Response rates were assessed in 15 patients who completed six or more cycles: Partial response and at least very good partial response were achieved in 53% and 20% of these patients, respectively. Thus far, no deaths, progression or therapy discontinuations due to toxicity have occurred.

Isatuximab (anti-CD38) A phase II study (NCT02960555) is exploring the efficacy of isatuximab (20 mg/kg intravenously) in high-risk SMM; the drug is administered at decreasing intervals up to 30 weeks.96 The median number of cycles given in the first 24 evaluable patients was 11.5 (range, 6-30). Five patients interrupted treatment, two because of progression to active MM. No deaths occurred during the study. Five grade 3 treatment-related adverse events [dyspnea, infusion related reactions, headache, neutropenia, urinary haematologica | 2021; 106(11)

tract infections] were observed, all of which resolved to baseline. Best responses included partial remission (42%), very good partial remission (17%), and complete remission with MRD negativity by multiparameter flow cytometry at 10-5 (5%). This study also evaluated healthrelated quality of life scores, suggesting that isatuximab may reduce anxiety and worry of progression to MM, with improvement of the score after six cycles. A phase III, randomized, multicenter study comparing isatuximab, lenalidomide + dexamethasone versus lenalidomide + dexamethasone in higher-risk SMM (according to IMWG criteria) within 5 years is about to start (NCT04270409). An initial safety run-in phase will confirm the recommended dose of isatuximab. Primary endpoints will include safety and efficacy (progressionfree survival), while pharmacokinetic and immunological studies, percentages, duration and quality of responses (including MRD), type of progression, second progression-free survival, overall survival, economic and healthrelated quality of life evaluations will be secondary objectives.

Elotuzumab (anti-SLAMF7/CS1) Elotuzumab monotherapy had modest activity in SMM patients (overall response rate, 10%,) in a phase II trial (NCT01441973).97 However, duration of response and the 2-year progression-free survival of 69% in some patients with high-risk characteristics were of some interest. In that study, no relationship emerged between baseline CD56dim NK cells and response. Another phase II trial (NCT02279394), tested elotuzumab, lenalidomide + dexamethasone in high-risk SMM.98 This combination induced partial remissions or better, including three complete remissions and 18 very good partial remissions, in 41 of 50 patients (84%). Genomic studies revealed that mutations occurring in genes involved in DNA repair were associated with poor response. No progression to symptomatic MM was observed and the toxicity profile was manageable, although some thromboembolic events occurred.

Pembrolizumab (anti-PD1) The checkpoint inhibitor pembrolizumab was investigated in 13 patients with intermediate/high-risk SMM (NCT02603887).99 After a median of eight cycles, 11 patients (85%) had stable disease, one patient progressed and one patient with 17p deletion and a high-risk geneexpression signature reached MRD negativity at 10-4 by next-generation sequencing in bone marrow which persisted after 27 months. Three patients discontinued the treatment due to immune-related adverse events.

Siltuximab (anti-interleukin-6) A randomized, double-blind, placebo-controlled study evaluated interleukin-6 blocking with siltuximab in 85 patients with high-risk SMM (NCT01484275).79 After a median follow-up of 29.2 months, the 1-year progression-free survival rate was 84.5% with siltuximab and 74.4% with placebo. The median progression-free survival was not reached with siltuximab whereas it was 23.5 months with placebo (P=0.057). Adverse events in the experimental arm were mainly of grade 2-3, the most common and serious ones being infections and urinary complications. Three deaths occurred with siltuximab and four with placebo. 2807

P. Musto et al.

Vaccines PVX-410 is a vaccine composed of a combination of four peptides, specifically targeting the highly overexpressed plasma cell antigens XBP1, CD138 and CS1/SLAMF7. A phase I/IIa multicenter, dose-escalation study (NCT01718899) accrued 22 patients with SMM at moderate or high risk of progression who received PVX410, with or without lenalidomide.100 The most common adverse events were mild-to-moderate injection site reactions and constitutional symptoms. PVX-410 was immunogenic as monotherapy and, more consistently, in combination with lenalidomide, as demonstrated by an increase in the percentage of PVX-410 tetramer and interferon-γ-positive specific CD3+CD8+ T lymphocytes, as well as by a persistent rise of vaccine-specific effector memory cells. In the PVX-410-alone cohort, three of 12 patients progressed, with a median time to progression of 36 weeks. In the combination cohort, five of 12 patients showed a clinical response, with one patient progressing and a median time to progression not reached. PVX-410 is also under investigation in SMM in combination with the selective histone-deacetylase inhibitor citarinostat ± lenalidomide in a phase I trial (NCT02886065). PD-L1 peptide vaccine, a new molecule targeting this immunological checkpoint, is also currently under investigation in SMM (NCT03850522).

Other trials A phase II trial with ibrutinib (NCT02943473), a Bruton tyrosine kinase inhibitor, was recently closed due to poor accrual and an unfavorable risk/benefit ratio in patients with high-risk SMM.

treatments, thus substantially accepting the philosophy that high-risk SMM “should” be treated. However, while some experts suggest that the Soutwest Oncology Group and Spanish trials should represent the current standard of care for patients with high-risk SMM,8,11 others are not convinced.10,14 Studies underway will likely show a response to therapy, more or less pronounced depending on the drugs used, with consequent improvement of progression-free survival. However, whether these often intensive treatments with novel agents will also have an extensive impact on overall survival of SMM patients will not be formally addressed and could remain unclear. The key issue remains to identify patients with high-risk SMM who “must” receive a treatment, because they will certainly have a significantly longer survival.

Recommendations for clinical practice and future research Which diagnostic procedures are necessary for diagnosing smoldering multiple myeloma? An adequate diagnostic work-up for SMM should include a hemogram and biochemistry with renal function and calcium levels, morphological and (if available) phenotypic quantification of clonal plasma cells in bone marrow smears and bone trephine biopsy, with cytogenetics by fluorescence in situ hybridization or a validated equivalent molecular method on purified plasma cells, evaluation of serum and urinary M-component, total serum immunoglobulins, serum involved/uninvolved FLC ratio and their absolute values. Diagnostic imaging should comprise low-dose whole-body CT and whole-body MRI, if low-dose whole-body CT is negative. Axial MRI or PETCT are reasonable alternatives, according to availability and specific diagnostic needs, as previously detailed.

General considerations

How should current predictive models be applied in the setting of smoldering multiple myeloma?

Despite the many ongoing and planned studies, only one randomized trial has so far demonstrated a significant survival benefit for early active therapy of high-risk SMM.82 Another recent trial showed an improvement in progression-free survival, but follow-up is still too short for adequate evaluation of overall survival.84 Both these studies raised some concerns. Their sample size was limited, with less than 100 patients in each arm. The Spanish study was conducted between 2007 and 2013,80 when some new MM drugs were not available, while bone involvement was assessed by a low-sensitivity technique such as plain radiography. Thus the survival benefit in some patients could have been due to treating active MM, rather than SMM. Furthermore, patients in the control group could start therapy only after meeting CRAB criteria, which may be difficult in patients with progressive deterioration of renal function or anemia, because they would start therapy before the cut-off is reached. On the other hand, in the US trial, the high discontinuation rate and the fact that the group achieving the most significant benefit with lenalidomide in terms of progression-free survival included only 25 patients could be concerns.84 Notwithstanding, all ongoing studies seem to take a survival benefit for granted, as they do not include an untreated control group or compare two arms with active

Prediction of progression to overt MM of SMM should be based on routine and universally applicable tests, considering, however, that proposed models may show significant discordances in identifying “true” high-risk SMM.101 All the clinical prognostic scores reported above, particularly the updated 2/20/20 model (endorsed by the IMWG and including variables that reflect both disease burden and biological features of SMM that can be measured by most centers), represent important tools that physicians must routinely use (whatever they select) for risk stratification, to reasonably predict the outcome of different patients with SMM, because their management should be riskadapted. Combination with appropriate imaging investigation is also recommended. Although more sophisticated genomic facilities may be available at specialized academic research institutions (where fluorescence in situ hybridization is replaced by tests such as whole exome sequencing), they are not mandatory, as not yet routinely used and validated for clinical purposes.

2808

Should patients with smoldering multiple myeloma be informed according to their different risk of evolution? Due to the heterogeneous behavior of SMM, appropriate information about their possible future clinical outcome should be given to patients with either lower- or haematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

higher-risk SMM, as defined by current risk models. The possibility of enrollment in a clinical trial, if available, should also be proposed.

How should smoldering multiple myeloma be monitored? Once diagnosed, SMM patients should be monitored according to the above-reported recommendations, taking into particular account the individual risk of progression. Changes in serum M-component represent the most simple, useful and worldwide available biomarker of progression over time. Serum FLC monitoring has also been demonstrated to be useful in this setting.102 However, for all SMM patients, it is important to follow the evolution of multiple different parameters that, taken together, may give better and comprehensive insight into the dynamics of the disease. Imaging should be periodically repeated, as previously detailed, and also performed if there is a biochemical or clinical suspicion of disease progression.

Which patients with smoldering multiple myeloma might benefit from early treatment? Regarding patients with lower-risk SMM, diagnosed according to current criteria, only active observation is recommended. With regard to early treatment of high-risk SMM, there is no consensus yet. Two prospective randomized trials have shown significant benefits from treatment with lenalidomide ± dexamethasone in these patients, but they were not registration studies and they were not presented to regulatory agencies.80,84 However, it should be considered that patients presenting with the coexistence of multiple risk factors, particularly increasing M-component levels or BMPC count or a significant decrease in hemoglobin concentration, high FLC ratio and/or high-risk cytogenetics, will further increase their risk of progression. For those cases, physicians may consider starting early treatment, with the intention of either delaying progression or even achieving a cure. However, it will be the individual physician’s responsibility to seek active risk/benefit discussion with their patients, also considering that health-related quality of life is an essential outcome parameter.103 The decision regarding treatment will also depend on whether such an unlicensed treatment approach falls within the legal framework of the national healthcare system. The Expert Panel agreed that therapy in these selected, very high-risk SMM patients, should be similar to that offered to patients with active MM, and that their treatment should be administered in a controlled setting, such as a clinical trial.104

What should be done in the near future to further improve the management of smoldering multiple myeloma? It is quite difficult to compare the results of the treatments that are currently being assessed in SMM, because of the substantial differences in times at which the studies were conducted, definitions of high-risk SMM, intensities of therapeutic approaches, and criteria/methods to evaluate response and progression across these studies. Thus, before definitively changing the current paradigms for the management of SMM, comparable future trials will have to be performed, aiming to define the following, relevant primary objectives: (i) to identify new predictive biomarkers (clinical, molecular/genomics, haematologica | 2021; 106(11)

immunological, microenvironmental, imaging) for further refining risk prediction and selecting SMM patients who may do well with observation (“Dr. Jekyll”) and those who require more stringent monitoring in order to establish the most appropriate moment to start treatment (“Mr. Hyde”). In this setting, possible racial diversities should also be considered, as they may have an impact on SMM biology;105 (ii) to assess the necessary balance between reduced risk of progression (and of consequent MM complications) with early treatment versus possible short- and long-term adverse effects, specifically deteriorating health-related quality of life, secondary primary malignancies and induction of refractory disease, elucidating, in particular, whether early treatment may select resistant clones or, the opposite, if delaying therapy may favor disease that is more resistant to future therapies; and (iii) to determine what intensity and which type of treatment are preferable in selected patients with high-risk SMM, i.e., short-term, intensive approaches with “curative” intent versus prolonged immunological control of the disease, for example by targeting immunity with memory to provide long-term surveillance, according to a “preventive” strategy. Both these approaches should have the primary objective of improving overall survival, without negatively affecting health-related quality of life. Disclosures PM has served as a member of advisory boards and/or received honoraria from Celgene, Janssen, Takeda, BristolMyers Squibb, Amgen, Novartis, Gilead, Jazz, Sanofi, Abbvie, and Glaxo-Smith-Kline. NB has received honoraria from Celgene, Takeda, Janssen, and Amgen; and served on an advisory board for Janssen. MK has received consultancy fees from AbbVie, Amgen, BMS/Celgene, GSK, Janssen, Karyopharm, Seattle Genetics, and Takeda; honoraria from BMS/Celgene, Janssen, and Takeda; research funding from Janssen (to his institution), BMS/Celgene (to his institution); and travel support from BMS/Celgene, Janssen, and Takeda. NVDD has received research support from Janssen Pharmaceuticals, Amgen, Celgene, Novartis, and BMS; and served on advisory boards for Janssen Pharmaceuticals, Amgen, Celgene, BMS, Takeda, Roche, Novartis, Bayer, and Servier. ET has received consultancy fees and honoraria from Amgen, BMS, Janssen, Celgene, Takeda, Genesis Pharma, GSK and Sanofi; and research support from Amgen, Janssen, Celgene, Genesis Pharma, GSK and Sanofi. FG has received honoraria from Amgen, Celgene, Janssen, Takeda, BMS, AbbVie, and GSK; and has served on advisory boards for Amgen, Celgene, Janssen, Takeda, BMS, AbbVie, GSK, Roche, Adaptive Biotechnologies, and Oncopeptides. HG has received grants and/or provision of investigational medicinal products from Amgen, BMS, Celgene, Chugai, Dietmar-Hopp-Foundation, Janssen, John Hopkins University, Sanofi; research support (institutions) from Amgen, BMS, Celgene, Chugai, Janssen, Incyte, Molecular Partners, Merck Sharp and Dohme (MSD), Sanofi, Mundipharma, Takeda, and Novartis; served on advisory boards for Adaptive Biotechnology, Amgen, BMS, Celgene, Janssen, Sanofi, and Takeda; and honoraria (for speakers bureaus) from Academy2 GmbH & Co. KG, Agentur Hogg Robinson Germany, Amgen, ArtTempi, Beupdated Helbig Consulting and Research AG Schweiz, BMS, Celgene, Chop GmbH, Chugai, Congress Culture Concept Dr. S. Stocker München, Connectmedia Warschau/Polen, Dr. Hubmann Tumorzentrum München, FomF GmbH, GlaxoSmithKline (GSK), GWT Forschung und Innovation Dresden, Institut für Versorgungsforschung in der 2809

P. Musto et al.

Onkologie GbR, Janssen, Kompetenznetz Maligne Lymphome (KML) e.V., MedConcept GmbH, Medical Communication Gmbh, Münchner Leukämie Labor Prof. Haferlach, New Concept Oncology, Novartis, Omnia Med Deutschland, Onko Internetportal dkg-web GmbH, Sanofi, STIL Forschungs GmbH, and Veranstaltungskonzept Gesundheit Mechernich. RH has received research funding from Janssen, Amgen, Celgene, BMS, Novartis and Takeda; served on advisory boards for Janssen, Amgen, Celgene, AbbVie, BMS, Novartis, PharmaMar and Takeda; and has received honoraria from Janssen, Amgen, Celgene, BMS, Pharma Mar and Takeda. EZ has received honoraria from and served on advisory boards for Janssen, BMS, Takeda, Sanofi, Oncopeptide, GSK, and Amgen. SZ has received research funding from Takeda and Janssen; and served on advisory boards for Celgene, Takeda, Janssen, Sanofi and Oncopeptides. MC has received honoraria from Janssen, Celgene, Amgen, BMS, Takeda, AbbVie, Sanofi, and Adaptive Biotechnologies, and is a member of speakers’ bureaus for Janssen and Celgene. MD has received consultancy fees and honoraria from Janssen, Celgene, Takeda, Amgen and BMS. HL has

References 1. Kyle RA, Greipp PR. Smoldering multiple myeloma. N Engl J Med. 1980;302(24):13471349. 2. Kyle RA, Child JA, Anderson K, et al. Criteria for the classification of monoclonal gammopathies, multiple myeloma and related disorders: a report of the International Myeloma Working Group. Br J Haematol. 2003;121(5):749-757. 3. Rajkumar SV, Dimopoulos MA, Palumbo A, et al. International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. Lancet Oncol. 2014;15 (12):e538-e548. 4. Caers J, Fernández de Larrea C, Leleu X, et al. The changing landscape of smoldering multiple myeloma: a European perspective. Oncologist. 2016;21(3):333-342. 5. Ravindran A, Bartley AC, Holton SJ, et al. Prevalence, incidence and survival of smoldering multiple myeloma in the United States. Blood Cancer J 2016;6(10):e486. 6. Kyle RA, Remstein ED, Therneau TM, et al. Clinical course and prognosis of smoldering (asymptomatic) multiple myeloma. N Engl J Med. 2007;356(25):2582-2590. 7. Rajkumar SV, Landgren O, Mateos M-V. Smoldering multiple myeloma. Blood. 2015;125(20):3069-3075. 8. Lonial S, Dhodapkar MV, Rajkumar SV. Smoldering myeloma and the art of war. J Clin Oncol. 2020;38(21):2363-2365. 9. Ahn IE, Mailankody S, Korde N, Landgren O. Dilemmas in treating smoldering multiple myeloma. J Clin Oncol. 2015;33(1):115-123. 10. Fonseca R, Gonzalez-Velez M. Treatment of smoldering multiple myeloma: expectant observation should still be the standard. Am Soc Clin Oncol Educ Book. 2020;40:1-7. 11. Joseph NS, Dhodapkar MV, Lonial S. The role of early intervention in high-risk smoldering myeloma. Am Soc Clin Oncol Educ Book. 2020;40:1-9. 12. Biran N, Vesole DH, Donato ML, et al. Treatment of smoldering myeloma: problems with study design as well as biological and clinical implications. J Clin Oncol. 2020;38(12):1367-1368. 13. Ho M, Patel A, Goh CY, Moscvin M, Zhang L, Bianchi G. Changing paradigms in diag-

2810

received research funding from Amgen, Takeda; and served on speaker’s bureau/advisory boards for Amgen, Takeda, Sanofi, Celgene-BMS, Seattle Genetics, and Janssen. MB has received honoraria from Sanofi, Celgene, Amgen, Janssen, Novartis, BMS, and AbbVie; served on advisory boards for Janssen and GSK; and has received research funding from Sanofi, Celgene, Amgen, Janssen, Novartis, BMS, and Mundipharma. M-VM has served on advisory boards for or received honoraria from Janssen, BMS, Celgene, Takeda, Amgen, Sanofi, Oncopeptides, GSK, Adaptive, Pfizer, Regeneron, Roche and Sea-Gen. PS has received honoraria and research funding from Amgen, Celgene, Janssen, SkylineDx, and Takeda; JSM has received consultancy fees from Amgen, BMS, Celgene, Janssen, MSD, Novartis, GSK Takeda, Sanofi, and Roche. Contributions PM reviewed the literature and wrote the first draft. All authors revised, provided comments and consensus according to the Delphi methodology, and approved the final draft before submission.

nosis and treatment of monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM) Leukemia. 2020;34(12):3111-3125. 14. Goodman AM, Kim MS, Prasad V. Persistent challenges with treating multiple myeloma early. Blood. 2021;137(4):456-458. 15. Bolli N, Sgherza N, Curci P, et al. What Is new in the treatment of smoldering multiple myeloma? J Clin Med. 2021;10(3):421. 16. Landgren O, Kyle RA, Pfeiffer RM, et al. Monoclonal gammopathy of undetermined significance (MGUS) consistently precedes multiple myeloma: a prospective study. Blood. 2009;113(22):5412-5417. 17. Morgan GJ, Walker BA, Davies FE. The genetic architecture of multiple myeloma. Nat Rev Cancer. 2012;12(5):335-348. 18. Boyle EM, Deshpande S, Tytarenko R, et al. The molecular make up of smoldering myeloma highlights the evolutionary pathways leading to multiple myeloma. Nat Commun. 2021;12(1):293. 19. Neben K, Jauch A, Hielscher T, et al. Progression in smoldering myeloma is independently determined by the chromosomal abnormalities del(17p), t(4;14), gain 1q, hyperdiploidy, and tumor load. J Clin Oncol. 2013;31(34):4325-4332. 20. Zandecki M, Laï JL, Geneviève F, et al. Several cytogenetic subclones may be identified within plasma cells from patients with monoclonal gammopathy of undetermined significance, both at diagnosis and during the indolent course of this condition. Blood. 1997;90(9):3682-3690. 21. Drach J, Angerler J, Schuster J, et al. Interphase fluorescence in situ hybridization identifies chromosomal abnormalities in plasma cells from patients with monoclonal gammopathy of undetermined significance. Blood. 1995;86(10):3915-3921. 22. Oliva S, De Paoli L, Ruggeri M, et al. A longitudinal analysis of chromosomal abnormalities in disease progression from MGUS/SMM to newly diagnosed and relapsed multiple myeloma. Ann Hematol. 2021;100(2):437-443. 23. Mailankody S, Kazandjian D, Korde N, et al. Baseline mutational patterns and sustained MRD negativity in patients with high-risk smoldering myeloma. Blood Adv.

2017;1(22):1911-1918. 24. Mikulasova A, Wardell CP, Murison A, et al. The spectrum of somatic mutations in monoclonal gammopathy of undetermined significance indicates a less complex genomic landscape than that in multiple myeloma. Haematologica. 2017;102(9):1617-1625. 25. Bolli N, Maura F, Minvielle S, et al. Genomic patterns of progression in smoldering multiple myeloma. Nat Commun. 2018;9(1): 3363. 26. Ledergor G, Weiner A, Zada M, et al. Single cell dissection of plasma cell heterogeneity in symptomatic and asymptomatic myeloma. Nat Med. 2018;24(12):1867-1876. 27. Walker BA, Wardell CP, Melchor L, et al. Intraclonal heterogeneity is a critical early event in the development of myeloma and precedes the development of clinical symptoms. Leukemia. 2014;28(2):384-390. 28. Zhao S, Choi M, Heuck C, et al. Serial exome analysis of disease progression in premalignant gammopathies. Leukemia. 2014;28(7):1548-1552. 29. Dutta AK, Fink JL, Grady JP, et al. Subclonal evolution in disease progression from MGUS/SMM to multiple myeloma is characterised by clonal stability. Leukemia. 2019;33(2):457-468. 30. Da Vià MC, Ziccheddu B, Maeda A, Bagnoli F, Perrone G, Bolli N. A journey through myeloma evolution: from the normal plasma cell to disease complexity. Hemasphere. 2020;4(6):e502. 31. Misund K, Keane N, Stein CK, et al. MYC dysregulation in the progression of multiple myeloma. Leukemia. 2020;34(1):322-326. 32. Oben B, Froyen G, Maclachlan KH, et al. Whole-genome sequencing reveals progressive versus stable myeloma precursor conditions as two distinct entities Nat Commun 2021 Mar 25;12(1):1861. 33. Maura F, Degasperi A, Nadeu F, et al. A practical guide for mutational signature analysis in hematological malignancies. Nat Commun. 2019;10(1):2969. 34. Maura F, Petljak M, Lionetti M, et al. Biological and prognostic impact of APOBEC-induced mutations in the spectrum of plasma cell dyscrasias and multiple myeloma cell lines. Leukemia. 2018;32(4): 1044-1048.

haematologica | 2021; 106(11)

EMN consensus on smoldering multiple myeloma

35. Shah V, Johnson DC, Sherborne AL, et al. Subclonal TP53 copy number is associated with prognosis in multiple myeloma. Blood. 2018;132(23):2465-2469. 36. Manzoni M, Marchica V, Storti P, et al. Application of next-generation sequencing for the genomic characterization of patients with smoldering myeloma. Cancers (Basel). 2020;12(5):1332. 37. Manier S, Sacco A, Leleu X, Ghobrial IM, Roccaro AM. Bone marrow microenvironment in multiple myeloma progression. J Biomed Biotechnol. 2012;2012:157496. 38. Bianchi G, Munshi NC. Pathogenesis beyond the cancer clone(s) in multiple myeloma. Blood. 2015;125(20):3049-3058. 39. Nakamura K, Smyth MJ, Martinet L. Cancer immunoediting and immune dysregulation in multiple myeloma. Blood. 2020;136(24): 2731-2740. 40. Das R, Strowig T, Verma R, et al. Microenvironment-dependent growth of preneoplastic and malignant plasma cells in humanized mice. Nat Med. 2016;22(11): 1351-1357. 41. Zavidij O, Haradhvala NJ, Mouhieddine TH, et al. Single-cell RNA sequencing reveals compromised immune microenvironment in precursor stages of multiple myeloma. Nat Cancer. 2020;1(5):493-506. 42. Wu V, Moshier E, Leng S, et al. Risk stratification of smoldering multiple myeloma: predictive value of free light chains and group-based trajectory modeling. Blood Adv. 2018;2(12):1470-1479. 43. Hillengass J, Usmani S, Rajkumar SV, et al. International myeloma working group consensus recommendations on imaging in monoclonal plasma cell disorders. Lancet Oncol. 2019;20(6):e302-e312. 44. Jamet B, Bailly C, Carlier T, et al. Imaging of monoclonal gammapathy of undetermined significance and smoldering multiple myeloma. Cancers (Basel). 2020;12(2):486. 45. Hillengass J, Moulopoulos LA, Delorme S, et al. Whole-body computed tomography versus conventional skeletal survey in patients with multiple myeloma: a study of the International Myeloma Working Group. Blood Cancer J. 2017;7(8):e599. 46. Gavriatopoulou M, Βoultadaki A, Koutoulidis V, et al. The role of low dose whole body CT in the detection of progression of patients with smoldering multiple myeloma. Blood Cancer J. 2020;10(9):93. 47. Wennmann M, Hielscher T, Kintzelé L, et al. Spatial distribution of focal lesions in wholebody MRI and influence of MRI protocol on staging in patients with smoldering multiple myeloma according to the new SLiM-CRAB criteria. Cancers (Basel). 2020;12(9):2537. 48. Cavo M, Terpos E, Nanni C, et al. Role of 18F-FDG PET/CT in the diagnosis and management of multiple myeloma and other plasma cell disorders: a consensus statement by the International Myeloma Working Group. Lancet Oncol. 2017;18(4):e206-e217. 49. Kyle RA, Durie BGM, Rajkumar SV, et al. Monoclonal gammopathy of undetermined significance (MGUS) and smoldering (asymptomatic) multiple myeloma: IMWG consensus perspectives risk factors for progression and guidelines for monitoring and management. Leukemia. 2010;24(6):11211127. 50. Cocito F, Mangiacavalli S, Ferretti VV, et al. Smoldering multiple myeloma: the role of different scoring systems in identifying high-risk patients in real-life practice. Leuk Lymphoma. 2019;60(12):2968-2974. 51. Pérez-Persona E, Vidriales M-B, Mateo G, et al. New criteria to identify risk of progres-

haematologica | 2021; 106(11)

sion in monoclonal gammopathy of uncertain significance and smoldering multiple myeloma based on multiparameter flow cytometry analysis of bone marrow plasma cells. Blood. 2007;110(7):2586-2592. 52. Dispenzieri A, Kyle RA, Katzmann JA, et al. Immunoglobulin free light chain ratio is an independent risk factor for progression of smoldering (asymptomatic) multiple myeloma. Blood. 2008;111(2):785-789. 53. Waxman AJ, Mick R, Garfall AL, et al. Classifying ultra-high risk smoldering myeloma. Leukemia. 2015;29(3):751-753. 54. Sørrig R, Klausen TW, Salomo M, et al. Smoldering multiple myeloma risk factors for progression: a Danish population-based cohort study. Eur J Haematol. 2016;97 (3):303-309. 55. Fernández de Larrea C, Isola I, Pereira A, et al. Evolving M-protein pattern in patients with smoldering multiple myeloma: impact on early progression. Leukemia. 2018;32(6): 1427-1434. 56. González-Calle V, Dávila J, Escalante F, et al. Bence Jones proteinuria in smoldering multiple myeloma as a predictor marker of progression to symptomatic multiple myeloma. Leukemia. 2016;30(10):2026-2031. 57. Ravi P, Kumar S, Larsen JT, et al. Evolving changes in disease biomarkers and risk of early progression in smoldering multiple myeloma. Blood Cancer J. 2016;6(7):e454. 58. Bustoros M, Kastritis E, Sklavenitis-Pistofidis R, et al. Bone marrow biopsy in low-risk monoclonal gammopathy of undetermined significance reveals a novel smoldering multiple myeloma risk group. Am J Hematol. 2019;94(5):E146E149. 59. Sanoja-Flores L, Flores-Montero J, Garcés JJ, et al. Next generation flow for minimallyinvasive blood characterization of MGUS and multiple myeloma at diagnosis based on circulating tumor plasma cells (CTPC). Blood Cancer J. 2018;8(12):117. 60. Aljama MA, Sidiqi MH, Lakshman A, et al. Plasma cell proliferative index is an independent predictor of progression in smoldering multiple myeloma. Blood Adv. 2018;2 (22):3149-3154. 61. Merz M, Hielscher T, Wagner B, et al. Predictive value of longitudinal whole-body magnetic resonance imaging in patients with smoldering multiple myeloma. Leukemia. 2014;28(9):1902-1908. 62. Zamagni E, Nanni C, Gay F, et al. 18F-FDG PET/CT focal, but not osteolytic, lesions predict the progression of smoldering myeloma to active disease. Leukemia. 2016;30(2):417-422. 63. Rajkumar SV, Gupta V, Fonseca R, et al. Impact of primary molecular cytogenetic abnormalities and risk of progression in smoldering multiple myeloma. Leukemia. 2013;27(8):1738-1744. 64. Khan R, Dhodapkar M, Rosenthal A, et al. Four genes predict high risk of progression from smoldering to symptomatic multiple myeloma (SWOG S0120). Haematologica. 2015;100(9):1214-1221. 65. Merz M, Hielscher T, Schult D, et al. Cytogenetic subclone formation and evolution in progressive smoldering multiple myeloma. Leukemia. 2020;34(4):1192-1196. 66. Visram A, Soof C, Rajkumar SV, et al. Serum BCMA levels predict outcomes in MGUS and smoldering myeloma patients. Blood Cancer J. 2021;11(6):120. 67. Dhodapkar MV, Sexton R, Waheed S, et al. Clinical, genomic, and imaging predictors of myeloma progression from asymptomatic monoclonal gammopathies (swog s0120).

Blood. 2014;123(1):78-85. 68. Lakshman A, Rajkumar SV, Buadi FK, et al. Risk stratification of smoldering multiple myeloma incorporating revised IMWG diagnostic criteria. Blood Cancer J. 2018;8(6):59. 69. Mateos MV, Kumar S, Dimopoulos MA, et al. International Myeloma Working Group risk stratification model for smoldering multiple myeloma (SMM). Blood Cancer J. 2020;10(10):102. 70. Hájek R, Sandecka V, Špička I, et al. Identification of patients with smouldering multiple myeloma at ultra-high risk of progression using serum parameters: the Czech Myeloma Group model. Br J Haematol. 2020;190(2):189-197. 71. Bustoros M, Sklavenitis-Pistofidis R, Park J, et al. Genomic profiling of smoldering multiple myeloma identifies patients at a high risk of disease progression. J Clin Oncol. 2020;38(21):2380-2389. 72. Zhao AL, Shen KN, Wang JN, Huo LQ, Li J, Cao XX. Early or deferred treatment of smoldering multiple myeloma: a metaanalysis on randomized controlled studies. Cancer Manag Res. 2019;11:5599-5611. 73. Hjorth M, Hellquist L, Holmberg E, Magnusson B, Rödjer S, Westin J. Initial versus deferred melphalan-prednisone therapy for asymptomatic multiple myeloma stage I - a randomized study. Eur J Haematol. 2009;50(2):95-102. 74. Riccardi A, Ucci G, Luoni R, et al. Treatment of multiple myeloma according to the extension of the disease: a prospective, randomised study comparing a less with a more aggressive cytostatic policy. Br J Cancer. 1994;70(6):1203-1210. 75. Riccardi A, Mora O, Tinelli C, et al. Longterm survival of stage I multiple myeloma given chemotherapy just after diagnosis or at progression of the disease: a multicentre randomized study. Br J Cancer. 2000;82(7): 1254-1260. 76. Witzig TE, Laumann KM, Lacy MQ, et al. A phase III randomized trial of thalidomide plus zoledronic acid versus zoledronic acid alone in patients with asymptomatic multiple myeloma. Leukemia. 2013;27(1):220225. 77. Musto P, Petrucci MT, Bringhen S, et al. A multicenter, randomized clinical trial comparing zoledronic acid versus observation in patients with asymptomatic myeloma. Cancer. 2008;113(7):1588-1595. 78. D'Arena G, Gobbi PG, Broglia C, et al. Pamidronate versus observation in asymptomatic myeloma: final results with long-term follow-up of a randomized study. Leuk Lymphoma. 2011;52(5):771-775. 79. Brighton TA, Khot A, Harrison SJ, et al. Randomized, double-blind, placebo-controlled, multicenter study of siltuximab in high-risk smoldering multiple myeloma. Clin Cancer Res. 2019;25(13):3772-3775. 80. Mateos MV, Hernández MT, Giraldo P, et al. Lenalidomide plus dexamethasone for highrisk smoldering multiple myeloma. N Engl J Med. 2013;369(5):438-447. 81. Mateos MV, Hernández MT, Giraldo P, et al. Lenalidomide plus dexamethasone versus observation in patients with high-risk smouldering multiple myeloma (QuiRedex): long-term follow-up of a randomised, controlled, phase 3 trial. Lancet Oncol. 2016;17(8):1127-1136. 82. Mateos M-V, Hernandez MT, Salvador C, et al. Over ten years of follow-up for phase II trial in smoldering myeloma at high risk of progression to myeloma: sustained TTP and OS benefit with RD versus no treatment. Hemasphere. 2020;294867:EP950.

2811

P. Musto et al. 83. Paiva B, Mateos MV, Sanchez-Abarca LI, et al. Immune status of high-risk smoldering multiple myeloma patients and its therapeutic modulation under LenDex: a longitudinal analysis. Blood. 2016;127(9):1151-1162. 84. Lonial S, Jacobus S, Fonseca R, et al. Randomized trial of lenalidomide versus observation in smoldering multiple myeloma. J Clin Oncol. 2020;38(11):1126-1137. 85. Korde N, Roschewski M, Zingone A, et al. Treatment with carfilzomib-lenalidomidedexamethasone with lenalidomide extension in patients with smoldering or newly diagnosed multiple myeloma. JAMA Oncol. 2015;1(6):746-754. 86. Kazandjian D, Hil E, Morrison C, et al. Background:. Treatment of high risk (HR) smoldering multiple myeloma (SMM) with carfilzomib, lenalidomide, and dexamethasone (KRd) followed by lenalidomide maintenance (-R): a phase 2 clinical and correlative study. Blood. 2020;136(1):43-45. 87. Mateos M-V, Martinez-Lopez J, Rodriguez Otero P, et al. Curative strategy (GEMCESAR) for high-risk smoldering myeloma (SMM): carfilzomib, lenalidomide and dexamethasone (KRd) as induction followed by HDT-ASCT, consolidation with KRd and maintenance with Rd. Blood. 2019;134(1):781. 88. Puig N, Contreras T, Paiva B, et al. Analysis of treatment efficacy in the GEM-CESAR trial for high-risk smoldering multiple myeloma patients: comparison between the standard and IMWG MRD criteria and QIPMS including FLC (QIP-FLC-MS). J Clin Oncol. 2020;38(15):8512. 89. Mailankody S, Salcedo M, Tavitian E, et al. Ixazomib and dexamethasone in high risk smoldering multiple myeloma: a clinical and correlative pilot study. J Clin Oncol. 2019;37(15):8051.

2812

90. Bustoros M, Liu C, Reyes K, et al. Phase II trial of the combination of ixazomib, lenalidomide, and dexamethasone in highrisk smoldering multiple myeloma. Blood. 2018;132(1):804. 91. Musto P, La Rocca F. Monoclonal antibodies in newly diagnosed and smoldering multiple myeloma: an updated review of current clinical evidence. Expert Rev Hematol. 2020;13(5):501-517. 92. Landgren CO, Chari A, Cohen YC, et al. Daratumumab monotherapy for patients with intermediate-risk or high-risk smoldering multiple myeloma: a randomized, openlabel, multicenter, phase 2 study (CENTAURUS). Leukemia. 2020;34(7):1840-1852. 93. Rajkumar SV, Voorhees PM, Goldschmidt H, et al. Randomized, open-label, phase 3 study of subcutaneous daratumumab (DARA SC) versus active monitoring in patients (Pts) with high-risk smoldering multiple myeloma (SMM): AQUILA. J Clin Oncol. 2018;36(15):TPS8062. 94. Kumar SK, Abdallah AO, Badros AZ, et al. Aggressive smoldering curative approach evaluating novel therapies (ASCENT): a phase 2 trial of induction, consolidation and maintenance in subjects with high risk smoldering multiple myeloma (SMM): Initial analysis of safety data. Blood. 2020;136(1):3536. 95. Nadeem O, Redd R, Stampleman LV, et al. A phase II study of daratumumab in patients with high-risk MGUS and low-risk smoldering multiple myeloma: first report of efficacy and safety. Blood. 2019;134(1):1898. 96. Manasanch EE, Jagannath S, Lee HC, et al. A multicenter phase II single arm trial of isatuximab in patients with high risk smoldering multiple myeloma (HRSMM). Blood. 2019;134(1):3116. 97. Jagannath S, Laubach J, Wong E, et al.

Elotuzumab monotherapy in patients with smouldering multiple myeloma: a phase 2 study. Br J Haematol. 2018;182(4):495-503. 98. Liu C, Ghobrial IM, Bustoros M, et al. Phase II trial of combination of elotuzumab, lenalidomide, and dexamethasone in highrisk smoldering multiple myeloma. Blood. 2018;132(1):154. 99. Manasanch EE, Han G, Mathur R, et al. A pilot study of pembrolizumab in smoldering myeloma: report of the clinical, immune, and genomic analysis. Blood Adv. 2019;3(15):2400-2408. 100. Nooka AK, Wang ML, Yee AJ, et al. Assessment of safety and immunogenicity of PVX-410 vaccine with or without lenalidomide in patients with smoldering multiple myeloma: a nonrandomized clinical trial. JAMA Oncol. 2018;4(12):e183267. 101. Hill E, Dew A, Morrison C, et al. Assessment of discordance among smoldering multiple myeloma risk models. JAMA Oncol. 2021;7(1):132-134. 102. Gran C, Luong V, Bruchfeld JB, et al. Dynamic follow-up of smoldering multiple myeloma identifies a subset of patients at high risk of progression. Am J Hematol. 2021;96(3):E63-E65. 103. Jean-Baptiste M, Gries KS, Lenderking WR, Fastenau J. Symptom burden and healthrelated quality of life impacts of smoldering multiple myeloma: the patient perspective. J Patient Rep Outcomes. 2020;4(1):95. 104.Dimopoulos MA, Moreau P, Terpos E, et al. Multiple myeloma: EHA-ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2021;32(3):309322. 105. Marinac CR, Ghobrial IM, Birmann BM, Soiffer J, Rebbeck TR. Dissecting racial disparities in multiple myeloma. Blood Cancer J. 2020;10(2):19.

haematologica | 2021; 106(11)

REVIEW ARTICLE

The possible role of mutated endothelial cells in myeloproliferative neoplasms

Ferrata Storti Foundation

Mirko Farina,1 Domenico Russo1 and Ronald Hoffman2 Unit of Blood Diseases and Bone Marrow Transplantation, Cell Therapies and Hematology Research Program, Department of Clinical and Experimental Sciences, University of Brescia, ASST Spedali Civili di Brescia, Brescia, Italy and 2Division of Hematology and Medical Oncology, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, NY, USA

ABSTRACT

Haematologica 2021 Volume 106(11):2813-2823

yeloproliferative neoplasms (MPN) are chronic, clonal hematologic malignancies characterized by myeloproliferation and a high incidence of vascular complications (thrombotic and bleeding). Although MPN-specific driver mutations have been identified, the underlying events that culminate in these clinical manifestations require further clarification. We reviewed the numerous studies performed during the last decade identifying endothelial cell (EC) dysregulation as a factor contributing to MPN disease development. The JAK2V617F MPN mutation and other myeloid-associated mutations have been detected not only in hematopoietic cells but also in EC and their precursors in MPN patients, suggesting a link between mutated EC and the high incidence of vascular events. To date, however, the role of EC in MPN continues to be questioned by some investigators. In order to further clarify the role of EC in MPN, we first describe the experimental strategies used to study EC biology and then analyze the available evidence generated using these assays which implicate mutated EC in MPN-associated abnormalities. Mutated EC have been reported to possess a pro-adhesive phenotype as a result of increased endothelial Pselectin exposure, secondary to degranulation of Weibel-Palade bodies, which is further accentuated by exposure to pro-inflammatory cytokines. Additional evidence indicates that MPN myeloproliferation requires JAK2V617F expression by both hematopoietic stem cells and EC. Furthermore, the reports of JAK2V617F and other myeloid malignancy-associated mutations in both hematopoietic cells and EC in MPN patients support the hypothesis that MPN driver mutations may first appear in a common precursor cell for both EC and hematopoietic cells.

Introduction The Philadelphia chromosome-negative myeloproliferative neoplasms (MPN) include polycythemia vera, essential thrombocythemia and primary myelofibrosis.1 These clonal hematopoietic stem cell (HSC) disorders are characterized by an increased rate of vascular complications including thrombotic and bleeding episodes.2,3 However, the mechanisms underlying these vascular events remains uncertain and have been the subject of considerable speculation and debate for decades.4,5 Recently, new insights into factors contributing to the development of thrombotic events in MPN patients have become available,6 including the role of endothelial cells (EC) that contain MPN driver mutations. Physiologically, EC participate in the maintenance of vascular integrity, and generate an anti-thrombotic surface.7 During the last decade, the JAK2V617F MPN driver mutation has been shown to be present in EC8,9 and their progenitors10–12 in some MPN patients, suggesting a link between mutated EC and the high incidence of vascular events. This concept and its implications remain controversial and its significance has been questioned by some investigators.9,11,13 The aim of this review is to analyze this evidence in a critical fashion and assess the validity of the link between EC and MPN pathobiology.

haematologica | 2021; 106(11)

Correspondence: MIRKO FARINA mirkfar@gmail.com Received: February 17, 2021. Accepted: June 28, 2021. Pre-published: July 29, 2021. https://doi.org/10.3324/haematol.2021.278499

2813

M. Farina et al.

Myeloproliferative neoplasms and vascular complications Vascular complications are the most common clinical sequelae and a major cause of morbidity and mortality in MPN patients.2,3 The incidence and the characteristic clinical presentations of vascular events in MPN patients are summarized in Table 1. Thrombotic events are often the initial manifestation of an MPN or may precede the diagnosis of the MPN. Thrombosis appears to be more common among patients with polycythemia vera than in those with essential thrombocythemia or primary myelofibrosis both at diagnosis3 and during follow up2 (Table 1). Bleeding episodes are less frequent than thrombotic events in MPN patients; and, contrary to thrombosis, occur primarily after the diagnosis of the MPN has been established14 (Table 1).

Factors predisposing to thrombosis in patients with myeloproliferative neoplasms Many features of a patient’s demographics are predictive of MPN-associated thrombotic complications15-17 including age, prior thrombotic events, an inflammatory state, and MPN-associated risk factors, such as degree of erythrocytosis, leukocytosis, and the presence of JAK2V617F. By contrast, individuals with calreticulin mutations have a lower risk of thrombosis than those with JAK2V617F.2 Notably, the frequency of the JAK2V617F variant allele influences the degree of thrombotic risk18 in patients with

essential thrombocythemia, while contradictory results were found in patients with polycythemia vera. Conventional cardiovascular risk factor (e.g., hypertension, hyperlipidemia, diabetes and smoking) are additional variables associated with an increased rate of thrombosis. Among factors predisposing to thrombosis, only age greater than 60 years and a prior history of a thrombotic event were validated as thrombotic risk factors in MPN patients, while conflicting results have been reported for other proposed predisposing factors.2,15,16 However, the presence of JAK2V617F as an MPN driver mutation has been confirmed as a predictor of additional thrombotic events in patients with essential thrombocythemia.15,16 The history of thrombotic events prior to a diagnosis of MPN may also be attributed to the presence in these patients of a clonal hematopoiesis of indeterminate potential (CHIP), involving JAK2V617F or calreticulin mutations prior to the development of a full blown MPN. Indeed, CHIP has been associated with an increased risk of coronary artery disease and stroke.19 In particular, JAK2V617F+ CHIP has been most frequently associated with an increased risk of developing cardiovascular diseases, thrombosis and coronary heart disease.19 Furthermore, Cordua et al.20 have shown that subjects with JAK2V617F or calreticulin CHIP frequently eventually develop a full-blown MPN. The underlying events that lead to thromboses in MPN patients remain the subject of investigation. Historically, the thrombotic tendency may be influenced, as outlined

Table 1. Incidence and main clinical characteristics of vascular events in patients with myeloproliferative neoplasms. THROMBOSIS Disease Molecular Main Incidence Type Clinical Ref. Incidence features Phenotype characteristics PV

JAK2 V617F (95%) JAK2 exon 12 (5%) Sub-clonal mutations in myeloid genes

Erythrocytosis. which can be associated with leukocytosis and thrombocytosis.

JAK2 V617F (60%) MPL exon 10 (5%) CALR exon 9 (20%) Triple negative (5-10%) Sub-clonal mutations in myeloid genes

Thrombocytosis. Sometimes patients presented with normal white blood cell counts. A reduced red blood cell count can also be observed

JAK2 V617F (60%) MPL exon 10 (5%) CALR exon 9 (20%) Triple negative (5-10%) Sub-clonal mutations in myeloid genes. (ASXL1, DMT3A, EZH2, IDH1/IDH2, SRSF2, or TP53 are associated with a worse outcome)

Splenomegaly (85%); Cytopenia: - 2/3 of patients had anemia at diagnosis; - 40 to 50% have leukocytosis - 13-32% have thrombocytosis

- At diagnosis: 28.6% - During follow up: 3.8 x 100 person/year (1.5 deaths per 100 person/year)

Both arterial and venous

2, 3, 17 Mild microcirculatory disturbances (headache, itching, buzzing)

3– 8% (usually after the diagnosis)

BLEEDINGS Clinical characteristics

14 - Minor bleeding (e.g. ecchymoses, gingival hemorrhage, menorrhagia and epistaxis)

2, 3, 17, 22 - At diagnosis: 20.7% Mainly - During follow up: arterial 2-4 x 100 person/year

Major arterial and venous thrombotic events (ischemic stroke, peripheral artery disease, splanchnic vein thromboses, cerebral sinus thromboses, myocardial infarction, and deep vein thromboses)

14, 29 3 - 18% (usually after the diagnosis)

- Major bleeding (e.g. intracranial hemorrhage, gastrointestinal bleeding, retroperitoneal bleeding) - Extreme thrombocytosis may cause bleeding due to development of an acquired Von Willebrand syndrome

19 - 56% (~12% in patients with pre-fibrotic myelofibrosis)

- Main cause of bleedings are Portal hypertension with esophageal varices, the use of anti-platelet and/or anti-coagulant therapy

2, 3, 4, 5 - At diagnosis: 9.5% - During follow up: 2.2 x 100 person/year

Both arterial and venous

Over-representation of thrombosis in unusual sites (portal system, Budd-Chiari syndrome, cerebral venous thrombosis)

4, 14, 29

PV: polycythemia; ET: essential thrombocythemia; MF: myelofibrosis; ref.: references.

2814

Ref.

haematologica | 2021; 106(11)

Role of mutated endothelial cells in MPN

below, by a combination of increased numbers of abnormal myeloid cells and the co-existence of a chronic inflammatory state.2 Recently, new evidence has shown a role for endothelial cells, which is the subject of this review.

Blood cell alterations and thrombotic tendency in myeloproliferative neoplasms The elevated number of red cells and the resultant increased hematocrit levels are well established to have pro-thrombotic effects.2 Under low shear rates an elevated hematocrit leads to increased blood viscosity, while at high shear rates, the increased red cell numbers disperse platelets toward the vessel walls, resulting in platelet activation. Finally, biochemical changes have been observed in red cell membranes both in patients with polycythemia vera and in those with essential thrombocythemia, causing red blood cell aggregation.2 In contrast to red blood cells, there are few studies on platelets directly correlating the degree of thrombocytosis with the rate of thrombosis in MPN patients.21 The impact of leukocytosis on thrombosis has been evaluated in numerous retrospective studies, but with discordant results. Several studies suggest that the adhesion of leukocytes to EC contributes to the development of thrombosis, especially the formation of venous thrombi.22 By contrast, it has been recently documented that persistent leukocytosis in polycythemia vera was associated with disease progression, rather than thrombosis.23 In general, neutrophils play a central role in generating the inflammatory response and in activation of the blood coagulation system through the release of proteolytic enzymes and reactive oxygen species and the increased expression of CD11b which activates or damages platelets, EC and coagulation proteins.2 Moreover, granulocytes in MPN patients produced an increased amount of neutrophil extracellular traps that initiate and propagate arterial and venous thrombosis.24,25 Mouse models have demonstrated that neutrophil extracellular traps are crucial in the development of thrombosis.32 Moreover, MPN blood cells are also qualitatively abnormal due to their procoagulant and proteolytic properties, secretion of inflammatory cytokines, and expression of cell adhesion molecules.2 In particular, activated platelets in MPN patients express P-selectin and tissue factor and secrete an increased number of platelet activation products.26

Inflammation and thrombosis In concert, inflammatory cytokines secreted by MPN cells and leukocyte-derived proteases damage the integrity of the normal vascular endothelium, leading to the acquisition of a pro-thrombotic phenotype in MPN patients. Specifically, EC overexpress adhesion receptors favoring the attachment of platelets, erythrocytes, and leukocytes to the vascular wall. In addition, MPN patients have increased levels of circulating procoagulant microparticles which are associated with activation of protein C.2

Endothelial cells and thrombosis In general, numerous insults occur in MPN patients, which perturb the integrity of the endothelium, resulting in a pro-adhesive and pro-coagulant EC surface. Over the last decade, increasing evidence has been provided indicating that JAK2-mutated MPN EC might also contribute to the MPN pro-thrombotic state.27,28 This evidence will be reviewed here. haematologica | 2021; 106(11)

Bleeding risk factors in patients with myeloproliferative neoplasms Risk factors for developing hemorrhagic events are less well understood. The JAK2V617F mutation has not only been related to the rate of thrombosis, but also to the rate of bleeding events.4 Furthermore, thrombocytopenia due to hypersplenism and/or progressive myelofibrosis may enhance the risk of bleeding.14 Paradoxically, extreme thrombocytosis is associated with bleeding due to the development of acquired von Willebrand syndrome.21 The type of MPN also appears to influence the hemorrhagic risk, with an increased incidence being associated with prefibrotic primary myelofibrosis as compared to essential thrombocythemia.29 In general, the effect of the administration of antiplatelet aggregating agents on bleeding events in MPN patients is debatable. These agents should however be used with caution in patients with extreme thrombocytosis and acquired von Willebrand syndrome, severe thrombocytopenia, or in those receiving oral anticoagulants. There are several possible factors that contribute to bleeding in MPN patients, including both disease-related factors (e.g., MPN subtypes, thrombocytopenia or extreme thrombocytosis, platelet storage pool defects with a downregulation of glycoproteins (GP)Ib and GPIIb/IIIa and therapy-related factors (e.g., use of antiplatelet and anticoagulant therapies,27 drug-induced thrombocytopenia due to ruxolitinib, fedratinib, interferon, busulfan or hydroxyurea).

Endothelial cell involvement in myeloproliferative neoplasms A significant increase in marrow and splenic microvascular density30 is a characteristic feature of MPN, particularly polycythemia vera and myelofibrosis. Moreover, neo-angiogenesis represents a hallmark of these diseases.31 Whether neo-angiogenesis in MPN is an epiphenomenon of the MPN pro-inflammatory milieu or a consequence of EC dysregulation due to the same pathogenic mechanism that leads to the hematopoietic cell proliferation32 remains controversial. It is important to realize that these two mechanisms are not mutually exclusive and could be operating in concert. In addition, increased serum levels of pro-angiogenic factors, such as vascular endothelial growth factor (VEGF), have been reported in MPN patients.33 It has been suggested that autocrine and paracrine signaling pathways lead to increased levels of VEGF, which may not only contribute to accelerated hematopoietic cell growth but may also contribute to the MPN-associated risk of thrombosis.34 The increased marrow and splenic microvessel density and neo-angiogenesis, together with the high incidence of vascular complications, has led some authors to hypothesize direct involvement of EC by the malignant process in MPN. The observation that EC and their precursors may harbor the JAK2V617F mutation supports this hypothesis.8,9,12,39,42,43 However, studying the contribution of EC to human disease development is challenging because endothelium cannot, for ethical reasons, be easily sampled from patients. This limitation has meant that most published papers providing support for the abovementioned hypothesis are based on in vitro studies dealing with circulating endothelial progenitors,32,34,37 and mature 2815

M. Farina et al.

EC.8,9 Moreover, some authors demonstrated that monocytes isolated from MPN patients are capable of generating cells that closely resemble EC, the so-called endothelial-like cells or angiogenic monocytes.38 Both in animal models and humans, angiogenic monocytes contribute to neo-vessel formation while assuming a mature EC phenotype.38 However, in humans it is currently thought that endothelial-like cells influence angiogenesis by secreting pro-angiogenic factors (paracrine effects), rather than participating directly in neovascularization.39 At present, the true origin of mutated EC in MPN patients remains the subject of debate. Where do these cells originate from? Can we be certain of their true EC nature? In the following sections we review the instruments that are presently used to study EC biology in order for the reader to better appreciate the challenges encountered in understanding the origins and consequences of JAK2-mutated EC.

Assays for endothelial cells and endothelial progenitor cells A growing number of assays have been utilized to study the origins of EC in MPN. It is impossible to evaluate the validity of such data without first understanding the nature of each of these assays as well as their strengths and limitations. We will describe each of the currently used assays below. Circulating endothelial progenitor cells (EPC) (Table 2) have the capacity to proliferate, migrate, and differentiate into cells belonging to the endothelial lineage, but do not acquire the characteristic features of mature EC. EPC are very rare peripheral blood cells (0.0001% of circulating nucleated cells).40 In both animal models and humans they have been reported to play a role in vascular repair and neo-angiogenesis.40 Asahara et al.40 initially reported the isolation of a putative EPC from human peripheral blood, on the basis of cell surface expression of CD34 (expressed by EC, as well as HSC) and Flk-1 (a receptor for VEGF2). These cells were capable of de novo blood vessel tube formation. Subsequently, Urbich and Dimmeler41 defined EPC as progenitors of EC that were capable of

clonal expansion with stem cell-like characteristics and had the capacity to differentiate into EC. Since these initial observations, there has been a great deal of debate concerning the definition and characterization of these progenitor cells. In addition, a variety of methods have been used to detect and characterize EPC, which has led to disparate results.49 Three main approaches have been used to identify and isolate EPC. One approach is to identify EPC using surface antigen expression with cytofluorimetry of circulating cells (Table 3). Unfortunately, the presently used cell surface markers, CD34, VEGFR2 (human KDR and mouse Flk-1) and CD133 do not unequivocally identify EPC.37 This approach allows EPC to be distinguished from mature circulating endothelial cells (CEC), since CD133 is a stem cell marker expressed by EPC but not by mature EC.43 A second method of assaying for EPC consists of plating human peripheral blood or cord blood low-density mononuclear cells in culture dishes coated with fibronectin in a commercially available culture medium rich in EC growth factors and fetal calf serum.44 After 4-5 days the non-adherent cells are removed and the adherent cells are examined for their ability to bind acetylated low-density lipoprotein and Ulex europaeus agglutinin 1 (a plant lectin). The putative EPC identified are called circulating angiogenic cells. These markers, however, lack specificity45 (numerous blood cells express the integrin receptors for fibronectin) and these cells typically do not form EC colonies in vitro.46 EPC identified in this manner are thought to contribute to neo-angiogenesis by secreting angiogenic factors (paracrine route).46 The third method to quantitate the numbers of EPC is based on the in vitro colony-forming capacity of cultured CD34+ cells. Two classes of EPC have been described, which are termed colony-forming unit-endothelial cells (CFU-EC) and endothelial colony-forming cells (ECFC). CFU-EC are assayed by plating CD34+ cells for 48 h in fibronectin-coated dishes and then replating the nonadherent cells and monitoring for the emergence of the EPC-derived colonies. These CFU-EC, however, fail to display any postnatal vasculogenic activity and are thought ultimately to be the cellular progeny of myeloid cells.45 Since this assay includes the adhesion of mononu-

Table 2. Main abbreviations referring to endothelial progenitor cells and mature endothelial cells, and brief definitions of the types of cells.

Abbreviation

Definition

EPC = endothelial progenitor cell

Endothelial progenitors that differentiate into endothelial cells and may become part of the newly formed vessel wall or favor angiogenesis by secretion of pro-angiogenic factors (paracrine effect). There are several ex vivo assays for EPC. Among the EPC, ECFC originate from peripheral blood mononuclear cells and are able to form large colonies of human CD45− cells after 1–3 weeks of incubation (once called late outgrowth endothelial cells, OEC), which have phenotypic and functional properties of endothelial cells. Indeed, they are able to generate new vessels in vivo and to generate endothelial colonies ex vivo, and are now considered the true precursor cells of endothelial cells. These are assayed by plating CD34+ cells for 48 h in fibronectin-coated dishes and then replating the non-adherent cells and monitoring for the emergence of the EPC-derived colonies. Because of the brief period of incubation ex vivo they were once called early outgrowth endothelial cells, (EOC). They were initially included as endothelial precursors, but they do not possess any postnatal vasculogenic activity and, therefore, are no longer considered true EPC. Bone marrow-derived immune cell populations (T cells and certain subsets of monocytes) that stimulate vascular regeneration and angiogenesis through a paracrine mechanism. Mature endothelial cells circulating in the peripheral blood, which are shed from vessel walls as a result of pathophysiological conditions that affect the endothelium. Monocytes that closely resemble endothelial cells and acquire endothelial cell surface markers.

ECFC = endothelial colony-forming cells

CFU-EC = colony-forming unit-endothelial cells

CAC = circulating angiogenic cells CEC = circulating endothelial cells ELC = endothelial-like cells 2816

haematologica | 2021; 106(11)

Role of mutated endothelial cells in MPN

clear cells in vitro, this approach may select for monocytes, expressing “endothelial-specific” markers.38 Another assay system identifies outgrowth EC. This assay identifies clonal ECFC capacity of EPC, which form large colonies of human CD45− cells after 1-3 weeks of incubation.45 The cells within these colonies are thought to be of EC origin because of their: EC morphology, expression of EPC/EC-related markers (CD31, CD105, CD144, CD146, VWF, and KDR)36 and spontaneous formation of human blood vessel tubes in vitro47 and in vivo (postnatal vasculogenesis).48 The ability of ECFC to display spontaneous vasculogenic properties and to remodel into arteries and veins in vivo distinguishes ECFC from all other EC precursor or progenitor cell types previously described.45 ECFC are likely the cell population that represents a true lineage-restricted EC progenitor cell.

Circulating endothelial cells CEC are mature differentiated EC that are shed from vessel walls as a result of pathophysiological conditions that affect the endothelium.49 CEC were first identified in the 1970s although more user-friendly techniques to isolate CEC have only recently become available.50 Prolonged or exaggerated activation by environmental stress leads to dysfunction and to irreversible loss of EC integrity with cell detachment, apoptosis and necrosis, which results in greater EC turnover and increased CEC levels in peripheral blood.50 CEC were initially identified using morphological criteria. Subsequently, objective methods to identify CEC with the application of immunofluorescence, and the use of antibodies against various EC markers, were introduced although these efforts have been hampered by the lack of reliable cell-specific markers.51 Recently a consensus definition of CEC has been reached,52 according to which CEC are large (>10 mm in length) CD146+ cells. CD146 (MUC18) is expressed by CEC but not by monocytes, granulocytes, platelets, megakaryocytes, T or B lymphoTable 3. Biological characteristics and immunophenotype of endothelial progenitor cells and circulating endothelial cells.

EPC CFU-EC

CAC Immunophenotype

CD34+/- * VEGFR2+ * CD133+ CD31+ CD146CD45+/Origin BM Proliferative capacity Replating ability In vitro tube formation +/In vivo de novo formation Paracrine augmentation + of angiogenesis Phagocytosis of bacteria +

CEC ECFC

CD34+ * CD34+ * CD34+ * VEGFR2+ * VEGFR2+ * VEGFR2+ * CD133+ CD133+ CD133CD31+ CD31+ CD31+ CD146CD146CD146+ CD45+/CD45CD45BM EC?/BM? EC + -/-/+ + -/+ +/+ + + +/NA +

*In common with hematopoietic stem cells. The main differences between CEC and EPC are shown in red. EPC: endothelial progenitors cells; CEC: circulating endothelial cells; CAC: circulating angiogenic cells; CFU-EC: colony-forming unit–endothelial cells; ECFC: endothelial colony-forming cells;VEGFR2: vascular endothelial growth factor receptor 2; BM: bone marrow; EC: endothelial cells; NA: not applicable.

haematologica | 2021; 106(11)

cytes.53 A battery of markers is now used to identify cells of endothelial origin, including CD31, CD105, and CD141.54 Notably, the absence of CD133 may also be used to distinguish CEC from EPC.55 Currently, CEC can also be isolated by immunomagnetic selection (CD146+ cells) or by flow cytometry. Notably, in 2008, Widemann et al.49 reported a hybrid assay that incorporated an algorithm combining immunomagnetic selection of CD146+ cells with flow cytometric quantification. In parallel, Terstappen’s group56 developed a semi-automatic method for the detection of CEC, also using a combination of iron microbeads and monoclonal antibodies. These assays overcome the lack of standardization and the variability in CEC detection associated with the methods previously described. Moreover, the true endothelial nature of the CEC obtained using this technology was confirmed by gene expression profiling studies.57 In healthy individuals, the endothelial layer lining blood vessels is continuously being renewed at a low replication rate of 0-1% per day since normal laminar flow suppresses EC apoptosis. CEC are rare cells, with as few as 0-10 CEC/mL being observed in healthy donors.58 By contrast, elevated levels of CEC have been reported in patients with various types of diseases, including cardiovascular,59 infectious,60 and immune disorrders,61 diabetes, chronic kidney disease,62 after hematopoietic stem cell transplantation,63 and cancer.64 Several pioneering studies have shown that raised CEC levels are also associated with specific tumor types, stage and prognosis,65 and can be used to monitor responses to chemotherapy.66 In addition, CEC have been proposed as a non-invasive marker of angiogenesis.67 In contrast to EPC, which are a proposed marker of regeneration and vessel proliferation, CEC serve as a marker of endothelial damage/dysfunction and reflect a pro-thrombotic tendency.68 Notably, the numbers of CEC are increased in MPN patients, regardless of their driver mutational status,69 highlighting the involvement of endothelium in these chronic hematologic neoplasms. CEC may provide a means to study mature EC that avoids laser microdissection or the limitations associated with performing the tedious and time-consuming EPC assays. However, a consensus on CEC phenotype and the origin of these cells is lacking and the possibility that EC or endothelial-like cells originate from monocytes remains.

JAK2V617F-positive endothelial cells in patients with myeloproliferative neoplasms (Figure 1) In 2009 Sozer et al.8 reported that mature EC captured by laser microdissection from the lumen of hepatic venules harbored the JAK2V617F mutation in three MPN patients with Budd-Chiari syndrome (Figure 1, on the right). Rosti et al. further confirmed the presence of JAK2V617F in micro-laser dissected EC from the splenic vein in MPN patients, but absence of the driver mutation in the ECFC residing in the spleen9 (Figure 1). Assayable MPN CFU-EC11,45,70 were first shown to be JAK2V617F+ while ECFC from these same patients were found to be JAK2V617F– (Figure 1). Only 3% of the ECFC colonies analyzed by Yoder et al.45 were JAK2V617F+. Interestingly, these mutated-ECFC were derived from the same patient, who presented with a thrombotic event and only later 2817

M. Farina et al.

developed classic hematologic signs of polycythemia vera. Notably, increased numbers of both CFU-EC32,34,70 and ECFC71 have been found in the blood of patients with MPN, regardless of their mutational status. The absence of the JAK2 mutation in ECFC from MPN patients was recently confirmed by Guy and colleagues.13 Teofili et al.12 however, reported that ECFC from patients with MPN were JAK2V617F+ (Figure 1). Almost half of the MPN patients studied were reported to have MPN-like genetic abnormalities in their ECFC, including either SOCS gene hypermethylation or the presence of JAK2V617F. Notably, mutated ECFC were detected only in patients with a history of thrombotic events.12 Moreover, the presence of JAK2V617F or other evidence of clonality in ECFC was associated with JAK/STAT pathway activation and significantly greater adhesion of mononuclear cells to mutated EC than normal ECFC.12 These reports support the hypothesis that EC and HSC may derive from a common progenitor cell, the “hemangioblast”,72 which results in mutated EC and myeloid cells in a subpopulation of patients with MPN. It must be said, however, that conclusive evidence unequivocally demonstrating the existence of the "hemangioblast" in vivo in higher vertebrates is lacking. Indeed, most of the published studies have been largely based on experiments that relied on the isolation, culture, and/or manipulation of cells in vitro,72,73 while various fate-mapping studies in the mouse, chick, and zebrafish have led to contradictory conclusions.74,75 Fate mapping in the zebrafish gastrula has indicated that the “hemangioblasts” are interspersed with hematopoietic and endothelial progenitors in the ventral-lateral mesoderm.76,77 In contrast, several other studies have suggested

that endothelial and hematopoietic lineages are independently derived from mesodermal cells.78,79 The discovery that MPN patients may share the JAK2V617F driver mutation has shed new light on this hypothesis. Moreover, some authors recently suggested that JAK2V617F, along with other myeloid malignancyassociated gene mutations, may be detected in CEC80 and HSC in patients with primary myelofibrosis. The concordance between mutations in HSC and CEC may further support the hypothesis of a common progenitor that generates these two subpopulations, but peer-reviewed studies are still required to confirm this hypothesis. Regardless of the presence of a common precursor, each of these observations supports the hypothesis that mutated EC in MPN represent a “neoplastic” vascular niche,81 which allows blood cell adhesion and tumor cell growth, as demonstrated using in vitro and in vivo assays.

Impact of JAK2V617F endothelial cells on hematopoiesis and vascular complications in myeloproliferative neoplasms (Figure 2) In vivo and in vitro models The observation that EC from some MPN patients were JAK2V617F+ stimulated the performance of additional studies exploring the possible functional consequences of JAK2-mutated EC. Etheridge et al.82 first described the critical role of JAK2V617F-mutated EC in the development of bleeding abnormalities using murine models. They used FF1 transgenic mice to express JAK2V617F in different cell lineages. In their model JAK2V617F was exclusively present in

Figure 1. Evidence for JAK2V617F mutated endothelial cells in patients with myeloproliferative neoplasms. JAK2V617F has been detected in both endothelial progenitors and mature endothelial cells. Studies in which a JAK2 mutation was detected in endothelial progenitor cells or mature endothelial cells are shown in bold. Positive colonies or patients are expressed on the total number of colonies or patients analyzed. CFU-EC: colony forming unit-endothelial cells, derived from nonadherent mononuclear cell culture (see the text); MNC: mononuclear cells; EC: endothelial cells; ECFC: endothelial colony-forming cells, derived from long-term adherent MNC culture (see the text); LCM: laser-capture microdissection; CEC: circulating endothelial cells; Drawn by C. Luzzani, Medical School, University of Milan

2818

haematologica | 2021; 106(11)

Role of mutated endothelial cells in MPN

EC, and the mice were characterized by dysfunctional hemostasis in response to injury, resembling the bleeding diathesis observed in MPN patients.82 One of the potential mechanisms proposed by Etheridge and colleagues was related to von Willebrand factor (VWF) regulation. More recently, using both an in vitro model of human JAK2V617F-mutated EC and an in vivo model of mice with endothelial-specific JAK2V617F expression, Guy et al.27 have shown that JAK2V617F+ EC in the absence of similarly mutated hematopoietic cells are associated with a higher rate of thrombosis due to a pro-adhesive phenotype as a result of increased endothelial P-selectin exposure, secondary to degranulation of Weibel-Palade bodies.27 Interestingly, these mice displayed a higher propensity for thrombosis in spite of having normal blood counts and normal rates of thrombin generation.27 In contrast, their EC were characterized by increased surface expression of P-selectin and VWF, both of which are contained within Weibel-Palade bodies. Moreover, the thrombotic tendency was accentuated by the creation of a pro-inflammatory milieu due to the administration of low doses of tumor necrosis factor-a.27 Furthermore, the pro-adhesive properties of the JAK2V617F-mutated EC were reversed by treatment with either a P-selectin blocking antibody or hydroxyurea.27 In addition, Poisson et al. showed an increased degree of arterial contraction in response to agents that promote vasoconstriction in mice with JAK2V617F+ HSC and EC.83 Castiglione et al.84 have reported that when JAK2V617F was expressed by both hematopoietic cells and EC in a murine model of MPN, the mice developed an MPN phe-

notype and a spontaneous age-related dilated cardiomyopathy with an increased risk of sudden death as well as a pro-thrombotic and vasculopathic phenotype. In contrast, mice expressing solely JAK2V617F in blood cells did not demonstrate any evidence of cardiac dysfunction or thrombosis, suggesting that expression of the MPN driver mutation in EC is required for the development of the cardiovascular disease phenotype. Moreover, the authors demonstrated that the JAK2V617F+ EC were associated with the development of a pro-inflammatory milieu. Finally, JAK2-mutated EC have been reported to respond to shear flow in a different manner than wild-type EC, leading to upregulation of EC adhesion molecules (platelet endothelial cell adhesion molecule and E-selectin). Guadall et al.28 have provided additional evidence that JAK2V617F+ EC possess pro-thrombotic properties. Using JAK2V617F+ and JAK2 wild-type induced pluripotent stem cells generated from an MPN patient and redirecting these cells towards the endothelial lineage, the authors observed that JAK2V617F+ EC had a greater proliferative capacity compared with wild-type EC. The numbers and fluorescence intensity of Weibel-Palade bodies as well as the expression of VWF and P-selectin were significantly greater and these effects were accompanied by greater accumulation of Pselectin at the cell surface of JAK2V617F+ EC than wildtype EC. The transcriptomic profile of these mutated cells revealed overexpression of transcripts for genes that are involved in inflammation and cell adhesion, extracellular matrix regulation, the generation of glycoproteins, and a variety of processes that occur in venous stenosis and thrombosis.

Figure 2. Effects of JAK2V617F expression in endothelial cells. The presence of the JAK2V617F mutation in endothelial cells (EC) has an impact on both (1) vascular complications and (2) the development of myeloproliferative neoplasms (MPN). Specifically, it affects bleeding82 (the carotid arteries of Tie2-Cre/FF1 mice expressing JAK2 mutations on both EC and hematopoietic stem cells (HSC) failed to occlude in response to ferric chloride, which normally induces occlusive thrombosis in murine carotid arterial); thrombosis (both in a mice model, due to enhanced P-selectin expression,27 and in an in vitro model of induced pluripotent stem cells28) and cardiovascular disease (mice expressing JAK2V617F had spontaneously dilated cardiomyopathy and an increased risk of sudden death84). Finally, JAK2-mutated EC affect MPN development, promoting JAK2 HSC expansion,85 and radioresistance.88 FeCl : ferric chloride; EC: endothelial cells; MPN: myeloproliferative neoplasms; iPS: induced pluripotent stem cells; HSC: hematopoietic stem cells;. WT: wild-type. 3

haematologica | 2021; 106(11)

2819

M. Farina et al.

Figure 3. Possible origin of JAK2-mutated endothelial cells. The documentation of JAK2 myeloproliferative neoplasm driver mutations in myeloid cells and endothelial cells (EC) suggests that in some individuals both cell types originate from a “hemangioblast”, which might serve as the cell of origin for myeloproliferative neoplasms during embryogenesis. On the other hand, JAK2-positive EC may derived from monocytes that resemble EC (endothelial-like cells) as well. EC: endothelial cells; EPC: endothelial progenitor cells; HSC: hematopoietic stem cells; ELC: endothelial-like cells.

Effects of JAK2V617F-positive endothelial cells on hematopoiesis JAK2V617F+ EC have also been shown to contribute not only to thrombo-hemorrhagic events, but also to MPNassociated myeloproliferation.85 JAK2V617F-bearing EC have been reported to promote the proliferation of JAK2mutated hematopoietic progenitor/stem cells over JAK2 wild-type ones in vitro. This proliferative advantage has been hypothesized to be due to activation of the thrombopoietin/MPL signaling axis.85 Subsequently, Zhan et al. provided in vivo evidence that JAK2V617F+ vascular niche cells promote JAK2V617F+ myeloid cell expansion, while inhibiting JAK2 wild-type hematopoiesis. Zhan et al. also reported that JAK2V617F+ HSC transplanted into wildtype recipient mice were incapable of developing an MPN phenotype in the absence of JAK2V617F+ vascular niche EC. Therefore, in this model MPN myeloproliferation requires JAK2V617F expression by both HSC and EC.86 However, there is evidence from mouse models indicating that the presence of the JAK2 mutation in HSC alone is sufficient to induce an MPN.87 In support of this role of mutated EC in MPN hematopoiesis, Lin and colleagues reported that JAK2V617F+ HSC were protected from lethal doses of irradiation by JAK2V617F+ vascular niche EC.88 These authors hypothesized that the relative resistance of MPN to radiation-based conditioning regimens used prior to allogeneic stem transplantation could be due to the presence of JAK2V617F+ EC within the patient’s bone marrow HSC vascular niche.88

Monocytes can assume the identity of endothelial cells Notably, some authors have reported that monocytes 2820

isolated from MPN patients resemble endothelial-like cells, accounting for the detection of MPN driver mutations in EC and hematopoietic cells (Figure 3). Leibundgut et al.32 initially reported that CD14+ monocytes were capable of generating JAK2V617F+ EC in vitro. Subsequently, Sozer and colleagues10 showed that human CD34+ cells, too, were capable of generating normal and JAK2V617F+ endothelial-like cells in vivo. These reports suggest that JAK2-mutated CD34+ cells and CD14+ monocytes (both elevated in MPN) may both transform to JAK2V617F+ endothelial-like cells. These observations have led to considerable confusion, suggesting to some investigators that monocytes can transition to EPC46 and then acquire an endothelial-like phenotype. However, a more plausible hypothesis is that monocytes can serve as circulating regulators of the angiogenic response and play a crucial role in neo-angiogenesis during wound healing, tissue ischemia, and tumorigenesis by secreting pro-angiogenic factors rather than by directly participating in neo-vessel formation or endothelial turnover.39,89

Do endothelial cells and hematopoietic stem cells share a common precursor cell in patients with myeloproliferative neoplasms? HSC and EC are both derived from the mesodermal layer during fetal development. Some authors have speculated that they may be derived from a common precursor cell, termed a “hemangioblast”. The term “hemangioblast” was initially coined by Murray in 1932,90 referring to a mass of cells derived from the primitive streak mesoderm containing both endothelial and blood cells. This term was meant to complement and contrast with the term ‘‘angioblast,’’ which was thought to be the source of vessels and endothelium.91 During the late haematologica | 2021; 106(11)

Role of mutated endothelial cells in MPN

1990s, the concept of the “hemangioblast” was developed, based on observations that single mesodermal cells isolated from mice had the potential to generate both blood cells and EC.92 Interestingly, in many species HSC appear as clusters attached to the endothelium that lines the ventral wall of the abdominal aorta during embryonic development; this observation has long implicated the hemogenic endothelium as the source of developing blood cells. Indeed, when EC isolated from mouse embryos are grown in culture, the hemogenic endothelium possesses the potential to develop into mature blood cells.93 During development this hemogenic endothelium, gives rise to HSC/hematopoietic progenitor cells that seed the fetal liver and the adult bone marrow.94 Lineage-tracing markers in mice have identified that definitive HSC arise in the aorta-gonad-mesonephric region of embryos from hemogenic endothelium which gives rise, by asymmetric division, to resident EC and HSC that are released into the blood and then colonize the liver.95 Peault’s team subsequently described the presence of definitive HSC in the aorta-gonad-mesonephric region of human embryos which were capable of colonizing adult xenografts and reported that definitive HSC were derived from hemogenic endothelium resembling those observed in mouse embryos.96 The relationship between HSC and hemogenic endothelium has been further clarified,94 based on continuous single-cell imaging which indicated that freely moving cells expressing blood-specific markers (CD45, CD41, CD11b) were generated from EC expressing vascular endothelial cadherin (VE-cadherin, also known as Cdh5).97 The reports discussed above showing that the JAK2V617F driver mutation8,9,11,12,37,98 and other myeloidassociated genes mutations80 may be present in both hematopoietic cells and EC in MPN patients have reinforced the evidence supporting the existence of a common precursor cell for both EC and hematopoietic cells. In addition, some authors have recently provided evidence that JAK2V617F may be acquired in utero99 or during childhood100 by MPN patients in whom JAK2V617F was the only or the first driver mutation. This finding indicates that the acquisition of JAK2V617F in MPN patients can occur in utero and is at least chronologically consistent with involvement of the “hemangioblast” by MPN driver

References 1. Spivak JL. Myeloproliferative neoplasms. N Engl J Med. 2017;376(22):2168-2181. 2. Barbui T, Finazzi G, Falanga A. Myeloproliferative neoplasms and thrombosis. Blood. 2013;122(13):2176-2184. 3. Rungjirajittranon T, Owattanapanich W, Ungprasert P, Siritanaratkul N, Ruchutrakool T. A systematic review and meta-analysis of the prevalence of thrombosis and bleeding at diagnosis of Philadelphia-negative myeloproliferative neoplasms. BMC Cancer. 2019;19(1):184. 4. Kc D, Falchi L, Verstovsek S. The underappreciated risk of thrombosis and bleeding in patients with myelofibrosis: a review. Ann Hematol. 2017;96(10):1595-1604. 5. Barbui T, Carobbio A, Cervantes F, et al. Thrombosis in primary myelofibrosis: incidence and risk factors. Blood. 2010;115 (4):778-782.

haematologica | 2021; 106(11)

mutations (Figure 3). Since the period when EC are hemogenic may be very brief and occurs very early during embryogenesis, the “hemangioblast” may acquire the MPN driver mutation in only a limited group of patients. These assumptions would support the observation that not all JAK2V617F MPN patients possess mutated EC.

Conclusions The findings summarized here indicate that mutated EC play multiple roles in the development of the clinical phenotype of MPN (Figures 2 and 3). The interaction between EC and MPN HSC creates microenvironmental niches which promote the predominance of the malignant MPN myeloid cells at the expense of the normal HSC. In addition, the documented MPN driver mutations in myeloid cells and EC suggest that in some individuals both cell types originate from a “hemangioblast” present during fetal development or which persists during adult life, and serves as the cell of origin of MPN. Further investigation using single-cell analysis of the putative MPN “hemangioblast” will be required to further confirm this hypothesis. A significant body of evidence indicates that JAK2V617F+ EC contribute to the thrombotic and bleeding tendencies of MPN patients. Additional work will also be required to assess the relative contribution of monocytes that resemble EC and mutated EPC to the prothrombotic MPN milieu. A likely scenario is that the contribution of these two types of EC to the prothrombotic tendency in MPN varies from patient to patient and may be determined in part by the vascular beds in which the thrombotic events occur. Disclosures No conflicts of interest to disclose. Contributions MF and RH conceived and wrote the manuscript. DR wrote the manuscript. All the authors approved the final version. Acknowledgments We acknowledge Carlo Luzzani, medical student at the University of Milan, for his help in drawing the figures provided in this review.

6. Bar-Natan M, Hoffman R. New insights into the causes of thrombotic events in patients with myeloproliferative neoplasms raise the possibility of novel therapeutic approaches. Haematologica. 2019; 104(1):3-6. 7. Michiels C. Endothelial cell functions. J Cell Physiol. 2003;196(3):430-443. 8. Sozer S, Fiel MI, Schiano T, Xu M, Mascarenhas J, Hoffman R. The presence of JAK2V617F mutation in the liver endothelial cells of patients with Budd-Chiari syndrome. Blood. 2009;113(21): 5246-5249. 9. Rosti V, Villani L, Riboni R, et al. Spleen endothelial cells from patients with myelofibrosis harbor the JAK2V617F mutation. Blood. 2013;121(2):360-368. 10. Sozer S, Ishii T, Fiel MI, et al. Human CD34+ cells are capable of generating normal and JAK2V617F positive endothelial like cells in vivo. Blood Cells Mol Dis. 2009;43(3):304-312. 11. Piaggio G, Rosti V, Corselli M, et al. Endothelial colony-forming cells from

patients with chronic myeloproliferative disorders lack the disease-specific molecular clonality marker. Blood. 2009;114 (14):31273130. 12. Teofili L, Martini M, Iachininoto MG, et al. Endothelial progenitor cells are clonal and exhibit the JAK2V617F mutation in a subset of thrombotic patients with Ph-negative myeloproliferative neoplasms. Blood. 2011;117(9):2700-2707. 13. Guy A, Danaee A, Paschalaki K, et al. Absence of JAK2V617F mutated endothelial colony-forming cells in patients with JAK2V617F myeloproliferative neoplasms and splanchnic vein thrombosis. Hemasphere. 2020;4(3):e364. 14. Kaifie A, Kirschner M, Wolf D, et al. Bleeding, thrombosis, and anticoagulation in myeloproliferative neoplasms (MPN): analysis from the German SAL-MPN-registry. J Hematol Oncol. 2016;9(1):18. 15. Guy A, Poisson J, James C. Pathogenesis of cardiovascular events in BCR-ABL1-nega-

2821

M. Farina et al. tive myeloproliferative neoplasms. Leukemia. 2021;35(4):935-955. 16. Hasselbalch HC, Elvers M, Schafer AI. The pathobiology of thrombosis, microvascular disease, and hemorrhage in the myeloproliferative neoplasms. Blood. 2121;137(16): 2152-2160. 17. De Stefano V, Za T, Rossi E, et al. Recurrent thrombosis in patients with polycythemia vera and essential thrombocythemia: incidence, risk factors, and effect of treatments. Haematologica. 2008;93(3):372-380. 18. Vannucchi AM, Pieri L, Guglielmelli P. JAK2 allele burden in the myeloproliferative neoplasms: Effects on phenotype, prognosis and change with treatment. Ther Adv Hematol 2011;2(1):21-32. 19. Jaiswal S, Natarajan P, Silver AJ, et al. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017;377(2):111-121. 20. Cordua S, Kjaer L, Skov V, Pallisgaard N, Hasselbalch HC, Ellervik C. Prevalence and phenotypes of JAK2 V617F and calreticulin mutations in a Danish general population. Blood. 2019;134(5):469-479. 21. Campbell PJ, MacLean C, Beer PA, et al. Correlation of blood counts with vascular complications in essential thrombocythemia: analysis of the prospective PT1 cohort. Blood. 2012;120(7):1409-1411. 22. Palandri F, Polverelli N, Catani L, Ottaviani E, Baccarani M, Vianelli N. Impact of leukocytosis on thrombotic risk and survival in 532 patients with essential thrombocythemia: a retrospective study. Ann Hematol. 2011;90(8):933-938. 23. Ronner L, Podoltsev N, Gotlib J, et al. Persistent leukocytosis in polycythemia vera is associated with disease evolution but not thrombosis. Blood. 2020;135(19):1696-1703. 24. Wolach O, Sellar RS, Martinod K, et al. Increased neutrophil extracellular trap formation promotes thrombosis in myeloproliferative neoplasms. Sci Transl Med. 2018;10(436):eaan8292. 25. Guy A, Favre S, Labrouche-Colomer S, et al. High circulating levels of MPO-DNA are associated with thrombosis in patients with MPN. Leukemia. 2019;33(10):2544-2548. 26. Jensen MK, De Nully Brown P, Lund BV, Nielsen OJ, Hasselbalch HC. Increased platelet activation and abnormal membrane glycoprotein content and redistribution in myeloproliferative disorders. Br J Haematol. 2000;110(1):116-124. 27. Guy A, Gourdou-Latyszenok V, Le Lay N, et al. Vascular endothelial cell expression of JAK2V617F is sufficient to promote a prothrombotic state due to increased P-selectin expression. Haematologica. 2019;104(1):7081. 28. Guadall A, Lesteven E, Letort G, et al. Endothelial cells harbouring the JAK2V617F mutation display pro-adherent and prothrombotic features. Thromb Haemost. 2018;118(09):1586-1599. 29. Finazzi G, Carobbio A, Thiele J, et al. Incidence and risk factors for bleeding in 1104 patients with essential thrombocythemia or prefibrotic myelofibrosis diagnosed according to the 2008 WHO criteria. Leukemia. 2012;26(4):716-719. 30. Boveri E, Passamonti F, Rumi E, et al. Bone marrow microvessel density in chronic myeloproliferative disorders: a study of 115 patients with clinicopathological and molecular correlations. Br J Haematol. 2008;140(2): 162-168 31. Barosi G, Rosti V, Massa M, et al. Spleen neoangiogenesis in patients with myelofi-

2822

brosis with myeloid metaplasia. Br J Haematol. 2004;124(5):618-625. 32. Oppliger Leibundgut E, Horn MP, Brunold C, et al. Hematopoietic and endothelial progenitor cell trafficking in patients with myeloproliferative diseases. Haematologica. 2006;91(11):1465-1472. 33. Tefferi A, Pardanani A. Myeloproliferative neoplasms. JAMA Oncol. 2015;1(1):97. 34. Massa M, Rosti V, Ramajoli I, et al. Circulating CD34+, CD133+, and vascular endothelial growth factor receptor 2-positive endothelial progenitor cells in myelofibrosis with myeloid metaplasia. J Clin Oncol. 2005;23(24):5688-5695. 35. Hill JM, Zalos G, Halcox JPJ, et al. Circulating endothelial progenitor cells, vascular function, and cardiovascular risk. N Engl J Med. 2003;348(7):593-600. 36. Ingram DA, Mead LE, Tanaka H, et al. Identification of a novel hierarchy of endothelial progenitor cells using human peripheral and umbilical cord blood. Blood. 2004;104(9):2752-2760. 37. Yoder MC, Mead LE, Prater D, et al. Redefining endothelial progenitor cells via clonal analysis and hematopoietic stem/progenitor cell principals. Blood. 2007;109(5):1801-1809. 38. Murdoch C, Muthana M, Coffelt SB, Lewis CE. The role of myeloid cells in the promotion of tumour angiogenesis. Nat Rev Cancer. 2008;8(8):618-631. 39. Dudley AC, Udagawa T, Melero-Martin JM, et al. Bone marrow is a reservoir for proangiogenic myelomonocytic cells but not endothelial cells in spontaneous tumors. Blood. 2010;116(17):3367-3371. 40. Asahara T, Murohara T, Sullivan A, et al. Isolation of putative progenitor endothelial cells for angiogenesis. Science. 1997;275 (5302):964-966. 41. Urbich C, Dimmeler S. Endothelial progenitor cells. Trends Cardiovasc Med. 2004;14 (8):318-322. 42. Basile DP, Yoder MC. Circulating and tissue resident endothelial progenitor cells. J Cell Physiol. 2014;229(1):10-16. 43. Sabatier F, Camoin-Jau L, Anfosso F, Sampol J, Dignat-George F. Circulating endothelial cells, microparticles and progenitors: key players towards the definition of vascular competence. J Cell Mol Med. 2009;13(3): 454-471. 44. Vasa M, Fichtlscherer S, Aicher A, et al. Number and migratory activity of circulating endothelial progenitor cells inversely correlate with risk factors for coronary artery disease. Circ Res. 2001;89(1):E1-7. 45. Yoder MC. Human endothelial progenitor cells. Cold Spring Harb Perspect Med. 2012;2(7):a006692. 46. Hirschi KK, Ingram DA, Yoder MC. Assessing identity, phenotype, and fate of endothelial progenitor cells. Arterioscler Thromb Vasc Biol. 2008;28(9):1584-1595. 47. Sieveking DP, Buckle A, Celermajer DS, Ng MKC. Strikingly different angiogenic properties of endothelial progenitor cell subpopulations. Insights from a novel human angiogenesis assay. J Am Coll Cardiol. 2008;51(6):660-668. 48. Melero-Martin JM, Khan ZA, Picard A, Wu X, Paruchuri S, Bischoff J. In vivo vasculogenic potential of human blood-derived endothelial progenitor cells. Blood. 2007;109(11):4761-4768. 49. Widemann A, Sabatier F, Arnaud L, et al. CD146-based immunomagnetic enrichment followed by multiparameter flow cytometry: a new approach to counting circulating

endothelial cells. J Thromb Haemost. 2008;6 (5):869-876. 50. Dignat-George F, Sampol J. Circulating endothelial cells in vascular disorders: new insights into an old concept. Eur J Haematol. 2000;65(4):215-220. 51. Shantsila E, Blann AD, Lip Gyh. Circulating endothelial cells: from bench to clinical practice. J Thromb Haemost. 2008;6(5):865-868. 52. Dignat-George F, Sampol J, Lip G, Blann AD. Circulating endothelial cells: realities and promises in vascular disorders. Pathophysiol Haemost Thromb. 2003;33(5-6):495-499. 53. Solovey AN, Gui L, Chang L, Enenstein J, Browne PV, Hebbel RP. Identification and functional assessment of endothelial P1H12. J Lab Clin Med. 2001;138(5):322-331. 54. Burger D, Touyz RM. Cellular biomarkers of endothelial health: microparticles, endothelial progenitor cells, and circulating endothelial cells. J Am Soc Hypertens. 2012;6(2):85-99. 55. Erdbruegger U, Haubitz M, Woywodt A. Circulating endothelial cells: a novel marker of endothelial damage. Clin Chim Acta. 2006;373(1-2):17-26. 56. Rowand JL, Martin G, Doyle GV, et al. Endothelial cells in peripheral blood of healthy subjects and patients with metastatic carcinomas. Cytometry A. 2007;71(2): 105-113. 57. Smirnov DA, Foulk BW, Doyle GV, Connelly MC, Terstappen LWMM, O’Hara SM. Global gene expression profiling of circulating endothelial cells in patients with metastatic carcinomas. Cancer Res. 2006;66(6):2918-2922. 58. Boos CJ, Lip GYH, Blann AD. Circulating endothelial cells in cardiovascular disease. J Am Coll Cardiol. 2006;48(8):1538-1547. 59. Werner N, Kosiol S, Schiegl T, et al. Circulating endothelial progenitor cells and cardiovascular outcomes. N Engl J Med. 2005;353(10):999-1007. 60. Peters K. Molecular basis of endothelial dysfunction in sepsis. Cardiovasc Res. 2003;60(1):49-57. 61. Arica DA, Akşan B, Örem A, Altinkaynak BA, Yayli S, Sönmez M. High levels of endothelial progenitor cells and circulating endothelial cells in patients with Behçet’s disease and their relationship to disease activity. An Bras Dermatol. 2019;94(3):320326. 62. Landray MJ, Wheeler DC, Lip GYH, et al. Inflammation, endothelial dysfunction, and platelet activation in patients with chronic kidney disease: the chronic renal impairment in Birmingham (CRIB) study. Am J Kidney Dis. 2004;43(2):244-253. 63. Almici C, Skert C, Bruno B, et al. Circulating endothelial cell count: a reliable marker of endothelial damage in patients undergoing hematopoietic stem cell transplantation. Bone Marrow Transplant. 2017;52(12):16371642. 64. Bertolini F, Shaked Y, Mancuso P, Kerbel RS. The multifaceted circulating endothelial cell in cancer: towards marker and target identification. Nat Rev Cancer. 2006;6(11):835845. 65. DePrimo SE, Bello C. Surrogate biomarkers in evaluating response to anti-angiogenic agents: focus on sunitinib. Ann Oncol. 2007;18 Suppl 10:x11-19. 66. Fürstenberger G, von Moos R, Lucas R, et al. Circulating endothelial cells and angiogenic serum factors during neoadjuvant chemotherapy of primary breast cancer. Br J Cancer. 2006;94(4):524-531. 67. Beerepoot LV, Mehra N, Vermaat JSP, Zonnenberg BA, Gebbink MFGB, Voest EE.

haematologica | 2021; 106(11)

Role of mutated endothelial cells in MPN

Increased levels of viable circulating endothelial cells are an indicator of progressive disease in cancer patients. Ann Oncol. 2004;15(1):139-145. 68. Woywodt A, Scheer J, Hambach L, et al. Circulating endothelial cells as a marker of endothelial damage in allogeneic hematopoietic stem cell transplantation. Blood. 2004;103(9):3603-3605. 69. Treliński J, Wierzbowska A, Krawczyńska A, et al. Circulating endothelial cells in essential thrombocythemia and polycythemia vera: correlation with JAK2-V617F mutational status, angiogenic factors and coagulation activation markers. Int J Hematol. 2010;91(5):792-798. 70. Sozer S, Wang X, Zhang W, et al. Circulating angiogenic monocyte progenitor cells are reduced in JAK2V617F high allele burden myeloproliferative disorders. Blood Cells Mol Dis. 2008;41(3):284-291. 71. Rosti V, Bonetti E, Bergamaschi G, et al. High frequency of endothelial colony forming cells marks a non-active myeloproliferative neoplasm with high risk of splanchnic vein thrombosis. PLoS One. 2010;5(12): e15277. 72. Cao N, Yao Z-X. The hemangioblast:from concept to authentication. Anat Rec (Hoboken). 2011;294(4):580-588. 73. Hirschi KK. Hemogenic endothelium during development and beyond. Blood. 2012;119 (21):4823-4827. 74. Ueno H, Weissman IL. Clonal analysis of mouse development reveals a polyclonal origin for yolk sac blood islands. Dev Cell. 2006;11(4):519-533. 75. Weng W, Sukowati EW, Sheng G. On hemangioblasts in chicken. PLoS One. 2007;2(11):e1228. 76. Vogeli KM, Jin S-W, Martin GR, Stainier DYR. A common progenitor for haematopoietic and endothelial lineages in the zebrafish gastrula. Nature. 2006;443 (7109):337-339. 77. Lee JD, Treisman JE. Sightless has homology to transmembrane acyltransferases and is required to generate active Hedgehog protein. Curr Biol. 2001;11(14):1147-1152. 78. Kinder SJ, Tsang TE, Quinlan GA, Hadjantonakis AK, Nagy A, Tam PP. The orderly allocation of mesodermal cells to the extraembryonic structures and the anteroposterior axis during gastrulation of the mouse embryo. Development. 1999;126 (21):4691-4701.

haematologica | 2021; 106(11)

79. Furuta C, Ema H, Takayanagi S-I, et al. Discordant developmental waves of angioblasts and hemangioblasts in the early gastrulating mouse embryo. Development. 2006;133(14):2771-2779. 80. Farina M, Bernardi S, Polverelli N, et al. Comparative somatic mutational profiling of CD34+ hematopoietic precursors (HSC) and circulating endothelial cells (CEC) in patients with primary myelofibrosis (PMF). Blood. 2019;134(Suppl_1):1684. 81. Teofili L, Larocca LM. Blood and endothelial cells: together through thick and thin. Blood. 2013;121(2):248-249. 82. Etheridge SL, Roh ME, Cosgrove ME, et al. JAK2V617F-positive endothelial cells contribute to clotting abnormalities in myeloproliferative neoplasms. Proc Natl Acad Sci U S A. 2014;111(6):2295-2300. 83. Poisson J, Tanguy M, Davy H, et al. Erythrocyte-derived microvesicles induce arterial spasms in JAK2V617F myeloproliferative neoplasm. J Clin Invest. 2020;130 (5):2630-2643. 84. Castiglione M, Jiang YP, Mazzeo C, et al. Endothelial JAK2V617F mutation leads to thrombosis, vasculopathy, and cardiomyopathy in a murine model of myeloproliferative neoplasm. J Thromb Haemost. 2020;18(12):3359-3370. 85. Lin CHS, Kaushansky K, Zhan H. JAK2V617F-mutant vascular niche contributes to JAK2V617F clonal expansion in myeloproliferative neoplasms. Blood Cells Mol Dis. 2016;62:42-48. 86. Zhan H, Kaushansky K. Functional interdependence of hematopoietic stem cells and their niche in oncogene promotion of myeloproliferative neoplasms: the 159th biomedical version of “it takes two to tango.” Exp Hematol. 2019;70:24-30. 87. Lundberg P, Karow A, Nienhold R, et al. Clonal evolution and clinical correlates of somatic mutations in myeloproliferative neoplasms. Blood. 2014;123(14):2220-2228. 88. Lin CHS, Zhang Y, Kaushansky K, Zhan H. JAK2V617F-bearing vascular niche enhances malignant hematopoietic regeneration following radiation injury. Haematologica. 2018;103(7):1160-1168. 89. De Palma M, Venneri MA, Galli R, et al. Tie2 identifies a hematopoietic lineage of proangiogenic monocytes required for tumor vessel formation and a mesenchymal population of pericyte progenitors. Cancer Cell.

2005;8(3):211-226. 90. Murray PDF. The development in vitro of the blood of the early chick embryo. Proc R Soc Lond B. 1932;111(773):497-521. 91. Sabin FR. Preliminary note on the differentiation of angioblasts and the method by which they produce blood-vessels, bloodplasma and red blood-ells as seen in the living chick. J Hematother Stem Cell Res. 2002;11(1):5-7. 92. Choi K, Kennedy M, Kazarov A, Papadimitriou JC, Keller G. A common precursor for hematopoietic and endothelial cells. Development. 1998;125(4):725-732. 93. Nishikawa S-I, Nishikawa S, Kawamoto H, et al. In vitro generation of lymphohematopoietic cells from endothelial cells purified from murine embryos. Immunity. 1998;8(6):761-769. 94. Lancrin C, Sroczynska P, Stephenson C, Allen T, Kouskoff V, Lacaud G. The haemangioblast generates haematopoietic cells through a haemogenic endothelium stage. Nature. 2009;457(7231):892-895. 95. Dzierzak E, Bigas A. Blood development: hematopoietic stem cell dependence and independence. Cell Stem Cell. 2018;22(5): 639-651. 96. Zambidis ET, Peault B, Park TS, Bunz F, Civin CI. Hematopoietic differentiation of human embryonic stem cells progresses through sequential hematoendothelial, primitive, and definitive stages resembling human yolk sac development. Blood. 2005;106(3):860-870. 97. Eilken HM, Nishikawa S-I, Schroeder T. Continuous single-cell imaging of blood generation from haemogenic endothelium. Nature. 2009;457(7231):896-900. 98. Helman R, Pereira W de O, Marti LC, et al. Granulocyte whole exome sequencing and endothelial JAK2V617F in patients with JAK2V617F positive Budd-Chiari syndrome without myeloproliferative neoplasm. Br J Haematol. 2018;180(3):443-445. 99. Williams N, Lee J, Moore L, et al. Phylogenetic reconstruction of myeloproliferative neoplasm reveals very early origins and lifelong evolution. bioRxiv 2020;2020: 374710. 100. Van Egeren D, Escabi J, Nguyen M, et al. Reconstructing the lineage histories and differentiation trajectories of individual cancer cells in myeloproliferative neoplasms. Cell Stem Cell. 2021;28(3):514-523.

2823

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(11)2824-2833

Acute Lymphoblastic Leukemia

Increments in DNA-thioguanine level during thiopurine-enhanced maintenance therapy of acute lymphoblastic leukemia Rikke Hebo Larsen,1 Cecilie Utke Rank,1,2 Kathrine Grell,1,3 Lisbeth Nørgaard Møller,1 Ulrik Malthe Overgaard,2 Peter Kampmann,2 Jacob Nersting,1 Matilda Degn,1 Stine Nygaard Nielsen,1 Helle Holst,1 Birgitte Klug Albertsen,4,5 Peder Skov Wehner,6 Michael Thude Callesen,6 Jukka Kanerva,7 Thomas Leth Frandsen,1 Bodil Als-Nielsen,1 Lisa Lyngsie Hjalgrim1 and Kjeld Schmiegelow1,8 Department of Pediatrics and Adolescent Medicine, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark; 2Department of Hematology, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark; 3Section of Biostatistics, Department of Public Health, University of Copenhagen, Copenhagen, Denmark; 4Department of Pediatrics and Adolescent Medicine, Aarhus University, Aarhus, Denmark; 5Department of Clinical Medicine, Aarhus University, Aarhus, Denmark; 6Department of Pediatric Hematology and Oncology, H.C. Andersen Children’s Hospital, Odense University Hospital, Odense, Denmark; 7HUS, Helsinki University Hospital, University of Helsinki, New Children’s Hospital, Helsinki, Finland and 8Institute of Clinical Medicine, Faculty of Medicine, University of Copenhagen, Copenhagen, Denmark 1

ABSTRACT

Correspondence: KJELD SCHMIEGELOW kschmiegelow@rh.dk Received: December 14, 2020. Accepted: April 28, 2021. Pre-published: May 27, 2021. https://doi.org/10.3324/haematol.2020.278166

2824

aintenance therapy containing methotrexate and 6-mercaptopurine is essential to cure acute lymphoblastic leukemia (ALL). Cytotoxicity is elicited by incorporation of thioguanine nucleotides into DNA (DNA-TG), and higher leukocyte DNA-TG is associated with increased relapse-free survival. As 6-thioguanine provides 6fold higher cytosolic levels of thioguanine nucleotides than does 6mercaptopurine, we added low-dose 6-thioguanine to methotrexate/6mercaptopurine maintenance therapy to explore if this combination results in significantly higher DNA-TG. The target population of the “Thiopurine Enhanced ALL Maintenance therapy” (TEAM) study was 30 patients with non-high-risk ALL, aged 1-45 years on methotrexate/6-mercaptopurine maintenance therapy receiving no other systemic chemotherapy. Incremental doses of 6-thioguanine were added to methotrexate/6-mercaptopurine maintenance therapy (starting 6-thioguanine dose: 2.5 mg/m2/day, maximum: 12.5 mg/m2/day). The primary endpoint was DNA-TG increments. Thirty-four patients were included, and 30 patients completed maintenance therapy according to the TEAM strategy. Of these 30 patients, 26 (87%) tolerated 10.0-12.5 mg/m2/day as the maximum 6-thioguanine dose. TEAM resulted in significantly higher DNA-TG levels compared to those in both TEAM patients before their inclusion in TEAM (on average 251 fmol/mg DNA higher [95% confidence interval: 160-341; P<0.0001]), and with historical patients receiving standard methotrexate/6-mercaptopurine maintenance therapy (on average 272 fmol/mg DNA higher [95% confidence interval: 147-398; P<0.0001]). TEAM did not increase myelotoxicity or hepatotoxicity. In conclusion, TEAM is an innovative and feasible approach to improve maintenance therapy and results in higher DNA-TG levels without inducing additional toxicity. It may therefore be an effective strategy to reduce the risk of ALL relapse through increased DNA-TG. This will be tested in a randomized ALLTogether-1 substudy.

Introduction Overall survival of patients with acute lymphoblastic leukemia (ALL) has improved immensely and now surpasses 90% in children and 70% in adults.1-4 However, cure rates after relapse have remained poor, and sustained effort to improve first-line ALL therapy further is crucial.5,6 Methotrexate/6-mercapto-

haematologica | 2021; 106(11)

TEAM: thiopurine-enhanced ALL maintenance therapy

purine based maintenance therapy is considered one of the most important phases of ALL therapy, but with an unmet need for improved treatment strategies.7 Dose adjustments are guided by white blood cell count or absolute neutrophil count, which have been related to relapse rates.7-9 However, both white blood cell and absolute neutrophil counts show pronounced variation with age, gender, circadian rhythms and ethnicity, and are therefore poor indicators of methotrexate/6-mercaptopurine treatment intensity.10-12 Cytotoxicity of 6mercaptopurine is mediated by thioguanine nucleotides,13 which mimic guanine nucleotides, and compete with these for incorporation into DNA. Incorporated thioguanine nucleotides (DNA-TG) can, when methylated, mismatch with thymidine (instead of cytosine), which activates the mismatch repair system. Methyl-thioguanine nucleotides will however continue to mismatch and this ultimately leads to cell death due to repetitive but futile activation of the mismatch repair system.13 Higher DNATG levels in circulating normal leukocytes during maintenance therapy were recently shown to be associated with reduced relapse rate,14 a finding that has subsequently been validated (Toksvang et al. in press). This suggests that DNA-TG levels in normal leukocytes are indicative of pharmacological events in malignant lymphoblasts. Adjustment of maintenance therapy to increase DNA-TG may therefore lead to improved relapse-free survival. However, attempts to obtain higher DNA-TG by solely incrementing methotrexate/6-mercaptopurine doses would generally fail, because of the complex pharmacokinetics of 6-mercaptopurine, and, furthermore, may increase the risk of serious myelotoxicity and hepatotoxicity.13,15-18 6-Thioguanine (6TG) also exerts cytotoxicity through formation of DNA-TG, leading to 6-7 times higher levels of cytosolic thioguanine nucleotides than those derived from 6-mercaptopurine at equipotent doses.16 Randomized studies have, however, demonstrated no benefit in overall survival, when 6-mercaptopurine is replaced by 6TG for maintenance therapy.19-22 This might reflect the inability of 6TG to generate higher DNA-TG than 6-mercaptopurine as other 6-mercaptopurine metabolites inhibit de novo purine synthesis, which promotes DNA-TG formation by reducing the levels of natural purines to compete with thioguanine nucleotides for DNA incorporation.11,13,14 As an alternative, we tested addition of very low 6TG doses to standard methotrexate/6-mercaptopurine maintenance therapy and explored whether this combination could achieve significantly higher DNA-TG levels without inducing additional toxicity.

captopurine maintenance therapy in the ALL2008 maintenance therapy substudy (ALL2008 MT substudy).14 Eligible patients were included, when they reached maintenance therapy phase II (maintenance-II) (Online Supplementary Appendix S1). Patients were included during the entire course of maintenance-II but had to have at least 3 months of remaining therapy at the first visit. An exclusion criterion was end-of-induction minimal residual disease-negative bone marrow, because DNA-TG levels are not associated with relapse risk in such patients.14 Patients with previous sinusoidal obstruction syndrome were also excluded, because of the association of this syndrome with 6TG at higher doses (Online Supplementary Appendix S1).19 The diagnosis of sinusoidal obstruction syndrome was made as defined by Schmiegelow et al.25 The TEAM study was registered at ClinicalTrials.gov: NCT02912676 and granted a EudraCT number: 2014-002248-42. It was approved by the Ethical Committee of the Capital Region of Denmark (H-3-2014-098), and the Danish Medicines Agency (2014-002248-42). The study was conducted according to the Declaration of Helsinki II and Good Clinical Practice guidelines (Online Supplementary Appendix S1).

Maintenance therapy in the NOPHO ALL2008 protocol Maintenance therapy in the NOPHO ALL2008 protocol was divided into two phases, maintenance-I and maintenance-II with weekly oral methotrexate and daily oral 6-mercaptopurine constituting the backbone of both phases and targeted to a white blood cell count of 1.5-3.0 x109/L. All therapy was discontinued 2.5 years after diagnosis. Patients with standard-risk ALL continued directly from maintenance-I to maintenance-II, whereas patients with intermediate-risk ALL received 6 weeks of delayed intensification before entering maintenance-II. Maintenance-I contained additional systemic chemotherapy as well as intrathecal chemotherapy, and patients with intermediate-risk ALL also received intrathecal chemotherapy during maintenance-II (Online Supplementary Appendix S1).

Outline of the trial therapy Incremental doses of 6TG were added to standard methotrexate/6-mercaptopurine maintenance therapy starting at a dose of 2.5 mg/m2/day and increased by 2.5 mg/m2/day biweekly until reaching a maximum 6TG dose of 12.5 mg/m2/day. The target level of DNA-TG was above 500 fmol/mg DNA (approximate mean DNA-TG at end of methotrexate/6-mercaptopurine-based maintenance therapy14). TEAM patients had the same white blood cell target of 1.5–3.0 x109/L as all other NOPHO ALL2008 patients. If DNA-TG and/or white blood cell count targets were not reached at the maximum 6TG dose (i.e., 12.5 mg/m2/day), 6mercaptopurine and/or methotrexate doses were adjusted (Online Supplementary Appendix S1).

Metabolite assessments Methods Study population and study design Patients aged 1-45 years with non-high-risk ALL (i.e., standard and intermediate risk) treated according to the Nordic Society of Pediatric Hematology and Oncology (NOPHO) ALL2008 protocol23,24 were eligible for the Thiopurine Enhanced ALL Maintenance therapy (TEAM) study. The TEAM study was designed as a phase I/II non-randomized clinical trial with increments in DNA-TG as the primary efficacy outcome. The DNA-TG levels documented were compared with DNA-TG levels from TEAM patients before TEAM and with historical patients who had received standard methotrexate/6-mer-

haematologica | 2021; 106(11)

DNA-TG in circulating normal leukocytes, thioguanine nucleotide level in erythrocytes (Ery-TGN) and erythrocyte level of methylated 6-mercaptopurine metabolites (Ery-MeMP) were quantified as previously described by Jacobsen et al.26 and Shipkova et al.27 (Online Supplementary Appendix S1).

Data analysis and statistics For each patient, median values were calculated for all outcome variables. Differences between patients’ medians of all outcomes during TEAM therapy and before TEAM followed a normal distribution and were compared using paired Welch t-tests. The patients’ medians of outcome variables during TEAM therapy and the historical data from the ALL2008 MT substudy had

2825

R.H. Larsen et al.

normal distributions, except for Ery-MeMP, and were compared using Welch two-sample t-tests, except for Ery-MeMP which was analyzed with a two-sample Wilcoxon rank sum test.

Results Study population The target study population of the TEAM study was 30 patients based on a power calculation (provided in Online Supplementary Appendix S2). A total of 34 patients were included in the TEAM study. The patients’ characteristics and demographics are summarized in Table 1. Inclusion was completed in December 2018, and the last TEAM patient finished therapy in March 2020. The study was closed for follow-up on August 31, 2020. The TEAM therapy period for data analysis was defined as the time period between 10 weeks after initial 6TG dosing (when patients would have had sufficient time to reach the potential maximum 6TG dose of 12.5 mg/m2/day) until discontinuation of TEAM therapy. The time period before TEAM was defined as 2 months prior to initiation of 6TG treatment. The historical data from the ALL2008 MT substudy (patients who received standard methotrexate/6-mercaptopurine maintenance therapy) included measurements from 10 weeks after the start of maintenance-II until discontinuation of antileukemic therapy. Of the total 34 patients, 32 patients received TEAM maintenance therapy for more than 10 weeks (Figure 1). Measurements from these 32 patients (denoted TEAM study population) in the period

from 10 weeks after initial 6TG dosing until cessation of antileukemic therapy (time period denoted TEAM therapy) are included in analyses comparing outcome variables during TEAM therapy with either before TEAM therapy or historical patients receiving standard methotrexate/6-mercaptopurine maintenance therapy. Two patients discontinued TEAM study participation after having received more than 10 weeks of TEAM maintenance therapy, as described Table 1. Demographics and patients’ characteristics.

Characteristic

Value

Sex, n (%) Male Female Age group, n (%) Children (<18 years at diagnosis) Adults (≥18 years at diagnosis) Age (years) at ALL diagnosis Median (range) Immunophenotype, n (%) Precursor B-cell T-cell WBC at ALL diagnosis, x109/L Mean (range) Risk group stratification (day 79), n (%) Standard risk Intermediate risk

22 (65) 12 (35) 29 (85) 5 (15) 3 (1–34) 31 (91) 3 (9) 30.4 (1.6 – 317) 16 (47) 18 (53)

ALL: acute lymphoblastic leukemia; WBC: white blood cell count.

Figure 1. Trial profile. Thirty-two patients received more than 10 weeks of TEAM maintenance therapy, and these patients constitute the TEAM study population. Data from these patients were used for comparison of outcome variables with historical patients (who received standard methotrexate/6-mercaptopurine maintenance therapy), and with TEAM patients before TEAM inclusion. The backgrounds for discontinuation of study participation (n=4 patients) are described in the manuscript. TEAM: thiopurine-enhanced ALL maintenance therapy.

2826

haematologica | 2021; 106(11)

TEAM: thiopurine-enhanced ALL maintenance therapy

in “Cessation of TEAM participation” below. The trial profile is illustrated in Figure 1. Of the total 34 patients, four patients discontinued study participation (see “Cessation of TEAM participation”). The remaining 30 patients completed maintenance-II according to the TEAM strategy. Of these, five patients received 3-6 months of TEAM maintenance therapy (from initiation of 6TG therapy until discontinuation of antileukemic therapy), 17 patients received 7-12 months, and eight patients received more than 12 months of TEAM maintenance therapy. A median of 27 DNA-TG measurements were available per patient (range, 8-45) in the period from initiation of 6TG until discontinuation of all therapy with an average of 1.8 samples (range, 0.7-3.0) per month. Of the 34 included patients, 32 were TPMT wild-type, when tested for the G460A and A719G variants, while two patients were heterozygous with one low activity TPMT variant. The median follow-up for all patients (from discontinuation of antileukemic therapy until August 31, 2020) was 20.9 months (interquartile range [IQR]: 14.4-29.3).

DNA-thioguaninine nucleotide levels A total of 645 DNA-TG measurements taken during TEAM therapy were available for analysis. The mean of the patients’ DNA-TG medians during TEAM therapy was 764 fmol/mg DNA (IQR, 577-890) and the patients’ medians during TEAM therapy varied 5-fold (Figure 2). DNA-TG medians from TEAM patients before their inclusion in the TEAM protocol are presented in Online Supplementary Appendix S2 and Online Supplementary Figure S1. DNA-TG levels (i.e., patients’ medians) were on average 272 fmol/mg DNA higher during TEAM therapy (95% confidence interval [95% CI]: 147-398; P<0.0001) than historical data from the ALL2008 MT substudy, in which patients received standard methotrexate/6-mercaptopurine maintenance therapy (Figure 3; Tables 2 and 3).14 If these results are entered into the regression model from the ALL2008 MT substudy concerning estimation of reduction in relapse haz-

ard rate with increasing DNA-TG levels,14 the DNA-TG increment with the TEAM strategy would have corresponded to a 59% reduction in the hazard rate of relapse (1[0.722.72]=59%). Furthermore, the average of DNA-TG measurements at the end of maintenance-II was approximately 500 fmol/mg DNA in the ALL2008 MT substudy,14 whereas TEAM patients had an average DNA-TG level of 933 fmol/mg DNA in the last month of maintenance-II. The number of patients and samples for the various outcomes from the ALL2008 MT substudy are provided in Online Supplementary Appendix S2. DNA-TG levels during TEAM therapy were on average 251 fmol/µg DNA higher (95% CI: 160-341; P<0.0001) than those in TEAM patients before participation in the TEAM study. Of the 32 patients who received more than 10 weeks of TEAM maintenance therapy, 27 patients obtained the DNA-TG target above 500 fmol/mg DNA during TEAM therapy. Five patients did not reach the target DNA-TG level. A detailed description of these patients is provided in Online Supplementary Appendix S2. During the entire course of the TEAM study, two patients experienced an episode of three consecutive measurements of DNA-TG above 1500 fmol/mg DNA, which is greater than the 99th percentile of the ALL2008 MT substudy DNATG distribution,14 without excessive myelotoxicity or hepatotoxicity.

Toxicities Treatment according to the TEAM strategy did not lead to increased myelotoxicity or hepatotoxicity, when compared with data from the ALL2008 MT substudy or when compared with data before TEAM.

Myelosuppression No significant differences were observed when white blood cell count, absolute neutrophil count and hemoglobin level were compared between the period of TEAM therapy and before TEAM (P=0.78, P=0.40, and P=0.10, respectively) (Figure 4A, B; Tables 2 and 3). No serious infections were

Figure 2. Median DNA-TG from each TEAM patient during TEAM therapy. Median DNA-TG (fmol/µg DNA) and interquartile range for each TEAM patient during TEAM therapy. A total of 32 patients received TEAM maintenance therapy for more than 10 weeks (denoted TEAM therapy). TEAM: thiopurine-enhanced ALL maintenance therapy; DNA-TG: thioguanine nucleotides incorporated into DNA.

haematologica | 2021; 106(11)

2827

R.H. Larsen et al.

observed. The platelet count was on average 19x109/L higher (95% CI: 3-35; P=0.02) during TEAM therapy than before TEAM (Figure 4C). No significant differences were observed, when white blood cell count, absolute neutrophil count and platelet count during TEAM therapy were compared with historical data from the ALL2008 MT substudy (P=0.09, P=0.07, and P=0.18, respectively) (Tables 2 and 3).

Hepatotoxicity No hepatic serious adverse events were observed, including no cases of sinusoidal obstruction syndrome. No signif-

icant differences were observed when the levels of alanine aminotransferase, coagulation factors II, VII and X, International Normalized Ratio (INR) and bilirubin were compared between the period of TEAM therapy and before TEAM (P=0.23, P=0.42, P=0.06, and P=0.95, respectively) (Figure 4D; Tables 2 and 3). Prior to inclusion in the TEAM protocol, one patient had experienced pronounced symptoms of 6-mercaptopurineinduced hypoglycemia, which disappeared during TEAM therapy. Another patient, prior to TEAM, had experienced recurrent severe hepatoxicity with alanine aminotransferase levels ranging between 522-7,470 U/L (i.e. 166 times

Figure 3. DNA-TG during TEAM therapy. DNA-TG (fmol/mg DNA) during TEAM therapy (red spline function line) compared with DNA-TG levels from the ALL2008 MT substudy (green spline function line),14 in which patients received standard methotrexate/6-mercaptopurine maintenance therapy. Each dot refers to one DNA-TG measurement during TEAM therapy (gray dots for adults). The mean of the DNA-TG patients’ medians from TEAM patients before their inclusion in TEAM is marked by a blue bar on the Y-axis. TEAM: thiopurine-enhanced ALL maintenance therapy; DNA-TG: thioguanine nucleotides incorporated into DNA.

Table 2. Mean of patients’ medians and 95% reference range for all outcomes. 6-mercaptopurine dose, mg/m2/day Methotrexate dose, mg/m2/week DNA-TG, fmol/mg Ery-TGN, nmol/mmol hgb Ery-MeMP, nmol/mmol hgb White blood cell count, x109/L Absolute neutrophil count, x109/L Platelet count, x109/L Hemoglobin, mmol/L Alanine aminotransferase, U/L Coagulation factors II-VII-X, IU/L International Normalized Ratio. Bilirubin, mmol/L

Before TEAM

TEAM therapy

ALL2008 MT substudy

53 (6–121) 19 (4–41) 530 (157–1,279) 240 (100–485) 8,462 (177–21,520) 3.1 (1.9–5.7) 1.8 (0.7–3.7) 247 (108–371) 7.6 (6.3–9.3) 139 (26–373) 0.7 (0.4–0.9) 1.2 (1.0–1.6) 11 (4–32)

45 (7– 79) 19 (5–38) 764 (273–1,402) 721 (339–1,396) 5,931 (142–14,385) 3.2 (2.2–5.5) 1.9 (1.0– 3.7) 261 (56–383) 7.7 (6.3–9.0) 118 (20–265) 0.7 (0.5–0.9) 1.2 (1.1 – 1.3) 12 (5– 29)

– – 492 (21–1,104) 231 (8–608) 12,032 (4,577–17,383)* 2.9 (1.7–4.0) 1.6 (0.5–2.7) 236 (117–359) – – – – –

Data before TEAM, during TEAM therapy and from the ALL2008 MT substudy (in which patients received standard methotrexate/6-mercaptopurine maintenance therapy14). Data in this table from “Before TEAM” and from “TEAM therapy” are from the 32 patients comprising the TEAM study population (i.e., patients who received more than 10 weeks of TEAM maintenance therapy). DNA-TG: level of thioguanine nucleotides incorporated into DNA; Ery-TGN: thioguanine nucleotide level in erythrocytes; Ery-MeMP: methylated mercaptopurine metabolite level in erythrocytes; – data not applicable. *Ery-MeMP from the ALL2008 MT substudy is reported as the median of the patients’ medians and interquartile range, as medians for this outcome were not normally distributed.

2828

haematologica | 2021; 106(11)

TEAM: thiopurine-enhanced ALL maintenance therapy

(mg/m2/week) were compared between TEAM therapy and before TEAM, (P=0.09 and P=0.99, respectively) (Tables 2 and 3).

upper normal level) and coagulation factors II, VII and X <0.5 IU/L causing repeated treatment interruptions. During TEAM therapy, no treatment interruptions occurred due to hepatotoxicity. Four patients developed symptoms of osteonecrosis after inclusion in the TEAM study.

6-Mercaptopurine/methotrexate metabolite levels Ery-TGN levels were on average 470 nmol/mmol hemoglobin higher (95% CI: 349–590; P<0.0001) during TEAM therapy than before TEAM (Online Supplementary Figure S2A), and on average 490 nmol/mmol hemoglobin higher (95% CI: 365-614; P<0.0001) when compared with data from the ALL2008 MT substudy (Tables 2 and 3) A plot of median Ery-TGN levels during TEAM therapy in relation to median DNA-TG level during TEAM therapy is provided in Online Supplementary Figure S3. Ery-MeMP tended to be lower during TEAM therapy than before TEAM, on average 1,948 nmol/mmol hemoglobin lower (95% CI: -4011 to 115, P=0.06), (Online Supplementary Figure S2B). Ery-MeMP was significantly lower during TEAM therapy when compared with data from the ALL2008 MT substudy (P=0.0001) with a difference in the median of patients’ levels of 5,363 nmol/mmol hemoglobin (Tables 2 and 3).

Maximum tolerated 6-thioguanine doses There was no apparent dose-response relation between maximum tolerated 6TG dose and median DNA-TG level during TEAM therapy (Figure 5). Of the 30 patients who completed therapy according to the TEAM strategy, 24 tolerated the maximum 6TG dose of 12.5 mg/m2/day. Two patients tolerated 5 mg/m2/day, two patients 7.5 mg/m2/day, and one patient 10 mg/m2/day as their maximum 6TG dosage, because higher 6TG doses led to recurring leukopenia/neutropenia. By decision of the treating physician, one patient received 10.0 mg/m2/day as maximum 6TG dose, as the white blood cell count therapy target was fulfilled with a median DNA-TG of 852 fmol/mg DNA during TEAM therapy, and tolerance of 12.5 mg/m2/day 6TG was never tested. Two patients were heterozygous with one low activity TPMT variant, and both tolerated 12.5 mg/m2/day 6TG as their maximum 6TG dosage with no unacceptable toxicities. The median DNA-TG level was 580 fmol/µg DNA and 1,349 fmol/mg DNA during 9 and 16 months of TEAM therapy for these two patients, respectively.

Cessation of TEAM participation Of the four patients who discontinued study participation, three discontinued by decision of their parents and/or treating physician, and one patient due to on-therapy leukemic relapse. The background for study discontinuation in one patient was recurring mild hyperbilirubinemia (1.5 times upper normal limit), with no other signs of sinusoidal obstruction syndrome leading to repeated therapy interruptions. The patient’s parents

Methotrexate and 6-mercaptopurine doses No significant difference was observed when 6-mercaptopurine dose (mg/m2/day) and methotrexate dose

Figure 4. Hematologic and hepatic parameters during TEAM therapy versus before TEAM. (A-D) Hematologic and hepatic parameters during TEAM therapy versus before TEAM. Each circle marks an individual patient’s median value during TEAM therapy versus before TEAM. The black line represents the diagonal. (A) Median white blood cell count (x109/L). (B) Median absolute neutrophil count (x109/L). (C) Median platelet count (x109/L). (D) Median alanine aminotransferase concentration (U/L). TEAM: thiopurine-enhanced ALL maintenance therapy; WBC: white blood cell count; ANC: absolute neutrophil count; TBC: thrombocyte count; ALAT: alanine aminotranscerase.

haematologica | 2021; 106(11)

2829

R.H. Larsen et al.

decided to opt out of TEAM, and the patient was switched to standard methotrexate/6-mercaptopurine maintenance therapy (during which therapy interruptions are not warranted until bilirubin level exceeds 3 times the upper limit of normal [NOPHO ALL2008 protocol guideline]). The patient continued to demonstrate mild hyperbilirubinemia on conventional methotrexate/6-mercaptopurine maintenance therapy for the remaining 11 months of antileukemic therapy. The background to withdrawal from the study in the second patient was recurring anemia without concurrent leukopenia or

thrombocytopenia. When this patient developed symptoms of osteonecrosis the parents took the patient off TEAM. The third patient who discontinued participation in the TEAM study had recurring elevated levels of alanine aminotransferase and P-ferritin reflecting previous multiple red blood cell transfusions. This patient had elevated alanine aminotransferase levels even at the starting 6TG dose of 2.5 mg/m2/day, but no other signs of liver dysfunction. The patient’s parents decided to opt out of TEAM after 2 months, as the elevated liver enzyme levels led to therapy interruptions (in the NOPHO ALL2008

Table 3. Results of comparison of outcomes. TEAM therapy versus before TEAM and TEAM therapy versus data from the ALL2008 MT substudy: mean difference and 95% confidence interval for all outcomes.

TEAM therapy versus before TEAM P-value 6-mercaptopurine dose, mg/m /day Methotrexate dose, mg/m2/week DNA-TG, fmol/mg Ery-TGN, nmol/mmol hb Ery-MeMP, nmol/mmol hb White blood cell count, x109/L Absolute neutrophil count, x109/L Platelet count, x109/L Hemoglobin, mmol/L Alanine aminotransferase, U/L Coagulation factors II-VII-X, IU/L International Normalized Ratio Bilirubin, mmol/L 2

-9 (-19 to 1) -0.02 (-2.8 to 2.7) 251 (160 to 341) 470 (349 to 590) -1948 (-4,011 to 115) 0.05 (-0.31 to 0.41) 0.13 (-0.19 to 0.45) 19 (2.7 to 35) 0.17 (-0.04 to 0.4) -13 (-35 to 9) 0.02 (-0.04 to 0.08) -0.08 (-0.15 to 0.002) 0.06 (-1.8 to 1.9)

P=0.09* P=0.99* P<0.0001* P<0.0001* P=0.06* P=0.78* P=0.40* P=0.02* P=0.10* P=0.23* P=0.42* P=0.06* P=0.95*

TEAM therapy versus ALL2008 MT sub-study P-value – – 272 (147 to 398) 490 (365 to 614) – 0.3 (-0.05 to 0.65) 0.3 (-0.02 to 0.57) 24 (-12 to 60) – – – – –

– – P<0.0001† P<0.0001† P=0.0001‡ P=0.09† P=0.07† P=0.18† – – – – –

Results of comparison of outcomes during TEAM therapy versus before TEAM, and TEAM therapy versus historical data from the ALL2008 MT sub-study (in which patients received standard methotrexate/6-mercaptopurine maintenance therapy14). Positive values indicate higher values during TEAM therapy, and negative values indicate lower values during TEAM therapy. For comparisons of all outcomes, data from the 32 patients comprising the TEAM study population were used (i.e., from patients who received more than 10 weeks of TEAM maintenance therapy). *Paired t-test. †Two-sample t-test. ‡Wilcoxon signed rank test. DNA-TG: level of thioguanine nucleotides incorporated into DNA; Ery-TGN: thioguanine nucleotide level in erythrocytes; Hb: hemoglobin; Ery-MeMP: methylated mercaptopurine metabolite level in erythrocytes; – data not applicable.

Figure 5. Median DNA-TG during TEAM therapy in relation to 6-thioguanine dose intensity. Median DNA-TG level (fmol/mg DNA) during TEAM therapy from the 30 patients, who completed therapy according to the TEAM strategy (blue dots represent individual patients) in relation to their individual maximum tolerated dose of 6-thioguanine. The black line indicates the TEAM therapy DNA-TG target of 500 fmol/mg DNA. TEAM: thiopurine-enhanced ALL maintenance therapy; DNA-TG: thioguanine nucleotides incorporated into DNA; 6TG: 6-thioguanine.

2830

haematologica | 2021; 106(11)

TEAM: thiopurine-enhanced ALL maintenance therapy

protocol elevated alanine aminotransferase does not in itself warrant therapy interruptions unless accompanied by an INR >1.5 or coagulation factors II, VII and X <0.5 IU/L). After discontinuation of 6TG, the patient continued to demonstrate elevated alanine aminotransferase levels on conventional methotrexate/6-mercaptopurine therapy. Two TEAM patients have been diagnosed with ALL relapse. Both patients had B-cell precursor ALL and were both initially assigned to intermediate-risk therapy due to a poor minimal residual disease response in their end-ofinduction bone-marrow evaluation. The original clone was identified at relapse in both patients. One patient was diagnosed with on-therapy isolated central nervous system relapse, after having received 4 months of TEAM therapy with a median DNA-TG during TEAM therapy of 1,561 fmol/mg DNA. The other patient was diagnosed with isolated bone-marrow relapse 18 months after completion of antileukemic therapy, having received a total of 11 months of TEAM maintenance therapy. Further evaluations of the leukemic clone at relapse revealed an ABLlike fusion, which was not identified at initial ALL diagnosis. Re-examination of the leukemic clone from the initial diagnosis confirmed that this aberration had already been present, which could have had an impact on the development of the relapse.

Discussion The TEAM strategy was demonstrated to be safe and to result in significantly higher DNA-TG levels compared with those occurring during standard methotrexate/6mercaptopurine maintenance therapy. The estimated reduction in relapse hazard rate of 59% on the TEAM protocol is theoretical, since the present study was not powered to address survival. However, the increase in DNA-TG on TEAM was significant from both clinical and statistical points of view. The TEAM study therefore introduces a completely novel dosing strategy for methotrexate/thiopurine-based maintenance therapy and could potentially lead to a significant improvement in relapse-free survival through increased DNA-TG.14 6TG is more easily converted into thioguanine nucleotides than 6-mercaptopurine and leads to 6- to 7fold higher Ery-TGN levels compared with 6-mercaptopurine at equipotent doses.13,16 Furthermore, methotrexate and methylated 6-mercaptopurine metabolites inhibit de novo purine synthesis, and concomitant administration therefore increases DNA-TG formation due to a reduced level of natural guanine to compete with thioguanine nucleotides for incorporation into DNA.11,13,28 Based on this synergy of cytotoxic mechanisms, the results of the TEAM study are, even with the very low 6TG dose, well explained. In contrast, attempts to increase DNA-TG by solely incrementing 6-mercaptopurine dose would primarily lead to higher levels of methylated 6-mercaptopurine metabolites rather than thioguanine nucleotides thus shifting cytotoxicity toward inhibition of de novo purine synthesis, but not necessarily higher DNA-TG.13,16 Accordingly, 6-mercaptopurine dose during maintenance therapy is not associated with DNA-TG level,14 and 6-mercaptopurine dose increments primarily increase the risk of significant hepatotoxicity and may even increase relapse risk.17,18 Of haematologica | 2021; 106(11)

note, the 6TG dose in TEAM did not appear to be associated with median DNA-TG level during TEAM therapy and, as with 6-mercaptopurine,14 the DNA-TG level cannot be predicted by 6TG dose intensity. Upon incorporation of thioguanine nucleotides into DNA, the DNA-TG will undergo random methylation, which will favor mismatching between methyl-thioguanine nucleotides and thymidine. This leads to activation of the mismatch repair system attempting to correct the mismatch.13 However, methyl-thioguanine nucleotides will continue to mismatch, which ultimately results in cell death due to repetitive but futile activation of the mismatch repair system.13 Tolerance of the increased DNA-TG levels in TEAM most likely reflects the fact that most of the thioguanine nucleotides in DNA are unmethylated and thus do not mismatch. One of the two patients who was diagnosed with ALL relapse had a median DNA-TG level of 1,561 fmol/mg DNA during TEAM therapy, which could reflect the presence of an MSH6 deletion causing thiopurine resistance.29 However, this was never evaluated, as the patient was referred for hematopoietic stem cell transplantation. The maximum 6TG dose in the TEAM study was preset at 12.5 mg/m2/day, as this would correspond to ~75 mg/m2 6-mercatopurine (the standard dose) with respect to Ery-TGN levels.13,16 A meta-analysis including three randomized clinical trials comparing methotrexate/6mercaptopurine and methotrexate/6TG based maintenance therapy demonstrated no difference in overall survival between recipients of these two therapy strategies.19-22 Although the latter will cause higher cytosolic levels of thioguanine nucleotides, the lack of concomitant inhibition of de novo purine synthesis mediated by methylated 6-mercaptopurine might explain these findings.13,16,1922 However, studies comparing DNA-TG levels during treatment with either 6-mercaptopurine or 6TG as the single thiopurine are lacking. Methotrexate/6TG-based maintenance therapy has previously been associated with an increased risk of sinusoidal obstruction syndrome.19 This syndrome was neither found nor expected among TEAM patients, as a recent systematic review showed that 6TG therapy was not associated with a risk of sinusoidal obstruction syndrome at 6TG doses below ~12.5 mg/m2/day.30 Higher Ery-MeMP has been associated with risk of hepatotoxicity, including a rise in aminotransferase levels18 and risk of hypoglycemia.31,32 The low Ery-MeMP in TEAM can therefore explain why alanine aminotransferase levels were generally low in the TEAM study, thus highlighting the TEAM strategy as a highly relevant alternative for patients with pronounced 6-mercaptopurine toxicity on conventional methotrexate/6-mercaptopurine maintenance therapy. As an alternative allopurinol has been used to ameliorate this propensity to hepatotoxicity as it shifts patients to a TPMT low-activity phenotype.33 However, TPMT heterozygocity is not associated with higher DNA-TG.34 In conclusion, DNA-TG level is a composite measure of upstream 6-mercaptopurine, methotrexate; 6TG metabolites and is associated with relapse risk,14 and TEAM may be an innovative and feasible approach to improve maintenance therapy by leading to higher DNA-TG levels. TEAM therefore represents a potentially effective strategy for reducing risk of ALL relapse without inducing additional toxicity. This will be tested in a randomized 2831

R.H. Larsen et al.

ALLTogether-1 substudy (EudraCT number: 2018001795-38). Disclosures KS has received speaker and/or advisory board honoraria from Jazz Pharmaceuticals and Servier; speaker fees from Amgen and Medscape; and an educational grant from Servier. JK has received consulting fees from Bayer. The authors declare that they have no other competing financial interests. The funding and supporting sources listed had no involvement in or restrictions regarding publication. Contributions RHL collected, analyzed, and interpreted data, served as principal investigator for the childhood cases and wrote and edited the manuscript; KS served as sponsor for the TEAM investigation, designed the study, interpreted data, and critically reviewed the manuscript; CUR served as the principal investigator for adults, collected and analyzed data, and critically reviewed the manuscript; LH, BAN, TLF, BKA, PSW, MTC, SNN, HH, and JK served as investigators for childhood cases, collected data and critically reviewed the manuscript; PK and UMO served as investigators for adult cases, collected data and critically reviewed the manuscript; JN and MD performed the analysis and quantification of DNA-TG, Ery-TGN and Ery-MeMP and critically

References 1. Toft N, Birgens H, Abrahamsson J, et al. Results of NOPHO ALL2008 treatment for patients aged 1-45 years with acute lymphoblastic leukemia. Leukemia. 2018;32(3): 606-615. 2. Hunger SP, Mullighan CG. Acute lymphoblastic leukemia in children. N Engl J Med. 2015;373(16):1541-1552. 3. Pieters R, de Groot-Kruseman H, Van der Velden V, et al. Successful therapy reduction and intensification for childhood acute lymphoblastic leukemia based on minimal residual disease monitoring: study ALL10 from the Dutch Childhood Oncology Group. J Clin Oncol. 2016;34(22):2591-2601. 4. Dores GM, Devesa SS, Curtis RE, Linet MS, Morton LM. Acute leukemia incidence and patient survival among children and adults in the United States, 2001-2007. Blood. 2012;119(1):34-43. 5. Malempati S, Gaynon PS, Sather H, La MK, Stork LC. Outcome after relapse among children with standard-risk acute lymphoblastic leukemia: Children's Oncology Group study CCG-1952. J Clin Oncol. 2007;25(36):58005807. 6. Oskarsson T, Soderhall S, Arvidson J, et al. Relapsed childhood acute lymphoblastic leukemia in the Nordic countries: prognostic factors, treatment and outcome. Haematologica. 2016;101(1):68-76. 7. Schmiegelow K, Nielsen SN, Frandsen TL, Nersting J. Mercaptopurine/methotrexate maintenance therapy of childhood acute lymphoblastic leukemia: clinical facts and fiction. J Pediatr Hematol Oncol. 2014;36(7): 503-517. 8. Arico M, Baruchel A, Bertrand Y, et al. The seventh international childhood acute lymphoblastic leukemia workshop report: Palermo, Italy, January 29-30, 2005. Leukemia. 2005;19(7):1145-1152. 9. Schmiegelow K, Nersting J, Nielsen SN, et al. Maintenance therapy of childhood acute

2832

reviewed the manuscript; LNM performed the statistical analyses, created the figures and critically reviewed the manuscript; KG critically reviewed the statistical analyses and the manuscript. All authors approved the final manuscript. Acknowledgments The authors thank the dedicated staff at the laboratory of Pediatric Oncology, Bonkolab, Copenhagen, Denmark for their valuable work with special recognition of the laboratory technicians and project nurses involved. Funding This work was supported by research grants from the Danish Cancer Society, Childhood Cancer Foundation (Denmark), Childhood Cancer Foundation (Sweden), Nordic Cancer Union, Otto Christensen Foundation, The Capital Region of Denmark, and The Research Foundation of Rigshospitalet, University of Copenhagen. This work is part of Childhood Oncology Network Targeting Research, Organization & Life expectancy (CONTROL), supported by the Danish Cancer Society (R-257-A14720) and the Danish Childhood Cancer Foundation (2019-5934). Data-sharing statement For original data, please contact Kjeld Schmiegelow: e-mail: kschmiegelow@rh.dk

lymphoblastic leukemia revisited - should drug doses be adjusted by white blood cell, neutrophil, or lymphocyte counts? Pediatr Blood Cancer. 2016;63(12):2104-2111. 10. Haus E, Smolensky MH. Biologic rhythms in the immune system. Chronobiol Int. 1999;16(5):581-622. 11. Nielsen SN, Grell K, Nersting J, Frandsen TL, Hjalgrim LL, Schmiegelow K. Measures of 6mercaptopurine and methotrexate maintenance therapy intensity in childhood acute lymphoblastic leukemia. Cancer Chemother Pharmacol. 2016;78(5):983-994. 12. Haddy TB, Rana SR, Castro O. Benign ethnic neutropenia: what is a normal absolute neutrophil count? J Lab Clin Med. 1999;133(1): 15-22. 13. Karran P, Attard N. Thiopurines in current medical practice: molecular mechanisms and contributions to therapy-related cancer. Nat Rev Cancer. 2008;8(1):24-36. 14. Nielsen SN, Grell K, Nersting J, et al. DNAthioguanine nucleotide concentration and relapse-free survival during maintenance therapy of childhood acute lymphoblastic leukaemia (NOPHO ALL2008): a prospective substudy of a phase 3 trial. Lancet Oncol. 2017;18(4):515-524. 15. Schmiegelow K, Forestier E, Kristinsson J, et al. Thiopurine methyltransferase activity is related to the risk of relapse of childhood acute lymphoblastic leukemia: results from the NOPHO ALL-92 study. Leukemia. 2009;23(3):557-564. 16. Erb N, Harms DO, Janka-Schaub G. Pharmacokinetics and metabolism of thiopurines in children with acute lymphoblastic leukemia receiving 6-thioguanine versus 6mercaptopurine. Cancer Chemother Pharmacol. 1998;42(4):266-272. 17. Schmiegelow K, Bjork O, Glomstein A, et al. Intensification of mercaptopurine/ methotrexate maintenance chemotherapy may increase the risk of relapse for some children with acute lymphoblastic leukemia. J Clin Oncol. 2003;21(7):1332-1339. 18. Nygaard U, Toft N, Schmiegelow K.

Methylated metabolites of 6-mercaptopurine are associated with hepatotoxicity. Clin Pharmacol Ther. 2004;75(4):274-281. 19. Escherich G, Richards S, Stork LC, Vora AJ. Meta-analysis of randomised trials comparing thiopurines in childhood acute lymphoblastic leukaemia. Leukemia. 2011;25(6): 953-959. 20. Vora A, Mitchell CD, Lennard L, et al. Toxicity and efficacy of 6-thioguanine versus 6-mercaptopurine in childhood lymphoblastic leukaemia: a randomised trial. Lancet. 2006;368(9544):1339-1348. 21. Harms DO, Gobel U, Spaar HJ, et al. Thioguanine offers no advantage over mercaptopurine in maintenance treatment of childhood ALL: results of the randomized trial COALL-92. Blood. 2003;102(8):27362740. 22. Stork LC, Matloub Y, Broxson E, et al. Oral 6-mercaptopurine versus oral 6-thioguanine and veno-occlusive disease in children with standard-risk acute lymphoblastic leukemia: report of the Children's Oncology Group CCG-1952 clinical trial. Blood. 2010;115 (14):2740-2748. 23. Toft N, Birgens H, Abrahamsson J, et al. Risk group assignment differs for children and adults 1-45 yr with acute lymphoblastic leukemia treated by the NOPHO ALL-2008 protocol. Eur J Haematol. 2013;90(5):404412. 24. Frandsen TL, Heyman M, Abrahamsson J, et al. Complying with the European Clinical Trials directive while surviving the administrative pressure - an alternative approach to toxicity registration in a cancer trial. Eur J Cancer. 2014;50(2):251-259. 25. Schmiegelow K, Attarbaschi A, Barzilai S, et al. Consensus definitions of 14 severe acute toxic effects for childhood lymphoblastic leukaemia treatment: a Delphi consensus. Lancet Oncol. 201617(6):e231-e239. 26. Jacobsen JH, Schmiegelow K, Nersting J. Liquid chromatography-tandem mass spectrometry quantification of 6-thioguanine in DNA using endogenous guanine as internal

haematologica | 2021; 106(11)

TEAM: thiopurine-enhanced ALL maintenance therapy

standard. J Chromatogr B Analyt Technol Biomed Life Sci. 2012;881-882:115-118. 27. Shipkova M, Armstrong VW, Wieland E, Oellerich M. Differences in nucleotide hydrolysis contribute to the differences between erythrocyte 6-thioguanine nucleotide concentrations determined by two widely used methods. Clin Chem. 2003;49(2):260-268. 28. Chabner BA, Allegra CJ, Curt GA, et al. Polyglutamation of methotrexate. Is methotrexate a prodrug? J Clin Invest. 1985;76(3):907-912. 29. Evensen NA, Madhusoodhan PP, Meyer J, et al. MSH6 haploinsufficiency at relapse contributes to the development of thiopurine resistance in pediatric B-lymphoblastic

haematologica | 2021; 106(11)

leukemia. Haematologica. 2018;103(5):830839. 30. Toksvang LN, Schmidt MS, Arup S, et al. Hepatotoxicity during 6-thioguanine treatment in inflammatory bowel disease and childhood acute lymphoblastic leukaemia: a systematic review. PLoS One. 2019;14(5): e0212157. 31. Halonen P, Salo MK, Makipernaa A. Fasting hypoglycemia is common during maintenance therapy for childhood acute lymphoblastic leukemia. J Pediatr. 2001;138(3): 428-431. 32. Melachuri S, Gandrud L, Bostrom B. The association between fasting hypoglycemia and methylated mercaptopurine metabolites in children with acute lymphoblastic

leukemia. Pediatr Blood Cancer. 2014;61(6): 1003-1006. 33. Cohen G, Cooper S, Sison EA, Annesley C, Bhuiyan M, Brown P. Allopurinol use during pediatric acute lymphoblastic leukemia maintenance therapy safely corrects skewed 6-mercaptopurine metabolism, improving inadequate myelosuppression and reducing gastrointestinal toxicity. Pediatr Blood Cancer. 2020;67(11):e28360. 34. Ebbesen MS, Nersting J, Jacobsen JH, et al. Incorporation of 6-thioguanine nucleotides into DNA during maintenance therapy of childhood acute lymphoblastic leukemiathe influence of thiopurine methyltransferase genotypes. J Clin Pharmacol. 2013;53 (6):670-674.

2833

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2834-2844

Chronic Lymphocytic Leukemia

Venetoclax plus bendamustine-rituximab or bendamustine-obinutuzumab in chronic lymphocytic leukemia: final results of a phase Ib study (GO28440) Stephan Stilgenbauer,1 Franck Morschhauser,2 Clemens-Martin Wendtner,3 Guillaume Cartron,4 Michael Hallek,5 Barbara Eichhorst,5 Mark F. Kozloff,6 Thomas Giever,7 Gerard Lozanski,8 Yanwen Jiang,9 Huang Huang,10 Daniela Soriano Pignataro,11 William Schary,12 Kathryn Humphrey,11 Mehrdad Mobasher9 and Gilles Salles13° 1

Department of Internal Medicine III, Ulm University, Ulm, Germany; 2University of Lille, Groupe de Recherche sur les Formes Injectables et les Technologies Associées, Lille, France; 3Munich Clinic Schwabing, Academic Teaching Hospital, Ludwig-MaximiliansUniversity (LMU), Munich, Germany; 4Department of Clinical Hematology, University Hospital of Montpellier, Montpellier, France; 5Department I of Internal Medicine, Center of Integrated Oncology Cologne-Bonn, University Hospital Cologne, Cologne, Germany; 6 Duchossois Center for Advanced Medicine, University of Chicago Medicine, Chicago, IL, USA; 7Department of Medicine, Medical College of Wisconsin, Milwaukee, WI, USA; 8 Department of Pathology, The Ohio State University, Columbus, OH, USA; 9Genentech, Inc., South San Francisco, CA, USA; 10F. Hoffmann-La Roche Ltd, Mississauga, Ontario, Canada; 11Roche Products Ltd, Welwyn Garden City, UK; 12AbbVie Inc, North Chicago, IL, USA and 13Hospices Civils de Lyon, Université de Lyon, Pierre-Bénite, France °Current address: Memorial Sloan Kettering Cancer Center, New York, NY, USA.

ABSTRACT

V Correspondence: STEPHAN STILGENBAUER stephan.stilgenbauer@uniklinik-ulm.de Received: June 9, 2020. Accepted: September 23, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.261107

2834

enetoclax (Ven), an orally administered, potent BCL-2 inhibitor, has demonstrated efficacy in chronic lymphocytic leukemia (CLL) in combination with rituximab (R) or obinutuzumab (G). Our aim was to investigate the addition of bendamustine (B) to these Ven-containing regimens in relapsed/refractory (R/R) or first-line (1L) CLL. This multi-arm, nonrandomized, open-label, phase Ib study was designed to evaluate the maximum tolerated dose (MTD) and safety/tolerability of Ven with BR/BG, with 3+3 dose-escalation followed by safety expansion. Patients received Ven (schedule A) or BR/BG first (schedule B) to compare safety and determine dose/schedule for expansion. Six Ven-BR/-BG cycles were to be administered, then Ven monotherapy until disease progression (R/R) or fixed-duration 1year treatment (1L). Overall, 33 R/R and 50 1L patients were enrolled. No dose-limiting toxicities were observed (doses 100–400 mg), and the MTD was not reached. Safety was similar between schedules; no tumor lysis syndrome occurred during dose-finding. Schedule B and Ven 400 mg were chosen for expansion. The most frequent grade 3–4 toxicity was neutropenia: R/R 64%, 1L Ven-BR 85%, 1L Ven-BG 55%. Grade 3–4 infection rate was: R/R 27%, 1L Ven-BR 0%, 1L Ven-BG 27%. During expansion, one clinical and two laboratory tumor lysis syndrome cases occurred. Fewer than half the patients completed six combination therapy cycles with all study drugs; rates of bendamustine discontinuation were high. Overall response rate was 91% in R/R and 100% in 1L patients (16 of 49 1L patients received Ven for >1 year). In conclusion, addition of bendamustine to Ven-R/-G increased toxicity without apparent efficacy benefit (clinicaltrial gov. Identifier: NCT01671904).

Introduction Treatment of chronic lymphocytic leukemia (CLL) has evolved in recent years, resulting in improved survival,1 with chemo-immunotherapy being the standardof-care over the past decade.2 However, agents targeting pathways involved in CLL cell proliferation and survival, such as B-cell receptor signaling3,4 and B-cell lymphoma-2 (BCL-2),5 are now standard treatment options.6–14 Combination of the selective selective BCL-2 inhibitor venetoclax (Ven) with

haematologica | 2021; 106(11)

Venetoclax with chemoimmunotherapy combinations

type II anti-CD20 antibody obinutuzumab (G; GA101) in a single-arm, phase Ib trial conferred high response rates with deep remission among patients with previously untreated (1L) or relapsed/refractory (R/R) CLL.13 Recently, Ven with rituximab (R) or G demonstrated impressive efficacy in phase III CLL trials,6-8 leading to approval of Ven-R and Ven-G in the R/R and 1L treatment settings, respectively.15,16 Bendabustine is an established chemotherapy agent in CLL that has shown clinical activity in combination with anti-CD20 antibodies.17-20 Whether the addition of a chemotherapy agent such as bendamustine to Ven-R or Ven-G could further improve outcomes in CLL has not yet been elucidated. The combination of Ven with bendamustine plus R (BR) produced greater growth inhibition in non-Hodgkin lymphoma (NHL) xenograft models than either Ven-R or BR alone,21 suggesting potential for increased clinical activity in B-cell malignancies. In addition, BCL-2 overexpression may be involved in resistance to the pro-apoptotic effects of chemoimmunotherapy, so the addition of Ven could overcome this and act as a chemosensitizer.22 We therefore evaluated a triplet combination of Ven with bendamustine and an anti-CD20 antibody (R or G) in 1L and R/R CLL.

Methods Study design and treatment This phase Ib, multi-arm, non-randomized, open-label study (clinicaltrials gov. Identifier: NCT01671904) was conducted at 11 sites across USA and Europe. Review boards at all institutions approved the protocol. Patients provided written informed consent. The study comprised two phases: dose-finding and safetyexpansion. Dose-finding, employing standard 3+3 dose escalation (Online Supplementary Table S1), was designed to include Ven doses from 100–600 mg daily with standard dose BR/BG (Figure 1; bendamustine: 90 mg/m2 (1L) or 70 mg/m2 (R/R) days (D)1–2 cycle (C)1–6; R: 375 mg/m2 D1 C1 then 500 mg/m2 D1 C2–6; G: 100 mg D1, 900 mg D2, 1000 mg D8 and D15 C1 then 1,000 mg D1 C2– 6). In order to mitigate tumor lysis syndrome (TLS) risk, Ven was initiated using a weekly ramp-up to target dose (Figure 1). TLS prophylaxis included hydration, a uric acid reducer, and hospitalization (Online Supplementary Table S2). Dose-finding compared two administration schedules for TLS risk mitigation during cycle 1 (Figure 1): schedule A (Ven ramp-up, followed by BR/BG) and schedule B (Ven introduced after 21-day BR/BG loading period). After each stage, an internal monitoring committee (IMC) and scientific overview committee (SOC) reviewed the data and provided dose/schedule recommendations for subsequent dose-finding stages and safety expansion (Online Supplementary Figure S1). Patients received six 28-day cycles of Ven-BR/-BG. R/R patients continued single-agent Ven until disease progression (PD), death, or unacceptable toxicity; 1L patients received 6 months of singleagent Ven for a total of 1-year treatment duration. Ven could be extended in 1L patients with detectable minimal residual disease (MRD) in bone marrow (BM) and/or partial response after 1 year of treatment. See the Online Supplementary Appendix for further details of study treatments and procedures.

Objectives Primary objectives were to identify the maximum tolerated dose (MTD) of Ven combined with BR, and evaluate safety/tolerability of Ven-BR in R/R and 1L CLL. Secondary objectives included efficacy evaluation of Ven-BR (including complete response

haematologica | 2021; 106(11)

[CR], overall response rate [ORR], duration of response, and progression-free survival [PFS]). Exploratory objectives were to determine the MTD of Ven with BG, safety and efficacy of Ven-BG, and the undetectable MRD (uMRD) rate with Ven-BR/-BG.

Patients Patients ≥18 years with a diagnosis of CLL according to the International Workshop on CLL (iwCLL) 2008 guidelines,23 in need of therapy, with an Eastern Cooperative Oncology Group (ECOG) performance status of 0–1 and adequate hematologic function (platelet count ≥75,000/mm3 or ≥30,000/mm3 if due to marrow involvement of CLL, and/or disease related immune thrombocytopenia; absolute neutrophil count ≥1,000/mm3; hemoglobin ≥9 g/dL) were eligible (Online Supplementary Table S3). Patients with R/R CLL must have received one to three prior lines of therapy.

Assessments Baseline molecular characteristics were assessed centrally (Online Supplementary Appendix). Safety and tolerability were assessed by incidence and type of dose-limiting toxicities (DLT) (Online Supplementary Table S4), AE (graded according to National Cancer Institute Common Terminology Criteria for Adverse Events v4.024), and serious AE, laboratory variables, and vital signs. TLS was classified according to Howard criteria.25 Efficacy was assessed by investigators according to iwCLL 2008 guidelines.23 Central MRD assessment was performed at Ohio State University, USA using five-colour flow cytometry according to the European Research Initiative on CLL principle.26

Statistical analyses Safety and efficacy analyses included all patients receiving ≥1 dose of any study drug. Peripheral blood (PB) MRD analyses reported landmark MRD rates after treatment completion in the 1L population and after antibody completion in the R/R population, given the difference in duration of treatment in these two populations. MRD analysis populations included all patients reaching the specified landmark time-point, plus those discontinuing the study earlier for AE, PD, or death. Given the long recruitment time, some patients had not reached the specific timepoint, therefore the intention-to-treat approach was not used. At each landmark assessment, the first evaluable PB MRD sample after the specified time-point was used. For BM MRD, due to limited sampling, best MRD response was reported, calculated in the efficacy population. Time-to-event analyses employed Kaplan–Meier methodology.27

Results Patients Thirty-three R/R and 50 1L patients were enrolled between January 2014 and June 2017 (Online Supplementary Figure S2). All R/R patients were enrolled in Ven-BR cohorts; 1L patients were enrolled in Ven-BR (n=27) or Ven-BG (n=23) cohorts. All 33 R/R patients were included in safety and efficacy analyses. One 1L patient enrolled in the Ven-BG cohort did not receive study drug, therefore 49 patients (27 Ven-BR, 22 Ven-BG) were included in analyses. Data cutoff was 17 August 2018; no further follow-up will be available. In R/R patients, median number of prior CLL therapies was 1 (range, 1–3); 79% had received fludarabine-based combinations and 6% Bruton’s tyrosine kinase inhibitors; none had received phosphoinositide-3-kinase inhibitors 2835

S. Stilgenbauer et al.

Table 1. Baseline characteristics.

Characteristic

Median age, years (range) Age ≤65 years, n (%) Male, n (%) ECOG PS 0–1, n (%) Rai stage, n (%) I II III IV Unknown Creatinine clearance <70 mL/min, n (%) Pre-treatment TLS risk, n (%) Low Medium High Cytogenetics, n (%)† del(17p) and/or TP53 mut‡ del(11q) Trisomy 12 No abnormalities del(13q) IGHV unmutated, n/N (%)§ Serum b -2 microglobulin, n/N (%) ≥3.5 mg/mL Prior therapies received, n (%) Fludarabine-based treatment Bendamustine or BR BTKis PI3Ki

R/R Ven-BR (n=33)*

1L Ven-BR (n=27)

1L Ven-BG (n=22)

62 (38–77) 22 (67) 20 (61) 32 (97)

65 (27–73) 15 (56) 14 (52) 27 (100)

64 (38–74) 12 (55) 15 (68) 22 (100)

5 (15) 2 (6) 7 (21) 15 (46) 4 (12) 8 (24)

2 (7) 4 (15) 8 (30) 5 (19) 8 (30) 11 (41)

4 (18) 6 (27) 4 (18) 0 8 (36) 5 (23)

7 (21) 16 (49) 10 (30)

2 (7) 19 (70) 6 (22)

7 (32) 10 (46) 5 (23)

13 (42) 6 (19) 4 (13) 2 (7) 6 (19) 20/30 (67) 8/16 (50)

2 (10) 2 (10) 3 (14) 1 (5) 13 (62) 10/20 (50) 5/8 (63)

3 (17) 3 (17) 2 (11) 1 (6) 9 (50) 15/21 (71) 5/8 (63)

26 (79) 4 (12) 2 (6) 0

0 0 0 0

*The Ven-BG cohort did not open in R/R patients. †FISH cutoffs for positivity: del(17p) >7%; del(11q) >6%; del(13q) >5.5%; trisomy 12 >2.5%. ‡A modified hierarchical model was used to maximize identification of the higher risk population due to missing samples for cytogenetic assessment. The del(17p)/TP53 mut subgroup included patients with a 17p deletion by FISH and/or TP53 mutation by NGS. §By NGS. Cutoff for positivity >5%. R/R: relapsed/refractory; Ven: venetoclax; B: bendamustine, R: rituximab; 1L: first-line, G: obinutuzumab; ECOG PS: Eastern Cooperative Oncology Group performance status; TLS: tumour lysis syndrome; mut: mutated; IGHV: immunoglobulin heavy-chain variable region, BTKi: Bruton’s tyrosine kinase inhibitor; PI3Ki: phosphoinositide 3-kinase inhibitor; FISH: fluorescence in situ hybridization; NGS: next-generation sequencing.

(Table 1). Among patients with available baseline samples for central testing, del(17p) and/or TP53 mutation was present in 42% of R/R and 13% of 1L patients; 67% of R/R and 60% of 1L patients had unmutated immunoglobulin heavy-chain variable region (IGHV) (Table 1).

Treatment exposure In total, 79% (26 of 33) of R/R, 96% (26 of 27) of 1L Ven-BR, and 95% (21 of 22) of 1L Ven-BG patients received Ven 400 mg. Fewer than half completed six cycles of the planned triple-drug combination (Online Supplementary Table S5). Overall, 16 R/R, 11 1L Ven-BR, and nine 1L Ven-BG patients completed six bendamustine cycles. R/R patients received a median of five (range, 1–6) bendamustine cycles and six (range, 1–6) cycles of R. 1L patients received a median of five (range, 1–6) bendamustine cycles; the median number of cycles of R or G received was six (range, 1–6). Median Ven treatment duration was 676 days (range, 48–1,649) in R/R, 371 days (range, 4–1,150) in 1L Ven-BR, and 336 days (range, 11–620) in 1L Ven-BG patients. Sixteen 1L patients, eight per arm, received Ven beyond 1 year (range, 381–1,150 days). Median relative dose intensity of Ven was 100% (range, 41–100) in R/R (n=30), 87% (range, 37–100) in 1L Ven-BR (n=26), and 100% 2836

(range, 33–100) in 1L Ven-BG (n=21) patients (Online Supplementary Appendix).

Safety During dose-finding, R/R patients were enrolled to 100mg (schedule A), 200-mg (schedule A), or 400-mg (schedule A or B) Ven cohorts; all 1L patients were included in the 400-mg (schedule A or B) Ven cohorts (Online Supplementary Table S6). The 600-mg Ven dose was not explored after review of the present study and programwide data, including phase Ib studies in CLL with Ven-R and Ven-G, where the recommended phase II dose of Ven was 400 mg.13,28 Eighteen patients received C1 treatment according to schedule A, and 21 patients according to schedule B (Online Supplementary Table S6). No DLT were observed with either treatment schedule or combination and the MTD was not reached in the doses explored. Ven 400 mg was selected as the recommended dose for expansion. There were no safety differences between schedules A and B and no TLS events were reported. After reviewing safety data from the dose-finding phase and programwide data, including from phase Ib studies with Ven-G13 and Ven-R in CLL,28 the IMC and SOC recommended schedule B (debulking with BR or BG followed by Ven), presuming increased practicality with mitigating risk of haematologica | 2021; 106(11)

Venetoclax with chemoimmunotherapy combinations

Figure 1. Treatment and dosing schedules. Schedule A: venetoclax followed by BR/BG. Schedule B: BR/BG followed by venetoclax. Schedule A with Ven-BR was explored in R/R patients before schedule B in the R/R and 1L populations. Data from schedule A provided safety guidance for subsequent dose-finding for patients in schedule B after a data review by the IMC and SOC. Venetoclax ramp-up: 3 weeks for the 100-mg cohort, 4 weeks for the 200-mg cohort, and 5 weeks for the 400-mg cohort; the treatment plan consisted of venetoclax plus BR or BG (6 x 28-day cycles) in combination with venetoclax, then single-agent venetoclax; each cohort continued treatment until PD, death, or unacceptable toxicity in R/R patients, or for a total of 1-year treatment duration in 1L patients (with potential for extension if BM was positive for MRD or patient had PR). Venetoclax ramp-up and maximum cohort dose are indicated by the blue arrows. BR/BG dosing schedule: bendamustine: 90 mg/m2 (1L) or 70 mg/m2 (R/R) D1–2 per cycle for six cycles; R: 375 mg/m2 (C1) then 500 mg/m2 (C2–6) D1 per cycle; G: 100 mg D1, 900 mg D2, 1,000 mg D8 and D15 C1 then 1,000 mg D1 (C2–6). Ven: venetoclax; B: bendamustine: R: rituximab; D: day; C: cycle; G: obinutuzumab; W: week; R/R: relapsed/refractory; 1L: first-line; IMC: internal monitoring committee; SOC: scientific overview committee; PD: disease progression; BM: bone marrow; MRD: minimal residual disease: PR: partial response.

TLS and a reduced number of high-risk TLS patients. All safety-evaluable patients reported ≥1 AE. All-grade infusion-related reaction events occurred in 12% of R/R and 45% of 1L patients (Table 2); all were grade 1–2 excepting for two grade 3 events with Ven-BG. Serious AE were reported in 52% of R/R and 53% of 1L patients (Online Supplementary Table S7); grade 3–4 AE occurred in 82% of R/R and 92% of 1L patients (Table 2). The most frequent grade 3–4 AE were neutropenia and thrombocytopenia (Table 2). Infections were mainly low grade and driven by upper respiratory tract and urinary tract infections (Table 2). Grade 3–4 infections occurred in 27% of R/R, 0% of 1L Ven-BR, and 27% of 1L Ven-BG patients. The frequency of grade 3–4 neutropenia was evenly dishaematologica | 2021; 106(11)

tributed between patients who received 1–4 or 5–6 bendamustine cycles (Online Supplementary Table S8). More grade 3–4 AE occurred during combination therapy versus monotherapy (Online Supplementary Table S9), with grade 3–4 neutropenia being reported in 65% of patients during combination therapy and 39% during monotherapy. Overall, 64%, 85%, and 59% of patients in the R/R, 1L Ven-BR, and 1L Ven-BG arms, respectively, received growth factors as prophylaxis and/or treatment (Online Supplementary Table S10); further details of growth factor treatment and response are not available. Three TLS cases were reported during safety expansion, all in patients receiving schedule B: one laboratory and one clinical TLS occurred in two R/R patients, whereas the 2837

S. Stilgenbauer et al. A

Figure 2. Minimal residual disease status by flow in peripheral blood and bone marrow in the (a) relapsed/refractory population and (b) untreated population. Discontinued includes patients who discontinued study due to PD, death, or AE (if applicable) before achieving the specified landmark time point; missing includes patients who reached the time-point but had no sample available for MRD analysis; undetermined includes patients with MRD level <10-4, but <200,000 leukocytes analyzed. uMRD: <1 CLL cell per 104 mononuclear cells; low-level MRD: ≥1 CLL cell per 104 mononuclear cells to <1 CLL cell per 104 mononuclear cells; high-level MRD: ≥1 CLL cell per 102 mononuclear cells. MRD: minimal residual disease; PB: peripheral blood; BM: bone marrow; R/R: relapsed/refractory; 1L first-line; R: rituximab; B: bendamustine; Ven: venetoclax; G: obinutuzumab; PD: disease progression; AE: adverse event; uMRD: undetectable MRD; CLL: chronic lymphocytic leukemia.

other laboratory TLS occurred in a 1L Ven-BG patient (Online Supplementary Table S11). Both laboratory TLS events occurred prior to initiation of Ven. The clinical TLS event occurred on day 29 after administration of Ven 50 mg (D1 of the second BR cycle); it was diagnosed due to clinical symptoms of hypotension and dyspnoea with hyperkalaemia (potassium: 7.6 mmol/L) and elevated phosphorus levels (2.97 mmol/L). Electrocardiogram data were not provided by the center. BR was permanently discontinued and single-agent Ven was re-introduced on study day 61 without further incidence of TLS. All TLS events resolved with standard-of-care measures; the two laboratory events did not lead to permanent discontinuation of any study drug. Bendamustine was permanently discontinued due to AE in 33% of R/R and 37% of 1L patients (Online Supplementary 2838

Table S12). The most common AE leading to permanent bendamustine withdrawal was neutropenia. Ven was interrupted and/or reduced due to AE in 67% of R/R and 82% of 1L patients, most frequently due to neutropenia (R/R, 36% of patients; 1L Ven-BR, 63%; 1L Ven-BG, 46%) (Online Supplementary Table S12). Ven was permanently discontinued due to AE in 27% of R/R and 29% of 1L patients (Online Supplementary Tables S12 and S13). Two deaths were reported: one due to stage 4, highgrade malignant hemangioendothelioma in a R/R patient, resulting in multiple organ failure on study day 144, and one due to hemorrhagic transformation of stroke on study day 83 in a 1L Ven-BR patient with history of hypertension and concomitant grade 3 treatmentrelated thrombocytopenia (platelet levels were normal at screening). haematologica | 2021; 106(11)

Venetoclax with chemoimmunotherapy combinations

Table 2. Treatment-emergent adverse events.

R/R Ven-BR (n=33)* All grades Grade 3–4

Any AE, n (%) 33 (100) AE occurring in ≥20% of patients, n (%)* Infections and infestations 28 (85) Neutropenia 21 (64) Nausea 18 (55) Thrombocytopenia 16 (49) Diarrhea 15 (46) Anemia 13 (39) Leukopenia 12 (36) Asthenia 12 (36) Pyrexia 11 (33) Fatigue 10 (30) Hypertension 7 (21) Cough 7 (21) Decreased appetite 7 (21) Dizziness 7 (21) Vomiting 6 (18) Rash 6 (18) Headache 5 (15) Constipation 5 (15) Infusion-related reaction 4 (12) Hyperuricemia 2 (6) Arthralgia 2 (6) Infection AE occurring in >5% of patients, n (%)* Bronchitis 10 (30) Urinary tract infection 8 (24) Nasopharyngitis 7 (21) Erysipelas 4 (12) Influenza 4 (12) Gastroenteritis 4 (12) Pneumonia 3 (9) Upper RTI 3 (9) Oral herpes 3 (9) Sinusitis 3 (9) Herpes zoster 3 (9) Conjunctivitis 2 (6) Herpes virus infection 1 (3) Pharyngitis 1 (3) Lung infection 0 Cytomegalovirus 0

27 (82)

1L Ven-BR (n=27) All grades Grade 3–4 27 (100)

25 (93)

1L Ven-BG (n=22) All grades Grade 3–4 22 (100)

20 (91)

9 (27) 21 (64) 0 8 (24) 6 (18) 4 (12) 7 (21) 1 (3) 0 3 (9) 4 (12) 0 0 0 0 0 0 0 0 1 (3) 0

20 (74) 24 (89) 18 (67) 16 (59) 13 (48) 13 (48) 5 (19) 8 (30) 13 (48) 8 (30) 1 (4) 6 (22) 3 (11) 2 (7) 7 (26) 7 (26) 8 (30) 8 (30) 8 (30) 4 (15) 3 (11)

0 23 (85) 2 (7) 10 (37) 1 (4) 3 (11) 3 (11) 1 (4) 0 0 0 0 0 0 1 (4) 0 0 0 0 2 (7) 0

16 (73) 12 (55) 16 (73) 15 (68) 11 (50) 8 (36) 2 (9) 9 (41) 8 (36) 5 (23) 3 (14) 3 (14) 7 (32) 0 4 (18) 5 (23) 7 (32) 4 (18) 14 (64) 5 (23) 5 (23)

6 (27) 12 (55) 0 11 (50) 2 (9) 2 (9) 2 (9) 0 1 (5) 1 (5) 1 (5) 0 0 0 1 (5) 0 0 0 2 (9) 1 (5) 1 (5)

1 (3) 2 (6) 0 1 (3) 0 1 (3) 2 (6) 0 0 0 0 0 1 (3) 0 0 0

3 (11) 4 (15) 6 (22) 0 1 (4) 0 1 (4) 3 (11) 2 (7) 1 (4) 1 (4) 2 (7) 2 (7) 3 (11) 2 (7) 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

2 (9) 4 (18) 4 (18) 0 1 (5) 1 (5) 0 5 (23) 2 (9) 2 (9) 0 0 0 1 (5) 0 2 (9)†

0 3 (14) 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Data include all investigator-reported adverse events (AE), regardless of relationship to study drug. AE occurring in ≥20% of patients are listed by MedDRA PT. Infection AE occurring in >5% patients are listed by MedDRA Systems Organ Class and PT. *Any-grade AE in any population or treatment arm. †Details of the diagnosis of these events were not reported, so it is not known whether they were symptomatic or detected via screening. R/R: relapsed/refractory; V: venetoclax; B: bendamustine; R: rituximab; 1L: first-line; G: obinutuzumab; RTI: respiratory tract infection; MedDRA: Medical Dictionary for Regulatory Activities; PT: preferred term; SOC: System Organ Class.

Efficacy Overall response rate was 91% (30 of 33) in R/R patients (including 42% [14 of 33] with complete response [CR] or CR with incomplete hematologic recovery [CRi]). All 1L patients responded (overall response rate 100%; including 44% [12 of 27] CR/CRi for Ven-BR, and 68% [15 of 22] CR/CRi for Ven-BG). Responses were similar regardless of cytogenetic status, IGHV status or whether patients received 1–4 or 5–6 bendamustine cycles (Table 3). Landmark PB uMRD rates were 58% (18 of 31) ≥12 months after last R dose in R/R patients, and 71% (15 of 21) and 89% (16 of 18) ≥3 months after last Ven dose with 1L Ven-BR and Ven-BG, respectively (Figure 2). These rates were observed regardless of whether patients received 1–4 or 5–6 bendamustine cycles and were maintained over time (Figure 2). haematologica | 2021; 106(11)

In the R/R population, after a median follow-up of 26 months (range, 24–31) from the last R dose, among patients who reached ≥24 months after the last R dose plus those who discontinued earlier, the PB uMRD rate was sustained at 37% (10 of 27). For the 1L population, ≥12 months after completion of all treatment (last Ven dose; median follow-up 14 months [range, 12–18]), the PB uMRD rate was sustained at 67% (12 of 18) for Ven-BR and 90% (9 of 10) for Ven-BG. 41–49% of patients had missing samples for BM MRD analysis. Among patients with samples available, the rate of uMRD as best MRD response in the BM was 53% (9 of 17), 69% (11 of 16), and 92% (11 of 12) in the R/R, 1L Ven-BR, and 1L Ven-BG arms, respectively. Among patients with PB and BM postbaseline paired samples from the same day, concordance between a patient’s MRD status determined from PB and 2839

S. Stilgenbauer et al.

BM was high and similar across the arms (Online Supplementary Table S14). MRD kinetics in individual patients are shown in Figure 3. Among 24 R/R patients who achieved PB uMRD in ≥1 assessment and had subsequent PB MRD assessment(s), nine converted to MRD positivity (low or high levels) in

two consecutive assessments, of whom three experienced PD as of the current follow-up. In 42 1L patients who achieved PB uMRD in ≥1 assessment and had subsequent PB MRD assessment(s), five converted to MRD positivity, only one of whom experienced PD. Median time to first MRD conversion (from the first PB uMRD result) was 360

Figure 3. Minimal residual disease kinetics in individual patients in the (a) relapsed/refractory population and (b) first-line population. Undetectable minimal residual disease (uMRD) was defined as <1 chronic lymphocytic leukemia (CLL) cell per 104 mononuclear cells in samples with a minimum of 200,000 leukocytes (<10-4). Low-level MRD was defined as between 1 CLL cell per 104 and 1 cell per 102 mononuclear cells (≥10-4–<10-2). High-level MRD was defined as ≥1 CLL cell per 102 mononuclear cells (≥10-2). MRD: minimal residual disease; R/R: relapsed/refractory; 1L: first-line; Ven: venetoclax; B: bendamustine; R: rituximab; PB: peripheral blood; BM: bone marrow; IGHV: immunoglobulin heavy-chain variable region; Tx: treatment; PD: disease progression; G: obinutuzumab; uMRD: undetectable MRD; CLL: chronic lymphocytic leukemia.

2840

haematologica | 2021; 106(11)

Venetoclax with chemoimmunotherapy combinations

days (range, 42–848) and 224 days (range, 77–600) in the R/R and 1L Ven-BR arms, respectively. For the patient in the 1L Ven-BG arm who experienced MRD conversion, time from first PB uMRD result to conversion was 324 days. PFS is shown in Figure 4. After a median observation time of 36.5 months (range, 1–54) and 21.6 months (range, 1–43) in the R/R and 1L populations, respectively, estimated 24-month PFS was 87% (95% Confidence Interval [CI]: 74–99) for R/R, 96% (95% CI: 89–100) for 1L Ven-BR, and 100% (95% CI: 100–100) for 1L Ven-BG patients. Seven PD, including two cases of Richter’s transformation to diffuse large B-cell lymphoma, were reported in the R/R population. Only one PD was observed with 1L Ven-BR. No PD were reported with 1L Ven-BG. Among the eight patients who progressed, four (all R/R) had baseline del(17p) and/or TP53 mutation.

Discussion In this phase Ib study, 400 mg Ven daily was selected, in combination with standard doses of BR/BG, for safety expansion in R/R and 1L CLL patients. While no MTD was reached during dose escalation up to 400 mg, both triplet combination regimens with bendamustine in the expansion phase showed increased toxicity versus Ven-R and Ven-G alone,6–8 leading to low tolerability, as seen with high rates of bendamustine discontinuation. The fact that fewer than half of all patients were able to complete the full six cycles of bendamustine (B) may at least partially explain the apparent lack of efficacy seen here in comparison with the efficacy of backbone Ven-R or Ven-G, despite the preclinical rationale. Neutropenia was the most important toxicity in triplet combinations in both 1L and R/R CLL populations. Rates of grade 3–4 neutropenia, however, were generally consistent with those observed with Ven-R in R/R CLL in the MURANO trial (58% [112 of 194]),7 with Ven-G in 1L CLL in CLL14 (53% [112 of 212]),8 and with ibrutinib plus BR

in R/R CLL in the HELIOS study (54% [154 of 287]).11,14 The small sample size in the 1L Ven-BR arm could have contributed to the numerically high rates of grade 3–4 neutropenia in this arm and limits the ability to make comparisons. Despite a high rate of neutropenia, infections were mainly low grade and driven by upper respiratory tract and urinary tract infections. In 1L populations, grade 3–4 infections appeared to be higher with Ven-BG (27% [6 of 22]) than Ven-BR (0% [0 of 27]), which is probably a reflection of the non-randomized nature of the study and the small numbers involved. However, a high rate of infections has been seen with BG combinations in CLL (21.4% grade ≥3 in the GREEN phase IIIb study),29 and in the randomized phase III GALLIUM study of BG versus BR in follicular lymphoma, grade 3–5 infection rates were also higher with BG than BR (26% vs. 20%).30 Grade 3–4 infection rates were higher in the present study than those seen in the CLL14 trial with Ven-G (17.5%)8 and in the CLL2-BAG trial during induction with G and Ven (19% [six of 31]).31 The grade 3–4 infection rate in the R/R population was higher than that seen with Ven-R in MURANO (18% [34 of 194]),7 but similar to the rate of grade ≥3 infection reported with ibrutinib plus BR in HELIOS (29% [83 of 287]).11 Direct between-study comparisons are difficult to interpret given differences in sample sizes, baseline characteristics, treatment duration, and follow-up. However, addition of bendamustine to Ven-R or Ven-G in the current study led to significantly increased infection rates and reduced the tolerability of these combinations. Increased toxicity with bendamustine-containing combinations with Ven is also seen in NHL,32,33 limiting the use of these combinations with Ven in future CLL treatment strategies. The higher CR rates observed with 1L Ven-BG than Ven-BR were consistent with previous demonstration of superior efficacy of G compared with R when used in combination with chemotherapy.34 Similarly, landmark (≥3 and ≥9 months after treatment) PB uMRD rates were higher in 1L Ven-BG than 1L and R/R Ven-BR patients. Around half of patients in all arms did not provide BM

Table 3. Response according to 2008 International Workshop on Chronic Lymphocytic Leukemia Guidelines in the untreated (1L) or relapsed/refractory populations.

Patients

Responses per cytogenetic status* Entire del(17p) and/or del(11q) Trisomy cohort TP53 mut 12

R/R patients, n (%) Ven-BR, n† 33 ORR 30 (91) CR/CRi 14 (42) PR 16 (49) 1L patients, n (%) Ven-BR, n 27 ORR 27 (100) CR/CRi 12 (44) PR 15 (56) Ven-BG, n 22 ORR 22 (100) CR/CRi 15 (68) PR 7 (32)

Responses per IGHV status

None

del(13q)

Responses per number of cycles of bendamustine received Mutated Unmutated 1–4 5–6

12 11 (85) 3 (23) 8 (62)

6 6 (100) 4 (67) 2 (33)

4 4 (100) 3 (75) 1 (25)

2 2 (100) 2 (100) 0

6 6 (100) 2 (33) 4 (67)

3 3 (100) 3 (100) 0

19 19 (100) 7 (37) 12 (63)

13 11 (85) 5 (39) 6 (46)

20 19 (95) 9 (45) 10 (50)

2 2 (100) 2 (100) 0 3 3 (100) 2 (67) 1 (33)

2 2 (100) 2 (100) 0 3 3 (100) 3 (100) 0

3 3 (100) 1 (33) 2 (67) 2 2 (100) 1 (50) 1 (50)

1 13 1 (100) 13 (100) 0 5 (39) 1 (100) 8 (62) 1 9 1 (100) 9 (100) 1 (100) 6 (67) 0 3 (33)

8 8 (100) 5 (63) 3 (38) 3 3 (100) 2 (67) 1 (33)

10 10 (100) 5 (50) 5 (50) 15 15 (100) 11 (73) 4 (27)

9 9 (100) 3 (33) 6 (67) 6 6 (100) 6 (67) 2 (33)

18 18 (100) 9 (50) 9 (50) 16 16 (100) 11 (69) 5 (31)

*Responses by cytogenetic abnormalities according to the modified hierarchical model in patients with samples available for cytogenetic assessment. †The Ven-BG cohort was not explored in R/R patients. CLL: chronic lymphocytic leukemia; R/R: relapsed/refractory; 1L: first-line; mut: mutated;Ven: venetoclax; B: bendamustine; R: rituximab ORR: overall response rate; CR: complete response; CRi :complete response with incomplete hematologic recovery; PR: partial response; G: obinutuzumab.

haematologica | 2021; 106(11)

2841

S. Stilgenbauer et al.

Figure 4. Progression-free survival. 1L: first-line; Ven: venetoclax; B: bendamustine; R: rituximab; G: obinutuzumab; R/R: relapsed/refractory.

samples for MRD assessment (an exploratory study endpoint), which affects the interpretability of BM uMRD rates. However, there was a high concordance between BM and PB results in paired samples. This was consistent with the concordance between PB and BM MRD determination seen with Ven treatment in the MURANO (86%) and CLL14 (87%) studies,6,35 suggesting that PB MRD status reflects BM MRD status in Ven-treated patients. High PB uMRD rates observed were consistent with those seen in phase II and III trials with Ven-R,6 Ven-G,8,35 and Ven plus ibrutinib36 and higher than reported previously with BG20 or BR alone in 1L patients,37 highlighting the efficacy of Ven-R and Ven-G alone, independent of the addition of bendamustine. However, cross-trial comparisons with this phase Ib study must be made with caution, and previous analyses were performed in the respective intention-totreat populations, as opposed to the subgroup of evaluable patients we had to use, and overall sample sizes were much larger. Re-emergence of MRD in PB, which was observed in a minority of patients, mainly in the R/R population, did not appear to be associated with immediate development of PD. Larger trials are needed to identify patients most likely to convert to MRD positivity, the impact and time from the MRD conversion on the appearance of clinical progression, and the feasibility of time-limited therapy in CLL. The 2-year PFS rate reported here in the R/R group with Ven-BR is similar to that seen with Ven-R in MURANO (85%),7 suggesting no benefit from the addition of bendamustine overall. In contrast, 2-year PFS in 1L patients was higher than that reported with Ven-G in CLL14 (88%);8 however, the two studies are very different and it should be noted that the CLL14 study recruited patients with co-existing conditions, whereas the population of the present study was relatively young, with a good performance status, and numbers were small. Also, 16 of 49 1L patients continued Ven beyond 1 year. 2842

A limitation of this study was the small sample size in each arm and non-randomized allocation to verify significant safety and efficacy differences across triplet combinations and across the number of bendamustine cycles received. In addition, most patients were <65 years old with creatinine clearance ≥70 mL/min, with a good performance status, which may have affected some of the study outcomes. 400 mg daily Ven given with standard-dose BR or BG increased toxicity compared with published data for VenR and Ven-G, without an apparent efficacy benefit for the addition of bendamustine. The question remains whether there is an optimal number of bendamustine cycles that would be beneficial for all or for a particular subgroup of CLL patients, or whether the addition of bendamustine (with a different dose/schedule) to Ven-R or Ven-G could improve clinical outcomes without impaired tolerability. However, considering the extent of the toxicity reported here, exploration of lower bendamustine doses is expected to have a limited role. Indeed, in the era of novel targeted agents replacing the standard chemo-immunotherapy regimens and demonstrating improved safety and efficacy profiles, there seems to be minimal need to add standard chemotherapy to novel regimens. Disclosures SS acts as a consultant, has received honoraria and research funding, sits on the board of directors or advisory committee of AbbVie, Gilead, GSK, Roche, Janssen, Novartis, Celgene, Amgen, Genentech; FM acts as a consultant for Epizyme, Roche/Genentech, Celgene, Gilead, has received honoraria from Roche/Genentech, Celgene, Gilead, BMS, Janssen, sits on the board of directors or advisory committee of Roche/Genentech, Celgene, Gilead, Servier, BMS, Janssen; CMW acts as a consultant, has received honoraria and research funding from Roche, Mundipharma, MorphoSys, Janssen, Gilead, AbbVie, Pharmacyclics, Genentech, GlaxoSmithKline, haematologica | 2021; 106(11)

Venetoclax with chemoimmunotherapy combinations

has received travel support from Roche, MorphoSys, Janssen, Gilead, AbbVie, Pharmacyclics, Genentech, GlaxoSmithKline; GC acts a consultant for Celgene, Roche, has received honoraria from Celgene, Sanofi, Roche, Janssen, Gilead; MH has received honoraria and research funding from AbbVie, Celgene, Gilead, Janssen, Mundipharma, Pharmacyclics, Roche; BE has received honoraria, research funding, and travel support from AbbVie, ArQule, BeiGene, Celgene, Gilead, Janssen, Mundipharma, Novartis, Roche; MFK acts as a consultant and sits on the board of directors or advisory committee of Genentech, Roche, AbbVie; TG has non conflicts of interest to disclose; GL has received research funding from Genentech, Stem Line, BI, Novartis, Beckman, Coulter; YJ and MM are employed by Genentech; YJ has equity ownership of Genentech; MM has ownership interests non-PLC in Roche; HH, DSP, and KH are employed by Roche; WS is employed by AbbVie; GS acts as a consultant for Novartis, Roche, has received honoraria from Novartis, Roche, Celgene, AbbVie, Acerta, Amgen, Epizyme, Gilead, Janssen, Merck, Morphosys, Pfizer, Servier, Takeda, BMS, sits on the advisory board of Celgene, Gilead, Janssen, Servier, BMS, and has received research funding from Roche, Celgene. Contributions SS, FM, C-MW, GC, MH, BE, MFK, GL, MM, and GS designed the study; SS, FM, C-MW, GC, MH, BE, MFK, TG, GL, and GS collected and assembled the data; SS, FM, C-MW, GC, MH, BE, MFK, GL, YJ, HH, DSP, WS, KH, MM, and

References 1. Gribben JG, O'Brien S. Update on therapy of chronic lymphocytic leukemia. J Clin Oncol. 2011;29(5):544-550. 2. Jain N, O'Brien S. Initial treatment of CLL: integrating biology and functional status. Blood. 2015;126(4):463-470. 3. O'Brien S, Furman RR, Coutre S, et al. Single-agent ibrutinib in treatment-naive and relapsed/refractory chronic lymphocytic leukemia: a 5-year experience. Blood. 2018;131(17):1910-1919. 4. Brown JR, Byrd JC, Coutre SE, et al. Idelalisib, an inhibitor of phosphatidylinositol 3-kinase p110delta, for relapsed/refractory chronic lymphocytic leukemia. Blood. 2014;123(22):3390-3397. 5. Stilgenbauer S, Eichhorst B, Schetelig J, et al. Venetoclax for patients with chronic lymphocytic leukemia with 17p deletion: results from the full population of a phase II pivotal trial. J Clin Oncol. 2018;36(19): 1973-1980. 6. Kater AP, Seymour JF, Hillmen P, et al. Fixed duration of venetoclax-rituximab in relapsed/refractory chronic lymphocytic leukemia eradicates minimal residual disease and prolongs survival: post-treatment follow-up of the MURANO phase III study. J Clin Oncol. 2019;37(4):269-277. 7. 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. 8. 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. 9. Burger JA, Sivina M, Jain N, et al. Randomized trial of ibrutinib vs ibrutinib plus rituximab in patients with chronic lymphocytic leukemia. Blood. 2019;

haematologica | 2021; 106(11)

GS analyzed the data and were involved in the writing process; SS, FM, C-MW, GC, MH, BE, MFK, TG, GL, YJ, HH, DSP, WS, KH, MM, and GS interpreted the data, participated in manuscript development, and gave final approval. Acknowledgments and Funding Special thanks go to the patients and their families, investigators, study coordinators, and support staff, and all GO28440 study team members. Venetoclax is being developed in collaboration between Genentech, Inc. and AbbVie. Genentech, Inc. and AbbVie provided financial support for the study and participated in the design, study conduct, and data analysis and interpretation. Third party medical writing and editorial assistance, under the direction of the authors, was provided by Kate Rijnen, a contract Medical Writer at Ashfield MedComms (Macclesfield, UK), an Ashfield Health company, and funded by F. HoffmannLa Roche Ltd. Data-sharing statement Qualified researchers may request access to individual patient level data through the clinical study data request platform (https://vivli.org/). Further details on Roche's criteria for eligible studies are available here (https://vivli.org/members/ourmembers/). For further details on Roche's Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see here (https://www.roche.com/ research_and_development/who_we_are_how_we_work/clinical_trials/our_commitment_to_data_sharing.htm

133(10):1011-1019. 10. 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. 11. 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):200-211. 12. Zelenetz AD, Barrientos JC, Brown JR, et al. Idelalisib or placebo in combination with bendamustine and rituximab in patients with relapsed or refractory chronic lymphocytic leukaemia: interim results from a phase 3, randomised, double-blind, placebo-controlled trial. Lancet Oncol. 2017;18(3):297-311. 13. Flinn IW, Gribben JG, Dyer MJS, et al. Phase 1b study of venetoclax-obinutuzumab in previously untreated and relapsed/refractory chronic lymphocytic leukemia. Blood. 2019;133(26):2765-2775. 14. Fraser G, Cramer P, Demirkan F, et al. Updated results from the phase 3 HELIOS study of ibrutinib, bendamustine, and rituximab in relapsed chronic lymphocytic leukemia/small lymphocytic lymphoma. Leukemia. 2019;33(4):969-980. 15. US Food and Drug Administration. FDA approves venetoclax for CLL and SLL. https://www.fda.gov/drugs/resourcesinformation-approved-drugs/fda-approvesvenetoclax-cll-and-sll. Accessed 22 May 2019. 16. US Food and Drug Administration. FDA approves venetoclax for CLL or SLL, with or without 17 p deletion, after one prior therapy. https://www.fda.gov/drugs/informationondrugs/approveddrugs/ucm61030

8.htm. Accessed 15 October 2018. 17. Brown JR, O'Brien S, Kingsley CD, et al. Obinutuzumab plus fludarabine/ cyclophosphamide or bendamustine in the initial therapy of CLL patients: the phase 1b GALTON trial. Blood. 2015;125(18): 2779-2785. 18. Flinn IW, Panayiotidis P, Afanasyev B, et al. A phase 2, multicenter study investigating ofatumumab and bendamustine combination in patients with untreated or relapsed CLL. Am J Hematol. 2016;91(9):900-906. 19. Michallet A-S, Aktan M, Hiddemann W, et al. Rituximab plus bendamustine or chlorambucil for chronic lymphocytic leukemia: primary analysis of the randomized, open-label MABLE study. Haematologica. 2018;103(4):698-706. 20. Stilgenbauer S, Leblond V, Foa R, et al. Obinutuzumab plus bendamustine in previously untreated patients with CLL: a subgroup analysis of the GREEN study. Leukemia. 2018;32(8):1778-1786. 21. Souers AJ, Leverson JD, Boghaert ER, et al. ABT-199, a potent and selective BCL-2 inhibitor, achieves antitumor activity while sparing platelets. Nat Med. 2013;19(2):202208. 22. Gibson CJ and Davids MS. BCL-2 antagonism to target the intrinsic mitochondrial pathway of apoptosis. Clin Cancer Res. 2015;21(22):5021-5029. 23. Hallek M, Cheson BD, Catovsky D, et al. Guidelines for the diagnosis and treatment of chronic lymphocytic leukemia: a report from the International Workshop on Chronic Lymphocytic Leukemia updating the National Cancer Institute-Working Group 1996 guidelines. Blood. 2008; 111(12):5446-5456. 24. Common Terminology Criteria for Adverse Events (CTCAE) Version 4.0. https://www.eortc.be/services/doc/ctc/CT

2843

S. Stilgenbauer et al.

CAE_4.03_2010-06-14_QuickReference_ 5x7.pdf. Accessed 30 January 2019. 25. Howard SC, Jones DP, Pui CH. The tumor lysis syndrome. N Engl J Med. 2011; 364(19):1844-1854. 26. Rawstron AC, Villamor N, Ritgen M, et al. International standardized approach for flow cytometric residual disease monitoring in chronic lymphocytic leukaemia. Leukemia. 2007;21(5):956-964. 27. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481. 28. Seymour JF, Ma S, Brander DM, et al. Venetoclax plus rituximab in relapsed or refractory chronic lymphocytic leukaemia: a phase 1b study. Lancet Oncol. 2017; 18(2):230-240. 29. LeBlond V, Aktan M, Ferra Coll CM, et al. Safety of obinutuzumab alone or combined with chemotherapy for previously untreated or relapsed/refractory chronic lymphocytic leukemia in the phase IIIb GREEN study. Haematologica. 2018;103(11):1889-1898. 30. Hiddemann W, Barbui AM, Canales MA, et

2844

al. Immunochemotherapy with obinutuzumab or rituximab for previously untreated follicular lymphoma in the GALLIUM Study: influence of chemotherapy on efficacy and safety. J Clin Oncol. 2018; 36(23):2395-2404. 31. Cramer P, von Tresckow J, Bahlo J, et al. Bendamustine followed by obinutuzumab and venetoclax in chronic lymphocytic leukaemia (CLL2-BAG): primary endpoint analysis of a multicentre, open-label, phase 2 trial. Lancet Oncol. 2018;19(9):1215-1228. 32. de Vos S, Swinnen LJ, Wang D, et al. Venetoclax, bendamustine, and rituximab in patients with relapsed or refractory NHL: a phase Ib dose-finding study. Ann Oncol. 2018;29(9):1932-1938. 33. Zinzani PL, Flinn IW, Yuen SLS, et al. Venetoclax–rituximab ± bendamustine vs bendamustine–rituximab in relapsed/refractory follicular lymphoma: CONTRALTO. Blood. 2020;136(23): 26282637. 34. Goede V, Fischer K, Busch R, et al. Obinutuzumab plus chlorambucil in

patients with CLL and coexisting conditions. N Engl J Med. 2014;370(12):11011110. 35. Fischer K, Ritgen M, Al-Sawaf O, et al. Quantitative analysis of minimal residual disease (MRD) shows high rates of undetectable MRD after fixed-duration chemotherapy-free treatment and serves as surrogate marker for progression-free survival: A prospective analysis of the randomized CLL14 trial. Blood. 2019; 134(suppl 1):36. 36. Hillmen P, Rawstron AC, Brock K, et al. Ibrutinib plus venetoclax in relapsed/refractory chronic lymphocytic leukemia: the CLARITY study. J Clin Oncol. 2019;37(30): 2722-2729. 37. Eichhorst B, Fink AM, Bahlo J, et al. First-line chemoimmunotherapy with bendamustine and rituximab versus fludarabine, cyclophosphamide, and rituximab in patients with advanced chronic lymphocytic leukaemia (CLL10): an international, openlabel, randomised, phase 3, non-inferiority trial. Lancet Oncol. 2016;17(7):928-942.

haematologica | 2021; 106(11)

ARTICLE

Chronic Lymphocytic Leukemia

Hodgkin lymphoma arising in patients with chronic lymphocytic leukemia: outcomes from a large multi-center collaboration Deborah M. Stephens,1 Ken Boucher,1 Elizabeth Kander,2 Sameer A. Parikh,3 Erin M. Parry,4 Mazyar Shadman,5 John M. Pagel,6 Jennifer Cooperrider,7 Joanna Rhodes,8 Anthony Mato,9 Allison Winter,10 Brian Hill,10 Sameh Gaballa,11 Alexey Danilov,12 Tycel Phillips,13 Danielle M. Brander,14 Sonali M. Smith,7 Matthew S. Davids,4 Kerry Rogers,2 Martha J. Glenn1 and John C. Byrd2 1

Division of Hematology and Hematologic Malignancies, University of Utah, Salt Lake City, UT; 2Division of Hematology, Ohio State University, Columbus, OH; 3Division of Hematology, Mayo Clinic, Rochester, MN; 4Division of Hematology, Dana Farber, Boston, MA; 5Division of Hematology, Fred Hutch, Seattle, WA; 6Division of Hematology and Oncology, Swedish Cancer Institute, Seattle, WA; 7Division of Oncology, University of Chicago, Chicago, IL; 8Division of Hematology, Northwell Health, New Hyde Park, NY; 9 Division of Hematology, Memorial Sloan Kettering Cancer Center, New York, NY; 10 Division of Hematology, Cleveland Clinic, Cleveland, OH; 11Division of Oncology, Jefferson University, Philadelphia, PA; 12Division of Hematology, City of Hope, Duarte, CA; 13 Division of Hematology, University of Michigan, Ann Arbor, MI and 14Division of Hematology, Duke University, Durham, NC, USA

Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2845-2852

ABSTRACT

hronic lymphocytic leukemia (CLL) patients who develop Hodgkin lymphoma (HL) have limited survival. No current therapeutic standard of care exists. We conducted a multi-center retrospective study of patients with Hodgkin transformation (HT) of CLL. Clinicobiologic characteristics, treatment type, and survival outcomes were analyzed and compared with historic case series. Ninety-four patients were identified. Median age at HT was 67 years (range, 38-85). Median time from CLL diagnosis to HT was 5.5 years (range, 0-20.2). Prior to HT, patients received a median of two therapies for CLL (range, 0-12). As initial therapy for HT, 61% (n=62) received ABVD-based regimens (adriamycin, bleomycin, vinblastine, and dacarbazine). Seven (7%) patients received hematopoietic cell transplantation (HCT) while in first complete remission (CR1). The median number of treatments for HT per patient was one (range, 0-5) with 59 (61%) patients only receiving one line of therapy. After HT, patients had a median follow-up of 1.6 years (range, 015.1). Two-year overall survival (OS) after HT diagnosis was 72% (95% Confidence Interval: 62-83). The patients who received standard ABVD-based therapy had a median OS of 13.2 years. Although limited by small sample size, the patients who underwent HCT for HT in CR1 had a similar 2-year OS (n=7; 67%) compared to patients who did not undergo HCT for HT in CR1 (n=87; 72%; P=0.46). In this multi-center study, HT patients treated with ABVD-based regimens had prolonged survival supporting the use of these regimens as standard of care for these patients.

Correspondence: DEBORAH M. STEPHENS deborah.stephens@hci.utah.edu Received: April 22, 2020. Accepted: September 24, 2020. Pre-published: October 5, 2020. https://doi.org/10.3324/haematol.2020.256388

Introduction Richter Transformation (RT) is the development of an aggressive lymphoma in the setting of chronic lymphocytic leukemia (CLL), occurring in 5-10% of CLL patients. Clinical outcomes in patients with RT are exceedingly poor with a median overall survival (OS) of 5-8 months from diagnosis.1 In the majority of RT cases, the CLL transforms into diffuse large B-cell lymphoma (DLBCL). Development of Hodgkin lymphoma (HL) in patients with CLL is a rare (<1%), but recognized form of RT.1,2 Based on prior data, the expected OS of CLL patients with HL transformation (HT) appears longer than CLL patients with transformation to DLBCL, with reported survival ranging from 0.8–3.9 years after diagnosis of HT.2-4 In contrast, HT patients have shorter OS (2-year OS of 30-40%) when

haematologica | 2021; 106(11)

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.

2845

Deborah M. Stephens et al.

compared with all patients with de novo HL (2-year OS >90%) after treatment with standard chemotherapy regimens, such as adriamycin, bleomycin, vinblastine, and dacarbazine (ABVD).3-7 There are limited published data about HT patients and most are small series reported from single institutions.3,810 To date, there are no clear recommendations for the management of these patients. Extrapolating from data in RT to DLBCL, some groups recommend aggressive therapy with hematopoietic cell transplantation (HCT) after achievement of first complete remission (CR1).3,11 As patients with CLL are typically elderly with comorbidities, undergoing a HCT is infrequently feasible for this patient population. In order to better understand the clinicobiologic features, treatment patterns, and clinical outcomes, we describe the largest reported series of HT patients based upon our inclusive multi-institutional clinical experience.

Methods After Institutional Review Board approval, CLL patients were retrospectively identified who also developed biopsy-proven, classical HL diagnosed between 2000 and 2018 at 13 United States tertiary cancer centers. The number of patients identified at each center is detailed in the Online Supplementary Table S1. Investigators from each site reviewed pathology records to confirm diagnosis. Clinicobiologic characteristics, treatment type, and survival outcomes for each patient were analyzed (collected variables are listed in the Online Supplementary Table S2). The International Prognostic Score (IPS) for HL and the Richter Scoring System (RSS, scoring validated previously on RT patients with DLBCL) was calculated for each patient at the time of HT diagnosis in patients where data were available (see the Online Supplementary Tables S3 and S4).1,12 Progression-free survival (PFS) was measured from the time of HT diagnosis until relapse, subsequent therapy for HT, or death. OS was measured from the time of HT diagnosis until time of death. PFS and OS estimates were calculated using the Kaplan-Meier method. The log-rank test was used to calculate differences in survival. Cox proportional hazards models and associated Wald tests were used to analyze the relationship between covariates and OS. Logistic regression was used to compare baseline characteristics of patients receiving full dose ABVD versus all other regimens. R statistical computing software version 3.2.1 (The R Foundation for Statistical Computing, Vienna, Austria 2015) was used for statistical analysis.

Results Patient characteristics Ninety-four CLL patients with HT were identified from 13 centers. Median age at initial CLL diagnosis was 60 years (range, 31-84) and 81% of the patients were men. At initial CLL diagnosis, 15% (11 of 81) were Rai 3-4, 67% (25 of 37) had an un-mutated IGHV gene, and 15% (nine of 61) had del(17p). Prior to HT diagnosis, patients had a median of two (range, 0-12) prior therapies for CLL. Seventeen (18%) had no prior CLL treatments. Fortythree (46%) and 25 (27%) patients had received purine analogue- and ibrutinib-based therapy prior to HT, respectively. Median time from CLL diagnosis to HT was 5.5 years (range, 0-20.2). Seven patients had simultaneous diagnosis of CLL and HL. 2846

Table 1. Baseline Characteristics at Hodgkin transformation of chronic lymphocytic leukemia.

Number

Median age at HT diagnosis, years (range) Male sex, no. (%) Ann Arbor Stage III/IV, no. (%) N = 83 ECOG performance status 0-1, no. (%) N = 80 B-symptoms, Yes, no. (%) N = 89 Bulky disease >10 cm, Yes, no. (%) N = 81 Pathology subtype, no. (%)* Type I HT Type II HT N = 56 Epstein Barr virus positive, Yes, no. (%) N = 72 Median WBC, x109/L (range) N = 73 Median ALC, x109/L (range) N = 72 Median hemoglobin, g/dL (range) N = 74 Median platelet, x109/L (range) N = 74 Elevated LDH, Yes, no. (%) N = 78 Median creatinine, g/dL (range) N = 72 Median number of CLL-directed therapies prior to HT, no. (range) CLL-directed therapies received prior to HT, no. (%) No therapy No cytotoxic chemotherapy+ Purine analogue-based therapy# Ibrutinib-based therapy Acalabrutinib-based therapy Venetoclax-based therapy Both cytotoxic chemotherapy+ and ibrutinib-basedtherapy Both purine analogue#- and ibrutinib-based therapy

67 (38 – 85) 76 (81) 72 (87) 54 (68) 58 (65) 9 (11) 33 (59) 23 (41) 41 (57)

7.4 (0.6-199.0) 1.8 (0.1-187.0)

11.0 (4.9-18.6) 185 (19-539) 40 1.0 (0.5-2.0) 2 (0-12)

17 (18) 27 (29) 43 (46) 25 (27) 2 (2) 1 (1) 15 (16) 12 (13)

ALC: absolute lymphocyte count; CLL: chronic lymphocytic leukemia; HT: Hodgkin transformation; LDH: lactate dehydrogenase; WBC: white blood cell; ECOG: Eastern Cooperative Oncology Group: *Type I: Hodgkin/Reed-Sternberg cells scattered in a background of CLL cells; Type II: typical classical Hodgkin lymphoma morphology showing Hodgkin/Reed-Sternberg cells in a polymorphous inflammatory background, largely segregated from CLL.20,26 +Cytotoxic chemotherapy regimens include bendamustine, chlorambucil, fludarabine, and pentostatin. #Purine analogue-based therapies include fludarabine and pentostatin.

Median age at HT was 67 years (range, 38-85; Table 1). Eighty-seven percent (72 of 83) had Ann Arbor stage III/IV, 15% had Eastern Cooperative Oncology Group (ECOG) performance status of >2, 65% (58 of 89) had Bsymptoms, 11% (nine of 81) had bulky disease (>10 cm), and 62% (36 of 58) had an IPS score of >3 (Table 1). Of 72 evaluable patient samples 41 (57%) tested positive for Epstein Barr virus (EBV). The median number of treatments for HT per patient was one (range, 0-5) with 59 (61%) patients only receiving one line of therapy. Details of regimens received can be found in the Online Supplementary Table S5. As initial therapy for HL, the haematologica | 2021; 106(11)

Prolonged survival for concurrent Hodgkin and CLL

Table 2. Baseline chronic lymphocytic leukemia (CLL) characteristics at time of CLL diagnosis: univariate Cox models for overall survival.

Covariate

Time from a CLL to HT Rai Stage 1 Rai Stage 2 IGHV Status Del(13q) Trisomy 12 Del(11q) Del(17p)

Reference Level

Tested Level

(per month) 0 0-1 U No No No No

1-4 2-4 M Yes Yes Yes Yes

N Estimate

Hazard Ratio 95% CI Low

95% CI High

1.00

1.01

1.00

0.26

81 81 37 59 44 59 60

0.48 0.39 0.23 0.15 0.24 0.86 0.38

2.14 1.76 3.52 1.36 2.59 5.76 4.57

0.48 0.39 0.23 0.15 0.24 0.86 0.38

0.98 0.63 0.88 0.16 0.70 0.10 0.67

CLL: chronic lymphocytic leukemia; HT: Hodgkin transformation of CLL; IGHV: Iimmunoglobulin variable heavy chain.

Table 3. Baseline characteristics at time of Hodgkin transformation of chronic lymphocytic leukemia: univariate Cox models for overall survival.

Covariate

Age at HT (years) HT subtype Ann Arbor Stage ECOG PS B symptoms LDH above normal Hgb (g/dL) Hgb (g/dL) WBC (x109/L) ALC (x109/L) ALC (x109/L) Platelet (x109/L) Creatinine (g/dL) Albumin (g/dL) ESR Lymph node (cm) EBV positive IPS Score+ RSS Score^

Reference Level

Tested Level

(per year) 1 1-2 1-2 N N < 12 < 10.5 < 15 < 0.6 <4 < 50 < 1.5 <4 < 50 < 10 N <4 0-1

2 3-4 3-4 Y Y ≥ 12 ≥ 10.5 ≥ 15 ≥ 0.6 ≥4 ≥ 50 ≥ 1.5 ≥4 ≥ 50 ≥ 10 Y 4+ 2-4

94 57 83 80 89 78 74 74 73 72 72 74 72 74 36 80 72 57 57

Estimate

Hazard Ratio 95% CI Low

95% CI High

1.03 0.67 1.26 1.13 1.85 4.45 0.75 0.75 1.47 1.25 1.97 -* 1.61 0.41 0.69 0.64 1.72 4.81 5.74

0.99 0.21 0.38 0.39 0.79 1.78 0.30 0.30 0.56 0.36 0.81 0.57 0.12 0.23 0.15 0.68 1.05 1.77

1.07 2.19 4.16 3.28 4.31 11.11 1.86 1.86 3.84 4.26 4.76 4.48 1.39 2.06 2.69 4.39 21.99 18.54

0.107 0.511 0.710 0.820 0.154 0.001 0.533 0.533 0.430 0.725 0.133 0.366 0.152 0.502 0.542 0.255 0.043 0.003

*Did not converge. The log-rank test (which is valid even if the Cox model does not converge) gives P=0.27: +See the Online Supplementary Table S3 for calculation of the International Prognostic Score. ^See the Online Supplementary Table S4 for calculation of the Richter Scoring System. ALC: absolute lymphocyte count; CLL: chronic lymphocytic leukemia; EBV: Epstein Barr virus: ESR: erythrocyte sedimentation rate; Hgb: hemoglobin; HT: Hodgkin transformation of CLL; IPS: International Prognostic Score; LDH: lactate dehydrogenase; PS: performance status; RSS: Richter Scoring System.

majority of patients (61%, n=62) received ABVD-based (n=48) or AVD-based regimens (n=14). Of these, antiCD20 monoclonal antibody was added in seven and BTK inhibitor was added in five (Online Supplementary Table S5). Ten (11%) received a brentuximab vedotin as part of initial therapy. Seven (7%) received an RCHOP-based regimen (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone). Six patients (6%) received no therapy for HT due to frailty. Nine (10%) patients received other regimens, which are detailed in the Online Supplementary Table S5. There were 25 patients who received ibrutinib (n=24) or acalabrutinib (n=1) as the CLL treatment immediately prior to the first HT treatment. Of the 25 patients, 17 patients discontinued BTK treatment during the first HT treatment and three of these resumed BTK inhibitor upon completion of HT haematologica | 2021; 106(11)

therapy. Of the nine patients who continued BTK inhibitor therapy concurrently with the first HT treatment, six received AVD, two received rituximab plus AVD, and one received brentuximab vedotin. Twentynine patients received salvage chemotherapy for HT. The majority (n=13) received only one salvage regimen for HT. The most common salvage regimens used were ifosfamide, carboplatin, and etoposide in seven patients and brentuximab vedotin in six patients. Online Supplementary Table S6 describes salvage chemotherapy regimens received by individual patients. Subsequent therapy included autologous HCT and allogeneic HCT in seven (7%) and eleven (12%) patients, respectively. Two (2%) and five (5%) patients received their autologous and allogeneic HCT while in CR1, respectively. Detailed toxicity data are not available for this patient subset. 2847

Deborah M. Stephens et al.

Figure 1. Overall survival for patients with Hodgkin transformation. (A) Overall survival for patients with Hodgkin transformation (HT) who received prior chronic lymphocytic leukemia (CLL)-directed treatment versus no prior CLL-directed treatment. (B) Overall survival for patients with HT who received prior purine analogue (PA)based therapy for CLL versus no prior PA-based therapy for CLL.

Survival After HT diagnosis, patients had a median follow-up of 1.6 years (range, 0.0–15.1). The median time between HT diagnosis and HT treatment was 15 days (interquartile range, 6.5–30.5 days). The median PFS was 21 months (95% Confidence Interval [CI]: 14-58). Estimated 2-year PFS was 48% (95% CI: 38-61; Online Supplementary Figure S1A). There was minimal difference in the estimated 2year PFS after censoring for subsequent HCT (2-year PFS =48.9%, 95% CI: 38.7– 61.8; Online Supplementary Figure S1B). The median OS was 65 months (95% CI: 34-infinity). Two-year OS after HT diagnosis was 72% (95% CI: 62-83). Five patients (5%) died within 2 months of HT diagnosis. At the time of initial CLL diagnosis, the baseline characteristics of Rai Stage, IGHV mutational status, or the presence of del(13)(q14), trisomy 12, del(11)(q22.3), or del(17)(p13.1) were not associated with differences in OS after HT (Table 2). Patients who received any CLL-directed therapy (n=80) prior to HT had a significantly lower estimated 2-year OS of 69% (95% CI: 58–82) compared with patients who did not receive any prior CLL-directed therapy (n=17; 93%; 95% CI: 82-100; P=0.02; Figure 1A). Patients who received purine-analogue-based therapy for CLL prior to HT had a significantly lower estimated 2year OS of 60% (95% CI: 46–79) compared with patients who did not receive purine-analogue-based CLL-directed therapy prior to HT (n=51; 83%; 95% CI: 73–96; P=0.009; Figure 1B). A more detailed depiction of estimated OS by prior CLL-directed therapy is shown in the Online Supplementary Figure S3. The available samples of those patients treated with purine-analogue-based therapy prior to HT (n=34) more frequently demonstrated positivity for EBV (OR 3.24; 95% CI: 1.1-9.9; P=0.02). Of the 25 patients who developed HT following treatment at any time with ibrutinib, the median time from initiation of ibrutinib to HT was 15.5 months (range, 1.2– 37.7). Ten of these patients had never received a standard 2848

chemoimmunotherapy regimen prior to HT. Compared with the patients who did not receive ibrutinib prior to HT, the estimated 2-year OS was slightly lower in the patients who did receive ibrutinib prior to HT (73% vs. 64%) although this finding was not statistically different (P=0.33). The available samples of those patients (n=22) treated with ibrutinib prior to HT tended to be less likely to be positive for EBV than samples from patients not previously treated with ibrutinib (OR 0.52; 95% CI: 0.2-1.6; P=0.21). At the time of HT diagnosis, the only characteristics associated with OS in univariate analysis were lactate dehydrogenase (LDH) above normal (Hazard Ratio [HR] 4.5; 95% CI: 1.8-11.1; P=0.001), IPS of ≥4 versus <4 (HR 4.8; 95% CI: 1.1-22.0; P=0.043) and RSS of ≥2 versus <2 (HR 5.7 95% CI: 1.8-18.5; P=0.003; Table 3; Figure 2A and B). In the patients who had available data to calculate the RSS (n=57), each increase of the RSS by one point resulted in an increased risk of death (HR 2.3; 95% CI: 1.3-3.9; P=0.004; Figure 2B). As compared to patients who received ABVD-based regimen (n=48) as first therapy for HT, those who received RCHOP (n=8) had a similar risk of death (HR 1.6, 95% CI: 0.4-5.7; P=0.48; Figure 3A). Conversely, those patients who received AVD-based (n=14; HR 3.0; P=0.04), BV-based (n=9; HR 3.6; P=0.05), or other (n=9; HR 7.67; P<0.001) regimens for first therapy for HT had an increased risk of death compared with patients who received ABVD-based regimen (n=48). This finding still remained significant after adjusting for age of patient at time of first HT therapy (Online Supplementary Table S7). Patients who received ABVD had a median OS of 13.2 years (95% CI: 4.8-infinity). Patients who received ABVD versus all other regimens were more likely to be less than 65 years of age (P=0.05), have a baseline absolute lymphocyte count (ALC) >0.6 (P=0.003), and have an IPS of <4 (P=0.03). There was no difference in OS based on the year that the patient received the first therapy for HT (P=0.81; haematologica | 2021; 106(11)

Prolonged survival for concurrent Hodgkin and CLL

Figure 2. Overall survival for patients with Hodgkin transformation. Overall survival for patients with Hodgkin transformation categorized by (A) International Prognostic Score (IPS) for Hodgkin lymphoma and (B) Richter Scoring System for Richter transformation.

Online Supplementary Figure S2). The patients who underwent HCT for HT in first complete remission had a similar 2-year OS (n=7; 67%; 95% CI: 38-100) to patients who did not undergo HCT for HT in first complete remission (n=87; 72%; 95% CI: 63–84; P=0.46; Figure 3B). Peritransplant mortality did not contribute to shortened survival in this group as the earliest death in this group occurred at 355 days post-transplant.

Discussion In this retrospective analysis, we describe the largest reported multi-center inclusive series of patients with HT from CLL. Clinical outcomes, including survival, in this series of patients with HT were higher than what has previously been reported for this patient population and strikingly similar to what is historically expected in elderly patients with de novo HL.3,4,6,7 Patients with HT who have received prior CLL-directed therapies (specifically purineanalogue-based treatments), elevated LDH, IPS ≥4, and RSS ≥2 are predicted to have a shorter OS. The OS of patients in this series was higher than what is expected in CLL patients who transform to DLBCL.1,13-15 Unlike patients with RT to DLBCL, the majority of patients only received one line of HL therapy. Only 20% went on to receive HCT and had similar OS to patients who did not receive HCT in CR1. This finding has impact in the standard management of patients with HT as these data challenge the recommendation of HCT in CR1. Patients in our series survived longer than historical series of RT to DLBCL (median OS 65 months vs. 5-8 months).13-17 It is unclear whether this difference is related to underlying disease biology or effectiveness of therapy. Prior molecular studies have revealed that patients with RT to DLBCL that is clonally related to the underlying CLL have much shorter survival than patients with RT to DLBCL that is clonally unrelated to the underlying CLL.18 haematologica | 2021; 106(11)

Therefore, clonally unrelated RT to DLBCL is treated as de novo DLBCL with expected outcomes more consistent with de novo DLBCL.19 Little is known about clonal relation of the HL and CLL cells in patients with HT. One group evaluated tissue samples of 33 HT cases.20 The HL cells were clonally related to the CLL cells in 14 cases and unrelated in 19 cases.20 Interestingly, they found no differences in baseline characteristics or OS with treatment when comparing patients with clonally related versus unrelated HT, which is in contrast to what was previously observed in patients with RT to DLBCL.20,21 Further molecular study of HT patient samples is needed to fully understand the underlying biology of HT, which may help to determine prognosis or guide therapy for these patients in the future. The dismal survival in patients with RT to DLBCL often prompts clinicians to implement more aggressive therapy for these patients. One group showed that RT patients with DLBCL who underwent allogeneic HCT in CR1 had longer survival than those patients in CR1 who did not receive further therapy or those patients who did not receive HCT until salvage therapy.1 Based on the data for patients with RT to DLBCL, it was extrapolated that patients with HT may require HCT when they achieve CR1.3,11 In our study, the majority of patients (61%) only received one line of HL therapy and only 20% went on to receive HCT (7% while in CR1). Although limited by small numbers, the patients who underwent HCT for HT in CR1 had a similar 2-year OS to patients who did not undergo HCT for HT in CR1. Our data indicate that these patients with HT can have prolonged OS after achieving response to first-line therapy for HT and may not require HCT in CR1. Therefore, our large series does not support the recommendation of HCT in CR1 for this patient population and has major implications in the management of these patients. In addition to our patient population demonstrating longer survival than previously published series of RT to 2849

Deborah M. Stephens et al.

Figure 3. Overall survival for patients with Hodgkin transformation. Overall survival for patients with Hodgkin transformation (HT) categorized by (A) first treatment for HT and (B) hematopoietic cell transplant (HCT) versus no stem cell transplant (SCT) in first complete remission (CR1). RCHOP: rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone; ABVD: adriamycin, bleomycin, vinblastine, and dacarbazine

DLBCL, the OS in our group of patients with HT was higher than what has been previously reported in historic series of patients with HT (2-year OS 72% vs. 30-40%).3,4 Without direct comparison of the baseline characteristics of the patients, it is possible that the historically diagnosed patients had higher risk features such as increased age or higher IPS or RSS compared with the patients in our group. Additionally, it is possible that the increased use of traditional HL treatment such as ABVD in our patients (60%) versus the prior patients (30%) could have contributed to the improved outcomes of our patients.3,4 As only 11% of our patients received targeted HL therapy with BV, it is unclear at this time if novel targeted HL therapy has improved the outcomes of these patients. In fact, in our review patients who received BV-based therapies (n=9) as initial treatment had a higher risk of death than those who received full dose ABVD therapy (n=48). This finding is likely biased by selection of less aggressive induction therapy in less fit patients (older and higher IPS) that were not felt to be able to tolerate full doses of ABVD. In contrast to prior reports indicating that patients with HT have a shorter survival than patients with de novo HL, these data from the largest multi-center study actually demonstrate a very similar survival compared to historic controls of patients with de novo HL when matched for age. In our study, the median age at time of HT was 67 years.3,4,6,7 A retrospective study described 95 patients diagnosed with HL over the age of 60 years.6 Similar to our population, the median age of the patients in this analysis was also 67 years and the majority of the patients received ABVD as initial HL treatment (n=67). The 2-year OS reported for this group was 73% and very comparable to the 72% found in our analysis of patients with HT.6 A subset analysis was performed on patients over the age of 60 years treated on the prospective phase III E2496 study, where patients were randomized to ABVD versus Stanford V regimen.7 The median age of patients receiving ABVD on this analysis was 66 years. In this group, the 2-year OS rate was approximately 80%, which is compa2850

rable to the 72% found in our current analysis.7 Based on these data, the survival of patients with HT seen in our analysis appear similar to patients with de novo HL of a similar age group. For older patients with de novo HL, clinical outcomes have improved over time.22 This finding was attributed to the adoption and development in expertise of ABVD administration and willingness of clinicians to offer combination chemotherapy to older patients. In much the same way, recent development of novel therapies, such as ibrutinib, have significantly improved survival for patients with CLL, a disease predominately of older patients.23,24 A single-center review found that RT to DLBCL was most likely to occur during the first 2 years of therapy (7% at 2 years).25 This is similar to our finding that of the 25 patients who developed HT following treatment with ibrutinib, the median time from initiation of ibrutinib to HT was 15.5 months. There was no statistically significant difference in 2-year OS between the patients who did and did not receive ibrutinib prior to HT. In five of our cases, clinicians opted to continue ibrutinib therapy to control the patient’s CLL while adding chemotherapy to treat the HL. At this time, the full effect of ibrutinib and other novel agents on risk for HT is unclear. Although prior ibrutinib therapy did not contribute to reduced survival, we found that prior purine-analoguebased therapies for CLL prior to HT led to reduced survival. This finding has been reported by other groups and is possibly attributed to immunosuppression resulting from purine-analogue therapy.8 In our limited data set, this attribution is supported by the fact that patients who received prior fludarabine-based therapy were more likely to have EBV-positive HT. Interestingly, other traditional baseline characteristics of CLL that have been associated with survival in CLL, such as Rai Stage, unmutated IGHV and karyotype, did not correlate with survival after HT. In addition to prior purine-analogue therapy, the major factors that were identified to predict poor survival for patients with HT were elevated LDH, IPS, and RSS. These haematologica | 2021; 106(11)

Prolonged survival for concurrent Hodgkin and CLL

factors may be useful in counseling patients with this diagnosis. Interpretation of our data is limited by the intrinsic nature of retrospective analyses. Although individual centers locally confirmed the diagnosis of HT, central review of samples was not feasible secondary to limited quantity of sample and a small number of patients from many centers. We did not search all patients with consecutive HL diagnoses at each center, but searched for CLL patients who subsequently developed HL. As such, it is possible this may have introduced selection bias to our results that may lead to overestimation of patient outcomes. Subset analyses of this patient population were limited secondary to small numbers of patients. Our study examined patients treated at tertiary referral cancer centers and results may not be generalizable to patients treated in the community setting. Despite these limitations, our series is notable as the largest reported cohort of this rare group of patients with HT. In summary, this series of patients with HT from CLL reveals similar survival to what is expected in patients with de novo HL when compared with age- and treatmentmatched HL patients in previously published studies. These data support that HT patients have longer survival compared with historic controls of patients with RT from DLBCL and do not require allogeneic HCT in CR1. Based on this series, HT patients should be treated with regimens used to treat de novo HL. The effects of novel targeted HL and CLL therapies on the outcomes of patients with HT is unclear at this time, and further study of this rare population, ideally in prospective clinical trials, is required to fully define the optimum management of these patients. Disclosures DS has received research funding from Acerta, Verastem, Juno, Karyopharm, and Gilead, as well consulting fees from Pharmacyclics, Jannsen, Karyopharm, Innate, and Genentech; EK has received consulting fees from AstraZeneca; SP has received honoraria and research funding from Pharmacyclics, Abbvie, and AstraZeneca, as well as research funding from MorphoSys, Janssen, and Gilead; MS has received research

References 1. Tsimberidou AM, O'Brien S, Khouri I, et al. Clinical outcomes and prognostic factors in patients with Richter's syndrome treated with chemotherapy or chemoimmunotherapy with or without stem-cell transplantation. J Clin Oncol. 2006;24(15):2343-2351. 2. Parikh SA, Habermann TM, Chaffee KG, et al. Hodgkin transformation of chronic lymphocytic leukemia: incidence, outcomes, and comparison to de novo Hodgkin lymphoma. Am J Hematol. 2015;90(4):334-338. 3. Tsimberidou AM, O'Brien S, Kantarjian HM, et al. Hodgkin transformation of chronic lymphocytic leukemia: the M. D. Anderson Cancer Center experience. Cancer. 2006;107(6):1294-1302. 4. Bockorny B, Codreanu I, Dasanu CA. Hodgkin lymphoma as Richter transformation in chronic lymphocytic leukaemia: a retrospective analysis of world literature. Br J Haematol. 2012;156(1):50-66. 5. Ramchandren R, Advani RH, Ansell SM, et al. Brentuximab vedotin plus chemotherapy in North American patients with newly

haematologica | 2021; 106(11)

funding from Genentech, Pharmacyclics, Gilead, Celgene, Mustan Biopharma, TG Therapeutics, Beigene, Acerta, as well as consulting fees from Qilu Puget Sound Biotherapeutics, Abbvie, Genentech, Verastem, and AstraZeneca; JP has received consulting fees from Pharmacyclics and Gilead; AM has received consulting fees and honoraria from Pharmacyclics, Abbvie, TG Therapeutics, as well as consulting fees from AstraZeneca, Celgene, Johnson & Johnson, and research funding from Acerta, Portola, and Regeneron; BH has received honoraria and consulting fees from Pharmacyclics, Abbvie, Seattle Genetics, Novartis, Genentech, and Pfizer, as well as research funding from Amgen and Genentech; AD has received research funding from Takeda, Aptose Biosciences, Gilead, Genentech, Verastem, Bayer, as well as consulting fees from Gilead, Genentech, AstraZeneca, Verastem, TG Therapeutics, and Bayer; TP has received consulting fees from Genentech, Gilead, Bayer, Seattle Genetics, and Pharmacyclics, as well as research funding from Pharmacyclics and Abbvie; DB has received research funding, consulting fees, and honoraria from Pharmacyclics, Abbvie, Genentech, Teva, TG Therapeutics, and Novartis, as well as research funding from Acerta, DTRM, and Gilead; SS has received consulting fees from BMS; MD has received consulting fees and research funding from Abbvie, Genentech, TG Therapeutics, AstraZeneca, and Verastem, as well as consulting fees from Merck, Celgene, Gilead, MEI Pharma, Janssen, Sunesis, and Roche, and research funding from BMS and Surface Oncology; KR has received consulting fees from Acerta and Pharmacyclics; JB has received research funding from Acerta, Genentech, Janssen, and Pharmacyclics. Contributions DS and JB conceived the research, recruited participants, analyzed data, and drafted the original manuscript; KB performed the statistical analysis. All remaining authors collected and analyzed data and reviewed the manuscript. Funding National Cancer Institute (R35 CA198183; K23 CA212271), Four Winds Foundation, Connie Brown CLL Fund, Kevin Sullivan Foundation, and the D. Warren Brown Foundation.

diagnosed stage III or IV Hodgkin lymphoma. Clin Cancer Res. 2019;25(6):17181726. 6. Evens AM, Helenowski I, Ramsdale E, et al. A retrospective multicenter analysis of elderly Hodgkin lymphoma: outcomes and prognostic factors in the modern era. Blood. 2012;119(3):692-695. 7. Evens AM, Hong F, Gordon LI, et al. The efficacy and tolerability of adriamycin, bleomycin, vinblastine, dacarbazine and Stanford V in older Hodgkin lymphoma patients: a comprehensive analysis from the North American intergroup trial E2496. Br J Haematol. 2013;161(1):76-86. 8. Parikh SA, Rabe KG, Ding W, et al. Hodgkin Transformation of chronic lymphocytic leukemia (CLL): Mayo Clinic experience. Blood. 2013;122(21):1642-1642. 9. Fayad L, Robertson LE, O'Brien S, et al. Hodgkin's disease variant of Richter's syndrome: experience at a single institution. Leuk Lymphoma. 1996;23(3-4):333-337. 10. Kazmierczak M, Kroll-Balcerzak R, Balcerzak A, et al. Hodgkin lymphoma transformation of chronic lymphocytic

leukemia: cases report and discussion. Med Oncol. 2014;31(1):800. 11. Kharfan-Dabaja MA, Kumar A, Hamadani M, et al. Clinical practice recommendations for use of allogeneic hematopoietic cell transplantation in chronic lymphocytic leukemia on behalf of the Guidelines Committee of the American Society for Blood and Marrow Transplantation. Biol Blood Marrow Transplant. 2016; 22(12): 2117-2125. 12. 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. 13. Tsimberidou AM, O'Brien SM, Cortes JE, et al. Phase II study of fludarabine, cytarabine (Ara-C), cyclophosphamide, cisplatin and GM-CSF (FACPGM) in patients with Richter's syndrome or refractory lymphoproliferative disorders. Leuk Lymphoma. 2002;43(4):767-772. 14. Robertson LE, Pugh W, O'Brien S, et al. Richter's syndrome: a report on 39 patients. J Clin Oncol. 1993;11(10):1985-1989.

2851

Deborah M. Stephens et al. 15. Dabaja BS, O'Brien SM, Kantarjian HM, et al. Fractionated cyclophosphamide, vincristine, liposomal daunorubicin (daunoXome), and dexamethasone (hyperCVXD) regimen in Richter's syndrome. Leuk Lymphoma. 2001;42(3):329337. 16. Tsimberidou AM, Kantarjian HM, Cortes J, et al. Fractionated cyclophosphamide, vincristine, liposomal daunorubicin, and dexamethasone plus rituximab and granulocyte-macrophage-colony stimulating factor (GM-CSF) alternating with methotrexate and cytarabine plus rituximab and GM-CSF in patients with Richter syndrome or fludarabine-refractory chronic lymphocytic leukemia. Cancer. 2003;97(7):1711-1720. 17. Harousseau JL, Flandrin G, Tricot G, Brouet JC, Seligmann M, Bernard J. Malignant lymphoma supervening in chronic lymphocytic leukemia and related disorders. Richter's syndrome: a study of 25 cases.

2852

Cancer. 1981;48(6):1302-1308. 18. Rossi D, Cerri M, Capello D, et al. Biological and clinical risk factors of chronic lymphocytic leukaemia transformation to Richter syndrome. Br J Haematol. 2008;142(2):202-215. 19. Wierda WG, Byrd JC, Abramson JS, et al. NCCN Guidelines insights: chronic lymphocytic leukemia/small lymphocytic lymphoma, version 2.2019. J Natl Compr Canc Netw. 2019;17(1):12-20. 20. Xiao W, Chen WW, Sorbara L, et al. Hodgkin lymphoma variant of Richter transformation: morphology, Epstein-Barr virus status, clonality, and survival analysiswith comparison to Hodgkin-like lesion. Hum Pathol. 2016;55:108-116. 21. Rossi D, Spina V, Deambrogi C, et al. The genetics of Richter syndrome reveals disease heterogeneity and predicts survival after transformation. Blood. 2011; 117(12): 3391-3401.

22. Brenner H, Gondos A, Pulte D. Ongoing improvement in long-term survival of patients with Hodgkin disease at all ages and recent catch-up of older patients. Blood. 2008;111(6):2977-2983. 23. 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. 24. Burger JA, Tedeschi A, Barr PM, et al. Ibrutinib as initial therapy for patients with chronic lymphocytic leukemia. N Engl J Med. 2015;373(25):2425-2437. 25. Woyach JA, Ruppert AS, Guinn D, et al. BTK. J Clin Oncol. 2017;35(13):1437-1443. 26. Ohno T, Smir BN, Weisenburger DD, Gascoyne RD, Hinrichs SD, Chan WC. Origin of the Hodgkin/Reed-Sternberg cells in chronic lymphocytic leukemia with "Hodgkin's transformation". Blood. 1998; 91(5):1757-1761.

haematologica | 2021; 106(11)

ARTICLE

Chronic Myeloid Leukemia

The effect of eltrombopag in managing thrombocytopenia associated with tyrosine kinase therapy in patients with chronic myeloid leukemia and myelofibrosis

Ferrata Storti Foundation

Mahran Shoukier, Gautam Borthakur, Elias Jabbour, Farhad Ravandi, Guillermo Garcia-Manero, Tapan Kadia, Jairo Matthews, Lucia Masarova, Kiran Naqvi, Koji Sasaki, Srdan Verstovsek and Jorge Cortes° Department of Leukemia, the University of Texas MD Anderson Cancer Center, Houston, TX, USA

Haematologica 2021 Volume 106(11):2853-2858

°Current address: Georgia Cancer Center, Augusta University, Augusta, GA, USA

ABSTRACT

pproximately 20-50% patients with chronic phase chronic myeloid leukemia (CML-CP) treated with tyrosine kinase inhibitors (TKI) or with myelofibrosis (MF) treated with ruxolitinib develop grade ≥3 thrombocytopenia needing treatment interruptions and dose reductions. We conducted a non-randomized, phase II, single-arm study to determine the efficacy of eltrombopag for patients with CML or MF with persistent thrombocytopenia while on TKI or ruxolitinib. Eltrombopag was initiated at 50 mg/day, with dose escalation up to 300 mg daily allowed every 2 weeks. Twenty-one patients were enrolled (CML=15, MF=6); with a median age of 60 years (range, 31-97 years). The median platelet count was 44x109/L (range, 3-49x109/L) in CML and 62x109/L (range, 21-75x109/L) in MF. After a median of 18 months (range, 5-77 months), 12 of 15 patients with CML achieved complete platelet response. The median peak platelet count among responders was 154x109/L (range, 74-893x109/L). Among CML patients five could re-escalate the TKI dose and nine improved their response. None of the six patients with MF had a sustained response. Therapy was generally well tolerated. One patient discontinued therapy due to toxicity (elevated transaminases). One patient with CML developed significant thrombocytosis (>1,000x109/L). Another CML patient developed non occlusive deep venous thrombosis in the right upper extremity without thrombocytosis, and one MF patient had myocardial infarction. Eltrombopag may help improve platelet counts in CML patients receiving TKI with recurrent thrombocytopenia. Further studies are warranted (clinicaltrials gov. Identifier: NCT01428635).

Introduction Tyrosine kinase inhibitors (TKI) are standard therapy for chronic myeloid leukemia (CML) and myelofibrosis (MF). Five TKI are currently approved for the treatment of CML in various stages, namely imatinib, nilotinib, dasatinib, bosutinib and ponatinib. Although these agents are generally well tolerated, some patients may develop adverse events, with myelosuppression being the most prominent.1-6 In most instances myelosuppression is grade 1 or 2 and requires no intervention. However, grade ≥3 thrombocytopenia (platelet ≤50x109/L) has been reported in 20% to 50% of patients. When this occurs, patients are usually managed with treatment interruption until platelets recovery (e.g., above 75x109/L) and dose reductions if thrombocytopenia recurs. Ruxolitinib is a JAK2 inhibitor used to manage splenomegaly and diseaseassociated symptoms in patients with MF.7 The dose limiting toxicity of ruxolitinib was thrombocytopenia8 and because of this the two pivotal phase III studies excluded patients with platelets ≤100x109/L. Still, thrombocytopenia was reported in 69% of patients, including 9% with grade ≥3.9 In patients with a platelet count of 50-

haematologica | 2021; 106(11)

Correspondence: JORGE CORTES jorge.cortes@augusta.edu Received: May 21, 2020. Accepted: September 3, 2020. Pre-published: September 14, 2020. https://doi.org/10.3324/haematol.2020.260125

2853

M. Shoukier et al.

100×109/L, grade ≥3 thrombocytopenia occurred in 56% of patients.10 Ruxolitinib-associated grade ≥3 thrombocytopenia is also typically managed with dose reductions or interruptions. Frequent dose interruptions and reductions might decrease TKI efficacy11,12 and may still not be sufficient to control thrombocytopenia. Patients who develop myelosuppression have a lower probability of achieving major or complete cytogenetic response (CCyR).11 For example, one study reported that patients treated with imatinib who developed grade ≥3 thrombocytopenia had a lower probability of CCyR compared to those who never developed thrombocytopenia (35% vs. 59%, P=0.02, respectively). Similarly, ruxolitinib efficacy is compromised with dose reductions and interruptions.12,13 In order to minimize dose reductions and interruptions, hematopoietic growth factors, filgrastim and erythropoietin stimulating agents (erythropoietin and darbepoetin) have been successfully used to manage neutropenia and anemia secondary to TKI in CML, respectively.14,15 Interleukin 11 (IL-11) was effective to manage thrombocytopenia associated with TKI in CML16 but use of this agent is associated with significant adverse events including fluid retention and cardiac arrhythmias. Eltrombopag is a non-peptide thrombopoietin receptor agonist that is effective and well tolerated among patients with immune thrombocytopenia, chronic hepatitis C-associated thrombocytopenia and severe aplastic anemia.17-19 It has also been safely used in acute myeloid leukemia without evidence of disease progression secondary to eltrombopag.20 Here, we report the results from a pilot trial investigating the use of eltrombopag in the management of TKIor ruxolitinib-associated thrombocytopenia among patients with CML and MF.

Methods Patients We conducted an open-label, non-randomized, phase II study of individualized dosing of eltrombopag. Eligible patients were aged 18 years or older with chronic phase CML receiving treatment with any Food and Drug Administration-approved TKI and experiencing grade ≥3 thrombocytopenia (platelets ≤50x109/L), or with MF receiving ruxolitinib and with platelets <100x109/L (since it is a dose-limiting toxicity and a label threshold for ruxolitinib), in either case after the first 3 months of therapy. Thrombocytopenia should have been either recurrent (i.e., be at least the second episode of thrombocytopenia) or have necessitated dose reductions of the TKI or ruxolitinib. All patients had signed an informed consent form approved by the Institutional Review Board, and the study was conducted in accordance with the Declaration of Helsinki.

Study design Eltrombopag was commenced at 50 mg with dose escalation allowed every 2 weeks to 100 mg, 150 mg, 200 mg, and 300 mg (a higher dose than per label considering the thrombocytopenia refractoriness on these patients and the intent to continue TKI/ruxolitinib) according to platelet response. For patients of East Asian ancestry, eltrombopag was commenced at 25 mg daily with dose escalation allowed every 2 weeks. The following guideline was used to adjust dosing of eltrombopag: if the platelet count was >200x109/L at any time, the daily dose was reduced by 25 mg and re-assessed in 2 weeks; if >400x109/L, therapy was withheld and platelets assessed twice weekly until platelet count <150x109/L;

2854

therapy could then be resumed with the daily dose reduced by 25 mg. If the platelet count >400x109/L after 2 weeks was at the lowest dose, therapy was permanently discontinued. TKI doses were adjusted at the discretion of the treating physician per standard practice. Liver function tests (LFT) (alanine aminotransferase [ALT], aspartate aminotransferase [AST], and bilirubin) were done before the initiation of eltrombopag, every 2 weeks during the dose adjustment phase and following the monthly establishment of a stable dose. When LFT abnormalities were identified, LFT were performed weekly until the abnormalities resolved or stabilized; if ALT/AST levels were ≥ three-times the upper limit of normal (ULN): therapy was withheld, we then repeated abnormal liver function tests within 3-5 days; if confirmed abnormal, we monitored LFT weekly until resolved, stabilized, or returned to baseline. If ALT/AST levels ≥ three-times the ULN and were progressive, persistent (≥4 weeks), accompanied by increased direct bilirubin, or accompanied by clinical signs of liver injury or evidence of hepatic decompensation, eltrombopag was permanently discontinued. Patients who experienced other clinically significant grade 3 or greater toxicity possibly related to eltrombopag, had eltrombopag interruption until toxicity resolved to grade 1 or less. Treatment then was resumed at the immediate lower dose level. Failure to achieve a platelet count ≥50x109/L or ≥100x109/L in CML and MF patients, respectively after 8 weeks of eltrombopag was considered as lack of response.

Statistical analysis Simon’s optimal two-stage design (Simon, 1989) was used to test the null hypothesis that the proportion of subjects with complete response is ≤0.10 versus the alternative that it is ≥0.30 (i.e., Po≤0.10 vs. Pa≥0.30) at alpha=0.05 with 80% power. The design resulted in an expected sample size of 15 and a probability of early termination of 0.736. The study was designed to study eltrombopag in ten patients in the first stage; the trial would be terminated if one or fewer achieved complete platelet response. Otherwise, the trial would go to the second stage, and 29 patients would be studied. If the total number of patients with complete platelet response were less than or equal to five, the drug would be deemed ineffective. The MF group was an exploratory group of ten patients to study the safety and activity of eltrombopag in patients with MF treated with ruxolitinib. We considered the activity promising if three or more patients out of ten achieved complete platelet response. For safety monitoring in the cohort with MF, accrual would stop if, at any time, four of ten patients encounter grade 3 or more nonhematological toxicity or progression to acute leukemia. As an additional safety procedure, we observed the first three MF patients on trial for at least 3 months before other patients were accrued.

Response definitions Complete platelet response was defined as platelet count ≥50x109/L for CML, and ≥100x109/L for MF that was sustained for ≥3 months while continuing TKI or ruxolitinib therapy or with sustained (≥3 months) re-escalation of TKI dose without recurrence of thrombocytopenia. Criteria for CML and MF response were previously defined.21,22 The target response was a complete response in at least 30% of patients.

Results Twenty-one patients were enrolled: 15 with CML and six with MF. Their median age was 60 years (range, 31-97) and their clinical characteristics are shown in Table 1. Median haematologica | 2021; 106(11)

The eltrombopag impact on thrombocytopenia in CML

duration of disease was 2.2 years (range, 0.5-29 years) for patients with CML and 2 years (range, 0.3-3.6 years) for patients with MF. At the time of enrollment, patients with CML were receiving the following TKI: dasatinib (n=5), ponatinib (n=4), nilotinib (n=3), bosutinib (n=2), and imatinib (n=1), 27% were receiving their first TKI, 27% the second TKI, 27% the third, and 19% the fourth or later TKI. The median platelet count was 44x109/L (range, 3-49x109/L) in patients with CML and 62x109/L (range, 21-75x109/L) in those with MF. Cytogenetic response for patients with CML at baseline were partial in three, minor in six, and none in six. Prior therapies in MF patients were an investigational JAK2 inhibitor, and interferon a-2 in one patient each. The median dose of ruxolitinib was 10 mg (range, 1030 mg) (Table 1). Eltrombopag dose distribution is summarized in Table 2. After a median duration of treatment of 18 months (range, 5-77 months), 12 of the 15 (80%) patients with CML achieved a complete platelet response with doses of eltrombopag of 50–300 mg per day. The median peak platelet count among responders was 154x109/L (range, 74893x109/L). The median time to best response was 6 months (range, 2.1-13 months). Ten patients had sustained platelet recovery after stopping eltrombopag. The median duration for sustained platelet response was 45 months (range, 3-69 months). The three patients who did not achieve a complete platelet response had only minor changes in platelet count while they were taking eltrombopag (from 3x109/L to 8x109/L, 19x109/L to 45x109/L, and from 42x109/L to 46x109/L, respectively). Two patients (one each of CML and MF) had improvement in hemoglobin of over 2 g/dL from baseline (from 8.2 g/dL to 10.6 g/dL, and from 9.4 g/dL to 11.4 g/dL, respectively), Hemoglobin improvement was sustained over 21.5 and 2 months respectively while patients were taking eltrombopag. Hemoglobin levels declined after interruption of eltrombopag. One patient with CML had an absolute neutrophil count recovery to >1x109/L (baseline neutrophils 0.71x109/L). Absolute neutrophil count improvement was sustained for >6 months while on eltrombopag. Absolute neutrophil count then declined after interruption of eltrombopag. The TKI doses and duration for patients with CML post enrollment are summarized in Table 1. Nine patients with CML experienced an improvement in the cytogenetic response during the observation period (all of them had sustained platelet recovery after stopping eltrombopag); one from none to complete, two from minor to complete, four from minor to partial, and two from partial to complete (Table 3). In five patients with CML the TKI dose was increased and maintained while continuing eltrombopag. Dasatinib daily dose was increased from 50 mg to 100 mg in three patients, nilotinib dose was increased in one patient form 150 mg twice daily to 200 mg twice daily, and one patient had an increase in ponatinib dose from 15 mg every other day to 15 mg daily. There were no TKI dose-limiting toxicities in patients who increased their TKI doses. The dose increase was associated with improvement in CML response in four of these five patients. In the five CML patients who had a cytogenetic response but did not have TKI dose escalation, the improvement in cytogenetic response was noticed while patients were on eltrombopag. Three CML patients had a switch in their TKI (Online Supplementary Table S1). All three of these patients had already some improvement in thrombocytopenia before switching their TKI, with the change indicated for haematologica | 2021; 106(11)

Table 1. Baseline characteristics.

Characteristics

N (%)

Median [range]

Age, years 60 [31-97] Male 12 (58) Disease CML 15 (71) MF 6 (29) Stage of CML at start of eltrombopag Chronic phase 15 (100) Disease duration prior to Eltrombopag initiation (years) CML 2.2 [0.5-29] MF 2 [0.3-3.6] Hemoglobin (all patients), g/dL 11.6 [8.2-13.4] WBC (all patients), x109/L 4.6 [1-71.3] Platelet count (CML), x109/L 44 [3-49] Platelet count (MF), x109/L 62 [21-75] TKI (CML patients) dose mg/d Dasatinib 5 (33) 50 [50-100] Ponatinib 4 (27) 11.25 [7.5-30] Nilotinib 3 (20) 300 [150-600] Bosutinib 2 (13) 300 Imatinib 1 (7) 400 Ruxolitinib (MF patients) 6 (100) 10 [10-30] Time on TKI before eltrombopag, years 2.1 [0.5-14] Time on ruxolitinib before eltrombopag, months 3 [3-18] Cytogenetic response prior to eltrombopag (CML) None 6 (40) Minor 6 (40) Partial 3 (20) CML: chronic myeloid leukemia; MF: myelofibrosis; TKI: tyrosine kinase inhibitor. TKI was on hold in six patients with CML at time of enrollment because of grade 3 thrombocytopenia; WBC: white blood cells.

Table 2. Eltrombopag dose distribution, mg per day (all patients).

N (%) Dose 0 25 50 100 150 200 275 300

Maximum dose

Dose at last follow-up

0 1 (5) 3 (14) 0 2 (10) 2 (10) 0 13 (61)

2 (10) 0 3 (14) 3 (14) 2 (10) 1 (5) 1 (5) 9 (42)

other non-hematologic adverse events in one patient and the inefficacy of the TKI in the other two patients. None of the six patients with MF responded (i.e., none had a sustained increase in platelet count to ≥100x109/L); minor upward transient variations in platelet counts were seen in three patients (from 21x109/L to 28x109/L, 41x109/L to 55x109/L and from 65x109L to 75x109/L, respectively). As of the date of this report, 20 patients were off study because of a lack of response (n=9), stem cell transplant (n=2), death (n=2), patient’s wish (n=1), adverse events (n=2), TKI discontinuation (n=1), loss to follow-up (n=1) 2855

M. Shoukier et al.

Figure 1. Platelet change from baseline to response. *Each blue bar reflects change in platelet count in a chronic myeloid leukemia patient, while each green bar reflects change in platelet count in a myelofibrosis patient.

and stable platelets (n=2). The two deaths on study were not related to treatment. One death was secondary to infectious complication in a patient with MF. The second death was secondary to hemorrhagic shock in a CML patient. This patient, treated with dasatinib, developed hepatosplenomegaly and ascites while on study but the etiology was not known. There was no evidence of portal vein thrombosis on CT abdomen/pelvis. Both eltrombopag and dasatinib were held and she had no platelets response. The platelet count was 10x109/L at the time of death due to severe gastrointestinal and genitourinary bleeding. Therapy was well tolerated in most patients, but two patients on ponatinib developed thrombotic events. Two months after eltrombopag discontinuation due to termination of the study, one patient with CML developed significant thrombocytosis (>1,000x109/L) with a white blood cell count of 9.5x109/L, 3% basophils, and 2% peripheral blast accompanied by headache and eye pain. Ophthalmoscopic examination was suggestive of bilateral plaques or thrombosis in the retinal vasculature but fluoroscopic evaluation did not reveal retinal vasculature blockage. Ponatinib was discontinued and thrombocytosis was managed with hydroxyurea. The aforementioned symptoms resolved. There was no cytogenetic response prior or after starting eltrombopag. Seven months after stopping eltrombopag, 2856

the patient had a persistent increase in blasts to 13% without a lack of hematologic response and she was then started on a clinical trial with an investigational TKI. Another CML patient developed non-occlusive deep venous thrombosis in the right upper extremity without thrombocytosis while on ponatinib 4 months after the study was terminated. One MF patient who had a history of coronary artery disease status post coronary artery bypass surgery developed myocardial infarction (MI) while on eltrombopag. This patent had then worsening increase in bone marrow fibrosis from grade 2 to grade 3 and was taken off study 40 days after MI. No further additional thrombotic/thromboembolic complications in CML and MF patients observed during or after the study (Online Supplementary Table S2). One patient (CML) discontinued therapy due to toxicity (elevation of liver function tests). Grade 3/4 toxicities irrespective of attribution listed in Table 4. One patient with MF had an increase in bone marrow fibrosis from grade 2 to grade 3. That patient had an increase in blast from 3% to 8% in the peripheral blood and an increase from 1% to 6% in the bone marrow while he was on study but with an improvement in hemoglobin. There was no change in the patient's disease other than this change in blast percentage. The patient was taken off study for lack of platelet response and later started on another clinical trial (PRM-151 + ruxolihaematologica | 2021; 106(11)

The eltrombopag impact on thrombocytopenia in CML

tinib). No progression of disease has been documented in any other patients. No clonal evolution was observed in patients with prolonged eltrombopag use.

Discussion Thrombocytopenia is a common adverse event in patients with CML and MF who are treated with TKI and ruxolitinib, respectively.10,23 In most instances, thrombocytopenia is transient, occurs early during treatment initiation, and can be successfully managed with transient treatment interruptions and occasionally dose adjustments. However, in some patients thrombocytopenia can be persistent and more severe requiring frequent treatment interruptions and dose reductions, which might adversely influence treatment outcome.11 To that end, rIL-11 was successfully used in CML patients for the management of TKI associated thrombocytopenia.16 The main limitation of use of rIL-11 in the management of chemotherapy-induced thrombocytopenia in solid malignancies was the narrow therapeutic window with significant fluid retention and occasional arrhythmias. However, at lower doses used in CML, it was well tolerated24,25 with grade 1 or 2 peripheral edema observed in six patients (43%). Eltrombopag is a second generation oral thrombopoietin receptor agonist that has induced improvement of thrombocytopenia in patients with immune-mediate thrombocytopenia (ITP) or aplastic anemia. The EXTEND trial demonstrated that long-term use of eltrombopag was effective in maintaining for more than 6 months platelet counts of 50×109/L or more and reducing bleeding in most patients with ITP. Addition of eltrombopag to immunosuppressive treatment also markedly increased overall and complete hematologic response rates in treatment-naive severe aplastic anemia.26 Here we describe the use of eltrombopag in the management of TKI-related thrombocytopenia in CML and MF. Our results suggest clinical benefit in most patients with CML with a generally favorable safety profile, although two patents (both on ponartinib) had thrombotic events. In contrast, no response was observed in patients with MF. Theoretical concerns about the use of eltrombopag in this setting include increase in marrow blasts and possible transformation to advanced phases, thrombotic events including portal vein thrombosis, and increase in marrow fibrosis. We did not observe any instance of transformation in our series, in concordance with pre-clinical and clinical data showing no evidence of worsening leukemia.20,27 There was also no increase in marrow fibrosis in CML patients. Our series is small so the lack of such events should be considered as preliminary but reassuring. The most common adverse event was LFT elevation, but these were generally transient, reversible and manageable with dose adjustments. However, in one case it led to discontinuation of eltrombopag because of recurrent transaminitis. Two patients who received ponatinib (50%) had thrombotic events while on eltrombopag, this might raise the precaution of using ponatinib in conjunction with eltrombopag in CML patients. Despite the median disease duration of 2.2 years and multiple TKI use in CML patients before enrollment, eltrombopag demonstrated clinical efficacy with complete platelet response of 80% (12 of 15). This compares favorably to what was reported with rIL-11.16 More important, haematologica | 2021; 106(11)

Table 3. Response to eltrombopag in chronic myeloid leukemia patients.

Before eltrombopag None Minor Minor Partial

Cytogenetic response (CML patients) On eltrombopag No. Complete Complete Partial Complete

1 2 4 2

TKI median percentage of standard dose Before eltrombopag On eltrombopag P 55

73.5

0.3

CML: chronic myeloid leukemia; TKI: tyrosine kinase inhibitor.

Table 4. Treatment emergent adverse events.

Event Any Elevated AST/ALT Fatigue Infection Diarrhea Rash Insomnia Hyperglycemia HTN HLD Pleural effusion Headache Hyperbilirubinemia Limb edema Blurred vision Peripheral neuropathy Thrombocytosis Chest pain Myocardial infarction Periorbital edema Non occlusive deep venous thrombosis

No. (%) Grade 3-4

9 (16) 7 (12) 7 (12) 4 (7) 3 (5) 3 (5) 2 (4) 2 (4) 2 (4) 2 (4) 2 (4) 2 (4) 2 (4) 2 (4) 2 (4) 1 (2) 1 (2) 1 (2) 1 (2) 1 (2)

5 (22) 2 (8) 6 (25) 1 (5) 2 (8) 2 (8) 2 (8) 1 (4) 1 (4) 1 (4) 1 (4) -

ALT: alanine aminotransferase; AST: aspartate aminotransferase; HTN: hypertension; HLD: hyperlipidemia.

nine patients (60%) had improvement in cytogenetic responses, likely the result of a more sustained therapy with TKI. Notably, as doses of eltrombopag were increased, LFT elevations were noted in some patients. Conversely, eltrombopag dose interruptions or reductions due to such events or to platelets reaching >200x109/L, occasionally resulted in a drop-in platelet counts. Thus, close monitoring and dynamic management is required, at least during the initial stages of therapy to obtain the maximum effect while maintaining safety. The lack of efficacy among patients with MF could be in part secondary to advanced disease, or possible antagonism between the two medications. Thrombopoietin agonist are dependent on JAK-stat pathway which is inhibited by ruxolitinib.28 Our study has several limitations. It was a small study, and it did not accrue to the target sample size of 29 patients due to slow enrollment making the observation preliminary and requiring confirmation. We also do not have evidence or investigation of any immune mechanisms associated 2857

M. Shoukier et al.

with thrombocytopenia, although we believe it is unlikely that these patients with CML had an immune mediated thrombocytopenia, and uncommon occurrence in this setting. In conclusion, our findings show that eltrombopag doses up to 300 mg may alleviate TKI-associated thrombocytopenia in some patients with CML. No similar benefit has been observed in patients with MF treated with ruxolitinib. Although generally safe, thrombotic events were noted that deserve further investigation, particularly when used in combination with ponatinib. Additional studies are warranted to confirm these observations.

Ariad; FR has received honoraria and is a member of the advisory board of Novarts; TK has received honoraria from Novartis; JC has received research support from BMS, Novartis, Ariad, Chemgenex, and Pfizer.

Disclosures MS, GG-M, JM, LM, KN, KS, and SV have no conflicts of interset to disclose; GB sit on the advisory board of Novartis; EJ has received consultancy honoraria from BMS, Novartis, Pfizer, and

Funding The study was supported in part by MD Anderson Cancer Center Support Grant CA016672 (PI: Dr. Ronald DePinho) and Novartis.

References 1. Kantarjian HM, Cortes JE, O'Brien S, et al. Imatinib mesylate therapy in newly diagnosed patients with Philadelphia chromosome-positive chronic myelogenous leukemia: high incidence of early complete and major cytogenetic responses. Blood. 2003;101(1):97-100. 2. Kantarjian H, Shah NP, Hochhaus A, et al. Dasatinib versus imatinib in newly diagnosed chronic-phase chronic myeloid leukemia. N Engl J Med. 2010; 362(24): 2260-2270. 3. Cortes JE, Jones D, O'Brien S, et al. Nilotinib as front-line treatment for patients with chronic myeloid leukemia in early chronic phase. J Clin Oncol. 2010; 28(3):392-397. 4. Cortes JE, Kim DW, Kantarjian HM, et al. Bosutinib versus imatinib in newly diagnosed chronic-phase chronic myeloid leukemia: results from the BELA trial. J Clin Oncol. 2012;30(28):3486-3492. 5. Cortes JE, Kim DW, Pinilla-Ibarz J, et al. A phase 2 trial of ponatinib in Philadelphia chromosome-positive leukemias. N Engl J Med. 2013;369(19):1783-1796. 6. Hochhaus A, Kantarjian HM, Baccarani M, et al. Dasatinib induces notable hematologic and cytogenetic responses in chronicphase chronic myeloid leukemia after failure of imatinib therapy. Blood. 2007; 109(6):2303-2309. 7. Incyte Corporation (2016) Jakafi (ruxolitinib) [package insert].https://www.jakafi. com/ pdf/prescribing-information.pdf. Accessed 1 March 2016. 8. Verstovsek S, Kantarjian H, Mesa RA, et al. Safety and efficacy of INCB018424, a JAK1 and JAK2 inhibitor, in myelofibrosis. N Engl J Med. 2010;363(12):1117-1127. 9. Harrison C, Kiladjian JJ, Al-Ali HK, et al. JAK inhibition with ruxolitinib versus best available therapy for myelofibrosis. N Engl J Med. 2012;366(9):787-798. 10. Talpaz M, Paquette R, Afrin L, et al. Interim analysis of safety and efficacy of ruxolitinib in patients with myelofibrosis and low platelet counts. J Hematol Oncol. 2013; 6(1):81. 11. Sneed TB, Kantarjian HM, Talpaz M, et al. The significance of myelosuppression dur-

2858

Contributions MS analyzed the data, and wrote the paper; GB designed, and performed research; JM performed research, and analyzed the data; JC designed, performed research, analyzed the data, and wrote the paper; and all authors contributed to data collection, reviewed and approved the manuscript, and shared final responsibility for the decision to submit.

ing therapy with imatinib mesylate in patients with chronic myelogenous leukemia in chronic phase. Cancer. 2004; 100(1):116-121. 12. Verstovsek S, Gotlib J, Gupta V, et al. Management of cytopenias in patients with myelofibrosis treated with ruxolitinib and effect of dose modifications on efficacy outcomes. Onco Targets Ther. 2013;7:1321. 13. Verstovsek S, Mesa RA, Gotlib J, et al. A double-blind, placebo-controlled trial of ruxolitinib for myelofibrosis. N Engl J Med. 2012;366(9):799-807. 14. Quintas-Cardama A, Kantarjian H, O'Brien S, et al. Granulocyte-colony-stimulating factor (filgrastim) may overcome imatinibinduced neutropenia in patients with chronic-phase chronic myelogenous leukemia. Cancer. 2004;100(12):2592-2597. 15. Cortes J, O'Brien S, Quintas A, et al. Erythropoietin is effective in improving the anemia induced by imatinib mesylate therapy in patients with chronic myeloid leukemia in chronic phase. Cancer. 2004;100(11):2396-2402. 16. Aribi A, Kantarjian H, Koller C, et al. The effect of interleukin 11 on thrombocytopenia associated with tyrosine kinase inhibitor therapy in patients with chronic myeloid leukemia. Cancer. 2008; 113(6):1338-1343. 17. Saleh MN, Bussel JB, Cheng G, et al. Safety and efficacy of eltrombopag for treatment of chronic immune thrombocytopenia: results of the long-term, open-label EXTEND study. Blood. 2013;121(3):537545. 18. McHutchison JG, Dusheiko G, Shiffman ML, et al. Eltrombopag for thrombocytopenia in patients with cirrhosis associated with hepatitis C. N Engl J Med. 2007; 357(22):2227-2236. 19. Townsley DM, Dumitriu B, Sheinberg P, Desmond P. Eltombopag added to standard immunosuppressin for aplastic anemia accelerates count recovery and increases response rates. Blood. 2015, 126(23)LBA-2LBA-2. 20. Platzbecker U, Wong RS, Verma A, et al. Safety and tolerability of eltrombopag versus placebo for treatment of thrombocytopenia in patients with advanced

myelodysplastic syndromes or acute myeloid leukaemia: a multicentre, randomised, placebo-controlled, double-blind, phase 1/2 trial. Lancet Haematol. 2015; 2(10):e417-426. 21. Baccarani M, Cortes J, Pane F, et al. Chronic myeloid leukemia: an update of concepts and management recommendations of European LeukemiaNet. J Clin Oncol. 2009; 27(35):6041-6051. 22. Tefferi A, Barosi G, Mesa RA, et al. International Working Group (IWG) consensus criteria for treatment response in myelofibrosis with myeloid metaplasia, for the IWG for Myelofibrosis Research and Treatment (IWG-MRT). Blood. 2006; 108(5):1497-1503. 23. Jain P, Kantarjian H, Alattar ML, et al. Longterm molecular and cytogenetic response and survival outcomes with imatinib 400 mg, imatinib 800 mg, dasatinib, and nilotinib in patients with chronic-phase chronic myeloid leukaemia: retrospective analysis of patient data from five clinical trials. Lancet Haematol. 2015;2(3):e118-128. 24. Bhatia M, Davenport V, Cairo MS. The role of interleukin-11 to prevent chemotherapy-induced thrombocytopenia in patients with solid tumors, lymphoma, acute myeloid leukemia and bone marrow failure syndromes. Leuk Lymphoma. 2007;48(1):9-15. 25. Cantor SB, Elting LS, Hudson DV, Rubenstein EB. Pharmacoeconomic analysis of oprelvekin (recombinant human interleukin-11) for secondary prophylaxis of thrombocytopenia in solid tumor patients receiving chemotherapy. Cancer. 2003;97(12):3099-3106. 26. Townsley DM, Scheinberg P, Winkler T, et al. Eltrombopag added to standard immunosuppression for aplastic anemia. N Engl J Med. 2017;376(16):1540-1550. 27. Roth M, Will B, Simkin G, et al. Eltrombopag inhibits the proliferation of leukemia cells via reduction of intracellular iron and induction of differentiation. Blood. 2012;120(2):386-394. 28. Erickson-Miller CL, Delorme E, Tian SS, et al. Preclinical activity of eltrombopag (SB497115), an oral, nonpeptide thrombopoietin receptor agonist. Stem Cells. 2009; 27(2):424-430.

haematologica | 2021; 106(11)

ARTICLE

Hematopoiesis

Characterization and evolutionary origin of novel C2H2 zinc finger protein (ZNF648) required for both erythroid and megakaryocyte differentiation in humans Daniel C. J. Ferguson,1* Juraidah Haji Mokim,1* Marjolein Meinders,1 Edmund R. R. Moody,2 Tom A. Williams,2 Sarah Cooke,1 Kongtana Trakarnsanga,3 Deborah E. Daniels,1,4 Ivan Ferrer-Vicens,1 Deborah Shoemark,1 Chartsiam Tipgomut,3 Katherine A. Macinnes,1,4 Marieangela C. Wilson,1 Belinda K. Singleton4,5 and Jan Frayne1,4 School of Biochemistry, University of Bristol, Bristol, UK; 2School of Biological Sciences, University of Bristol, Bristol, UK; 3Department of Biochemistry, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand; 4NIHR Blood and Transplant Research Unit in Red Blood Cell Products, University of Bristol, Bristol, UK and 5Bristol Institute for Transfusion Sciences, National Health Service Blood and Transplant (NHSBT), Bristol, UK 1

Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2859-2873

*DCJF and JHM contributed equally as co-first authors.

ABSTRACT

uman ZNF648 is a novel poly C-terminal C2H2 zinc finger (ZnF) protein identified amongst the most dysregulated proteins in erythroid cells differentiated from induced pluripotent stem cells. Its nuclear localization and structure indicate it is likely a DNA-binding protein. Using a combination of ZNF648 overexpression in an induced pluripotent stem cells line and primary adult erythroid cells, ZNF648 knockdown in primary adult erythroid cells and megakaryocytes, comparative proteomics and transcriptomics we show that ZNF648 is required for both erythroid and megakaryocyte differentiation. Orthologues of ZNF648 were detected across Mammals, Reptilia, Actinopterygii, in some Aves, Amphibia and Coelacanthiformes suggesting the gene originated in the common ancestor of Osteichthyes (Euteleostomi or bony fish). Conservation of the C-terminal ZnF domain is higher, with some variation in ZnF number but a core of at least six ZnF conserved across all groups, with the N-terminus recognisably similar within but not between major lineages. This suggests the N-terminus of ZNF648 evolves faster than the C-terminus, however this is not due to exon-shuffling as the entire coding region of ZNF648 is within a single exon. As for other such transcription factors, the N-terminus likely carries out regulatory functions, but showed no sequence similarity to any known domains. The greater functional constraint on the ZnF domain suggests ZNF648 binds at least some similar regions of DNA in the different organisms. However, divergence of the N-terminal region may enable differential expression, allowing adaptation of function in the different organisms.

Correspondence: JAN FRAYNE Jan.Frayne@Bristol.ac.uk Received: April 22, 2020. Accepted: September 15, 2020. Pre-published: October 5, 2020. https://doi.org/10.3324/haematol.2020.256347

Introduction The C H zinc finger (ZnF) proteins represent one of the largest families of regulatory proteins in humans, involved in a variety of cellular activities including development, cell differentiation, genome integrity and tumour suppression (reviewed by Iuchi, 2001).1 The C H motif represents the classical ZnF DNA-binding domain (reviewed by 2), with 675-700 C H ZnF genes identified in the human genome, a large proportion of which are transcription factors.3 Such C H ZnF transcription factors, including GATA 1 and 2, FOG and KLF1, play important roles in differentiation and development of red blood cells (reviewed by Kim and Bresnick4 and Siatecka and Bieker)5. 2

haematologica | 2021; 106(11)

2859

D.C.J. Ferguson et al.

The generation of red blood cells in vitro as an alternative transfusion product is a goal for research groups and blood services globally. In particular, for patients with rare blood group phenotypes for who matched donor blood is difficult to source, and for those requiring regular transfusions who are at risk of alloimmunisation. Various types of stem cells have been used for in vitro erythroid culture systems. Of these, adult peripheral blood (PB) and umbilical cord blood (CB) CD34+ cells differentiate efficiently along the erythroid pathway with high rates of enucleation (60-95%).6-8 However, the erythroblasts have restricted expansion potential using current systems and are thus not a sustainable system, with repeat collections of stem cells required for successive cultures. In contrast, embryonic stem cells (ESC) and induced pluripotent stem cells (iPSC) have the potential to provide an inexhaustible source of progenitors for the generation of large numbers of erythroid cells. Of these, iPSC are particularly attractive as they can be derived from easily accessible adult cells, and without the associated ethical issues of ESC. However, iPSC-derived erythroid cells have historically exhibited terminal differentiation defects and severely impaired enucleation,9-12 although more recent advancements in culture techniques have demonstrated degrees of improvements.13-15 We originally sought to identify proteins potentially involved in the terminal differentiation defect of human iPSC-derived erythroid cells, comparing the proteome of erythroid cells differentiated from three iPSC lines, originating from different cellular origins, with those differentiated from adult PB CD34+ cells.12 The most differentially expressed proteins were γ-, ε- and b-globin, as expected due to the known differences in the globin expression profiles of these cells. However, further interrogation of datasets revealed a novel, previously unstudied ZnF protein, ZNF648, amongst the most differentially expressed proteins, >20-fold lower in both early and late iPSC-derived erythroid cells compared to respective adult erythroid cells. Human ZNF648 is a poly C H ZnF protein which we show is essential for both erythroid and megakaryocyte differentiation. As conservation of protein sequences across species can indicate functional importance, we also explored the conservation and profile of ZNF648 through evolution. ZNF648 originated in the common ancestor of Osteichthyes (Euteleostomi or bony fish). However, conservation of the C-terminal ZnF domain is higher, with some variation in ZnF number but a core of at least six ZnF conserved across all groups, with the Nterminus recognizably similar within, but not between, major lineages. However, unlike the evolution of many C H ZnF transcription factors, this is not due to exonshuffling, as the entire coding region of ZNF648 is within a single exon. If, as with other transcription factors, the N-terminal region of ZNF648 contains regulatory domains, these are potentially novel as no sequence similarity to any known domains was found. The greater functional constraint on the ZnF domain suggests ZNF648 binds at least some similar regions of DNA in the different organisms. However, divergence of the N-terminal region may enable differential expression and hence control of gene expression required for the different environmental conditions of the various organisms. 2

2860

Methods Cell culture and transduction Adult CD34+ cells were isolated from Leukocyte Reduction System (LRS) cones, with informed consent from all donors, and used in accordance with the Declaration of Helsinki and approved by the National Health Service National Research Ethics Committee (reference number 08/H0102/26) and the Bristol Research Ethics Committee (reference 12/SW/0199). Adult CD34+ erythroid and megakaryocyte, K562 and HiDEP-1 cultures were performed as described previously.6,16,17 Lentiviral transduction for protein overexpression was performed with pXLG3 construct as described previously.18 Knockdown studies using pLKO.1 short hairpin RNA (shRNA) plasmid TRCN0000107710 (ZNF648 shRNA), TRCN0000107714 (ZNF648 shRNA2) or a scrambled control (Scr) shRNA were as described previously18 (all designed by the Broad Institute and purchased from Open Biosystems, GE Dharmacon, Lafayette, CO, USA). Dead cell removal was carried out with a dead cell removal kit (Miltenyi Biotech Ltd).

Megakaryocyte ploidy DNA content was measured on day 14 cultured cells. Cells were incubated with anti-CD41 antibody for 30 min at 4°C, washed, fixed in 75% EtOH, stained with propidium iodide and measured using the MACSquant VYB Analyser.

Quantitative polymerase chain reaction analysis of ZNF648 transcript levels Total RNA from cell pellets was extracted using the RNeasy Kit according to the manufacturer’s instructions (Qiagen, Hilden, Germany). RNA quantity and purity were determined using NanoDrop Lite (NanoDrop Technologies, Wilmington, DE, USA), and RNA integrity was assessed by determining the RNA 28S/18S ratio. RNA (500 ng) was reverse-transcribed into cDNA using Oligo(dt)18 primers (Thermo Scientific, Vilnus, Lithuania) and SuperScript IV (Invitrogen, Vilnus, Lithuania). The cDNA products were amplified by quantitative polymerase chain reaction (qPCR) using the SYBR select master mix (Applied Biosystems, Vilnus, Lithuania). All reactions were carried out in triplicate. Real-time qPCR (RT-qPCR) was run in QuantStudio™ 3 Real-Time PCR System (Applied Biosystems). Primers used to detect ectopic ZNF648-GFP: forward 5’-GTGGAAATGTCTGGGAAAGC, reverse 5’-CAATTTGTGTGCGAGACCAC; primers used to detect endogenous ZNF648: forward 5’AGCGTGAGAGACAGAGACACC, reverse 5’-GGATACCTGGGAAATGCAGA. Primers for PABPC1: forward 5’AGCTGTTCCCAACCCTGTAATC, reverse 5’-GGATAGTATGCAGCACGGTTCTG. All primers were synthesized by Sigma-Aldrich. Results were normalized to PABPC1 levels. The threshold cycle (Ct) was determined and the relative gene expression was expressed using the 2-DDCt method.

Transcriptomics and proteomics Transcriptomics using Human Genome U133 Plus 2.0 arrays and tandem mass tag (TMT) proteomics were performed as described previously.12,19,20

Sequence retrieval and analysis Annotated ZNF648 (ENSG00000179930) sequences were retrieved from the Ensembl database,21 with BLASTP and TBLASTN searches used to identify other ZNF648 homologues from RefSeq22 including sequences for which the protein had not been annotated from the genome. In order to identify more divergent ZNF648 homologues and trace the evolution of the

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

family, more sensitive Hidden Markov Model (HMM) searches were used against a local database of metazoan proteomes using HMMER3. ExPASY prosite23 was used to find where the C-terminal cluster of C H motifs started and separated the N-terminal region and C-terminal region. All-versus-all BLASTP searches24 were used to determine pairwise percentage identities of both the N and C terminal regions. These results were visualized using heatmaps generated in R25 with ggplot26 and Viridis (https://github.com/sjmgarnier/viridis). 2

GFP, with bands of the expected size detected by both antibodies, but no protein detected in control cells (Online Supplementary Figure S2). Overexpression of ZNF648 in K562 cells showed ZNF648 exclusively localized to the nucleus by both confocal microscopy and immunoblot of cell fractions (Figure 1B and C), supporting the hypothesis that ZNF648 is a DNA-binding protein.

Overexpression of ZNF648 in erythroid cells reduces proliferation and advances differentiation

Phylogenetics Sequences were aligned with Mafft using the most accurate lINS-i mode.27 We inferred a maximum likelihood phylogeny of ZNF648 (949 aligned sites/1134 positions/42 taxa) using the best-fitting LG+C60+G+F28,29 model in IQ-Tree 1.6.10. We used 1,000 ultrafast bootstraps30,31 to assess support.

Results Structure of ZNF648 The gene for human ZNF648 is mapped to chromosome 1q25.3. It has 2 exons and 1 intron with the entire open reading frame located in the first half of exon 2. The gene codes for a protein of 568 amino acids with 10 C H ZnF arranged adjacent to each other at the C-terminus, a typical architecture for a human ZnF protein. NCBI BLAST (Basic Local Alignment Search Tool; https://blast.ncbi.nlm.nih.gov/) search revealed no homology of ZNF648 with other human transcription factors. About 40% of human ZnF superfamily members have an N-terminal KRAB domain (Krüppel-Associated Box), which can confer transcriptional repression by recruiting KAP-1 (reviewed by Emerson and Thomas32). However homology modeling indicated ZNF648 does not contain this domain, a BTB/POZ effector33 or SCAN domain.34 In considering a role for ZNF648 as a DNA binding protein, tandemly arranged C H ZnF are often connected by short linker regions which play an important role in conferring high-affinity binding to the DNA.35 These linker regions are made up of seven to eight amino acids, with the consensus sequence TGEKP identified in 50% of C H ZnF proteins35 with variations of this sequence also verified.36,37 The presence of this region is often used as a predictor for the DNA binding property of a protein. Interrogation of the ZNF648 sequence revealed the consensus linker sequence between zinc fingers 2 and 3, with alternatively recognised motifs between ZnFs 1 and 2, 5 and 6, 6 and 7, 7 and 8 (Online Supplementary Figure S1). The position of these linkers suggests two potential DNA-binding clusters, firstly ZnF 1, 2 and 3 and secondly ZnF 5, 6, 7 and 8. It is therefore predicted that ZNF648 binds to DNA. 2

Expression profile of ZNF648 during erythropoiesis Adult PB CD34+ cells were cultured in our erythroid culture system which supports efficient erythroid differentiation, with yields of up to 95% reticulocytes on day 19.6,8 Analysis of ZNF648 transcripts throughout differentiation by qPCR showed a slight upward trend, but levels did not reach significance at any time point (Figure 1A). Endogenous ZNF648 protein was below the level of detection by immunoblot. ZNF648 antibodies (sc-249727; Santa Cruz and ab170269; Abcam) were therefore validated by transduction of cells with ZNF648 and ZNF648haematologica | 2021; 106(11)

In order to begin to explore the role of ZNF648 in erythropoiesis we initially expressed exogenous ZNF648 in HiDEP-1 cells, an erythroid cell line created by immortalising erythroid cells differentiated from iPSC.38 HiDEP-1 cells exhibit many of the same properties as iPSCderived erythroid cells, e.g., impaired erythroid differentiation, defective enucleation and fetal/embryonic globin expression.17,38 As with iPSC-derived erythroid cells, the level of ZNF648 protein is substantially lower (approximately 20-fold) in HiDEP-1 cells compared to adult erythroid cells (Frayne unpublished comparative proteomic data for HiDEP-1 vs. primary adult erythroid cells; ZNF648 quantified from 33 peptide spectra matches [PSM]). HiDEP-1 cells were transduced with ZNF648, ZNF648GFP or with a green fluorescent protein (GFP) control construct. Transduction efficiency was over 90%. Expression was confirmed by immunoblot (Figure 2A), with the expected size bands for the untagged and GFPtagged proteins observed. Quantification from abundance values following TMT labeling of tryptic peptides and analysis by nano-LC MS/MS gave an approximately 10-fold increase in ZNF648 (from 23 unique peptides and 75 PSM). Confocal analysis of the ZNF648-GFP transduced cells showed exclusive nuclear localisation (Figure 2B), as for the transduced K562 cells. Cells transduced with ZNF648 and ZNF648-GFP had equivalent viability, but a reduced proliferation rate (∼80% reduction by day 10, P<0.01 for both controls vs. ZNF648 and vs. ZNF848GFP; Figure 2C) and accelerated differentiation; at day 12 significantly fewer polychromatic normoblasts (P<0.001) and significantly more orthochromatic normoblasts (P<0.01) compared to controls (Figure 2D). Enucleation rates were not increased. We also analyzed the effect of ZNF648 overexpression on cultured adult erythroid cells. Erythroid cells differentiated from adult PB CD34+ cells at day 3 in culture were transduced with ZNF648-GFP. GFP+ cells were isolated on day 8 and maintained in our erythroid culture system thereafter. As with HiDEP-1, overexpression of ZNF648 reduced the proliferation rate (e.g., by ∼50% on day 15; Online Supplementary Figure S3A), without affecting viability, and advanced the rate of differentiation, with significantly fewer basophilic normoblasts (P<0.05) and significantly more orthochromatic normoblasts and reticulocytes (both P<0.05) on day 13 of culture compared to controls (Online Supplementary Figure S3B). However, a major issue with such ectopic expression studies is the non-physiological levels of ZNF648 achieved. As the expression of endogenous ZNF648 is below the level of detection by antibody-dependant assays, this likely masks the true role of ZNF648 in erythropoiesis. We therefore took the alternative approach of knocking down ZNF648 in adult erythroid cells. 2861

D.C.J. Ferguson et al.

Figure 1. Expression and localisation of ZNF648 during erythroid differentiation. (A) Expression of ZNF648 transcript levels at days 3, 5, 7, 9 and 11 in culture analysed by quantitative polymerase chain reaction (qPCR), n=2. (B) Western blot of whole cell lysate, cytoplasmic fraction, and nuclear fraction of control K562 cells and K562 cells transduced with ZNF648-GFP probed with antibody to green fluorescent protein (GFP). Antibody to a-tubulin was used as a control for cytoplasmic fraction protein loading. Molecular weight (MW) markers shown on left hand side. (C) images of K562 cells transduced with ZNF648-GFP (i) DAPI staining (ii) ZNF648-GFP fluorescence (iii) Merged image of panels i and ii. Images obtained using a Leica SP5 confocal microscope. Magnification 400x.

2862

Figure 2. Exogenous expression of ZNF648 in the induced pluripotent stem cells line HiDEP-1 impedes proliferation and advances differentiation. HiDEP-1 cells were transduced with ZNF648, ZNF648-GFP or a control green fluorescent protein (GFP) construct. (A) Western blot of HiDEP-1 cells probed with ZNF648 antibody (Abcam). Molecular wieght (MW) markers shown on left hand side (B) Images of control and HiDEP-1 cells transduced with ZNF648-GFP; DAPI nuclear staining in blue. (C) Relative fold expansion of ZNF648 and ZNF648-GFP expressing HiDEP-1 cells compared to untransduced (UT) and GFP control cells. *GFP vs. ZNF648, +GFP vs. ZNF648-GFP. **/++ P<0.01; ***/+++ P<0.001, n=2 each for ZNF648 and ZNF648-GFP ± standard deviation, t-test. (D) proportion of HiDEP-1 cells at each stage of differentiation present at day 12 of culture. ProE: proerythroblast; BasoE: basophilic erythroblast; PolyE: polychromatic erythroblast; OrthoE: orthochromatic erythroblast; Retics: reticulocytes. **P<0.01, ***P<0.001, 2-way ANOVA. N=3 ± standard deviation.

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

ZNF648 knockdown impedes erythroid cell differentiation Erythroid cells differentiated from adult PB CD34+ cells were transduced with a ZNF648 shRNA or with a Scr control shRNA at day 3 in culture. Following puromycin selection, dead cells were removed, and surviving cells seeded at the same number and density (schematic of protocol shown in Figure 3A). Knockdown (KD) of ZNF648 was verified by qPCR which showed an average of ∼85% reduction in transcripts compared to control (Figure 3B).

We also obtained ZNF648 protein abundance values from TMT labeled tryptic peptides analyzed by nanoscale liquid chromatography coupled to tandem mass spectrometry (nano-LC MS/MS) (see below). ZNF648 protein level was reduced by ∼70% (data from 11 PSM). Morphological analysis showed no difference in the population of seeded cells between ZNF648 KD and Scr control cultures, with the majority of cells having the appearance of pro-erythroblasts (Online Supplementary Figure S4). Cells were maintained in our erythroid culture

Figure 3. ZNF648 knockdown impedes erythroid differentiation. Erythroblasts differentiated from adult peripheral blood (PB) CD34+ cells at day 3 in culture were transduced with scrambled (Scr) short hairpin RNA (shRNA) as control or ZNF648 shRNA, puromycin selected and seeded at same cell density along with untransduced cells (UT) that served as a further control. (A) Schematic of experimental design. (B) quantitative polymerase chain reaction (qPCR) of ZNF648 transcript levels normalised to Scr control. (C) Relative fold expansion during differentiation compared to cell number at day 8 in culture. (D) Viability of cells during differentiation assessed by trypan blue exclusion. (E) Proportion of cells at different stages of differentiation at day 19 in culture. ProE: proerythroblast; BasoE: basophilic erythroblast; PolyE: polychromatic erythroblast; OrthoE: orthochromatic erythroblast; Retics: reticulocytes. (F) Reticulocyte yield at day 19 of culture. (G) Morphology of cells at day 19 of culture stained with MayGrünwald-Giemsa, representative images of 3 independent cultures. *P<0.05, **P<0.01, ***P<0.001, ttest, n=3 ± standard devation.

haematologica | 2021; 106(11)

2863

D.C.J. Ferguson et al.

system with counts performed and samples taken for analysis throughout differentiation (day 8, 11, 13, 15, 17 and 19 of culture). ZNF648 KD had a striking, negative effect on the expansion and viability of cells, with significantly fewer cells recorded at all time points from day 13 (P<0.05 to P<0.001, n=3), due to both decreased cell expansion rate and significantly increased cell death (P<0.05 to P<0.01, n=3; Figure 3C and D). However, a proportion of cells do survive, possibly due to higher levels of remaining ZNF648, but their differentiation was impaired, with significantly more (P<0.05, n=3) early erythroid cells (basophilic and polychromatic normoblasts) still present at day 19 of culture compared to control cultures (Figure 3E). In addition, although the percentage of orthochromatic normoblasts appears similar between the controls and ZNF648 KD cultures at day 19 this is due to a combination of more control cells having already enucleated (with ∼30% more reticulocytes; Figure 3E) and significantly more ZNF648 KD cells dying (Figure 3D; P<0.05, n=3), possibly due to impeded enucleation. The reticulocyte yield, calculated as percentage of enucleation multiplied by the cumulative fold expansion,39 is also significantly decreased (P<0.001, n=3) in the ZNF648 KD cultures (Figure 3F). High levels of debris were clearly visible in the ZNF648 KD cultures due to the increased cell death (Figure 3G). We also analyzed four other ZNF648 shRNA. Unlike the above ZNF648 shRNA which targets the 3’ non-coding region of ZNF648, these shRNA all have complementary sequences within the ZNF648 coding region. We were therefore able to determine their efficiency by evaluating reduction in the level of ectopically expressed ZNF648 on immunoblot (Supplementary Figure 5A and B). Three of the shRNA did not reduce ZNF648 levels. However, ZNF648 shRNA2 reduced ectopic ZNF648 protein ∼2-fold (also see Online Supplementary Figure 6A). In order to compare endogenous protein levels, we again used abundance values from mass spectrometry (TMT labelled tryptic peptides analysed by nano-LC MS/MS) of adult erythroid cells transduced with Scr or ZNF648 shRNA2, which showed ∼20% reduction in endogenous ZNF648. Hence, ZNF648 shRNA2 reduces ZNF648 levels, but not to the same magnitude as the above ZNF648 shRNA. We therefore used ZNF648 shRNA2 in our KD protocol. Expansion of cells transduced with ZNF648 shRNA2 was reduced (Online Supplementary Figure 6B), however, the magnitude decrease was less (e.g., at day 15, 4.5-fold less) than for ZNF648 shRNA, in line with the less efficient KD of ZNF648 by ZNF648 shRNA2. Terminal differentiation was also impaired, with significantly more orthochromatic normoblasts (P<0.05, n=3; Online Supplementary Figure S6C) still present at day 19 of cultures compared to controls in which significantly more cells had already enucleated (P<0.05, n=3; Online Supplementary Figure S6C). The reticulocyte yield of ZNF648 shRNA2 cultures at day 19 was also significantly lower (P<0.05) than control cultures (Online Supplementary Figure S6D). The effect of ZNF648 shRNA2 therefore parallels that of the first ZNF648 shRNA but the effects are less pronounced due to the smaller magnitude decrease in ZNF648 levels.

Rescue of ZNF648 knockdown improves phenotype We also rescued ZNF648 KD cells by co-transduction of ZNF648 shRNA and ZNF648 coding region tagged with 2864

GFP (Online Supplementary Figure S7); as shown above (Figure 2) ZNF648-GFP and untagged ZNF648 have the same effect on proliferation and viability. Co-transduction was required due to death of cells immediately following ZNF648 KD. As ZNF648 shRNA targets the 3’ non-coding region of ZNF648 it cannot target the ZNF648 transgene. The increase in ZNF648 transcript levels following transduction is shown in the Online Supplementary Figure S7C (day 8 panel), with densitometry values from ZNF648 western blot showing >50-fold increase in ZNF648 protein level. A schematic of the protocol is shown in the Online Supplementary Figure S7A, with GFP+ cells isolated from the co-transduced population by fluorescence-activated cell sorting (FACS) and cells from all cultures seeded at the same density in erythroid culture. The viability of the ZNF648 KD population again declined to <50% of control cells by day 13 of culture (Online Supplementary Figure S7B). In contrast, viability of the rescued population mirrored that of the Scr shRNA control for the first 3 days, but declined to a greater extent thereafter, although was higher than that of the ZNF648 KD population to day 19. The lower viability of the rescued culture seen by day 13 compared to control may be due to a more rapid loss of ZNF648. As the ZNF648 transgene lacks a 3’ non-coding region it is less stable than endogenous transcripts, decreasing its half-life. This is exacerbated as differentiation continues, as the nuclei condense, and transcriptional activity reduces. A reduced level of ZNF648-GFP mRNA and protein is indeed seen by day 13 of culture (Online Supplementary Figure S7C and D). ZNF648 rescue also significantly improved the differentiation potential of the cells (Online Supplementary Figure S7E and F), with no basophilic or polychromatic normoblasts remaining at day 19 in culture, in contrast to the ZNF648 KD cultures, along with ~30% more reticulocytes. There was also notably less debris in the rescue than ZNF648 KD cultures, more clearly seen at day 11 (Online Supplementary Figure S7F). However, the enucleation rate of the rescued population was not fully recovered, being still ~35% lower than the control. Again, this is likely due to physiological levels of ZNF648 not being achieved. Nevertheless, the data clearly supports that the observed phenotype following ZNF648 KD is due to reduced levels of ZNF648.

Perturbation of the transcriptome and proteome following ZNF648 knockdown. Analysis of transcriptome In order to determine the mechanism by which loss of ZNF648 impedes erythroid differentiation, and results in reduced viability, we initially looked at transcript levels for key erythroid transcription factors and proteins following ZNF648 KD in adult erythroid cells on day 8 of culture (5 days post transduction with ZNF648 shRNA) by PCR. Expression of transcription factors GATA1, FOG1, KLF1 and BCL11A-XL were unchanged, as was AHSP. However, there was a decrease in expression of HBB (b-globin), SCL4A1 (Band 3), GYPA (glycophorin A) and CA1 (carbonic anhydrase 1) (Online Supplementary Figure S8), possibly indicating some defect or delay in differentiation even in these early cells, despite no obvious morphological differences compared to control cells. There was also reduced expression of E2F2, consistent with decreased cell expansion rate.40 haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

For a more in-depth and global analysis of the effect of ZNF648 KD, comparative transcriptomic analysis was performed. Following transduction of adult erythroid cells with Scr or ZNF648 shRNA, total RNA was extracted at day 8 from three independent cultures and used to screen human genome arrays. Following statistical analysis, 299 probe ID (representing 208 unique genes) were down-regulated ≥2-fold and 124 probe ID (representing 84 unique genes) were up-regulated ≥2-fold upon ZNF648 KD (Online Supplementary Table S1). All genes with increased or decreased expression ≥2fold were entered into the Reactome Knowledgebase (https://reactome.org) to reveal significantly over-represented pathways in the datasets.41,42 Identified pathways, sorted by P-value, are provided unless too few proteins (<5) were assigned or pathway association was not significant (P>0.05). Of the genes with decreased expression, 153 were found in Reactome with the data demonstrating signifi-

cant over-representation of megakaryocyte genes, with almost all identified pathways associated with megakaryocyte or platelet regulation and function (Figure 4A). Of the genes with increased expression, 47 were found in Reactome but no pathways were significantly over-represented.

Analysis of proteome Transcript abundance does not directly correlate with protein abundance, due to substantial post-transcriptional, translational and protein degradation regulation.43 We therefore investigated the change in proteome following ZNF648 KD to further, and more directly inform how ZNF648 KD disrupts erythropoiesis, helping delineate its role in this process. TMT comparative proteomics was performed on adult erythroid cells on day 8 in culture following transduction with Scr or ZNF648 shRNA, as for the transcriptomic analysis above. A total of 7,868 unique proteins were

Figure 4. Megakaryocyte-associated proteins and pathway classification of transcripts and proteins decreased in level following ZNF648 knockdown. Total RNA and whole cell protein lysates were prepared from erythroid cells differentiated from adult peripheral blood (PB) CD34+ cells transduced with scrambled (Scr) or ZNF648 short hairpin RNA (shRNA), at day 8 in culture. RNA was used to screen human genome arrays, with transcripts decreased in level by ≥2-fold following ZNF648 knockdown (KD) entered into Reactome. Data from three independent KD experiments (A). Tryptic peptides were prepared from cell lysates and labeled with Tandem Mass Tags followed by nanoscale liquid chromatography coupled to tandem mass spectrometry (nano-LC MS/MS) with (B) relative decrease in level of megakaryocyte associated proteins following ZNF648 KD, identified amongst the 123 unique proteins decreased by ≥2-fold. Data from two independent KD experiments ± standard deviation, and (C) Reactome analysis of the proteins decreased in level. In (A) and (C) dark shades indicate genes/proteins apportioned to the pathway shown on the left. Pathways shown with minimum probability of over representation in the dataset, corrected for false discovery rate of P<0.006 for transcripts and P<1.42x10 -5 for proteins.

haematologica | 2021; 106(11)

2865

D.C.J. Ferguson et al.

quantified, with 123 unique proteins decreased in level by ≥2-fold following ZNF648 KD (Online Supplementary Table S2A). In keeping with the transcriptomic data, interrogation of the proteomic dataset revealed a striking number of these 123 proteins are known megakaryocyte proteins (14%; Figure 4B). Following analysis with Reactome, almost all assigned pathways were again associated with megakaryocyte or platelet regulation and function (Figure 4C; 86 of the 123 proteins found in Reactome). Of note, six of the ten proteins with greatest magnitude decrease in level were megakaryocyte-associated proteins (Online Supplementary Figure S9). These proteins were all found to decrease in level during normal erythropoiesis (Online Supplementary Figure S10), hence their decrease on ZNF648 KD is not due to delayed differentiation. Megakaryocyte proteins in Figure 4B were also decreased on transduction of cells with ZNF648 shRNA2 but by 2030% compared to 50-70% with ZNF648 shRNA, in line with the magnitude of ZNF648 KD by the two shRNA. 175 unique proteins were found to be increased in level following ZNF648 KD (Online Supplementary Table 2B), 119 of which were found in Reactome. Interrogation of

the pathways assigned (Figure 5) revealed proteins typically expressed by cells of the monocyte lineage (e.g., apolipoproteins, complement cascade and regulatory components, SERPINS) as well as immunoglobulin chains normally synthesized by B cells, although some of the above are also expressed by granulocytes and megakaryocytes. Consistent with the PCR data, no change in the level of GATA1, FOG1, KLF1, BCL11A-XL and AHSP, as well as key erythroid proteins ITGA4, CD71, CD44, a-Spectrin or b-Spectrin was detected in the transcriptomic or proteomic datasets. A decreased level of b-globin and glycophorin A was detected in both datasets, and of Band 3 and CA1 in the proteomic dataset, in line with the PCR data. The raw data for the transcriptomic analysis did also show a decrease for E2F2, SCL4A1 and CA1 but levels were variable and therefore did not pass the statistical cut-offs. Overall, the data demonstrate a distorted genetic readout on ZNF648 KD, with altered expression of genes normally associated with cells of other haematopoietic lineages. In particular, there was a striking effect on proteins associated with megakaryocyte or platelet function.

Figure 5. Pathway classification of proteins increased in level following ZNF648 knockdown. Tandem Mass Tag labeling and nanoscale liquid chromatography coupled to tandem mass spectrometry (nano-LC MS/MS) was performed on whole cell lysates from erythroid cells transduced with ZNF648 or scrambled (Scr) (control) short hairpin RNA (shRNA) at day 8 in culture. Proteins increased in level following ZNF648 knockdown were entered into Reactome. Dark red indicates proteins apportioned to the pathway shown on the left. Top 20 pathways shown with minimum probability of over representation in the dataset, corrected for false discovery rate, of P<2.11x10-5.

2866

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

Proteins increased in level following ZNF648 knockdown are also increased in erythroid cells differentiated from three induced pluripotent stem cell lines As ZNF648 is reduced in level in erythroid cells differentiated from our three iPSC lines, we cross referenced the most abundantly upregulated proteins (≥4-fold; 112 proteins) following ZNF648 KD with our previously published proteomic dataset of erythroid cells differentiated from the iPSC lines compared to differentiation stage matched normal adult erythroid cells.12 Of the 112 proteins, 69 were found in the iPSC proteomic dataset, and of these 46 (67%; Online Supplementary Table S3) were also increased in the iPSC-derived erythroid cells, with the same pathways assigned by Reactome (Online Supplementary Figure S11). Proteins that increased in level by <4-fold did not show a correlative increase in the iPSC-derived erythroid cells. Analysis of all proteins increased in the iPSC-derived erythroid cells (≥2-fold; see Trakarnsanga et al.12) for datasets) highlighted dysregulation of the same proteins and pathways as above, but also proteins associated with other pathways, showing greater dysregulation in these cells (Online Supplementary Figure S12). However, there was no correlation between proteins decreased in level following ZNF648 KD and those decreased in iPSCderived versus adult erythroid cells, suggesting differences in the regulation or programming of iPSC-derived erythroid cells compared to adult cells.

ZNF648 knockdown impedes megakaryocyte differentiation We compared the abundance of ZNF648, as well as abundance of the megakaryocyte proteins that decreased in level in erythroid cells following ZNF648 KD, in erythroblasts and megakaryoblasts. For this we took advantage of available RNA sequencing (RNAseq) data for normal erythroblasts and megakaryoblasts differentiated from CD34+ cells accessible via the Bloodspot website.44,45 The expression of ZNF648 was an order of magnitude lower in erythroblasts compared to megakaryoblasts (see also PCR analysis of ZNF648 transcript levels in erythroblasts and megakaryoblasts, Online Supplementary Figure S13). Similarly, the expression of all 17 megakaryocyte proteins shown in Figure 4B were between 2- to 8-fold lower in erythroid cells. We confirmed expression of some by PCR (Online Supplementary Figure S13). Higher levels of ITGA2B, PF4, GP9 and ITGB3 in megakaryocytes compared to erythroblasts have been reported previously.46 As a control, we looked at the expression of erythroid markers Band 3, glycophorin A, ITGA4, CD36, CD44, a-Spectrin and b-Spectrin in the Bloodspot dataset, all of which were at a higher level in erythroblasts than megakaryoblasts as expected. Finally, using the same ZNF648 shRNA, we knocked down ZNF648 in megakaryoblasts to determine if ZNF648 also plays a regulatory role in this lineage. Adult CD34+ cells transferred to megakaryocyte culture medium were transduced on day 1 with ZNF648 shRNA or with the control Scr shRNA, followed by puromycin selection. Untransduced cells served as a further control (schematic of experimental design shown in Figure 6A). As expected, there was similar loss of CD34 expression with time in all cultures (Figure 6B). In order to determine the effect of ZNF648 KD on megakaryopoiesis, megakaryocyte membrane protein marker expression haematologica | 2021; 106(11)

was assessed during differentiation. There was no significant difference between control and ZNF648 KD cultures from day 3 through to day 8, although the level of CD42b was consistently lower in the latter. However, on day 12 the levels of CD61, CD41 and CD42b were significantly lower in the ZNF648 KD cultures, indicating a defect in megakaryocyte differentiation (Figure 6B). In order to further assess megakaryocyte differentiation, we measured the levels of CD61 and CD41 as indicators of commitment to the megakaryocyte lineage, and CD61 and CD42b as indicators of mature megakaryocytes on day 14 of culture. As seen in Figure 6C both megakaryocyte commitment and maturity were significantly reduced following ZNF648 KD. Finally, a unique characteristic of megakaryocytes is their capacity to become polyploid. We therefore measured the ploidy status of CD61+ cells. There was a significant shift to lower ploidy status for the ZNF648 KD cells (Figure 6D). This is also seen in the cell images where fewer large cells are present in ZNF648 KD cultures (Fig. 6E, upper panel) and those detected have less nuclear content (Figure 6E, lower panel), indicating the cells are not able to mature properly to large polyploid megakaryocytes. Overall, the data support a role for ZNF648 in megakaryocyte differentiation.

ZNF648 sequence conservation and evolution origins Conservation of protein sequences across species can indicate functional importance. Having established a role for ZNF648 in humans, we next investigated the conservation and profile of ZNF648 through evolution to gain further information on this novel protein. We used BLAST and HMMER sequence similarity searches to identify homologues of ZNF648 across eukaryotes, then inferred phylogenetic trees to determine orthology relationships and pinpoint the evolutionary origin of the ZNF648 gene. Among mammals, ZNF648 is conserved (Online Supplementary Table S4). However, the C-terminus, which contains ten tandemly arranged C H ZnF motifs in all, is more highly conserved than the N-terminus (94-100% amino acid sequence identity in the C-terminus, human residues 279-568, compared to 50-98% N-terminal sequence identity, residues 1-278). The N-terminal region was however conserved between closely related family members (Online Supplementary Table S5). Broadening our search beyond mammals, we detected orthologues of ZNF648 across Reptilia, in some Amphibia and Aves, in Coalacanthiformes and across the bony fish (Actinopterygii) but not outside this group, suggesting that the gene originated in the common ancestor of Osteichthyes (Euteleostomi) after their evolutionary divergence from gnathostomes. As in mammals, ZNF648 C-terminal conservation (28-100%) is higher than N-terminal conservation (21-97%), with the N-terminus recognisably similar within, but not between, major lineages (Figure 7). This suggests that the N-terminus of ZNF648 evolves faster than the C-terminus, perhaps due to reduced or lineage-specific functional constraints. We hypothesize that the C-terminus (containing the ZnF motifs) evolves slowly because it performs the same function, binding to a DNA target motif, while the N-terminus may vary in function across lineages. We used sensitive profile-profile sequence searches but were unable to detect any significant similarity between the N-terminus of the ZNF648 protein family and any characterized data2

2867

D.C.J. Ferguson et al.

Figure 6. ZNF648 knockdown impedes megakaryocyte differentiation. Isolated adult peripheral blood CD34+ cells were transferred to megakaryocyte differentiation medium and transduced with scrambled (Scr) short hairpin RNA (shRNA) as control or ZNF648 shRNA followed by puromycin selection. Untransduced cells served as a further control. (A) Schematic of experimental design and expression profile of CD34, CD41, CD61 and CD42b during normal megakaryopoiesis, (B) expression of membrane markers CD34, CD41 (platelet glycoprotein IIb), CD61 (platelet glycoproteins IIIa) and CD42b (GPIb a) analysed by flow cytometry with antibodies CD41PE, CD61-APC (both from Biolegend), CD34-VB BD and CD42b-PE BD (both from Pharming). Data was acquired with a MacsQuant VYB Analyser using a plate reader, n=6. (C) Proportion of CD41/CD61 and CD61/CD42b positive cells on day 14 of culture, n=6 ± standard error of the mean. (D) Ploidy status of cells at day 14 of culture, n=3. (E) Morphology of cells at day 14 of culture stained with May-Grünwald-Giemsa. Upper panel 100x magnification, lower panel 400x magnification. Images representative of three independent cultures. *P<0.05, **P<0.01, ***P<0.001, Student’s t-test.

2868

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

Figure 7. Pairwise sequence identity and phylogeny of representative ZNF648 orthologues across Osteichthyes. (A) (i) C-terminus (human residues 279-568;), (ii) N-terminus (human residues 1278). Warm colors indicate higher pairwise percentage identity. The C-terminus (containing the ZnF motifs) is conserved across Osteichthyes, but the N-terminus is much less conserved, showing high identity only within major lineages.

haematologica | 2021; 106(11)

2869

D.C.J. Ferguson et al.

base sequences, suggesting that this represents a novel functional domain; in other ZnF proteins, the non-ZnF region of the protein acts as an effector domain, for example binding to other proteins to bring about transcription repression.32 Notwithstanding, proteins or regions with low sequence similarity can still have homologous function via adoption of a similar structure47 so we analysed the Nterminal half of ZNF648 from a range of species to see if residues shared a common fold. All sequences were aligned using Clustal Omega48 and a secondary structure prediction carried out by PsiPred.49 From this secondary structure alignment, confidence that the N-terminal half

of these proteins share a common fold is low. However, there are two significant patches of sequence with shared charge similarity between all mammalian sequences but also found in reptilia and coelacanth (Online Supplementary Figure S14), which suggests that even if they do not share the same fold they may have targeting or protein-protein interaction sites in common. In the C-terminus, the number of detected C H zinc finger motifs varies among taxa, from e.g., 11 in Reptilia, ten in mammals to seven in Aves (see schematic in the Online Supplementary Figure S15). All outside the mammalian clade have an additional conserved C H ZnF at the N-terminal end of the ZnF domain, lost in mammals. Of 2

Figure 8. ZNF648 phylogenetic tree. A maximum likelihood phylogeny of ZNF648 was inferred (949 aligned sites/1134 positions/42 taxa) using the best-fitting LG+C60+G+F model in IQ-Tree 1.6.10. Branch supports are 1,000 ultrafast bootstraps. ZNF648 is conserved across the major lineages of Osteichthyes (Euteleostomi), but several lineage-specific losses have occurred. The topology of the tree is congruent with our understanding of species relationships, suggesting that the gene traces back to the common ancestor of Osteichthyes.

2870

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

potential interest only Reptilia and Coelacanth have all 10 conserved mammalian ZnF, which along with the conserved patches of charge similarity in their N-terminal region suggests a possible closer functional relationship between ZNF648 in these species. While the ZNF648 gene has a broad phylogenetic distribution in bony fishes, it has been lost repeatedly and independently in a number of descendant lineages, particularly in many lineages of birds, and in the common ancestor of all Lepidosaurians (snakes, lizards and tuatara). The patchy distribution of ZNF648 orthologues across these groups appears to be the result of independent losses rather than gains (for example, by de novo origin of a new C H ZnF-motif containing sequence) because the retained sequences follow the species tree (Figure 8). Phylogenetic trees inferred separately for the C-terminal ZnF domain and N-terminal region (Online Supplementary Figure S16A and B) are consistent with the hypothesis that both domains have evolved on the same underlying tree, indicating that while N-terminal sequence conservation is low the two domains are likely to have evolved together as a unit since the common ancestor of bony fish. 2

Discussion Our data show that ZNF648 is required for both erythroid and megakaryocyte differentiation in humans, with the order of magnitude higher level of ZNF648 in megakaryoblasts suggesting differential levels of ZNF648 may be required by the two lineages. In addition, the striking reduction in abundance of proteins associated with the megakaryocyte lineage following ZNF648 KD in erythroid cells supports a function for these proteins in erythropoiesis. Of interest, the level of transcription factor Fli1 was found to be decreased in erythroid cells in both the transcriptomic and proteomic datasets following ZNF648 KD. Fli1 is critical for megakaryopoiesis,50 with cross-antagonism between FLI-1 and KLF1, a transcription factor essential for erythroid differentiation,5 implicated in the control of erythroid versus megakaryocyte lineage differentiation.51 However, Fli1 mutant mice have aberrant red blood cell development, as well as severely impeded megakaryopoiesis,52 suggesting a requirement for Fli1 or its relative level also in erythropoiesis. Fli1 directly regulates the expression of many megakaryocyte proteins, including ITGA2B, GP1BA, GP9 and PF4,53,54 all decreased in level on ZNF648 KD in erythroid cells. Therefore, ZNF648 may, at least in part, exert its effect upstream of Fli1. Furthermore, ZNF648 may suppress differentiation to other haematopoietic lineages, as the level of proteins normally expressed by e.g., monocytes is increased following its KD, contributing to the distorted genetic readout, defective phenotypes and cell death observed. Interrogation of RNAseq datasets via the Bloodspot website44,45 showed ZNF648 is also expressed by neutrophils, monocytes, CD8+ T cells and hematopoietic stem cells as well as megakaryocytes and erythroid cells. Megakaryocytes have the highest level of expression, with the other cell types having similar levels to erythroid cells. We do not yet know how ZNF648 functions, but its specific nuclear localisation and presence of canonical linker sequences between ZnFs 1-3 and 5-8 support it as a DNAhaematologica | 2021; 106(11)

binding transcriptional regulator. It is known that binding of adjacent poly C H ZnF proteins to DNA is carried out by 24–75% of the ZnF, with two to three successive fingers responsible for the specific DNA motif interaction, with other fingers facilitating binding. For example, TFIIIA with nine ZnF binds its DNA recognition motif via fingers 1-3, with finger 5 and 7-9 also interacting with the DNA. Similarly, analysis of the linker sequences between ZnF in ZNF648 suggest specific DNA recognition by fingers 1-3 as the consensus linker sequence, inducing highest binding affinity, is found between fingers 2 and 3, but that fingers 5-8 also interact. The remaining ZnF in such proteins may interact with other molecules, including RNA and proteins (reviewed by Iuchi1). Notwithstanding, it has also been suggested that different groupings of adjacent ZnF in poly-ZnF proteins bind different specific sites in the genome, thus independently regulating a transcriptional programme.32 Alternatively, ZNF648 could interact with DNA indirectly via interaction with other factors, act as a scaffold or modulate the binding of other DNA-binding proteins. Future work elucidating the target DNA-binding site(s) and interacting partners for ZNF648 will help reveal the specific function of this protein. In order to further determine the functional significance of ZNF648 we explored its sequence conservation across a wide range of species and through evolution, identifying a clear common ancient ancestor, that of Osteichthyes (Euteleostomi/bony fish). This is in contrast to the large number of poly-ZnF proteins encoded in the human genome which are postulated to be a recent invention, the ancestral size of this gene family being small with rapid expansion on the primate lineage; a substantial proportion of the proteins having no mouse ortholog.32 The biological function of a large majority of these proteins is still unknown. Instead ZNF648 may have occurred with a much earlier expansion of C H -ZnF proteins in metazoans.55 There has been clear differential divergence within the ZNF648 sequence through evolution. The C-terminal sequential C H ZnF domain is highly conserved across mammals, reptilians, actinopterygii, in a member of the Sarcopterygii order Coelacanth and the amphibian Caecilian. In contrast the remaining N-terminal region of the protein is much less conserved; it is recognizably similar within each of the major groups surveyed (mammals, reptiles, actinopterygians), with greater conservation between more closely related members, but similarity between groups is low. ZNF648 orthologues were clearly distinguishable from other C H ZnF proteins in our phylogenetic analyses, indicating that the conserved sequences we detected across Osteichthyes do not represent other, more distantly related ZnF families. Rapid evolution of the N-terminal regions of poly C H ZnF proteins, often involving loss or gain of regulatory KRAB and SCAN domain, has been observed previously for mammalian poly C H ZnF proteins and predicted to be due to an exon-shuffling mechanism.56 In such proteins the ZnF domain is often located within a single exon, with the N-terminal domain on separate exon(s) and separated from the ZnF exon often by a large intron.57,58 However, analysis of the intron/exon organisation of ZNF648 in a range of organisms (mammals, reptilians and actinopterygii) showed the entire coding sequence is within a single exon. It is therefore likely that the common ancestor had a “complete” ZNF648 gene (i.e., with both the N-terminal 2

2871

D.C.J. Ferguson et al.

region and C-terminal ZnF domain), but that the N-terminal region has evolved more quickly than the C-terminal domain. Greater functional constraint on the DNA interacting ZnF domain suggests ZNF648 binds at least some similar regions of DNA in the different organisms. However, divergence of the N-terminal region, which likely contains binding sites for proteins that regulate ZNF648 function but has a potentially novel functional domain, having no sequence similarity to known regulatory domains, may enable differential expression of genes in the different organisms. This could essentially create control points connecting cellular requirements to DNA-binding, and hence control of gene expression required for the different environmental conditions of the various organisms. Finally, ZNF648 in humans is also expressed in other non-hematopoietic cell types (Online Supplementary Figure S17) indicating a broader regulatory role. Hence, although all organisms with a ZNF648 gene have red blood cells and platelets (or the more primitive thrombocytes in lower vertebrates) with ZNF648 possibly involved in regulating the differentiation of these cells throughout evolution, it may additionally or alternatively be involved in other processes. In conclusion, in humans the novel C H ZnF protein ZNF648 is involved in the regulation of both erythroid and megakaryocyte differentiation. Functional importance of ZNF648 is further implied by its maintenance through evolution from the common ancestor of Osteichthyes, with conservation within the ZnF domain but divergence of the N-terminal region allowing adaptation of function in the different organisms. 2

Disclosures No conflicts of interest to disclose.

References 1. Iuchi S. Three classes of C2H2 zinc finger proteins. Cell Mol Life Sci. 2001;58(4):625635. 2. Klug A. The discovery of zinc fingers and their development for practical applications in gene regulation and genome manipulation. Q Rev Biophys. 2010;43(1):1-21. 3. Vaquerizas JM, Kummerfeld SK, Teichmann SA, et al. A census of human transcription factors: function, expression and evolution. Nat Rev Genet. 2009;10(4):252-263. 4. Kim SI, Bresnick EH. Transcriptional control of erythropoiesis: emerging mechanisms and principles. Oncogene. 2007;26(47):67776794. 5. Siatecka M, Bieker JJ. The multifunctional role of EKLF/KLF1 during erythropoiesis. Blood. 2011;118(8):2044-2054. 6. Griffiths RE, Kupzig S, Cogan N, Mankelow TJ, et al. Maturing reticulocytes internalize plasma membrane in glycophorin A-containing vesicles that fuse with autophagoBlood. exocytosis. before somes 2012;119(26):6296-6306. 7. Timmins NE, Athanasas S, Gunther M, et al. Ultra-high-yield manufacture of red blood cells from hematopoietic stem cells. Tissue Eng Part C Methods. 2011;17(11):1131-1137. 8. Kupzig S, Parsons SF, Curnow E, et al. Superior survival of ex vivo cultured human reticulocytes following transfusion into mice. Haematologica. 2017;102(3):476-483. 9. Dias J, Gumenyuk M, Kang H, et al.

2872

Contribution The project was conceived and supervised by JF; experiments were designed by JF, DCJF, JHM, MM and BKS Those for evolution analysis by ERRM and TAW; experimental work was carried out by DCJF, JHM, ERRM, MM, SC, DED, CT, K.T, IFW and BKS; data analysis was carried out by DCJF, JHM, ERRM, TAW, MM, SC, DS, KM, MCW, IFV, BKS and JF; figure preparation was by DCJF, JHM, ERRM, TAW, MM, SC, DS, BKS and JF wrote the manuscript; DCJF, JHM, TAW, DED, DS, KM and BKS edited the manuscript. Acknowledgements The authors would like to thank Dr Kate Heesom, Director of the Bristol University Proteomic Facility, UK for performing Mass Spectrometry, the University of Bristol Genomics Facility, Dr Lee Carpenter, Oxford for providing the iPSC, Profs Yukio Nakamura and Ryo Kurita, RIKEN BioResource Research Center, Japan for the HiDEP-1 cells. Funding The work was funded by the Government of Brunei Darussalam via an In-Service Training Scholarship to JHM, The Wellcome Trust (grant numbers 087430/Z/08 and 102610), BrisSynBio a BBSRC/EPSRC Synthetic Biology Research Centre (BB/L01386X/1) and NIHR Blood and Transplant Research Unit (NIHR BTRU) in Red Cell Products (IS-BTU-1214-10032). This manuscript presents independent research funded by the National Institute for Health Research (NIHR). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. T.A.W. is funded by a Royal Society University Research Fellowship (UF140626). ERR is funded by a Royal Society Fellows Enhancement Award to TAW (RGF\EA\180199)

Generation of red blood cells from human induced pluripotent stem cells. Stem Cells Dev. 2011;20(9):1639-1647. 10. Dorn I, Klich K, Arauzo-Bravo MJ, et al. Erythroid differentiation of human induced pluripotent stem cells is independent of donor cell type of origin. Haematologica. 2015;100(1):32-41. 11. Lapillonne H, Kobari L, Mazurier C, et al. Red blood cell generation from human induced pluripotent stem cells: perspectives for transfusion medicine. Haematologica. 2010;95(10):1651-1659. 12. Trakarnsanga K, Wilson MC, Griffiths RE, et al. Qualitative and quantitative comparison of the proteome of erythroid cells differentiated from human iPSCs and adult erythroid cells by multiplex TMT labelling and nanoLC-MS/MS. PLoS One. 2014;9(7): e100874. 13. Hansen M, Varga E, Aarts C, et al. Efficient production of erythroid, megakaryocytic and myeloid cells, using single cell-derived iPSC colony differentiation. Stem Cell Res. 2018;29:232-244. 14. Olivier EN, Zhang S, Yan Z, et al. PSC-RED and MNC-RED: albumin-free and lowtransferrin robust erythroid differentiation protocols to produce human enucleated red blood cells. Exp Hematol. 2019;75:31-52. 15. Razaq MA, Taylor S, Roberts DJ, et al. A molecular roadmap of definitive erythropoiesis from human induced pluripotent stem cells. Br J Haematol. 2017;176(6):971983.

16. Salunkhe V, Papadopoulos P, Gutierrez L. Culture of megakaryocytes from human peripheral blood mononuclear cells. Bioprotocol. 2015;5(21):e1639. 17. Trakarnsanga K, Wilson MC, Lau W, et al Induction of adult levels of beta-globin in human erythroid cells that intrinsically express embryonic or fetal globin by transduction with KLF1 and BCL11A-XL. Haematologica. 2014;99(11):1677-1685. 18. Satchwell TJ, Hawley BR, Bell AJ, et al. The cytoskeletal binding domain of band 3 is required for multiprotein complex formation and retention during erythropoiesis. Haematologica. 2015;100(1):133-142. 19. Singleton BK, Burton NM, Green C, et al. Mutations in EKLF/KLF1 form the molecular basis of the rare blood group In(Lu) phenotype. Blood. 2008;112(5):2081-2088. 20. Trakarnsanga K, Griffiths RE, Wilson MC, et al. An immortalized adult human erythroid line facilitates sustainable and scalable generation of functional red cells. Nat Commun. 2017;8:14750. 21. Cunningham F, Achuthan P, Akanni W, et al. Ensembl 2019. Nucleic Acids Res. 2019;47 (D1):D745-D751. 22. O'Leary NA, Wright MW, Brister JR, et al. Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation. Nucleic Acids Res. 2016;44(D1):D733-745. 23. de Castro E, Sigrist CJ, Gattiker A, et al. ScanProsite: detection of PROSITE signature matches and ProRule-associated functional

haematologica | 2021; 106(11)

ZNF648 regulatory role in erythropoiesis

and structural residues in proteins. Nucleic Acids Res. 2006;34(Web Server issue):W362365. 24. Camacho C, Coulouris G, Avagyan V, et al. BLAST+: architecture and applications. BMC Bioinformatics. 2009;10:421. 25. Team RC. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2017. 26. Wickham H. ggplot2. Wiley interdisciplinary reviews: computational statistics. 2011;3(2):180-185. 27. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30(4):772-780. 28. Lartillot N, Philippe H. A Bayesian mixture model for across-site heterogeneities in the amino-acid replacement process. Mol Biol Evol. 2004;21(6):1095-1109. 29. Le SQ, Gascuel O. An improved general amino acid replacement matrix. Mol Biol Evol. 2008;25(7):1307-1320. 30. Hoang DT, Chernomor O, von Haeseler A, et al. UFBoot2: improving the ultrafast bootstrap approximation. Mol Biol Evol. 2018;35(2):518-522. 31. Hoang DT, Vinh LS, Flouri T, et al. MPBoot: fast phylogenetic maximum parsimony tree inference and bootstrap approximation. BMC Evol Biol. 2018;18(1):11. 32. Emerson RO, Thomas JH. Adaptive evolution in zinc finger transcription factors. PLoS Genet. 2009;5(1):e1000325. 33. Albagli O, Dhordain P, Deweindt C, et al. The BTB/POZ domain: a new protein-protein interaction motif common to DNA- and actin-binding proteins. Cell Growth Differ.1995;6(9):1193-1198. 34. Edelstein LC, Collins T. The SCAN domain family of zinc finger transcription factors. Gene. 2005;359:1-17. 35. Wolfe SA, Nekludova L, Pabo CO. DNA recognition by Cys2His2 zinc finger proteins. Annu Rev Biophys Biomol Struct. 2000;29:183-212. 36. Brayer KJ, Segal DJ. Keep your fingers off my DNA: protein-protein interactions mediated by C2H2 zinc finger domains. Cell Biochem Biophys. 2008;50(3):111-131.

haematologica | 2021; 106(11)

37. Liu Q, Segal DJ, Ghiara JB, et al, 3rd. Design of polydactyl zinc-finger proteins for unique addressing within complex genomes. Proc Natl Acad Sci U S A. 1997;94(11):5525-5530. 38. Kurita R, Suda N, Sudo K, et al. Establishment of immortalized human erythroid progenitor cell lines able to produce enucleated red blood cells. PLoS One. 2013;8(3):e59890. 39. Daniels DE, Downes DJ, Ferrer-Vicens I, et al. Comparing the two leading erythroid lines BEL-A and HUDEP-2. Haematologica. 2020;105(8):e389-e394. 40. Pilon AM, Arcasoy MO, Dressman HK, et al. Failure of terminal erythroid differentiation in EKLF-deficient mice is associated with cell cycle perturbation and reduced expression of E2F2. Mol Cell Biol. 2008;28(24):7394-7401. 41. Fabregat A, Korninger F, Viteri G, et al. Reactome graph database: efficient access to complex pathway data. PLoS Comput Biol. 2018;14(1):e1005968. 42. Fabregat A, Sidiropoulos K, Viteri G, et al. Reactome diagram viewer: data structures and strategies to boost performance. Bioinformatics. 2018;34(7):1208-1214. 43. Vogel C, Marcotte EM. Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat Rev Genet. 2012;13(4):227-232. 44. Bagger FO, Kinalis S, Rapin N. BloodSpot: a database of healthy and malignant haematopoiesis updated with purified and single cell mRNA sequencing profiles. Nucleic Acids Res. 2019;47(D1):D881-D885. 45. Chen L, Ge B, Casale FP, et al. Genetic drivers of epigenetic and transcriptional variation in human immune cells. Cell. 2016;167(5):1398-1414. 46. Macaulay IC, Tijssen MR, Thijssen-Timmer DC, et al. Comparative gene expression profiling of in vitro differentiated megakaryocytes and erythroblasts identifies novel activatory and inhibitory platelet membrane proteins. Blood. 2007;109(8):3260-3269. 47. Koonin EV, Galperin MY. Sequence - evolution - function: computational approaches in Comparative Genomics. Boston: Kluwer Academic; 2003.

48. Sievers F, Wilm A, Dineen D, et al. Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega. Mol Syst Biol. 2011;7:539. 49. Jones DT. Protein secondary structure prediction based on position-specific scoring matrices. J Mol Biol. 1999;292(2):195-202. 50. Jackers P, Szalai G, Moussa O, et al. Etsdependent regulation of target gene expression during megakaryopoiesis. J Biol Chem. 2004;279(50):52183-52190. 51. Starck J, Cohet N, Gonnet C, Sarrazin S, et al. Functional cross-antagonism between transcription factors FLI-1 and EKLF. Mol Cell Biol. 2003;23(4):1390-1402. 52. Kawada H, Ito T, Pharr PN, et al. Defective megakaryopoiesis and abnormal erythroid development in Fli-1 gene-targeted mice. Int J Hematol. 2001;73(4):463-468. 53. Pang L, Xue HH, Szalai G, et al. Maturation stage-specific regulation of megakaryopoiesis by pointed-domain Ets proteins. Blood. 2006;108(7):2198-206. 54. Wang X, Crispino JD, Letting DL, et al. Control of megakaryocyte-specific gene expression by GATA-1 and FOG-1: role of Ets transcription factors. EMBO J. 2002;21(19):5225-5234. 55. Najafabadi HS, Garton M, Weirauch MT, et al. Non-base-contacting residues enable kaleidoscopic evolution of metazoan C2H2 zinc finger DNA binding. Genome Biol. 2017;18(1):167. 56. Tadepally HD, Burger G, Aubry M. Evolution of C2H2-zinc finger genes and subfamilies in mammals: species-specific duplication and loss of clusters, genes and effector domains. BMC Evol Biol. 2008;8:176. 57. Bellefroid EJ, Marine JC, Ried T, et al. Clustered organization of homologous KRAB zinc-finger genes with enhanced expression in human T lymphoid cells. EMBO J. 1993;12(4):1363-1374. 58. Huntley S, Baggott DM, Hamilton AT, et al. A comprehensive catalog of human KRABassociated zinc finger genes: insights into the evolutionary history of a large family of transcriptional repressors. Genome Res. 2006;16(5):669-677.

2873

ARTICLE Ferrata Storti Foundation

Hemostasis

Targeting shear gradient activated von Willebrand factor by the novel single-chain antibody A1 reduces occlusive thrombus formation in vitro Thomas Hoefer,1* Akshita Rana,2* Be’eri Niego,2 Shweta Jagdale,1,2 Hugo J. Albers,3,4 Elizabeth E. Gardiner,5 Robert K. Andrews,2 Andries D. van der Meer,3 Christoph E. Hagemeyer1,2# and Erik Westein1,2#

Haematologica 2021 Volume 106(11):2874-2884

1 Baker Heart and Diabetes Institute, Melbourne, Victoria, Australia; 2Australian Center for Blood Diseases, Monash University, Melbourne, Victoria, Australia; 3Applied Stem Cell Technologies, University of Twente, Enschede, the Netherlands; 4BIOS Lab-on-aChip, University of Twente, Enschede, the Netherlands and 5ACRF Department of Cancer Biology and Therapeutics, John Curtin School of Medical Research, Australian National University, Canberra, New South Wales, Australia

*TH and AR contributed equally as co-first authors. # CEH and EW contributed equally as co-senior authors.

ABSTRACT

Correspondence: ERIK WESTEIN erik.westein@monash.edu Received: February 20, 2020. Accepted: September 11, 2020. Pre-published: September 21, 2020.

ntraluminal thrombus formation precipitates conditions such as acute myocardial infarction and disturbs local blood flow resulting in areas of rapidly changing blood flow velocities and steep gradients of blood shear rate. Shear rate gradients are known to be pro-thrombotic with an important role for the shear-sensitive plasma protein von Willebrand factor (VWF). Here, we developed a single-chain antibody (scFv) that targets a shear gradient specific conformation of VWF to specifically inhibit platelet adhesion at sites of shear rate gradients (SRG) but not in areas of constant shear. Microfluidic flow channels with stenotic segments were used to create SRG during blood perfusion. VWF-GPIba interactions were increased at sites of SRG compared to constant shear rate of matched magnitude. The scFv-A1 specifically reduced VWF-GPIba binding and thrombus formation at sites of SRG but did not block platelet deposition and aggregation under constant shear rate in upstream sections of the channels. Significantly, the scFv A1 attenuated platelet aggregation only in the later stages of thrombus formation. In the absence of shear, direct binding of scFv-A1 to VWF could not be detected and scFVA1 did not inhibit ristocetin induced platelet agglutination. We have exploited the pro-aggregatory effects of SRG on VWF dependent platelet aggregation and developed the shear gradient-sensitive scFv-A1 antibody that inhibits platelet aggregation exclusively at sites of SRG. The lack of VWF inhibition in non-stenosed vessel segments places scFV-A1 in an entirely new class of anti-platelet therapy for selective blockade of pathological thrombus formation while maintaining normal hemostasis.

https://doi.org/10.3324/haematol.2020.250761

Introduction ©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.

2874

Atherothrombotic events precipitating myocardial infarction and ischemic stroke are the most common causes of death worldwide.1 These events are triggered by intraluminal plaque rupture causing an inappropriate platelet response which leads to occlusion of the vessel lumen.2 As a preventative measure, prescribed antiplatelet drugs such as aspirin and clopidogrel target major platelet activation pathways such as the TXA2- and ADP-P2Y12-pathways, thereby reducing platelet reactivity.3 While this is a clinically proven strategy to reduce cardiovascular events, this approach can also cause serious side effects, most notably bleeding, as the targeted pathways also play a critical role in normal hemostasis thereby partly offsetting the benefits gained from these drugs.4-6 Thus, a strategy to uncouple thrombosis from hemostasis to target one without the other is an urgent unmet clinical need. Platelet aggregates formed during a hemostatic response typically do not grow

haematologica | 2021; 106(11)

Targeting VWF to inhibit occlusive thrombus formation

beyond 50% of the vessel lumen7,8 a restriction that is insufficient to markedly change the shear rates at the site of injury.9 However, studies into the hemodynamic regulation of thrombus formation have led to the concept that rapidly changing shear conditions at the site of a cardiovascular insult contribute significantly to the pathological thrombotic response.10,11 Plaque rupture with its release of highly thrombogenic content may cause the formation of a large thrombus that reduces the vessel lumen which in turn causes local steep increases in blood shear rates, also termed shear rate gradients (SRG). In recent years, the effects of SRG flow patterns on platelet reactivity and von Willebrand factor (VWF) unfolding and activation have been increasingly studied.11-13 VWF is a large blood-borne protein critical to platelet adhesion in areas of high shear such as in arterioles or larger arteries with luminal constriction. In normal circulation, VWF has a globular conformation, which unfolds under high shear stress thereby exposing binding epitopes, most importantly the platelet GPIbα binding domain A1. This unfolding of VWF is referred hereafter as VWF activation. It is thought that SRG are one to two orders of magnitude more efficacious in unfolding VWF, compared to constant shear.14,15 Indeed, we and others have previously demonstrated that SRG strongly unfold and subsequently activate VWF and that shear gradient dependent platelet aggregation mechanisms promote arterial thrombosis in a VWF-dependent fashion.15,16 Given the central role of VWF in platelet adhesion at arterial shear rates, and even more so under pathological flow conditions (i.e., SRG), we hypothesized that activated (i.e., unfolded) VWF is a potential target for a new class of anti-thrombotic therapy. Here we investigated the effect of SRG on VWF-mediated platelet adhesion and thrombus formation. Moreover, we report on the generation of the single-chain antibody (scFv) A1, designed to target shear activated VWF. Here, we developed a scFv based on the sequence of parent immunoglobulin (Ig) G antibodies raised against the monomeric 39/34-Kd fragment of VWF which encompasses the VWF-A1 domain and a proximal N-terminal sequence to VWF-A1. Our scFv targets a shear gradient specific conformation of VWF to inhibit VWF-platelet binding specifically at sites of SRG but not in areas of constant shear. This antibody specifically inhibits VWF-platelet interactions at pathological shear but not under physiological shear regimens and may therefore form the basis of a new and safer class of anti-thrombotic therapy.

Methods A more detailed description of the methods is published in the Online Supplementary Appendix.

Single-chain antibody generation A single-chain antibody was designed by combining the variable regions of the light and heavy chains of two selected antibodies from a panel of parent IgG antibodies that we raised against a 34/39 kDa fragment of VWF.17

Blood sampling Blood samples were collected in trisodium citrate (0.32% weight/volume [w/v] final) and immediately processed for further

haematologica | 2021; 106(11)

use. The study was approved by the Alfred Hospital’s (Melbourne, Australia) ethics committee no 67/15.

Microfluidic chips and in vitro flow perfusions In-house designed polydimethylsiloxane (PDMS) microfluidic chips (Figure 1A) contained channels of 52 mm height with a semicircular stenotic section, creating 80% lumen reduction. Chips were manufactured as previously reported.16 Citrated whole blood was incubated with DiOC6 (0.5 mg mL-1) (Sigma-Aldrich) and drawn through the microfluidic channels by a programmable syringe pump (Legato 130, KD Scientific) over a coated surface of 100 mg/mL collagen type III or type I.

Computational fluid dynamics and calculations Fluid dynamics were simulated with a COMSOL Multiphysics 4.2 laminar flow module applied to three-dimensional meshes. For all simulations, no-slip boundary conditions and Newtonian fluids were assumed. Both the computational fluid dynamical simulation and the calculations of fluid dynamical and geometrical parameters were carried out with the assumption that the channels and vessels were filled with Newtonian fluids. This assumption is valid because the shear rates are in a regime in which the non-Newtonian behavior of blood is limited.18

Platelet agglutination Platelet agglutination in response to 0.75 mg mL-1 ristocetin (Helena Biosciences, Gateshead, UK) was performed for 10 minutes (min) on an AggRam system from Helena Laboratories, USA. Isolated platelets (150 or 300 mL at 3x108/mL) and platelet poor plasma (50 mL) were incubated with 2, 2.5, 4, 5, 40 and 80 mg mL-1 scFv A1.

Western blot Denatured full-length VWF (Mybiosource, USA) was electrophoresed at 0.1, 10, 100 ng and 1 mg using 5% SDS-polyacrylamide gel at 125 V for 90 min. The protein was transferred onto PVDF membrane at 90 V for 90 min followed by incubation with 1 mg mL-1 scFv A1 overnight and the membrane was probed with horse-radish peroxidase (HRP)-conjugated detection antibody (1:2,500, Sigma). Later, the membrane was stripped and reprobed with 1 mg mL-1 polyclonal sheep anti-VWF antibody (Abcam), followed by donkey anti-sheep HRP-conjugated secondary antibody (1:2,500, R&D Systems, USA).

Enzyme-linked immunosorbent assay Purified full-length VWF or isolated A1 domain was coated at 0.1, 1 and 10 mg mL-1 in the presence of ristocetin (0.75 mg mL-1) and blocked with 2% bovine serum albumin followed by incubation with 5 mg mL-1 of scFv A1 or sheep anti-VWF antibody and detection by HRP-conjugated secondary antibody

BLItz Bio-layer interferometry was performed on a BLItz System (Pall Forte Bio, USA) using protein-A biosensors tagged with VWF-A1 domain (130 mg mL-1). Remaining binding sites were blocked by exposure to a polyclonal mouse IgG for 5 min. Sequential adhesion of scFv-A1 at 0, 20, 40 and 100 mg mL-1 was monitored for 5 min per concentration.

Results Shear rate gradients activate von Willebrand factor (VWF) and reduce platelet rolling velocity on VWF/collagen type III As previously reported16 exacerbated thrombus forma2875

T. Hoefer et al.

tion at atherosclerotic geometries is dependent on the degree of intraluminal sidewall protrusion which creates SRG. Thus, to elucidate the underlying mechanism of exacerbated thrombus formation under SRG, platelet adhesion and VWF-activation were studied at sites of SRG utilizing a microfluidic platform. These channels exhibit a stenotic feature which resembled rheological parameters in larger carotid arteries in humans (i.e., slope of stenosis, wall shear rates at the inlet versus stenotic region, and maximal flow elongational rate) while keeping required blood volume low.16 Channels incorporated a semi-circular stenotic geometry of 600 µm in length to reduce the chan-

nel width by 80 % from 300 mm to 60 mm (Figure 1A). Computational fluid dynamic analysis indicated that the wall shear rate was symmetrically distributed around the stenosis, with an 8-fold higher shear rate in the apex region compared with the pre-stenotic segment. At an input wall shear rate of 300 s-1, platelets traveling through the stenosis are therefore exposed to an increase in shear rate to 2,400 s-1 and a shear gradient (Ds-1) of approximately 7.1 s-1 mm-1 followed by an identical shear decrease in the outlet of the stenosis (Figure 1B). Platelet rolling velocities, a measure of VWF-GPIba bond strength, were reduced at SRG (assessed in a 140 mm zone around the

Figure 1. Shear rate gradients increase von Willebrand factor-GPIb bond strength and reduce platelet rolling velocity. (A) Complex flow dynamics were studied in microfluidics exhibiting a semi-circular stenotic site with a diameter of 600 µm which reduced the channel width from 300 mm to 60 µm. (B) Computational fluid dynamics identified the stenosis apex as a zone with an 8-fold increase in shear compared to the inlet, followed by a negative but equal drop in shear rate in the stenosis outlet. (C) Abciximab-treated (20 mg mL-1), citrated whole blood, labelled with 0.5 mg mL-1 DiOC6 was perfused over a collagen III-coated surface to promote binding of blood borne von Willebrand factor (VWF) under flow and subsequent platelet adhesion. Input shear rate was set to 300 s-1 which then increased to 2,400 s-1 in the apex of the stenosis. Control channels with straight walls were used to generate a constant shear rate of 2,400 s-1. Data represented as mean ± standard error of them mean; n=8; *P<0.05 (unpaired t-test). (D) Platelet rolling velocity on VWF-collagen III was assessed at shear rate gradients (SRG) (platelets rolling through the apex of the stenotic site) or constant shear (upstream of the stenotic site) using particle tracking software. Platelet rolling velocities were divided into three bins (0-3 mm s-1; >3-6 mm s-1; >6 mm s-1) and presented as relative frequencies. Data represented as mean ± standard error of the mean (SEM); n=5; *P<0.05 (multiple unpaired t-tests). (E) Surface coverage of abciximab-treated, citrated whole blood, labeled with DiOC6 perfused over a collagen III-coated surface. Input shear rates ranged from 125 s-1 up to 8,000 s-1 for stenotic channels, resulting in peak shear rates of 1,000 s-1 to 36,000 s-1 in the apex of the stenosis; or matched shear rates in a parallel control channel exhibiting a constant shear rate throughout. (F) Collagen type III-coated channels were incubated either statically or perfused at 1,000 s-1 or 10,000 s-1 with platelet poor plasma in the absence of SRG to allow VWF adhesion to collagen. Abciximab-treated, citrated whole blood, labeled with DiOC6 was perfused over the differentially coated collagen III/VWF matrix as described in (B). Platelet rolling velocity on VWF/collagen III was assessed at SRG (platelets rolling through the apex of the stenotic site) or constant shear (upstream of the stenotic site) for all three coating regimens using particle tracking software. Data represented as mean ± standard error of the mean (SEM); n=3; *P<0.05 (multiple unpaired t-tests) (G) Abciximab-treated, citrated whole blood, labeled with DiOC6 was perfused over a collagen III matrix, through 300 µm wide channels incorporating semi-circular stenotic segments which reduced the channel width to 60 µm. The diameters of semi-circular stenotic elements ranged from 600 mm and 1,000 mm to 2,000 µm. Whereas Δ shear of all three stenotic segments was equal, levels of elongational flow differed. (H) Platelet rolling velocity on VWF-collagen III was assessed the same conditions as in (C) under SRG (platelets rolling through the apex of the stenotic site) or constant shear (upstream of the stenotic site) using particle tracking software. Data represented as mean ± SEM; n=3; *P<0.05 (Dunnett’s multiple comparison one-way ANOVA).

2876

haematologica | 2021; 106(11)

Targeting VWF to inhibit occlusive thrombus formation

apex of the stenosis with a maximal shear rate of 2,400 s-1) compared to matched constant wall shear rate of 2,400 s-1 (3.9±0.2s-1 vs. 5.4±0.7s-1) (Figure 1C), with 52.9±2.5% of slow-rolling platelets at SRG compared to 36±5.7% at constant shear (Figure 1D). This finding indicates that shear rate gradients reduce platelet translocation, presumably through the unfolding of the VWF-A1 domain leading to stronger VWF-GPIb interactions. In order to confirm this finding, platelet surface coverage, a measure of VWF-platelet interaction was measured in stenotic channels over a wide range of input shear rates ranging from 125 s-1 to 8,000 s-1, resulting in shear rates of 1,000 s-1 to 36,000 s-1 in the stenosis apex and SRG of approximately 3 mms-1 to 1,000 mms-1. SRG facilitated higher levels of transient platelet adhesion, with maximal platelet adhesion of 20.1±3.4 % at 2,500 s-1 and sustained platelet adhesion up to 7,000 s-1, (7.89±1.86%; P=0.0462) while platelet adhesion rapidly decreased under constant shear from a maximum of 19.3±1.9 % at 875 s-1 and was abolished above 2,000 s-1 (13.6±2.0%; P=0.0341) (Figure 1E). In order to elucidate the parameters critical for platelet-VWF interaction under SRG, VWF deposition onto collagen type III was tested under either static conditions or under flow in the absence of SRG. Following VWF deposition onto collagen, the coated strips were overlaid with micro-channels featuring straight or stenotic sections. Platelet rolling velocities at SRG or constant shear

rate (CSR) were assessed at an effective wall shear rate of 2,400 s-1. The rolling velocity of platelets on collagen strips that were statically incubated with VWF was higher (SRG: 5.03±0.31 mm s-1; CSR: 5.44±0.2 mm s-1) than those in channels incubated with VWF at a shear rate of 1,000 s-1 (SRG: 4±0.26 mm s-1; CSR: 3.95±0.21 mm s-1, P<0.05) (Figure 1F) indicating a role for flow in VWF deposition to collagen. However, when VWF was bound to collagen before platelet perfusion, the SRG specific increase of the GPIba-VWF bond strength was lost (Figure 1F), suggesting that VWF deposition needs to occur under real-time SRG to have subsequently increased engagement of GPIba. In order to test the effect of the steepness of the stenosis gradient on platelet rolling velocity, channels containing different stenotic segment lengths were manufactured. These ranged from 600 to 2,000 mm in length, resulting in the same peak shear rate in the apex, but different SRG, ranging from 7.1 s-1 mm-1 to 3.2 s-1 mm-1 (Figure 1G). Interestingly, platelet rolling velocities correlated with the magnitude of SRG. Rolling velocities were negatively correlated with the stenosis gradient, producing the slowest rolling velocity (4.32±0.23 mm s-1) when traveling through a 600 mm stenosis and the highest (5.89±0.35 mm s-1) when traveling through a 2,000 mm stenosis. However, the rolling velocity in the absence of SRG (i.e., under constant shear) was still higher (6.72±0.28 mm s-1) (Figure 1H). In order to confirm that previously observed platelet-

Figure 2. Platelet adhesion to collagen III is von Willebrand factor-GPIb dependent. (A) Blood, containing DiOC6-labelled platelets, was treated with increasing concentrations of OS-1. Blood samples were drawn through collagen III coated channels at 1,500 s-1. Fluorescence intensity was measured and expressed as relative platelet adhesion. (B) Citrated whole blood treated with 20 mg mL-1 Abciximab and 3 mM OS-1 was perfused through collagen III-coated microfluidics at constant shear rates ranging from 125 s-1 to 2,000 s-1. Data represented as mean ± standard error of the mean (SEM); n=4; *P<0.05 (Dunnett’s multiple comparison oneway ANOVA). (C) Abciximab-treated (20 mg mL-1) platelets were treated with DAPT (30 mM aspirin plus 300 mM 2-MeSAMP) or control and drawn through stenotic or straight channels. Platelet rolling velocity on collagen III bound von Willebrand factor (VWF) was assessed under constant shear (C) and at shear rate gradients (SRG) (D) using particle tracking software. Rolling velocities were divided into three bins (0-3 mm s-1; >3-6 µm s-1; >6 mm s-1) and presented as relative frequencies (C, D) Data represented as mean ± standard error of the mean; n=3; *P<0.05 (multiple unpaired t-tests).

haematologica | 2021; 106(11)

2877

T. Hoefer et al.

surface interactions in our experimental setup were exclusively mediated through VWF-GPIba, blood was incubated with the GPIba inhibitor OS-1.19 Blood perfusion at a shear rate of 1,500 s-1 resulted in platelet adhesion of 74.2±19.5% surface area coverage which was concentration-dependently reduced by OS-1. Platelet adhesion was completely abolished in the presence of 3 mM OS-1 (Figure 2A). Furthermore, inhibition of platelet GPIba by OS-1 (3 mM) caused a sharp decrease in numbers of adherent platelets from above 500 s-1 (53±4 platelets per field) to 2,000 s-1 (11±4 platelets per field) (Figure 2B). Having demonstrated that the observed interaction was GPIba-VWF dependent with only minor if any contribution of direct collagen III interaction above 500 s-1, we tested whether platelet activation played a role in the observed differential rolling velocity. Thus, platelets were treated with aspirin plus the P2Y12 inhibitor 2-MeSAMP. The drug treatment did not affect rolling velocity at sites of SRG (Figure 2D), nor at constant shear (Figure 2C), highlighting that the observed rolling velocity on collagen III is solely determined by the biophysical interaction of GPIba with VWF independent of platelet activation.

Single-chain antibody A1 specifically inhibits shear rate gradients-dependent platelet-von Willebrand factor interaction SRG have been shown to strongly activate VWF, which promotes platelet deposition at pathological shear rates

thereby exacerbating thrombus formation. Specifically blocking SRG-mediated VWF-activation would potentially open up a new avenue of limiting excessive thrombus growth at the latter stages of atherothrombosis, thereby keeping the blood vessels patent. In order to investigate whether SRG-activated VWF is a potential therapeutic target, we produced several scFv, based on the sequences of the variable heavy and light chain regions derived from a panel of monoclonal antibodies raised against a 34/39 kDa VWF fragment incorporating the VWF A1 domain. The new scFv were specifically tested for their capacity to inhibit VWF-platelet interactions selectively at sites of SRG. ScFv A1 reduced plateletVWF adhesion specifically in the stenosis inlet, where shear gradients are the greatest from 6,187±1,097 platelets/mm2 to 5,158±1,032 platelets/mm2; P=0.0013 (Figure 3A). No differences in platelet adhesion were observed at the apex (3,934±892 platelets/mm2 control; 3,805±783 platelets/mm2 scFv A1, P=0.7932); nor in the stenosis outlet (5,042±1,167 platelets/mm2 control; 4,475±1,199 platelets/mm2 scFv A1; P=0.07). Surprisingly, scFv A1 showed an opposite effect in the straight section under constant shear, leading to increased platelet-deposition (2,047±232 platelets/mm2 control; 3,096±690 platelets/mm2 scFv A1; P=0.0195) (Figure 3A). Platelet rolling velocities were largely unaffected by the addition of scFv A1 except for the stenosis apex where a mild increase in rolling velocity was observed upon addition of

Figure 3. Single-chain antibody A1 differentially inhibits von Willebrand factor-platelet interaction at constant shear or gradients of shear. (A) Abciximab-treated (20 mg mL-1), citrated whole blood, labeled with 0.5 mg ml-1 DiOC6 containing single-chain antibody (scFv) A1 (5 mg mL-1) or control was perfused over a collagen III-coated surface as described in Figure 1B. Platelet adhesion was assessed at the stenosis inlet, apex, outlet and upstream areas using particle tracking software. Data represented as matched individual data points and mean ± standard error of the mean (SEM); n=5; *P<0.05. (C) Relative von Willebrand factor (VWF) amount corrected for the level per platelet in platelet aggregates under shear rate gradient (SRG) and constant shear. Data represented as mean ± SEM; n=3-4; (multiple t-test).

2878

haematologica | 2021; 106(11)

Targeting VWF to inhibit occlusive thrombus formation

scFv A1 (4.41±0.1 mm s-1 control; 4.6±0.1 mm s-1 scFv A1; P=0.038) (Figure 3B). The relative amount of VWF per platelet deposited on aggregates in areas of SRG and constant shear was not reduced by scFv A1 (Figure 3C). Taken together these data suggest that scFv A1 interferes with a soluble fraction unfolded VWF which prevents VWF from engaging with surface-bound platelets.

Single-chain antibody A1 specifically inhibits shear rate gradient-exacerbated thrombus formation Taking the findings of the effect of scFv A1 on platelet rolling velocity and adhesion further into a more complex scenario, we tested its capacity to specifically inhibit SRGmediated thrombus formation. Blood containing scFv A1 (5 mg mL-1) or control antibody was drawn through stenosis channels coated with collagen type I. ScFv A1 had no effect on platelet deposition at constant shear in the straight channel sections (Figure 4A, left panel). In order to highlight the differences in growth rate between constant shear and SRG, the aggregate sizes over time were nor-

malized to those measured at 1 min into the flow (set as t=0 in the graph) and defined as relative thrombus growth. Relative thrombus growth under constant shear ranged from 2.11±0.23 versus 1.9±0.36 at 2,000 s-1 to 1.48±0.16 versus 1.62±0.2 at 1,500 s-1 (Figure 4B, left panels). However, under SRG conditions, scFv A1 significantly reduced thrombus formation from 1.85±0.28 to 1.17±0.07 (P<0.0001) at 2,000 s-1 and from 1.58±0.22 to 1.2±0.09 (P<0.001) at 1,500 s-1 (Figure 4B, right panels). Inhibition of thrombus formation by scFv A1 was absent at a dose of 2.5 mg mL-1 whereas 5 mg mL-1 and 10 mg mL-1 showed inhibition and to a similar degree (Online Supplementary Figure S1). Next, the inhibitory effects of ScFv A1 were compared to the platelet GPIba inhibitor OS-1 after 15 minutes of blood flow. Addition of 0.1 µM OS-1, a concentration that allowed residual platelet adhesion as determined in Figure 2A, inhibited platelet deposition under both constant and shear gradient conditions (Figure 4C). At 2,000 s-1 constant shear OS-1 reduced thrombus formation from while scFv

Figure 4. Single-chain antibody A1, but not OS-1 selectively inhibits thrombus formation in stenotic channel segments, a site of shear rate gradient, but not at constant shear rate. Confocal microscopy images of platelet aggregates (A) and quantification (B) of aggregate growth. Citrated whole blood, labeled with 0.5 mg mL-1 DiOC6, containing 5 mg mL-1 single-chain antibody (scFv) A1 (black) or control (red), was perfused over a collagen type I matrix for 10 minutes at an effective shear rate of 1,500 and 2,000 s-1 in the stenosis apex (shear rate gradient [SRG]) or upstream (constant shear). Note the inhibitory effect of scFv A1 at SRG but not under constant shear. Data represented as mean ± standard error of the mean (SEM); n=3-4; *P<0.05 (two-way ANOVA). (C) Confocal images of platelet aggregate formation as in (A) in the presence of the GPIb inhibitor OS-1 (0.1 mM) or control. (D) Quantification of platelet aggregate growth after 15 minutes at sites of SRG or at constant shear in the presence of scFv A1 (5 mg/mL), OS-1 (0.1 mM) or control (D). Data represented as mean ± SEM; n=3-6; *P<0.05 (two-way ANOVA).

haematologica | 2021; 106(11)

2879

T. Hoefer et al.

A1 had no effect. A similar pattern, although weaker, was observed at 1,500 s-1. In contrast, at SRG (300-2,000 s-1) and (187-1,500 s-1), both OS-1 and scFv A1 inhibited thrombus formation. These data highlight the shear selective nature of ScFV A1 compared to a generic GPIb-VWF inhibitor.

Single-chain antibody A1 does not interact with von Willebrand factor under no shear conditions Next we investigated whether scFv A1 could also inhibit VWF-GPIb binding under zero shear conditions. VWF dependent platelet agglutination was tested in a ristocetininduced platelet agglutination test (RIPA). Since the exposure of VWF to ristocetin or shear reportedly exposes similar epitopes within VWF,20 we hypothesized that scFv A1 may inhibit ristocetin stimulated platelet agglutination. However, scFv A1 up to 80 mg mL-1 (16 times higher than used in the microfluidic flow device) did not result in delayed or reduced agglutination of isolated platelets by ristocetin (0.75 mg mL-1) in the presence of plasma VWF. However, when plasma VWF was preincubated with scFv A1 and ristocetin to facilitate their interaction before platelet binding, a minor inhibitory effect of scFv A1 on platelet agglutination was observed (Online Supplementary Figure S2A). Western blotting of denatured VWF with scFv A1 also did not show any complex formation (Online Supplementary Figure S2B). Similarly, scFv A1 did not bind to immobilized full-length VWF or isolated VWF A1 domain in an enzyme-linked immunosorbent assay test (Online Supplementary Figure S2C). Consistent with this result, we also did not observe any binding between immobilized VWF full length or VWF A1 domain with scFv using the BLItz binding assay (Online Supplementary Figure S2D).

Single-chain antibody A1 dampens thrombus formation in the latter stages only During the growth of an intraluminal thrombus, the local shear environment becomes progressively more complex. Here we set out to mimic this highly dynamic process and characterize the inhibitory effects of scFv A1 within this process. Large platelet aggregates were allowed to form in straight channels and create their own local shear gradient environment during thrombus growth. Whole blood was perfused at 1,000 s-1 over discrete patches of coated collagen type I surrounded by albumin coated areas (Figure 5A). The presence of the albumin-collagen interface generates a sharp front of aggregates which grow to significant heights relative to the 52 mm high channels and as a result create a local steep shear gradient. After an initial phase of homogenous platelet accumulation across the entire collagen patch, platelet aggregation progressively increased at the front area of the patch (2.0±0.2-fold increase; front over rear) (Figure 5B). This effect could not be explained by depletion of platelets at the boundary layer as it was observed on sequential patches within a single channel. Next, we tested scFv A1 for its capacity to inhibit thrombus formation in this progressive shear gradient flow model. ScFv A1 selectively attenuated platelet aggregation at the front area of the collagen patch (Figure 5C) however, scFv A1 did not reduce platelet aggregation in the rear area where shear rates were constant and SRG were absent. Computational fluid dynamics (CFD) analysis of a representative experiment revealed that platelet aggregates at the front area of the patch experienced increased surface 2880

shear rates compared to those in the back area of the patch (Figure 5D; upper panel), concomitant with platelet deposition patterns. As expected, scFv A1 reduced platelet deposition at the front area resulting in reduced aggregate surface shear rates (Figure 5D; lower panel). The inhibitory effect of scFv A1 was observed even though the calculated maximum SRG, expressed as s-1/mm, was lower in this flow model (4.9 and 3.1 s-1/mm for control and scFv A1 respectively) compared to the SRG calculated in the microfluidic stenosis channels used in Figures 1-4 (7.1 s1 /mm) (Figure 5D). Time-lapse confocal microscopic analysis of platelet aggregate formation across the collagen patch revealed a lag phase of up to 180 seconds where small aggregates formed throughout the patch, followed by a growth phase at the front area of the patch where scFv A1 was inhibitory (Figure 5E). Indeed, quantitative analysis showed that scFv A1 inhibited platelet aggregation only at the front of the patch but not at the rear (Figure 5F).

Discussion In this study, we demonstrate that SRG exacerbate VWF-dependent platelet aggregation through increased VWF-GPIba bond strength and that this process can selectively be inhibited using an antibody strategy that targets a SRG-sensitive epitope within the VWF A1-domain. Thus, SRG are a potential novel drug target for the prevention of occlusive thrombus formation. The blood flow dynamics in areas of flow restriction, stenosis or mural thrombus formation have long been known to create gradients in shear rate, or SRG. However, the activating effects of SRG on VWF-dependent platelet aggregation are still largely unknown. We have previously demonstrated the potential of micro-shear gradients in promoting thrombus formation.10 Following on from this finding, we have shown exacerbated thrombus formation in the outlet of stenotic vessel sections to be VWF dependent.16 While high constant shear in the range of approximately 5,000 s-1 to 8,000 s-1 is required to unfold and thereby activate VWF,14 SRG dramatically reduce this shear range by 10-fold. This means SRG mediated platelet aggregation can occur in virtually all vessel beds where a flow constriction occurs. Previous studies investigating the role of the VWF-GPIba axis in platelet adhesion and aggregation used various approaches ranging from mathematical simulations,21,22 single molecule force probe pulling,23,24 VWF multimer analysis25 to functional platelet adhesive behavior under flow conditions.26-28 However, it is difficult to establish functional links between these studies and, in isolation, they provide limited insight into the overall mechanism of VWF-mediated platelet aggregation under shear conditions, particularly SRG. In our study, we used CFD analysis for the characterization of our microfluidics setup, linking the shear profile to functional readouts such as platelet aggregate formation, platelet adhesion and rolling velocity as a direct consequence of SRG induced VWF unfolding and activation. First, we investigated the effects of SRG on individual platelet-surface interactions on a collagen type III matrix because this type of collagen shows high affinity for VWF, resulting in high VWF density, while causing mild platelet activation. In order to prevent engagement of aIIbb3 with VWF during platelet adhesion/aggregation, whole blood haematologica | 2021; 106(11)

Targeting VWF to inhibit occlusive thrombus formation

was preincubated with the aIIbb3-inhibitor abciximab. Next, we investigated the effects of SRG on platelet aggregation on a collagen type I matrix because this type of collagen induces strong platelet activation and subsequent aggregate formation. Most importantly, the use of collagen I or III as a physiological VWF substrate, rather than glass-immobilized VWF, ensured correct conformation of shear-immobilised VWF.29,30 Our data show that platelet rolling velocities at SRG are lower than at matched constant shear, suggesting that SRG at the inlet of the stenosis, lead to a rapid unfolding of VWF beyond the degree expected under matched constant shear. This unfolding promotes the formation of catch bonds between VWF and GPIba and reduces the

rolling velocity of platelets on the adhesive surface either due to a higher degree of unfolding per VWF molecule or more unfolded VWF molecules in general. In analogy to tether formation in platelets,28 this feature of SRG may arrest a platelet sufficiently long to facilitate firm adhesion to the adhesive surface or a neighbouring platelet, thereby exacerbating thrombus formation under SRG. Common antiplatelet drugs - aspirin plus a P2Y12 inhibitor - did not prevent the pro-adhesive phenotype of platelets under SRG conditions, highlighting that platelet activation does not play a key role in SRG-mediated platelet adhesion and that current antiplatelet therapy would be unable to inhibit exacerbated thrombosis which is induced by mechanical hyper-activation of VWF by SRG.

Figure 5. Figure continued on the following page.

haematologica | 2021; 106(11)

2881

T. Hoefer et al.

Figure 5. Single-chain antibody A1 specifically inhibits thrombus formation at sites of shear rate gradients but not at constant shear rate. Citrated whole blood, labelled with 0.5 mg mL-1 DiOC6, containing 5 mg ml-1 single-chain antibody (scFv) A1 or control, was drawn at 1,000 s-1 through flow channels exhibiting a patch of perpendicularly-coated collagen type I (100 mg mL-1) on the bottom surface, creating a BSA-collagen interface. (A) Z-slices of confocal microscopy images were binarized and compiled to z-projections showing thrombus height for control (top panel) and scFv A1-loaded blood (bottom panel). (B) Relative fluorescence intensities of platelet aggregates on sequential collagen patches in the same flow channel. Data represented as mean ± standard error of the mean (SEM); n=3; *P<0.05 (multiple t-tests). (C) Relative fluorescence intensities of platelet aggregates in the front, center and rear areas of the collagen patch after 10 minutes of flow. Data represented as mean ± SEM; n=3; *P<0.05 (multiple t-tests). (D) Computational fluid dynamics (CFD) analysis of shear rates present at the surface of platelet aggregates in control blood and scFv A1-treated blood, expressed as a heat map and numerical. Maximal shear rate gradients are shown for the microfluidic stenosis channels used in Figures 1 to 4 and at the surface of the aggregates shown in the heat map. (E) Graphical presentation of the platelet fluorescence distribution over the entire collagen patch at various timepoints. Data is mean of n=3 flows. (F) Quantification of time-lapse confocal microscopy images showing reduced platelet deposition in the presence of scFv A1 at the front area but not the rear area of the patch. Data represented as mean ± SEM; n=3; *P<0.05 (two-way ANOVA).

Others have identified the GPIba-VWF axis as a potential target and various drug candidates are currently under investigation.31 In contrast to our SRG-specific approach, these drugs take a systemic approach, thereby potentially affecting hemostasis. We sought an approach that would allow inhibition of VWF specifically at sites of SRG, a feature of atherothrombosis, without affecting VWF activation under constant shear, which prevails during normal hemostasis. Our blocking strategy contained the novel scFv A1, which was based on the sequences from a previously developed human monoclonal antobody panel raised in mice against a 39/34-kDa VWF fragment (Leu-480/Val481–Gly-718) which encompasses the A1 domain and a sequence N-terminal to that.17 The basis for the generation 2882

of our scFv A1 involved combining the variable heavy chain of an antibody which recognizes a linear epitope in VWF-A1 with the variable light chain of another antibody which recognises a conformation-specific epitope in the VWF A1 domain.17 Assessing single platelet binding to the adhesive surface, our scFv A1 showed selective platelet-VWF inhibition at stenotic sites where SRG are prevalent compared to areas of constant shear suggesting that the scFv A1 did not systemically block the VWF-GPIba interaction but rather specifically at the stenosis and in a transient manner. Similarly, under constant shear conditions platelet aggregation on collagen-bound VWF in the presence of scFv A1 was not inhibited. Taken together, these results suggest that scFv A1 targets the transiently present hyperactive haematologica | 2021; 106(11)

Targeting VWF to inhibit occlusive thrombus formation

form of VWF at the site of a stenosis. In agreement with this, the presence of hyperactive VWF at sites of SRG has previously been reported.16 The activation state of VWF during the flow immobilization to collagen or immobilized platelets appeared to modulate the subsequent recruitment of additional platelets. When VWF was deposited onto collagen before the flow experiment, SRG did not exert this pro-adhesive effect on platelets, indicating that the activation state of VWF in the fluid phase at the moment of deposition onto collagen or immobilized platelets is a key determinant in the pro-thrombotic phenotype of SRG. Similar observations have recently been reported by Receveur et al.32 who used a flow model which introduced SRG of the same magnitude as those used in this study. VWF unfolding in an accelerating flow field is likely to be very transient, with apparent rapid refolding in areas downstream of SRG where shear rates resume a constant level again. Therefore, the binding of scFv A1 to VWF is an elusive event that could only be achieved under shear gradient forces in the microfluidic flow assays. Indeed, we were unable to establish evidence of scFv A1 binding to VWF in other assays with no shear being present, likely due to the epitope for scFv A1 being shielded or not in a correct conformation. Since SRG are thought to be mechanistically similar to very high constant shear, we aimed to mimic the shear unfolding of VWF with the biochemical modulator ristocetin. This antibiotic is known to modulate the VWF-GPIb interaction, binding similar epitopes to those exposed by elevated shear.20 After optimizing experimental conditions for platelet agglutination, scFv A1 inhibited ristocetin-induced platelet agglutination in a minor way suggesting that ristocetin may expose epitopes in VWF that are recognized by scFv A1. Although ristocetin-treated VWF exposes similar epitopes as shear stress, we cannot claim that the topology of the epitopes is the same under both conditions, particularly because we create SRG conditions whereas the literature is based on constant shear rates. This likely explains why the inhibitory effect of our scFv-A1 in the ristocetin induced aggregation assays was only modest and requires further investigations. One limitation of this study is its exclusive in vitro approach to address these questions. The scFv A1 antibody recognizes human VWF and thus would require a fully humanized mouse model of the VWF-GPIb axis to study the efficacy of the antibody in an in vivo setting. These in vivo approaches will be addressed in subsequent studies. Therapeutic development would include full humanization of the scFv replacing structural elements

References 1. Wang H, Wolock TM, Carter A, et al. Estimates of global, regional, and national incidence, prevalence, and mortality of HIV, 1980-2015: the Global Burden of Disease Study 2015. Lancet HIV. 2016;3(8):e361-387. 2. Robbie L, Libby P. Inflammation and atherothrombosis. Ann N Y Acad Sci. 2001;947:167-179; discussion 179-180. 3. Hoefer T, Armstrong PC, Finsterbusch M, et al. Drug-free platelets can act as seeds for aggregate formation during antiplatelet therapy. Arterioscler Thromb Vasc Biol. 2015;35(10):2122-2133.

haematologica | 2021; 106(11)

that are murine based. Current humanization technologies are based on decades of experience with antibody scaffolds and in the majority of cases the antibodies keep their properties. However, there is always a small risk that the properties will change and the antibody might become less affine or specific. In case this is observed further affinity maturation might be required which ultimately could make the antibody more potent. A scFv typically clears the circulation in 1/2 hour compared to days of a full IgG antibody. The ultimate clinical application (acute treatment of thrombosis vs. prophylaxis) will determine which format is optimal. There might be scope to a fast and slow clearing antibody alone or in combination as required based on the desired pharmacokinetics. In conclusion, we have shown that pro-thrombotic effects of SRG, which are known to lead to “hyperactivation” of VWF32-35 can be site-specifically inhibited with our scFv A1 antibody without interfering with the VWF-GPIba interaction under normal flow. We speculate that our targeting strategy could provide an important cornerstone in future antithrombotic therapy which mechanically decouples thrombosis from hemostasis and therefore does not contribute to an increased bleeding tendency, one of the main culprits in current antiplatelet drug development. Disclosures The molecular design and sequence of said single-chain antibody scFv A1 is subject to a patent application (provisional patent no. 2020902148) and is therefore not disclosed in this manuscript. Contributions TH, AR, HA and EW performed research; TH, AR, BN, SJ, HA, AM, EW performed data analysis; EG, RA provided critical reagents and supervision; TH, AR and EW wrote the manuscript; CH and EW supervised the study. Acknowledgments We are grateful to Monash micro imaging for their technical assistance for the use of confocal microscopes. We thank P. Lenting (INSERM, U1176, France) for providing the plasmids encoding the VWF-FL and VWF-A1 domains. Funding This work was supported by grants from the National Health and Medical Research Council and the National Heart Foundation. HA was funded by the European Research Council under the Advanced Grant ‘VESCEL’ Program (grant no. 669768).

4. Antithrombotic Trialists' Collaboration. Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ. 2002; 324(7329):71-86. 5. Wallentin L, Becker RC, Budaj A, et al. Ticagrelor versus clopidogrel in patients with acute coronary syndromes. N Engl Med. 2009;361(11):1045-1057. 6. Wiviott SD, Braunwald E, McCabe CH, et al. Prasugrel versus clopidogrel in patients with acute coronary syndromes. N Engl J Med. 2007;357(20):2001-2015. 7. oude Egbrink MG, Tangelder GJ, Slaaf DW, Reneman RS. Thromboembolic reaction

following wall puncture in arterioles and venules of the rabbit mesentery. Thromb Haemost. 1988;59(1):23-28. 8. Sixma JJ, Wester J. The hemostatic plug. Semin Hematol. 1977;14(3):265-299. 9. Bark DL, Jr., Ku DN. Wall shear over high degree stenoses pertinent to atherothrombosis. J Biomech. 2010;43(15):2970-2977. 10. Nesbitt WS, Westein E, Tovar-Lopez FJ, et al. A shear gradient-dependent platelet aggregation mechanism drives thrombus formation. Nat Med. 2009;15(6):665-673. 11. Springer TA. von Willebrand factor, Jedi knight of the bloodstream. Blood. 2014; 124(9):1412-1425. 12. Zhu S, Herbig BA, Li R, et al. In microfluidi-

2883

T. Hoefer et al. co: recreating in vivo hemodynamics using miniaturized devices. Biorheology. 2015; 52(5-6):303-318. 13. Zheng Y, Chen J, Lopez JA. Flow-driven assembly of VWF fibres and webs in in vitro microvessels. Nat Commun. 2015; 6:7858. 14. Schneider SW, Nuschele S, Wixforth A, et al. Shear-induced unfolding triggers adhesion of von Willebrand factor fibers. Proc Natl Acad Sci U S A. 2007;104(19):7899-7903. 15. Sing CE, Alexander-Katz A. Elongational flow induces the unfolding of von Willebrand factor at physiological flow rates. Biophys J. 2010;98(9):L35-37. 16. Westein E, van der Meer AD, Kuijpers MJ, et al. Atherosclerotic geometries exacerbate pathological thrombus formation poststenosis in a von Willebrand factor-dependent manner. Proc Natl Acad Sci U S A. 2013;110(4):1357-1362. 17. De Luca M, Facey DA, Favaloro EJ, et al. Structure and function of the von Willebrand factor A1 domain: analysis with monoclonal antibodies reveals distinct binding sites involved in recognition of the platelet membrane glycoprotein Ib-IX-V complex and ristocetin-dependent activation. Blood. 2000;95(1):164-172. 18. Chien S, Usami S, Dellenback RJ, et al. Blood viscosity: influence of erythrocyte aggregation. Science. 1967;157(3790):829831. 19. Benard SA, Smith TM, Cunningham K, et al. Identification of peptide antagonists to glycoprotein Ibalpha that selectively inhibit von Willebrand factor dependent platelet aggregation. Biochemistry. 2008; 47(16):4674-4682. 20. Dong JF, Berndt MC, Schade A, et al.

2884

Ristocetin-dependent, but not botrocetindependent, binding of von Willebrand factor to the platelet glycoprotein Ib-IX-V complex correlates with shear-dependent interactions. Blood. 2001;97(1):162-168. 21. Sing CE, Selvidge JG, Alexander-Katz A. Von Willlebrand adhesion to surfaces at high shear rates is controlled by long-lived bonds. Biophys J. 2013;105(6):1475-1481. 22. Shiozaki S, Takagi S, Goto S. Prediction of molecular interaction between platelet glycoprotein Iba and von Willebrand factor using molecular dynamics simulations. J Atheroscler Thromb. 2016;23(4):455-464. 23. Wijeratne SS, Botello E, Yeh HC, et al. Mechanical activation of a multimeric adhesive protein through domain conformational change. Phys rev Lett. 2013; 110(10):108102. 24. Aponte-Santamaria C, Huck V, Posch S, et al. Force-sensitive autoinhibition of the von Willebrand factor is mediated by interdomain interactions. Biophys J. 2015; 108(9):2312-2321. 25. Groot E, Fijnheer R, Sebastian SA, de Groot PG, Lenting PJ. The active conformation of von Willebrand factor in patients with thrombotic thrombocytopenic purpura in remission. J Thromb Haemost. 2009; 7(6):962-969. 26. Savage B, Sixma JJ, Ruggeri ZM. Functional self-association of von Willebrand factor during platelet adhesion under flow.Proc Natl Acad Sci U S A. 2002;99(1):425-430. 27. Maxwell MJ, Dopheide SM, Turner SJ, Jackson SP. Shear induces a unique series of morphological changes in translocating platelets: effects of morphology on translocation dynamics. Arterioscler thromb Vasc Biol. 2006;26(3):663-669.

28. Dopheide SM, Maxwell MJ, Jackson SP. Shear-dependent tether formation during platelet translocation on von Willebrand factor. Blood. 2002;99(1):159-167. 29. Kang I, Raghavachari M, Hofmann CM, Marchant RE. Surface-dependent expression in the platelet GPIb binding domain within human von Willebrand factor studied by atomic force microscopy. Thromb Res. 2007;119(6):731-740. 30. Raghavachari M, Tsai H, Kottke-Marchant K, Marchant RE. Surface dependent structures of von Willebrand factor observed by AFM under aqueous conditions. Colloids Surf B Biointerfaces. 2000;19(4):315-324. 31. Firbas C, Siller-Matula JM, Jilma B. Targeting von Willebrand factor and platelet glycoprotein Ib receptor. Exp Rev Cardiovasc Ther. 2010;8(12):1689-1701. 32. Receveur N, Nechipurenko D, Knapp Y, et al. Shear rate gradients promote a bi-phasic thrombus formation on weak adhesive proteins, such as fibrinogen in a VWFdependent manner. Haematologica. 2020;105(10):2471-2483. 33. Springer TA. Biology and physics of von Willebrand factor concatamers. J Thromb Haemost. 2011;9(Suppl 1):S130-143 34. Vergauwe RM, Uji IH, De Ceunynck K, et al. Shear-stress-induced conformational changes of von Willebrand factor in a water-glycerol mixture observed with single molecule microscopy. J Phys Chem B. 2014;118(21):5660-5669. 35. Colace TV, Diamond SL. Direct observation of von Willebrand factor elongation and fiber formation on collagen during acute whole blood exposure to pathological flow. Arterioscler Thromb Vasc Biol. 2013;33(1):105-113.

haematologica | 2021; 106(11)

ARTICLE

Iron Metabolism & its Disorders

Intravenous iron preparations transiently generate non-transferrin-bound iron from two proposed pathways

Ferrata Storti Foundation

Maciej W. Garbowski,1-3 Sukhvinder Bansal,2 John B. Porter,1 Claudio Mori,4 Susanna Burckhardt4 and Robert C. Hider2,3 1

University College London (UCL) Cancer Institute, Hematology Department, London, UK; 2King’s College London (KCL), Institute of Pharmaceutical Science, London, UK; 3 London Metallomics Consortium, London, UK and 4Vifor Pharma Group, Glattbrugg, Switzerland

ABSTRACT

Haematologica 2021 Volume 106(11):2885-2896

ntravenous iron-carbohydrate complex preparations (IVIP) are noninterchangeable pro-drugs: their pharmacokinetics (PK) varies determined by semi-crystalline iron core and carbohydrate shell structures, influences pharmacodynamics (PD) and thus efficacy and safety. Examining PK/PD relationships of three IVIP we identify a two-pathway model of transient non-transferrin-bound iron (NTBI) generation following single dose administration. Twenty-eight hypoferremic non-anemic patients randomized to 200 mg iron as ferric carboxymaltose (Fe-carboxymaltose), iron sucrose (Fe-sucrose), iron isomaltoside 1000 (Fe-isomaltoside-1000), n=8/arm, or placebo, n=4, on a 2-week PK/PD study, had samples analysed for total serum iron, IVIP-iron, transferrin-bound iron (TBI) by high-performance liquid chromatography in combination with inductively coupled plasma mass spectrometry (HPLC-ICP-MS), transferrin saturation (TSAT), serum ferritin (s-Ferritin) by standard methods, NTBI and hepcidin as published before. IVIP-dependent increases in these parameters returned to baseline in 48-150 hours (h), except for s-Ferritin and TSAT. NTBI was low with Fe-isomaltoside-1000 (0.13 mM at 8 h), rapidly increased with Fe-sucrose (0.8 mM at 2 h, 1.25 mM at 4 h), and delayed for Fe-carboxymaltose (0.57 mM at 24 h). NTBI area-under-curve (AUC) were 7-fold greater for Fe-carboxymaltose and Fe-sucrose than for Fe-isomaltoside-1000. Hepcidin peak time varied, but not AUC or mean levels. s-Ferritin levels and AUC were highest for Fe-carboxymaltose and greater than placebo for all IVIP. We propose two mechanisms for the observed NTBI kinetics: rapid and delayed NTBI appearance consistent with direct (circulating IVIP-to-plasma) and indirect (IVIP-to-macrophage-to-plasma) iron release based on IVIP plasma half-life and s-Ferritin dynamics. IVIP generate different, broadly stability- and PK-dependent, NTBI and s-Ferritin signatures, which may influence iron bioavailability, efficacy and safety. Longer-term studies should link NTBI exposure to subsequent safety and efficacy parameters and potential clinical consequences.

Introduction Non-transferrin-bound iron (NTBI) collectively refers to a heterogeneous group of plasma iron species that are not bound to transferrin, ferritin or heme typically present with transferrin saturation (TSAT) >75%.1,2 In iron overload conditions these are thought to comprise ferric citrate species, mostly albumin-bound, some of which are more redox-active than others.3 NTBI species, present in iron-overloaded patients, are responsible for the pattern of atypical tissue iron distribution seen under such conditions, including extrahepatic (endocrine and myocardial) hemosiderosis.4 However, the intravenous iron-carbohydrate complex preparations (IVIP) themselves are not considered to represent NTBI, or the redox-active, weakly-bound species, although they can become the source of both. Following IVIP infusion, in principle, NTBI could be generated by either a rapid iron egress from

haematologica | 2021; 106(11)

Correspondence: MACIEJ GARBOWSKI maciej.garbowski@ucl.ac.uk Received: February 20, 2020. Accepted: September 3, 2020. Pre-published: September 14, 2020. https://doi.org/10.3324/haematol.2020.250803

2885

M.W. Garbowski et al.

macrophages following primary uptake into the macrophage system or by intravascular iron release from the circulating IVIP prior to macrophage uptake. The structure of the IVIP being administered is likely to influence the levels and duration of both NTBI components. These are largely un-described however, and are compared here for three different IVIP. IVIP have become increasingly used in the treatment of iron deficiency within the context of a wide range of diseases.5 All of these preparations are pro-drugs6,7 for bioavailable iron but their exact mode of iron delivery is unclear.5 This remains poorly recognized despite their widespread use.5 It has been suggested that, depending on the nature of the carbohydrate shell, some IVIP partly decompose in plasma before macrophage uptake; their subsequent endolysosomal degradation releases iron for transient storage or export to plasma.8 Available IVIP include iron sucrose (Fe-sucrose), ferric carboxymaltose (Fe-carboxymaltose), sodium ferric gluconate, iron isomaltoside 1000 (Fe-isomaltoside-1000), ferumoxytol, and low molecular weight iron dextran. Depending on the carbohydrate shell type, these preparations can be classified as non-dextran-based and dextran/dextran-based complexes.9 Non-dextran-based complexes exhibit a correlation between molecular weight (MW) distribution and complex stability, i.e., higher MW complexes are more stable and have lower labile iron content than lower MW complexes. In contrast, dextran/dextran-based complexes are all very stable independent of their MW.5,10 Here, we selected two non-dextran-based IVIP with different MW distributions, Fe-carboxymaltose (Ferinject®, 145-155 kDa) and Fe-sucrose (Venofer®, 42-44 kDa), as well as a dextran-based complex Fe-isomaltoside-1000 (Monofer®, 63-69 kDa).5 The purpose of this study was to examine the NTBI profiles in the context of other pharmacokinetics/pharmacodynamics (PK/PD) parameters under identical conditions (200 mg iron dose administered over 10 minutes [min]), in hypoferremic, otherwise healthy, subjects, to better understand their mode of action. Previous studies showed that all three preparations are known to transiently increase TSAT with decay half-lives t1/2=23 hours (h) (Fe-isomaltoside-1000), t1/2=8 h (Fe-carboxymaltose), and t1/2=5 h (Fe-sucrose),11 possibly with associated NTBI.12,13 Crucially however, appropriate NTBI and TSAT methods are necessary that distinguish between IVIP-Fe, TBI and NTBI. We used a highly sensitive and specific bead-based NTBI assay that is robust to transferrin.1,14 This means that due to the hexadentate nature of the assay chelator on the beads (CP851) iron is neither removed from ferrotransferrin nor donated (shuttled) to apotransferrin, thus minimizing nonspecific overestimation or underestimation of NTBI, respectively.1,14 Different IVIP display different iron pharmacokinetics, dependent on the structure of the semi-crystalline iron core and the type of polysaccharide ligand.7,15 Such properties may influence IVIP efficacy and safety,11,16 but also inform the mechanisms of iron delivery and IVIP differences. PK, stability, and the amount of weakly-bound iron, define the maximal single dose that can be administered.5 This is relevant, as the increased IVIP use is also linked to the fact that, with new preparations, high doses can be given in a short amount of time. Therefore, following a single dose of 200 mg iron, we monitored, among others, the PK of the IVIP, TSI (TBI, 2886

NTBI and IVIP-Fe), TSAT, hepcidin and s-Ferritin levels over a period of 2 weeks. Although there have been some direct comparisons between different IVIP,17–19 none have compared the above six parameters in man, over a 2-week period. This approach has allowed us to identify two proposed pathways of NTBI generation in IVIP treatment.

Methods Study design and patient population Our work is part of an exploratory phase I single-center, singleblind, randomized, placebo-controlled study to describe and characterize the PK and PD of Fe-carboxymaltose, Fe-sucrose, and Feisomaltoside-1000 in hypoferremic non-anemic subjects. Twentyeight subjects (hemoglobin [Hb] ≥13 g/dL for males, and ≥12 g/dL for females, fasting s-Ferritin<30 µg/L and TSAT<20% measured at 08:00-09:00 a.m.) randomized into one of three treatment groups (n=8/arm), or one placebo control group (n=4/arm), received a single intravenous 200 mg iron dose of Fe-carboxymaltose, Fe-sucrose or Fe-isomaltoside-1000, or saline solution (placebo) over 10 minutes. Blood samples, processed for serum and batch-frozen at -80ºC, were taken on day -1 at 08:00-09:00, then 4 and 12 h later (baseline iron profile); on day 1 immediately predose at 08:00-09:00 (time: 0 min), 10 min (immediately post-dose), 20 min, and 40 min, then 1, 2, 4, 6, 8, 12, 16, 24 h, then 36, 48, 72, 96, 120, 144, and 312-336 h post-dose. The study was approved by a local Ethics Committee and all patients gave informed consent on entering the study.

Serum iron profiles and s-Ferritin levels Vifor Pharma performed all assays except as stated. TSI was measured using a validated inductively coupled plasma optical emission spectroscopy method. Unsaturated iron binding capacity (UIBC) was measured by a photometric colorimetric test (Beckman Coulter UIBC Metabolite assay). TBI was calculated as TBI[mmol/L]=(transferrin[g/l]x25.12)-UIBC[mmol/L] where transferrin concentration was measured immuno-turbidimetrically (Beckman Coulter Transferin assay). Conversion of TBI units: TBI[mg/mL]=TBI[mmol/L]x55.845[µg/mmoL]x0.001[l/mL]. TSAT was calculated as TSAT[%]=100–(3.98×UIBC[mmol/L]/transferrin[g/L]). UIBC values below the lower limit of quantification were imputed as 0 in the TBI formula. S-Ferritin was measured with a chemiluminescent microparticle immunoassay (Architect systems), hemoglobin and reticulocytes using standard methods, and soluble transferrin receptor-1 (sTfR1) by particle-enhanced immunonephelometry (N Latex sTfR assay). IVIP-Fe was determined subtracting TBI from TSI.20

Non-transferrin-bound iron We used the previously published bead-NTBI method1,14 with some modifications (see the Online Supplementary Appendix). Serum samples or buffered solutions incubated with the CP851 beads were assayed by flow-cytometry (Beckman Coulter CytoFLEX, CytExpert software, KCL) and quantitated against ferric nitrilotriacetate (Fe-NTA) standards (Figure 1). Limit of blank was 0.51 nM, limit of detection 14.6 nM, limit of quantitation 30 nM; intra-assay and inter-assay precision coefficient of variation (CV) were 1.28% and 4.34%, respectively (for further details see the Online Supplementary Appendix).

Plasma hepcidin Plasma hepcidin (KCL) was measured in serum samples using tandem mass spectrometry;21 for detailed method see the Online Supplementary Appendix.

haematologica | 2021; 106(11)

NTBI following intravenous iron administration

Figure 1. Bead-non-transferrin-bound iron assay. (A) Side-scatter-area versus forward-scatter-area plot, gate P1 identifies a population of CP-851 beads. (B) Sidescatter-area versus side-scatter-height plot, gate P3 excludes doublets: (C) Histogram of FITC-area from combination of P1 and P3 gates, shown with an example statistics (right), median of the distribution represents the bead fluorescence: (D) All 28 standard curves (each in triplicate) plotted together with mean and standard deviation (SD) shown (n=84), 4-parameter logistic curve fitted r2=0.97. (E) Example standard curve with mean and SD shown (triplicates), logM units were transformed by exponentiation and multiplied by 1.0x106 to obtain results in mmol/L. (F) Overlay histograms of FITC-area, as in C above, for buffer (0 mM ferric nitrilotriacetate [Fe-NTA] standard), control serum, a patient sample in duplicate extrapolated at 1.02 mM non-transferrin-bound iron (NTBI) and 10 mM Fe-NTA standard, see inset. (G) overlay of FITC-area histograms showing a standard curve example. Limit of Blank (95th percentile of eight replicates, LoB) was 0.51nM, Limit of detection (LoB-1.654SD(low standard), LoD) 14.6 nM, Limit of quantitation (LoQ) 30 nM, intra-assay and inter-assay precision coefficients of variation were 1.28% and 4.34% respectively (28 consecutively run assays over 82 days, with triplicate controls).

Inductively-coupled plasma mass-spectrometry serum analyses A parallel set of serum samples (baseline, 4-312 h) was also analyzed by inductively-coupled plasma mass-spectrometry serum analyses (HPLC-ICPMS) at KCL to measure IVIP-Fe and examine the chromatographic behavior of IVIP. Perkin Elmer Flexar HPLC coupled to a Perkin Elmer NexION 350 D ICPMS was used with Syngistix and Chromera operating software (Figure 2A). Sample preparation, instrumentation, and measurement of IVIP iron in plasma, are detailed in the Online Supplementary Appendix.

Statistical analysis All data is presented as mean ± standard deviation, unless otherwise stated. Analysis of variance (ANOVA) with a post-test Holm-Sidak’s multiple comparisons test was used to compare means between treatments. PK parameters were derived from non-compartmental analysis for all subjects using Full Analysis Set (cmax, half-life, and tmax are shown). Area-under-curve (AUC) and nonlinear regression analysis using global fitting was performed on GraphPad Prism (version 6.0). A P-value<0.05 was considered statistically significant.

Results

(FO) were then injected into HPLC-ICPMS. IVIP chromatograms in 20% serum are shown as subtraction plots, with control chromatograms subtracted (Figure 2B to D). Ferrotransferrin peak appears at 13.17 min as a result of rapid iron exchange with apotransferrin (stoichiometric amount of apotransferrin was subtracted) either during the short sample processing at the bench and/or waiting time in the auto-sampler, or directly on the column. The post-subtraction IVIP-Fe recovery was 28 mM for Fe-carboxymaltose, 32 mM for Fe-isomaltoside-1000, and 23.5 mM for Fe-sucrose. Low recovery for Fe-sucrose appears likely due to multiple serial dilutions in ammonium acetate buffer (pH=7.4) to obtain 30 mM, with some precipitation or re-speciation (Fe-sucrose pH 10.5-11.1). The subtraction chromatograms show that Fe-isomaltoside-1000 particularly is distributed very widely (peak at 11 min, 126 kDa), while Fe-carboxymaltose and Fe-sucrose have comparable species distribution peaking at 9.2 min (160 kDa) and 8.9 min (165 kDa), respectively. Notably, these elution times and therefore MW ranges do not correspond to the values obtained in the presence of buffers alone.5 This difference results most likely from the in vivo formation of the protein corona22 and probably aggregation between IVIP cores and plasma proteins.

Chromatographic behavior of intravenous iron-carbohydrate complex preparations in serum

Patient baseline characteristics

In order to examine the chromatographic behavior of the IVIP in serum, buffered solutions of Fe-carboxymaltose, Fe-sucrose, Fe-isomaltoside-1000 (30 mM) were mixed with normal human serum (80/20 vol/vol) and the mixtures were incubated in vitro for 1 h on the bench top. Samples spiked with the internal standard ferrioxamine

Patients were balanced between arms for all the baseline parameters (Table 1) with the exception of sex. Screening s-Ferritin ranged between 2.85-27.54 mg/L (percentiles 25th, 50th, 75th: 8.49, 11.05, 17.78 mg/L), TSAT: 5.11-19.62% (9.48, 14.00, 17.24%), and hemoglobin: 12.0-15.4 g/dL (12.5, 13.0, 13.4 g/dL).

haematologica | 2021; 106(11)

2887

M.W. Garbowski et al.

Table 1. Baseline parameters.

Blood test

FCM, n=8

IIM, n=8

IS, n=8

Placebo, n=4

All, n=28

Females C-Reactive Protein [mg/L] Serum Ferritin [mg/L] Hemoglobin [g/L] Hepcidin [ng/mL] Iron [mmol/L] NTBI [mmol/L] Reticulocytes [109/L] Soluble Transferrin Receptor [mg/L] Transferrin [g/L] Transferrin Saturation [%] UIBC [mmol/L]

7 1.47±2.18 14.24±8.8 126.5±5.42 1.07±1.2 10.34±5.11 0.0±0.0 61.8±10.88 1.49±0.41 3.0±0.4 13.81±7.05 56.68±9.3

7 0.85±0.84 10.94±4.77 125.5±9.42 1.36±2.17 13.04±5.89 0.0±0.0 56.81±13.99 1.3±0.13 2.9±0.28 18.07±8.41 50.95±8.29

7 0.96±0.65 15.05±10.39 128.37±11.23 3.27±2.52 9.46±4.75 0.0±0.0 53.75±18.51 1.26±0.45 2.84±0.18 13.55±7.57 53.97±8.01

4 0.34±0.16 8.26±4.33 123.25±4.64 0.67±0.76 12.69±5.68 0.0±0.0 60.97±29.43 1.47±0.25 3.1±0.2 16.65±8.28 58.35±10.28

25 0.98±1.29 12.67±7.89 126.28±8.28 1.72±2.16 11.2±5.26 0.0±0.0 57.95±16.67 1.36±0.34 2.94±0.29 15.36±7.6 54.51±8.72

FCM: ferric carboxymaltose; IIM: iron isomaltoside-1000; IS: iron sucrose; NTBI: non-transferrin bound iron; UIBC: unsaturated iron binding capacity.Values given as mean ± standard deviation.

Table 2. Selected pharmacokinetic parameters of intravenous iron-carbohydrate complex preparations.

Biomarker IVIP iron content

Total serum iron

Transferrin-bound iron

PK parameter

Ferric carboxymaltose n=8

Iron isomaltoside 1,000 n=8

Iron sucrose n=8

cmax, mean±SD [mmol/L] tmax, mean (min-max) [h] t½, mean±SD [h] AUC0-inf [h*mmol/L] cmax, mean±SD [mmol/L] tmax, mean (min-max) [h] t½, mean±SD [h] AUC0-inf [h*mmol/L] cmax, mean±SD [mmol/L] tmax, mean (min-max) [h] t½, mean±SD [h] AUC0-inf [h*mmol/L]

1.16±0.09 0.34 (0.33-0.7) 6.82±1.93 12.39±1.2 1.18±0.1 0.33 (0.18-0.7) 11.14±3.42 14.8±1.22 49.13±7.19 24 (4-24) 11.45±4.88 2.09±0.42

1.29±0.1 0.67 (0.33-6) 20.3±2.27 36.76±4.9 1.3±0.1 0.68 (0.33-6) 19.21±2.97 38.5±4.13 38.34±10.8 8.05 (6-24) 15.35±4.44 1.72±0.49

0.85±0.17 0.35 (0.33-0.37) 3.43±1.55 2.59±0.5 0.89±0.21 0.34 (0.18-0.37) 7.5±1.76 4±0.53 51.27±4.68 5 (4-24) 6.82±1.38 1.37±0.28

IVIP: intravenous iron-carbohydrate complex preparations; PK: pharmacokinetics; TSI: total serum iron; TBI: transferrin-bound iron; cmax: maximal plasma concentration; tmax: time of maximal plasma concentration; t½: plasma half-life; AUC0-inf,: area-under-curve from zero to infinity.

Comparison of intravenous iron-carbohydrate complex iron preparations between treatments There were striking differences in biomarker timecourses between IVIP, as shown in Figure 3A to C. For the IVIP-Fe (as TSI-TBI), the elimination rate was slowest, while cmax was highest with Fe-isomaltoside-1000, followed by Fe-carboxymaltose and Fe-sucrose (at median tmax 0.67, 0.34, 0.35 h, respectively), see Table 2. The cmax was comparable within 95% Confidence Interval [CI] for all IVIP (spanning 1 mM value): 1.29±0.1 mM for Fe-isomaltoside-1000, 1.16±0.09 mM for Fe-carboxymaltose, and 0.85±0.17 mM for Fe-sucrose, Figure 4A, Table 2. The IVIP-Fe AUC differed between treatments (P<0.0001), with Fe-isomaltoside-1000 14-fold higher than Fe-sucrose and 3-fold higher than Fe-carboxymaltose, Fe-carboxymaltose 4.8-fold higher than Fe-sucrose (P<0.0001 for all pairs). The between-patient average IVIP-Fe half-life was 10-fold longer for Fe-isomaltoside-1000 (20.3±2.27 h) and twice longer for Fe-carboxymaltose (6.82±1.93 h) than for Fe-sucrose (3.43±1.55 h), Table 2. The chromatographic behavior of IVIP was also monitored using HPLC-ICPMS at KCL (see Methods). Subtraction chromatograms (4hbaseline) show clear differences in IVIP-Fe profiles (Figure 2B), each presenting a distinct chromatographic behavior: Fe-isomaltoside-1000 has a broad peak eluting at 11.3 min, Fe-carboxymaltose at 9.36 min and Fe-sucrose at 8.97 min 2888

with a single principal peak each. These ex vivo 100% serum profiles compare well with the in vitro 20% serum profiles, Figure 2C (discussed).

Total serum iron changes between treatments The time-courses of total serum iron (TSI) closely follow the kinetics of IVIP-Fe, especially for Fe-isomaltoside1000. For Fe-carboxymaltose and Fe-sucrose there is an early separation of IVIP-Fe from the trajectory of TSI (Figure 3A to C, green and red profiles, also compare Figure 4A and B). This is a consequence of their half-life being much shorter, and thus IVIP-Fe decaying to sufficiently low values for the TBI component now to dominate TSI. TSI cmax is comparable, within 95% CI, across treatments, and TSI tmax closely corresponds to IVIP-Fe tmax, Figure 4B, Table 2. The TSI AUC differed (P=0.0001) between treatments, with Fe-isomaltoside-1000 9.6-fold higher than Fe-sucrose and 2.6-fold higher than Fe-carboxymaltose, while Fe-carboxymaltose 3.7-fold higher than Fe-sucrose (P<0.0001 for all pairs).

Transferrin saturation and non-transferrin-bound iron changes between treatments TSAT was lower with Fe-isomaltoside-1000 reaching 81.04±16.52% at 24 h (range 60.5-100%) versus early full saturation with Fe-sucrose (already at 2 h 95.24±8.87%, haematologica | 2021; 106(11)

NTBI following intravenous iron administration

Figure 2. Chromatography of the intravenous iron-carbohydrate complex preparations using HPLC-ICPMS at KCL. (A) An example serum sample high-performance liquid chromatography in combination with inductively coupled plasma mass spectrometry (HPLC-ICP-MS) run with integration of chromatograms within the Chromera software. (B) The subtraction chromatograms of intravenous iron-carbohydrate complex preparations (IVIP) in serum samples (ex vivo), baseline subtracted from the 4 hour (h) time-point. Variable abundance is a function of the given IVIP half-life at 4 h; transferrin peak (at 13.2 minutes [min]) indicates the iron exchange from IVIP to apotransferrin. The precision of subtraction is validated by the overlay of the internal standard (not shown) converging on 15 mM FO (subtraction of 5 mM from 20 mM). (C) The subtraction chromatograms of IVIP (30 mM) incubated in vitro with 20% serum for 1 h with control chromatograms subtracted. Distinct chromatograms for Fe-isomaltoside-1000, Fe-sucrose and Fe-carboxymaltose identified. Small iron signal at 13.2 min is ferrotransferrin testifying to exchange of labile iron between polymers and apotransferrin. (D) Comparison of in vitro (thick) and ex vivo (thin) IVIP chromatograms from B and C using two y-axes for peak size adjustment. At 4 h post-infusion patient samples show distinct shift of the peak elution time, which indicates that principal IVIP species changes with time i.e., that larger species are preferentially removed from plasma. Fe-carboxymaltose shows greatest shift of peak time.

haematologica | 2021; 106(11)

2889

M.W. Garbowski et al.

Figure 3. Aggregate profiles of key iron metabolism parameters. Intravenous iron-carbohydrate complex preparation (IVIP) is infused at time 0 hours (h) (red arrow), with time <0 h indicating baseline values, see legend for graph explanation: non-transferrin-bound iron (NTBI) (black) is plotted on the right y-axis, all other variables – on the log axis (left y-axis); note hepcidin concentration is multiplied by 10 to enable plotting on the log axis. (A) Fe-isomaltoside-1000, n=8; (B) Fe-sucrose, n=8; (C) Fe-carboxymaltose, n=8; (D) placebo, n=4.

range 79.2-100%), and later full saturation with Fe-carboxymaltose at 16 h (97.00±8.49, range 76-100%), Figure 4C. TSAT AUC barely differed between treatments; for Fe-carboxymaltose versus Fe-isomaltoside-1000 by only 32% (P=0.048) with other comparisons statistically insignificant. NTBI was relatively low with Fe-isomaltoside-1000 (peak 0.13±0.27 mM at 8 h), but appeared high notably rapidly with Fe-sucrose (0.79±0.72 mM at 2 h, peak 1.25±0.61 mM at 4 h), and was delayed with Fe-carboxymaltose (peak at 0.58±0.43 mM at 24 h), Figure 4D. The first appearance of NTBI associated with relatively low TSAT, likely related to the rate of iron release from the IVIP. The peak NTBI values were associated with full TSAT (or near-saturation in Feisomaltoside-1000 group) and the disappearance of NTBI precedes (occurs before) the normalization of TSAT, compare Figure 4C and D. In contrast to TSAT AUC, the NTBI AUC differed significantly (P=0.01) between groups on average, with Fe-carboxymaltose nearly 7-fold higher than Fe-isomaltoside-1000 (P=0.04), Fe-sucrose versus Fe-isomaltoside-1000 nearly 9-fold higher (P=0.01) while for Fesucrose versus Fe-carboxymaltose only 1.15-fold higher (P=not significant).

Hepcidin changes between treatments Hepcidin peaks earlier with Fe-sucrose to 68.6±34.7 2890

ng/mL at 24 h, than with Fe-isomaltoside-1000 to 54.3±15.2 ng/mL at 36 h and than with Fe-carboxymaltose to 58.9±25.2 ng/mL at 48 h. All hepcidin AUC were similar (P=0.54), Figure 4E. This is in keeping with similar exposure of hepatocytes to TSAT (transferrin-Fe2) between treatments: see above. Furthermore, it indicates the NTBI exposure (AUC) differences do not influence the exposure to hepcidin.

s-Ferritin changes between treatments s-Ferritin showed significant differences between treatments, increasing faster with time for both Fe-sucrose and Fe-carboxymaltose and slower for Fe-isomaltoside-1000. Ferritin increase from ~12 mg/L at baseline was greatest for Fe-carboxymaltose to ~200 mg/L at 72 h with an exponential constant k=0.028±0.013 mg/L*h, followed by Fesucrose to ~150 mg/L at 36 h k=0.076±0.022 mg/L*h, and Fe-isomalt side-1000 to ~102 mg/L at 96 h, k=0.021±0.014 mg/L*h (Figure 4F, Figure 5A and C). Ferritin AUC differed between groups (P<0.0001), Fe-carboxymaltose being 75% higher (P=0.007) than Fe-isomaltoside-1000 and 63% higher (P=0.01) than Fe-sucrose, with comparable AUC for Fe-isomaltoside-1000 and Fe-sucrose (P=0.98). All IVIP had greater ferritin AUC versus placebo: 9-, 10-, and 16fold for Fe-isomaltoside-1000, Fe-sucrose and Fe-carboxyhaematologica | 2021; 106(11)

NTBI following intravenous iron administration

Figure 4. Aggregate data of intravenous iron-carbohydrate complex iron preparations, total serum iron, transferrin saturation, non-transferrin-bound iron, hepcidin and ferritin for all patients. Data shown in aggregate for treatment groups: placebo (blue), Fe-carboxymaltose (red), Fe-isomaltoside-1000 (black), Fe-sucrose (green). (A) intravenous iron-carbohydrate complex iron preparations (IVIP-Fe) plasma concentrations (log scale in the inset), estimated from the pre-transferrin chromatographic peak area-under-curve (AUC); (B) total serum iron (log scale in the inset); (C) transferrin saturation (TSAT); (D) non-transferrin-bound iron (NTBI), x-axis limited to 48 hours (h) as no appearance of NTBI after 48 h was observed; (E) plasma hepcidin (10* ng/mL), i.e., measured hepcidin values are lower by a factor of 10; (F) serum ferritin during the initial 48 h for direct comparison with NTBI in D.

maltose, respectively, Figure 5B. s-Ferritin iron content changes between treatments and the erythropoietic response are presented in the Online Supplementary Appendix.

Discussion NTBI typically appears as the primary pathological culprit of parenchymal hemosiderosis23 in chronic iron overload conditions: hereditary hemochromatosis,24 thahaematologica | 2021; 106(11)

lassemias,25 sickle cell anemia,26 rare anemias,27,28 myelodysplasia29,30 and myeloablation.31,32 Different NTBI species and other factors33 may determine variable distribution of tissue hemosiderosis.4 NTBI levels vary by disease, degree of transfusion dependence, and the assays used to detect them,2,34 but can be as high as 8.5 mM.35 In contrast, NTBI appearance is transient after oral or IV iron administration. In studies of single dose oral iron at doses equivalent to 100 mg of ferrous sulphate, NTBI concentrations measured by various methodologies ranged from 1 to 6 mM.36–39 In this study, maximal NTBI concentrations 2891

M.W. Garbowski et al. A

Figure 5. Serum ferritin behavior on study. (A) Cumulative serum ferritin profiles for treatment groups, fitted using global fitting with a custom function (plateau followed by mono-exponential association followed by mono-exponential decay), standard error of the mean, global fit R square 0.84; (B) box-and-whisker plots of ferritin area-under-curve (AUC) (baseline 312 hours); (C) box-and-whisker plots of ferritin mono-exponential association rate constant Kassoc; (D) relationships of the rate of increase in serum ferritin to the plasma half-life of the intravenous iron-carbohydrate complex preparations (IVIP).

ranged from 0.13 to 1.25 mM (Fe-sucrose>Fe-carboxymaltose>Fe-isomaltoside-1000), being at the lower end of concentrations observed after oral administration. We also measured labile plasma iron40 which reflects the activity of redox-active NTBI subspecies, but its levels were lower than measured NTBI concentrations (0.83, 0.15 and 0.18 mM for Fe-sucrose, Fe-isomaltoside-1000 and Fe-carboxymaltose, respectively, data not shown). NTBI appearance associated with oral or IV iron administration has been associated with acute increases in non-specific biomarkers of oxidative stress in clinical studies.41–44 However, several epidemiologic studies and a recent prospective clinical trial have not shown that IVIP administration is associated with adverse cardiovascular outcomes.45,46 Analyses of large dialysis patient datasets have shown modest associations with increased infections at higher doses, however, this was not observed in the recent prospective clinical trial in chronic kidney disease patients.47,48 Although NTBI appearance following oral36–39,49 and intravenous iron administration of different formulations is described,17,18,50 here we present a unique study directly 2892

comparing three IVIP (Fe-isomaltoside-1000, Fe-sucrose, Fe-carboxymaltose). We compared six associated iron metabolism parameters over 2 weeks with complex kinetics unique to each formulation identified for the first time. We reported increases in IVIP-Fe, TSI, TSAT, s-Ferritin, hepcidin, and NTBI, which returned to baseline within 2 weeks, or sooner, except for s-Ferritin and TSAT (Figure 3 to 5). Based on the corresponding AUC, Fe-isomaltoside1000 resulted in the highest TSI exposure but the lowest TSAT, s-Ferritin, and NTBI. Fe-sucrose resulted in the lowest TSI exposure, intermediate TSAT and s-Ferritin exposure but highest NTBI exposure. Fe-carboxymaltose resulted in intermediate TSI and NTBI exposure, but highest TSAT and s-Ferritin exposure. Increased hepcidin levels were similar for the three IVIP. Thus, the exposure to bioavailable iron, as judged from s-Ferritin and TSAT AUC, was highest for Fe-carboxymaltose, followed jointly by Fesucrose and Fe-isomaltoside-1000. For the latter two preparations, kinetics of iron bioavailability differed in that the short-term bioavailability rate was markedly greater for Fesucrose due to its known more limited stability profile (Figure 5A green vs. black curve).6 Although head-to-head haematologica | 2021; 106(11)

NTBI following intravenous iron administration

Figure 6. Schematic model of intravenous iron-carbohydrate complex preparations-dependent generation of non-transferrin-bound iron from two compartments. Before intravenous iron-carbohydrate complex preparations (IVIP) are eventually taken up by macrophages, small proportion of circulating IVIP may directly release loosely bound iron as non-transferrin-bound iron (NTBI) or release iron at sufficiently high rate to generate NTBI despite apotransferrin (ApoTf) presence in hypoferremic state (Pathway I: direct release from IVIP). Following macrophage uptake, ferroportin (Fpn)-dependent release of NTBI (Pathway II: Macrophage-mediated release) is the default NTBI route. iron, Tf-Fe2: transferrin-Fe2; Fpn: ferroportin; SF: serum ferritin.

comparison studies between Fe-carboxymaltose, Fesucrose, and Fe-isomaltoside-1000 looking at clinically meaningful outcomes are lacking, we speculate that relatively higher iron bioavailability for Fe-carboxymaltose could be responsible for the positive effect shown on outcomes in heart failure in comparison with oral iron.51–53 This is the first report of IVIP-Fe kinetics measured directly for Fe-isomaltoside-1000, Fe-sucrose and Fe-carboxymaltose rather than indirectly using TSI as a proxy for IVIP-Fe. This matters particularly for Fe-carboxymaltose and Fe-sucrose where the separation between IVIP-Fe and TSI trajectories is particularly apparent (Figure 3B and C); thus their t1/2 is twice longer with the TSI than the IVIP-Fe method (Table 2). This study shows that, for Fecarboxymaltose and Fe-sucrose, the assumption that IVIPFe=TSI is incorrect, especially for lower doses. We also confirmed relative differences in the IVIP half-lives (Fe-isomaltoside-1000>Fe-carboxymaltose>Fe-sucrose) although the absolute values differed from those published previously,5 most likely due to the lower doses used. Plasma hepcidin increments after IVIP injection have been described.18,54,55 Here we show hepcidin peak following the TSAT peak by 20-24 h with all IVIP (Figure 4C and E). This likely reflects the negative feedback between transferrin-Fe2 and hepcidin expression. In hepatocytes TfR1-TfR2-HFE-BMPR interaction positively regulates BMPR signaling to hepcidin transcription.56,57 This peakto-peak lag between TSAT and hepcidin is shorter for Fesucrose (18-22 h). One explanation is that additional positive signaling to hepcidin expression occurs earlier with Fe-sucrose due to higher peak NTBI concentrations. If hepatocyte iron is increased, e.g., rapidly via the NTBI route, this may be sensed by the BMP6 pathway that uphaematologica | 2021; 106(11)

regulates hepcidin expression, and supplements the transferrin-Fe2-dependent TfR2 signaling. We relied on three parameters established to address this subject, namely on the differences between IVIP in the peak time and the magnitude of the rise and fall of NTBI levels relative to the peak time, on the magnitude and the rate of the rise and fall of s-Ferritin, as well as including the mutual temporal relationships between them. We propose two mechanisms for the generation of the kinetic profiles of NTBI, Figure 4D and F. The Fe-sucrosegenerated NTBI profile features two distinct kinetic components: a rapid onset and an early peak at 4 h followed by decay and a secondary peak/shoulder at 16 h disappearing at 36 h (individual profiles are bi- or tri-modal, but the peak times and magnitudes vary across patients possibly due to the effect of protein corona, data not shown). The first peak represents predominantly the rapid release into plasma of weakly-bound iron directly from circulating complexes, because the said 4 h peak occurs within the first plasma half-life of Fe-sucrose (henceforth Pathway I). This can occur even when transferrin is not fully saturated and is dependent on rate of release of iron from the formulation, the iron-ligand speciation, and the kinetic binding equilibrium of transferrin.38,43,58 In contrast, because the second peak/shoulder occurs after four half-lives, at which time over 90% of Fe-sucrose has been taken up by macrophages, that 16 h NTBI peak predominantly represents indirect, ferroportin-mediated iron efflux from macrophages that fully saturates transferrin (henceforth Pathway II), Figure 6. With Fe-isomaltoside-1000, given its long plasma halflife, the small NTBI peak at 8 h thus represents direct iron 2893

M.W. Garbowski et al.

release from circulating IVIP, while any indirect release after the first half-life is evidently at much slower macrophage-mediated release rate at which transferrin does not become saturated, and consequently NTBI is virtually absent. Fe-carboxymaltose initially follows the Fe-isomaltoside1000 NTBI profile corresponding to the direct generation of NTBI, but displays a starkly different behavior reflecting the indirect macrophage iron release with a late NTBI peak at 24 h, at which time approximately four plasma half-lives have already elapsed. This difference between Fe-isomaltoside-1000 and Fe-carboxymaltose may partially be accounted for by differences in plasma half-lives, i.e., by the amount of IVIP having been hitherto taken up by macrophages (>90% of Fe-carboxymaltose and <60% of Fe-isomaltoside-1000). As for the macrophage-mediated uptake process, chromatography of the IVIP shows the average size of IVIP decreasing with time. This effect is more marked for Fecarboxymaltose and Fe-sucrose than for Fe-isomaltoside1000 (shift in peak elution time, Figure 2D; Online Supplementary Figure S4). This is best interpreted by macrophages preferentially removing the larger IVIP species and so the smaller species are likely to persist longer in plasma. The macrophage processing efficiency is higher for Fecarboxymaltose because, according to the s-Ferritin AUC differences (Figure 5B), 75% more iron per 200 mg dose is extracted from Fe-carboxymaltose than Fe-isomaltoside1000. Plausibly, some unprocessed Fe-isomaltoside-1000 complexes remain in macrophages and are non-bioavailable (at least within 2 weeks, as here). This could be resolved if ferritin were measured sequentially until it reached baseline to enable total s-Ferritin AUC comparison between IVIP. This approach would confirm that the complete s-Ferritin AUC represents the fraction of bioavailable iron at a given dose processed by macrophages, which is not reduced by the fraction that escapes via NTBI route (Pathway I) to parenchymal cells. When NTBI is absent or relatively low as in the case of Feisomaltoside-1000, the difference in complete s-Ferritin AUC between e.g., Fe-carboxymaltose and Fe-isomaltoside-1000 could mean that a sizeable proportion of the Feisomaltoside-1000 iron remains unprocessed in macrophages. Whether that is due to inherent structural properties of some IVIP or to dysregulation of iron metabolism e.g., high hepcidin in haemodialysis18,59,60 or both, remains to be determined. Nevertheless, the relative magnitude of iron shunted away from IVIP via the NTBI plasma compartment to parenchymal cells (hepatocytes) is small (estimated approximately 1,000-times smaller) as compared with the iron flux directed to the erythron from IVIP via the TBI compartment, see the Online Supplementary Figure S6. The proposed NTBI generation Pathways I and II can be corroborated by comparing NTBI profiles with the rate of s-Ferritin increase (Figure 4D vs. Figure 4F). As s-Ferritin reports macrophage iron content,61,62 at 16 hours after Fesucrose injection it is already half-maximal (s-Ferritin=~80 mg/L), at which time we observe a second NTBI peak/shoulder of ~0.7 mM (indicating high rate iron egress exceeding the transferrin iron binding capacity). The fact that for Fe-sucrose at 4 h the ~1.3 mM NTBI peak (and ~0.8 mM at 2 h) does not correspond to any s-Ferritin increase indicates that the NTBI cannot have arisen from 2894

the macrophage compartment, which thus confirms the existence of two separate NTBI generation pathways. IVIP typically become a source of iron for metabolic pathways after they have been processed by macrophages.16 Macrophage uptake rate can be inferred from plasma half-life of IVIP, whereas endolysosomal IVIP degradation rate following macrophage uptake can be inferred from the rate of s-Ferritin increase. s-Ferritin increases in plasma as a marker of the iron stored in macrophage cellular ferritin, which undergoes turnover via the transient labile iron pool that regulates ferritin mRNA translation via IRP-IRE system.61 Importantly, the faster the IVIP disappears from plasma, the faster s-Ferritin increases (compare IVIP-Fe t1/2 and the exponential association constant of s-Ferritin increase, Figure 5D), suggesting that the different physicochemical characteristics of the IVIP that influence the half-life6,7 are likely to be the fundamental reason for this relationship and hence for the differential iron bioavailability. Thus, Fe-isomaltoside-1000 is a relatively inert polymer taking longer to be removed by macrophages and longer to release bioavailable iron (60 h longer than Fe-sucrose and 24 h longer than Fe-carboxymaltose), Figure 5A. Significantly, the s-Ferritin decay rate is similar across IVIP (Figure 5A) suggesting a common mechanism of iron egress (macrophage ferroportin), once released from IVIP within endolysosomes. This egress of iron (hence its bioavailability) can be inferred from the fall of macrophage iron stores as reported by falling s-Ferritin. Similarly, albeit less clearly, IVIP half-life correlates with exposure to NTBI (NTBI AUC), but without associating with either the exposure to TSAT or to hepcidin, Online Supplementary Figure S2A and C. This is consistent with NTBI generation resulting from IVIP rapidly loading the macrophage compartment and iron rapidly egressing (Pathway II) against a hitherto relatively low hepcidin level (thus open ferroportin) where that hepcidin level later increases in response to increased plasma transferrin-Fe2. The rapidity of NTBI appearance with Fe-sucrose versus Fe-carboxymaltose or Fe-isomaltoside-1000 strongly suggests that ambient iron is immediately made available in plasma for exchange with transferrin until its saturation (Pathway I) before further iron is released via macrophage ferroportin. Rapid NTBI release in vitro from buffered Fe-sucrose was abrogated following 1 h incubation with normal serum, confirming the released NTBI as apotransferrin-exchangeable (Online Supplementary Figure S5). Thus, Fe-sucrose is more labile than the other IVIP, consistent with our clinical observations, Figure 4D. This in vitro NTBI behavior may result from rapid reconstitution of alkaline Fe-sucrose solution (pH 10-11) in plasma (pH 7.4) with speciation changes enabling iron donation to the CP851 chelator (or, by extension, an endogenous chelator in vivo e.g., plasma citrate) more rapidly than with Fe-carboxymaltose or Fe-isomaltoside-1000. Furthermore, the protein corona could make a contribution to the stability of the iron core, and macrophage processing of that is unique to each IVIP formulation. Late onset of NTBI appearance with Fe-carboxymaltose strongly suggests that its iron becomes available to macrophage ferroportin much later, and that the initial rapid release from the formulation directly is negligible (Online Supplementary Figure S5). The crystalline (i.e., more stable) akaganeite-like form of Fe-carboxymaltose iron may be digested differently within endolysosomes than haematologica | 2021; 106(11)

NTBI following intravenous iron administration

the less well defined structure of Fe-sucrose iron.5 Although Fe-isomaltoside-1000 and Fe-carboxymaltose both possess akaganeite structures, the core size in Fe-isomaltoside-1000 is smaller.5 Thus, differential stability between these two preparations may be explained by different carbohydrate shells or different core sizes of the crystalline iron (hence more significant differences in surface area).6 The results of this study should be considered in the context of some limitations. Firstly, the IVIP were all administered at 200 mg doses of elemental iron, which was required to accurately evaluate the complex pharmacokinetic profiles of each formulation. Thus, following the successful model developed in this work, doses that are clinically administered should be studied in the future. It is possible that larger doses may increase NTBI generation by Pathway II or that AUC of s-Ferritin greatly increases if there is a transition from linear to zero-order elimination. Secondly this study did not directly compare different NTBI assays and evaluate their association with the pharmacokinetic and pharmacodynamic profiles of the ironcarbohydrate nano-medicines, however, this is the subject of ongoing work from our group. In conclusion, IVIP PK and PD support a two-pathway model of NTBI release. Although all IVIP are iron sources, the rate and extent of iron bioavailability differs, following over a 2-week period the sequence Fecarboxymaltose>Fe-sucrose>Fe-isomaltoside-1000. This strengthens the notion of IVIP as pro-drugs that should not be considered interchangeable. The potential clinical

References 1. Garbowski MW, Ma Y, Fucharoen S, Srichairatanakool S, Hider R, Porter JB. Clinical and methodological factors affecting non-transferrin-bound iron values using a novel fluorescent bead assay. Transl Res. 2016;177:19-30.e5. 2. de Swart L, Hendriks JCM, van der Vorm LN, et al. Second international round robin for the quantification of serum non-transferrin-bound iron and labile plasma iron in patients with iron-overload disorders. Haematologica. 2016;101(1):38-45. 3. Evans RW, Rafique R, Zarea A, et al. Nature of non-transferrin-bound iron: studies on iron citrate complexes and thalassemic sera. J Biol Inorg Chem. 2008;13(1):57-74. 4. Porter JB, Garbowski M. The pathophysiology of transfusional iron overload. Hematol Oncol Clin North Am. 2014; 28(4):683-701. 5. Neiser S, Rentsch D, Dippon U, et al. Physico-chemical properties of the new generation IV iron preparations ferumoxytol, iron isomaltoside 1000 and ferric carboxymaltose. BioMetals. 2015;28(4):615635. 6. Geisser P. Why different iron(III)-oxyhydroxide complexes have different PK/PD characteristics and specific reactivities. J Pharm Nanotechnol. 2016;4(1):14-18. 7. Bhandari S, Pereira DIA, Chappell HF, Drakesmith H. Intravenous irons: from basic science to clinical practice. Pharmaceuticals. 2018;11(3):1-20. 8. Koskenkorva-Frank TS, Weiss G, Koppenol WH, Burckhardt S. The complex interplay of iron metabolism, reactive oxygen species, and reactive nitrogen species: Insights into the potential of various iron

haematologica | 2021; 106(11)

consequences of NTBI on long-term iron bioavailability, efficacy and safety remain unknown and require further studies. Disclosures MG declares consultancy with Vifor Pharma, Imara; SBa declares no conflict of interest; JP declares honoraria from Celgene, Agios, Protagonist, LaJolla, Silence Therapeutics, bluebird bio, Vifor Pharma, and Cerus, as well as consultancy with Celgene, Agios and Novartis; CM declares employment by Vifor Pharma; SBu declares former employment by Vifor Pharma, RH declares research funding from Renapharma AB, and Vifor Pharma. Contributions MG performed the sample analyses, analysed the data and made the figures, designed and wrote the manuscript; SBa performed sample assays; JP co-wrote the manuscript, CM co-wrote the manuscript; SBu designed and oversaw the clinical trial, and co-wrote the manuscript; RH oversaw the sample analysis, codesigned and co-wrote the manuscript. All authors approved the manuscript. Acknowledgments The authors would like to thank Andrew Cakebread who operated the HPLC-ICPMS at KCL and Simon Cheesman who provided the intravenous iron preparations. Vifor Pharma provided blood samples and additional study data. JP would like to acknowledge UCL Biomedical Research Center for Cardiometabolic Program Support.

therapies to induce oxidative and nitrosative stress. Free Radic Biol Med. 2013;65:1174-1194. 9. Neiser S, Koskenkorva TS, Schwarz K, Wilhelm M, Burckhardt S. Assessment of dextran antigenicity of intravenous iron preparations with enzyme-linked immunosorbent assay (ELISA). Int J Mol Sci. 2016;17(7):1185. 10. Jahn MR, Andreasen HB, Fütterer S, et al. A comparative study of the physicochemical properties of iron isomaltoside 1000 (Monofer®), a new intravenous iron preparation and its clinical implications. Eur J Pharm Biopharm. 2011;78(3):480-491. 11. Geisser P, Burckhardt S. The pharmacokinetics and pharmacodynamics of iron preparations. Pharmaceutics. 2011;3(1):1233. 12. Rangel ÉB, Espósito BP, Carneiro FD, et al. Labile plasma iron generation after intravenous iron is time-dependent and transitory in patients undergoing chronic hemodialysis. Ther Apher Dial. 2010; 14(2):186-192. 13. Espósito BP, Breuer W, Slotki I, Cabantchik ZI. Labile iron in parenteral iron formulations and its potential for generating plasma nontransferrin-bound iron in dialysis patients. Eur J Clin Invest. 2002;32(s1):4249. 14. Ma Y, Podinovskaia M, Evans PJ, et al. A novel method for non-transferrin-bound iron quantification by chelatable fluorescent beads based on flow cytometry. Biochem J. 2014;463(3):351-362. 15. Danielson BG. Structure, chemistry, and pharmacokinetics of intravenous iron agents. J Am Soc Nephrol. 2004;15(SUPPL. 2):93-98. 16. Geisser P, Baer M, Schaub E. Structure/his-

totoxicity relationship of parenteral iron preparations. Arzneimittelforschung. 1992; 42(12):1439-1452. 17. Pai AB, Conner T, McQuade CR, Olp J, Hicks P. Non-transferrin bound iron, cytokine activation and intracellular reactive oxygen species generation in hemodialysis patients receiving intravenous iron dextran or iron sucrose. BioMetals. 2011;24(4):603-613. 18. Kitsati N, Liakos D, Ermeidi E, et al. Rapid elevation of transferrin saturation and serum hepcidin concentration in hemodialysis patients after intravenous iron infusion. Haematologica. 2015;100(3):e80-83. 19. Danielson BG, Salmonson T, Derendorf H, Geisser P. Pharmacokinetics of iron(III)hydroxide sucrose complex after a single intravenous dose in healthy volunteers. Arzneimittelforschung. 1996;46(6):615621. 20. Food and Drug Administration (FDA). Draft guidance on iron sucrose. 2012;20122013. 21. Bansal SS, Halket JM, Fusova J, et al. Quantification of hepcidin using matrixassisted laser desorption/ ionization timeof-flight mass spectrometry. Rapid Commun Mass Spectrom. 2009;23(11): 1531-1542. 22. Nguyen VH, Lee B-J. Protein corona: a new approach for nanomedicine design. Int J Nanomedicine. 2017;12:3137-3151. 23. Brissot P, Ropert M, Le Lan C, Loréal O. Non-transferrin bound iron: a key role in iron overload and iron toxicity. Biochim Biophys Acta. 2012;1820(3):403-410. 24. Loreal O, Gosriwatana I, Guyader D, Porter J, Brissot P, Hider RC. Determination of non-transferrin-bound iron in genetic hemochromatosis using a new HPLC-

2895

M.W. Garbowski et al. based method. J Hepatol. 2000;32(5):727733. 25. al-Refaie FN, Wickens DG, Wonke B, Kontoghiorghes GJ, Hoffbrand AV. Serum non-transferrin-bound iron in b-thalassaemia major patients treated with desferrioxamine and L1. Br J Haematol. 1992; 82(2):431-436. 26. Inati A, Musallam KM, Wood JC, SheikhTaha M, Daou L, Taher AT. Absence of cardiac siderosis by MRI T2* despite transfusion burden, hepatic and serum iron overload in Lebanese patients with sickle cell disease. Eur J Haematol. 2009;83(6):565-571. 27. Porter JB, Walter PB, Neumayr LD, et al. Mechanisms of plasma non-transferrin bound iron generation: insights from comparing transfused diamond blackfan anaemia with sickle cell and thalassaemia patients. Br J Haematol. 2014;167(5):692696. 28. Porter JB, Lin KH, Beris P, et al. Response of iron overload to deferasirox in rare transfusion-dependent anaemias: equivalent effects on serum ferritin and labile plasma iron for haemolytic or production anaemias. Eur J Haematol. 2011;87(4):338348. 29. Santini V, Girelli D, Sanna A, et al. Hepcidin levels and their determinants in different types of myelodysplastic syndromes. PLoS One. 2011;6(8):e23109. 30. Cortelezzi a, Cattaneo C, Cristiani S, et al. Non-transferrin-bound iron in myelodysplastic syndromes: a marker of ineffective erythropoiesis? Hematol J. 2000;1(3):153158. 31. Bradley SJ, Gosriwitana I, Srichairatanakool S, Hider RC, Porter JB. Non-transferrin-bound iron induced by myeloablative chemotherapy. Br J Haematol. 1997;99(2):337-343. 32. Sahlstedt L, Ebeling F, von Bonsdorff L, Parkkinen J, Ruutu T. Non-transferrinbound iron during allogeneic stem cell transplantation. Br J Haematol. 2001; 113(3):836-838. 33. Garbowski MW, Evans P, Vlachodimitropoulou E, Hider R, Porter JB. Residual erythropoiesis protects against myocardial hemosiderosis in transfusiondependent thalassemia by lowering labile plasma iron via transient generation of apotransferrin. Haematologica. 2017; 102(10): 1640-1649. 34. Garbowski MW, Ma Y, Fucharoen S, Srichairatanakool S, Hider R, Porter JB. Clinical and methodological factors affecting non-transferrin-bound iron values using a novel fluorescent bead assay. Trans Res. 2016;177:19-30.e5. 35. Porter JB, Cappellini MD, Kattamis A, et al. Iron overload across the spectrum of nontransfusion-dependent thalassaemias: role of erythropoiesis, splenectomy and transfusions. Br J Haematol. 2016;176(2):288-299. 36. Schmann K, Kroll S, Romero-Abal ME, et al. Impact of oral iron challenges on circulating non-transferrin-bound iron in

2896

healthy guatemalan males. Ann Nutr Metab. 2012;60(2):98-107. 37. Schümann K, Solomons NW, Orozco M, Romero-Abal ME, Weiss G. Differences in circulating non-transferrin-bound iron after oral administration of ferrous sulfate, sodium iron EDTA, or iron polymaltose in women with marginal iron stores. Food Nutr Bull. 2013;34(2):185-193. 38. Dresow B, Petersen D, Fischer R, Nielsen P. Non-transferrin-bound iron in plasma following administration of oral iron drugs. Biometals. 2008;21(3):273-276. 39. Hutchinson C, Al-Ashgar W, Liu DY, Hider RC, Powell JJ, Geissler CA. Oral ferrous sulphate leads to a marked increase in pro-oxidant nontransferrin-bound iron. Eur J Clin Invest. 2004;34(11):782-784. 40. Esposito BP, Breuer W, Sirankapracha P, Pootrakul P, Hershko C, Cabantchik ZI. Labile plasma iron in iron overload: Redox activity and susceptibility to chelation. Blood. 2003;102(7):2670-2677. 41. Erichsen K, Ulvik RJ, Grimstad T, Berstad A, Berge RK, Hausken T. Effects of ferrous sulphate and non-ionic iron-polymaltose complex on markers of oxidative tissue damage in patients with inflammatory bowel disease. Aliment Pharmacol Ther. 2005;22(9):831-838. 42. Khalid S, Shaikh F, Moeen S. Decreased activity of glutathione peroxidase with oral ferrous iron administration: a cause of oxidative stress. Pak J Pharm Sci. 2018; 31(2):405-409. 43. Pai AB, Boyd AV, McQuade CR, Harford A, Norenberg JP, Zager PG. Comparison of oxidative stress markers after intravenous administration of iron dextran, sodium ferric gluconate, and iron sucrose in patients undergoing hemodialysis. Pharmacotherapy. 2007; 27(3):343-350. 44. Kuo KL, Hung SC, Lee TS, Tarng DC. Iron sucrose accelerates early atherogenesis by increasing superoxide production and upregulating adhesion molecules in CKD. J Am Soc Nephrol. 2014;25(11):2596-2606. 45. Kshirsagar AV, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Brookhart MA. Intravenous iron supplementation practices and short-term risk of cardiovascular events in hemodialysis patients. PLoS One. 2013;8(11):e78930. 46. Macdougall IC, White C, Anker SD, et al. Intravenous iron in patients undergoing maintenance hemodialysis. N Engl J Med. 2019;380(5):447-458. 47. Brookhart MA, Freburger JK, Ellis AR, Wang L, Winkelmayer WC, Kshirsagar AV. Infection risk with bolus versus maintenance iron supplementation in hemodialysis patients. J Am Soc Nephrol. 2013; 24(7):1151-1158. 48. Macdougall IC, Bhandari S, White C, et al. Intravenous iron dosing and infection risk in patients on hemodialysis: a prespecified secondary analysis of the PIVOTAL trial. J Am Soc Nephrol. 2020;31(5):1118-1127. 49. Schümann K, Solomons NW, Romero-Abal

ME, Orozco M, Weiss G, Marx J. Oral administration of ferrous sulfate, but not of iron polymaltose or sodium iron ethylenediaminetetraacetic acid (NaFeEDTA), results in a substantial increase of nontransferrin-bound iron in healthy iron-adequate men. Food Nutr Bull. 2012;33(2):128136. 50. Scheiber-Mojdehkar B. Non-transferrinbound iron in the serum of hemodialysis patients who receive ferric saccharate: no correlation to peroxide generation. J Am Soc Nephrol. 2004;15(6):1648-1655. 51. Van Veldhuisen DJ, Ponikowski P, Van Der Meer P, et al. Effect of ferric carboxymaltose on exercise capacity in patients with chronic heart failure and iron deficiency. circulation. 2017;136(15):1374-1383. 52. Lewis GD, Malhotra R, Hernandez AF, et al. Effect of oral iron repletion on exercise capacity in patients with heart failure with reduced ejection fraction and iron deficiency the IRONOUT HF randomized clinical trial. JAMA. 2017;317(19):1958-1966. 53. Ambrosy AP, Lewis GD, Malhotra R, et al. Identifying responders to oral iron supplementation in heart failure with a reduced ejection fraction: a post-hoc analysis of the IRONOUT-HF trial. J Cardiovasc Med. 2019;20(4):223-225. 54. Gaillard CA, Bock AH, Carrera F, et al. Hepcidin response to iron therapy in patients with non-dialysis dependent CKD: an analysis of the FIND-CKD trial. PLoS One. 2016;11(6):e0157063. 55. Zehra A, Saleh Abdullah SM, Saboor M, Moinuddin. Effect of Intravenous iron supplementation on hepcidin levels in iron deficient pregnant females in second and third trimester. Indian J Hematol Blood Transfus. 2017;33(3):396-401. 56. Lin L, Valore E V, Nemeth E, Goodnough JB, Gabayan V, Ganz T. Iron transferrin regulates hepcidin synthesis in primary hepatocyte culture through hemojuvelin and BMP2/4. Blood. 2007;110(6):2182-2189. 57. Ganz T. Hepcidin and iron regulation, 10 years later. Blood. 2011;117(17):4425-4433. 58. Lee DH, Ding YL, Jacobs DR, et al. Common presence of non-transferrinbound iron among patients with type 2 diabetes. Diabetes Care. 2006;29(5):10901095. 59. Wish JB. Assessing iron status: beyond serum ferritin and transferrin saturation. Clin J Am Soc Nephrol. 2006;1(Suppl 1):S4S8. 60. Wish JB, Aronoff GR, Bacon BR, et al. Positive iron balance in chronic kidney disease: how much is too much and how to tell? Am J Nephrol. 2018;47(2):72-83. 61. Rouault TA, Stout CD, Kaptain S, Harford JB, Klausner RD. Structural relationship between an iron-regulated RNA-binding protein (IRE-BP) and aconitase: Functional implications. Cell. 1991;64(5):881-883. 62. Torti FM, Torti SV. Regulation of ferritin genes and protein. Blood. 2002; 99(10): 3505-3516.

haematologica | 2021; 106(11)

ARTICLE

Iron Metabolism & its Disorders

The TMPRSS6 variant (SNP rs855791) affects iron metabolism and oral iron absorption – a stable iron isotope study in Taiwanese women

Ferrata Storti Foundation

Simone Buerkli,1* Sung-Nan Pei,2,3,4* Shu-Chen Hsiao,5 Chien-Te Lee,2 Christophe Zeder,1 Michael B. Zimmermann1 and Diego Moretti1 1

Laboratory of Human Nutrition, Institute of Food Nutrition and Health, Department of Health Science and Technology, Swiss Federal Institute of Technology (ETH Zurich), Zurich, Switzerland; 2Department of Internal Medicine, Kaohsiung Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Kaohsiung, Taiwan; 3 Department of Hematology Oncology, E-Da Cancer Hospital, Kaohsiung, Taiwan; 4 College of Medicine, I-Shou University, Kaohsiung, Taiwan and 5Department of Pharmacy, Kaohsiung Chang Gung Memorial Hospital, Kaohsiung, Taiwan ° Current address: Swiss Distance University of Applied Sciences, Department of Health, Regensdorf/Zurich, Switzerland.

Haematologica 2021 Volume 106(11):2897-2905

*SB and SNP contributed equally as co-first authors.

ABSTRACT

enome wide studies have associated TMPRSS6 rs855791 (2321 C>T) with iron status and hepcidin. It is unclear whether this polymorphism affects iron absorption. We administered standardized ricebased test meals containing 4 mg of labeled 57Fe or 58Fe as FeSO4 on alternate days in non-anemic Taiwanese women (n=79, 44 TT variant, 35 CC variant). Fractional iron absorption was measured by erythrocyte incorporation of the tracers 14 days after administration. Compared to the CC variant, iron and transferrin saturation were lower (P=0.001; P<0.001, respectively) and serum hepcidin/transferrin saturation and serum hepcidin/serum iron ratios were higher (P=0.042; P=0.088, respectively) in the TT variant. Serum hepcidin did not differ between the groups (P=0.862). Geometric mean (95% Confidence Interval [CI]) fractional iron absorption, corrected to a serum ferritin of 15 mg/L, was 26.6% (95% CI: 24.0-29.5) in the CC variant and 18.5% (95% CI: 16.2-21.1) in the TT variant (P=0.002). Overall, predictors of iron absorption were: serum ferritin (P<0.001); genetic variant (P=0.032); and hepcidin (P<0.001). In the models by variant, in the CC variant the model explained 6771% of variability in absorption and serum ferritin was the only significant predictor (P<0.001); while in the TT variant, the model explained only 3543% of variability, and hemoglobin (P=0.032), soluble transferrin receptor (P=0.004) and hepcidin (P<0.001) were significant predictors. Women with the TMPRSS6 rs855791 (2321 C>T) polymorphism show altered iron homeostasis which affects oral iron absorption and may increase their risk for iron deficiency. The trial was registered as clinicaltrials gov. Identifier: NCT03317873, and funded by the Kaohsiung Chang-Gung Memorial Hospital, Kaohsiung, Taiwan, (grant CMRPG8F0721) and ETH Zurich, Switzerland.

Introduction In the absence of a physiological iron excretion mechanism, long-term iron balance in humans is determined by dietary iron absorption. Systemically, iron absorption is controlled by hepcidin (Hep), a peptide hormone synthesized in hepatocytes1 that regulates iron export from cells via its interaction with ferroportin.2 Hep is synthesized in response to increasing body iron in a homeostatic feedback loop, involving iron sensing of iron-saturated transferrin by transferrin receptors (Tfr1 and Tfr2) and associated proteins (HFE, hemojuvelin) initiating a cascade involving bone morphogenic protein (BMP) receptor activation.3,4 The TMPRSS6 gene encodes the transmembrane serine protease matriptase-2, which interacts with hemojuvelin, modulating the Hep activation pathway.5 Consistently with this regulatory model, nonsense mutations in TMPRSS6 cause

haematologica | 2021; 106(11)

Correspondence: DIEGO MORETTI diego.moretti@alumni.ethz.ch Received: June 25, 2020. Accepted: September 23, 2020. Pre-published: October 5, 2020. https://doi.org/10.3324/haematol.2020.264556

2897

S. Buerkli et al.

iron refractory iron deficiency anemia (IRIDA), due to inappropriately elevated Hep levels.6 The ratio of serum Hep/transferrin saturation (Hep/TS) may be useful to differentiate subjects with IRIDA from subjects with chronic iron deficiency (ID),7 consistent with a disrupted feedback loop between TS and Hep. Common genetic variants of TMPRSS6, are associated with erythrocyte parameters in human genome wide association studies.8-11 The single nucleotide polymorphism (SNP) rs855791 (2321 C>T) of TMPRSS6 has a population frequency of ≈0.5 in Caucasians,10,12 ≈0.6 in Japanese13 and ≈0.2-0.1 in African Americans.10,12 It causes a non-synonymous substitution near the catalytic and active site of the protease,10 with a strong association with iron status, erythrocyte parameters,8-10,12,14-17 Hep18 and ratios of Hep to iron indices.19,20 T-allele variants in the rs855791 are associated with an increased risk for ID and iron deficiency anemia (IDA).16,17 In a case control study in Taiwan, homozygotes for the SNP rs855791 CC had a lower prevalence of IDA, compared to subjects with the CT or TT variant.21 In European populations, variants (TT) in rs855791 are associated with lower TS and serum ferritin (SF), higher Hep, and higher ratios of Hep to iron indices.18-20 In first time blood donors, the TT variant was associated with larger decreases in SF and hemoglobin (Hb) after multiple donations, suggesting an impaired capacity to replenish stores following donation.22 ID is considered the most prevalent nutritional deficiency worldwide and one of the leading causes of anemia among non-pregnant and pregnant women.23 While iron status and dietary composition are the main determinants of iron absorption, individual factors other than iron status have been estimated to account for ≈50% of the variance in iron absorption.24 Furthermore, a strong familial tendency in iron absorption has been reported in mother child pairs using stable iron isotopes;25,26 this could be due to genetic, epigenetic or shared environmental mechanisms. The genetic determinants of iron status and Hep metabolism in humans, including the effect of mutations in TMPRSS6, are poorly understood. The study aim was to compare iron absorption, Hep and other indices of iron metabolism in iron-sufficient Taiwanese women carrying the TT or the CC variant of the rs855791 SNP in the TMPRSS6 gene. We hypothesized that the TT variant would be associated with higher serum Hep concentrations, higher ratios of Hep to iron indices, and lower iron absorption at comparable iron status.

ii) SF 30–120 mg/L and iii) CRP <5 mg/L. The ethical committees of ETH Zurich in Switzerland and the Chang Gung Memorial Foundation Institutional Review Board in Taiwan approved the study. All participants provided written informed consent, and the study was registered as clinicaltrials gov. Identifier: NCT03317873. On study days 1 and 3 (D1, D3), we administered two standardized rice test meals to fasting participants, labeled with 4 mg iron (57Fe, and 58Fe) as labeled ferrous sulfate (FeSO4). A detailed description of the test meal administration and the preparation of stable iron isotopes, can be found in the Online Supplementary Appendix.

Laboratory analyses We determined fractional iron absorption (FIA) based on the shift in the enrichment ratio of stable iron isotopes into the erythrocytes on D17. We performed the analyses by inductively coupled plasma mass spectrometry (MC-ICP-MS, Neptune; Thermo Finnigan) as previously described.27 We calculated the amounts of 57 Fe, and 58Fe isotopic labels in blood on D17 on the basis of the shift in iron isotope ratios and on the estimated amount of iron circulating in the body.28 We corrected the FIA for SF to the cutoff for ID (15 mg/L),29 and to 50 mg/L as a level representing sufficient iron stores with a modification of the Cook et al. formula,30 as described in the Online Supplementary Appendix. Procedures such as the assessment of menstrual blood loss, and laboratory measurements such as genotyping, measurement of erythrocyte parameters, CRP, acute phase protein a-1-acid glycoprotein (AGP), SF, serum iron (SFe), total iron-binding capacity (TIBC), Hep, and soluble transferrin receptor (sTfR) are also described in the Online Supplementary Appendix.

Sample size calculation We based the sample size calculation on a design with two repeated measurements with a compound symmetry covariance structure. Based on previous studies from the Human Nutrition Laboratory, using log-transformed data, we assumed an intra-individual correlation of 0.7, and a standard deviation of 0.235. A difference of 30% in iron absorption was considered relevant. Therefore, we planned to recruit 40 subjects per variant, with 80% power and a = 0.05, it allows two dropouts per group. Due to the imbalanced distribution of the minor allele in the Taiwanese population, and difficulties enrolling the planned number of CC subjects, we made a protocol amendment to include 35 CC and 45 TT subjects. This unbalanced distribution results in an estimated power of 75%.

Data and statistical analysis Methods Subjects The study was performed at the Kaohsiung Chang Gung Memorial Hospital (K-CGMH) in Taiwan, between February 2018 and February 2019. The flow of study participants is shown Figure 1. We invited apparently healthy females, with no known history of thalassemia or anemia, aged between 20 to 45 years for screening, assessed their medical history and measured body weight and height, complete blood count, SF and the rs855791 genotype. Inclusion criteria are described in the Online Supplementary Appendix. All participants that were homozygous in the rs855791 (TT or CC), and fulfilled all inclusion criteria were recalled 1 week before the first test meal administration, where we assessed Hb, SF, C-reactive protein (CRP), and menstrual blood losses. Study inclusion criteria were: i) Hb > 120 g/L;

2898

We used IBM SPSS statistics (Version 24) for statistical analysis. After testing for normality, we used log-transformed data further analysis if not normally distributed. Normally distributed data is presented as means ± standard deviation (SD), transformed normal data as geometric mean with the 95% Confidence Interval (95% CI), non-normal data as median and the interquartile range (IQR). Means or medians of red cell parameters, are based on the concentrations measured on D1. Means, medians, or geometric means of CRP, AGP, SF, SFe, TIBC, TS, sTfR, BIS, Hep, Hep/TS, Hep/SF, and FIA are based on concentrations measured on D1 and D3. We tested between-group differences for i) normally distributed variables using independent samples t-test and for ii) not normally distributed variables using Mann-Whitney U test, as well as iii) differences in CRP, AGP, SF, SFe, TIBC, TS, sTfR, BIS, Hep, Hep/TS, Hep/SF, and FIA by linear mixed models (LMM), with subjects’ code as a random intercept, the corresponding variable as a dependent variable and the genotype as a fixed effect. We

haematologica | 2021; 106(11)

A TMPRSS6 polymorphism affects iron absorption

Figure 1. Study flow chart.

assessed Pearson’s correlations and differences between the coefficients with with the Fisher Z-transformation. We assessed predictors of iron absorption with LMM using subjects’ code as a random intercept, FIA as a dependent variable, and the genotype, Hb, SF, TS, sTfR, Hep and PBAC as fixed factors. We performed a backward linear regression to assess a minimal adequate model, and we fitted the variables in a LMM. Statistical significance was defined as P<0.05.

women with TT variants and 15 with CC variants received iron supplements. After the iron supplementation period, 20 subjects with the TT and 12 subjects with the CC variant were included into the study. Finally, 35 subjects with CC and 45 with the TT variant fulfilled all study inclusion criteria and were enrolled (Figure 1). One woman with the TT variant left the study after study D3, thus, 79 women completed the study.

Iron indices Results Subjects We screened 296 women and identified 93 women carrying the TT variant and 66 with the CC variant, while 137 were excluded as heterozygotes (Figure 1). Of the identified subjects, 33 women with the TT variant and 30 with CC variant met all inclusion criteria. Thirty-four haematologica | 2021; 106(11)

SF concentrations were balanced between the two variants, while SFe was lower in the TT compared to the CC variant (P=0.001; Table 1). Similarly, TS was lower (P<0.001), and TIBC higher (P=0.086) in the TT variant (Table 1). While Hep did not differ between the groups (P=0.862), the Hep/TS ratio (P=0.042) and the Hep/SFe ratio (P=0.088) were 28% and 25% higher in the TT variant, respectively (Table 1). None of the subjects had sys2899

S. Buerkli et al.

Table 1. Subject characteristics of Taiwanese women with the homozygous CC and TT variants of the rs855791 in TMPRSS6. n Age, y* Weight, kg* Height, cm* CRP, mg/L† AGP, g/L* RBC, million/mL‡ Hb, g/dL HCT, %‡ MCV, fL/cell‡ MCH, pg/cell‡ SF, mg/L† SFe, mg/dL† TIBC, mg/dL TS, %† sTfR, mg/L† BIS, mg/kg BW Hep, nM† Hep/SFe, pmol/mg† Hep/TS, pM/%† Hep/SF, pmol/mg† PBAC*

35 34 ± 6 54.6 ± 4.8 160 ± 4 0.271 (0.213, 0.346) 0.426 ± 0.0962 4.51 (4.19-4.59) 13.3 ± 0.6 40.0 (38.4-40.9) 89.8 (88.3-92.0) 30.0 (29.2-31.0) 45.1 (41.0, 49.7) 114.5 (105.2, 124.7) 315.6 ± 29.7 36.5 (33.3, 39.9) 4.14 (3.97, 4.32) 7.13 ± 1.62 2.10 (1.80, 2.46) 183.4 (160.3, 209.8) 57.6 (50.8, 65.4) 46.5 (40.7, 53.2) 126 ± 63

45 36 ± 7 54.2 ± 5.2 160 ± 5 0.388 (0.306, 0.491) 0.455 ± 0.113 4.58 (4.37-4.74) 13.3 ± 0.7 40.3 (38.7-41.5) 88.4 (86.4-90.6) 29.6 (28.8-30.0) 47.0 (43.5, 50.9) 90.3 (82.6, 98.7) 327.3 ± 33.5 27.7 (25.5, 30.2) 4.33 (4.12, 4.55) 7.12 ± 1.80 2.06 (1.79, 2.37) 227.8 (193.3, 268.4) 74.2 (63.0, 87.4) 43.7 (38.4, 49.8) 171 ± 99

P 0.436§ 0.717§ 0.780§ 0.125|| 0.206|| 0.032¶ 0.762§ 0.349¶ 0.008¶ 0.018¶ 0.626|| 0.001|| 0.086||| <0.001|| 0.351|| 0.978|| 0.862|| 0.088|| 0.042|| 0.537|| 0.015§

Anthropometrics, red blood cell (RBC) indices and hemoglobin (Hb) were assessed on day (D) 1, inflammation, iron parameters, plasma hepcidin concentration were assessed on D1 and D3. AGP: acute phase protein a-1-acid glycoprotein; BIS: body iron stores; CRP: C-reactive protein; Hb: hemoglobin; HCT: hematocrit; Hep: plasma hepcidin; MCH: mean corpuscular hemoglobin; MCV: mean corpuscular volume; PBAC: pictorial blood-loss assessment chart; SF: serum ferritin; Sfe: serum iron; sTfR: soluble transferrin receptor; TS: transferrin saturation; y: years. ∗Means ± standard deviation (SD); †geometric means (95% Confidence Interval [CI]); ‡medians (interquartile range [IQR]); §differences were assessed by two-sided independent t-test; ||differences were assessed by fitting linear mixed models with genotype as fixed effect, participants as the random effects, and the corresponding variable as dependent variable; ¶differences were assessed by Mann-Whitney U test.

temic inflammation, during the study period (Table 1). The menstrual blood loss scores (PBAC) was higher in the TT variant (P=0.015, Table 1).

more pronounced (P=0.004) in the CC variant (CC: r=0.81, P<0.001, TT: r=-0.45, P=0.002).

Predictors of fractional iron absorption Fractional iron absorption The uncorrected FIA on D1 and D3 within variant did not differ (Table 2; Figure 2) but the Pearson’s correlation between days (D1 and D3) FIA was stronger in the CC (r=0.86) than in the TT variant (r=0.67; for both, P<0.001; Figure 2). The mean uncorrected FIA of D1 and D3 of the TT variant, was 7.96% (95% CI: 6.87-9.22), and in the CC variant was 6.50% (95% CI:5.54-7.62) (P=0.160; Figure 2). FIA corrected to a SF concentration of 15 mg/L was significantly lower in the TT 18.5% (95% CI: 16.2-21.1), compared to the CC variant 26.6% (95% CI: 24.0-29.5) (P=0.002; Table 2; Figure 2). When corrected to a SF of 50 mg/L, the TT variant had significantly higher FIA than the CC variant: 7.59% (95% CI: 6.66-8.66), compared to 5.70% (95% CI: 5.15-6.31) (P=0.012).

Correlation of fractional iron absorption, iron indices and hepcidin The Pearson’s correlation between FIA and SF was more pronounced in the CC variant (r=-0.79, P<0.001) than in the TT variant (r=-0.45, P=0.002) and there was a difference in the strength of the correlation between groups (P<0.001, Figure 3). Fractional absorption was correlated with TS only in the CC variant (CC: r=-0.45, P=0.006; TT r=-0.14, P=0.360) and the correlation coefficients tended to differ (P=0.070). The correlation of FIA with Hep was 2900

Genetic variant (b=-0.346, P=0.032) was a significant predictor of overall FIA along with SF (b=-0.393, P<0.001), and Hep (b=-0.312, P<0.001), (R2adjusted=0.468), Table 3. Stepwise deletion removed TS, PBAC and CRP from the model (R2adjusted=0.469). In the prediction model by variant, in the CC variant only SF was significantly associated with FIA (b=-0.696, P<0.001), explaining 67% of the variability in iron absorption (R2adjusted=0.669) (Table 4). In contrast, in the TT variant, Hep (b=-0.353, P<0.001), sTfR (b=0.317, P=0.004), Hb (b=-0.252, P=0.023), and TS (b=0.199, P=0.011) were associated with FIA (Table 4), but with a substantially lower coefficient of determination (R2adjusted=0.375), explaining 38% of the variability. In the minimal adequate model (Table 5), for the CC variant, significant predictors are SF (b=-0.667, P<0.001) and Hep (b=-0.217, P=0.002) (R2adjusted=0.688). For the TT variant, significant predictors are Hep, (b=-0.411, P<0.001) sTfR, (b=0.320, P=0.003) and Hb (b=-0.226, P=0.038) (R2adjusted=0.356, Table 5).

Discussion Our study shows that the TMPRSS6 rs855791 TT variant is associated with lower iron absorption in an overall model controlling for other iron status indicators. At a haematologica | 2021; 106(11)

A TMPRSS6 polymorphism affects iron absorption

Table 2. Fractional iron absorption from rice meals in CC and TT variants of the rs855791.

CC variant FIAD1, % FIAD3, % FIAD1 & D3, % FIAD1 & D3, SF15corr, %||

TT variant

6.50 (5.14, 8.22)* 6.49 (5.17, 8.15)* 6.50 (5.54, 7.62) 26.6 (24.0, 29.5)

P †

7.99 (6.39, 9.98) 7.93 (6.50, 9.68)† 7.96 (6.87, 9.22) 18.5 (16.2, 21.1)

0.206‡ 0.183‡ 0.160§ 0.002§

Values are the geometric means and the 95% Confidence Interval (CI). D1: study day 1; D3: study day 3; FIA: fractional iron absorption; SF15corr: serum ferritin correction to a concentration of 15 µg/L. ∗differences between study day one and three were assessed by paired t test P=0.984. †differences between study day one and three were assessed by paired t test P=0.935. ‡differences between the two variants were assessed by two-sided independent t-test. §differences between the two variants were assessed by fitting linear mixed models with genotype as fixed effect, participants as the random effects, and the corresponding variable as dependent variable. ||correction was done using the formula: log(FIAC) = log(FIAO) + a * log(SFC/SFO), with aCC = -1.28, aTT = -0.74.

standardized SF concentration of 15 mg/L, iron absorption was significantly lower in the TT variant. The TT variant also had lower TS and SFe and higher Hep/TS ratios, suggesting an altered interplay of SFe, hepcidin, iron stores and the regulation of dietary absorption compared to the CC variant. Similarly, known predictors of iron absorption explained much less of the variability in iron absorption in the TT variant. To our knowledge, this is the first study comparing dietary iron absorption using stable iron isotopes among the common SNP rs855791 of the TMPRSS6, which has been associated with iron status and red blood cell parameters in various genome wide association studies and large cross-sectional studies.10,12,15-17 In humans, inter-subject absorption of nonheme iron shows a wide variation in healthy young women. Zimmermann et al. reported a variation from 1% to 58% in iron absorption from standardized test meals labeled with 4 mg Fe as stable isotopes.31 Some of this variation is haematologica | 2021; 106(11)

Figure 2. Correlation of the inter-individual fractional iron absorption (A and B) and fractional iron absorption (C-D) in rs855791 variants. (A and B) Fractional iron absorption (FIA) measured from identical rice test meals on study day one and three, separated by variant in the TMPRSS6 rs855791, in (A) the CC variant (○ n=35), and in (B) the TT (□ n=44). The Pearsons correlation factors are: 0.86, and 0.67 for the CC and TT, respectively (both, **P<0.001). (C and D) Each point represents the mean of the FIA on day 1 (D1) and D3 from two identical rice meals, the line represents the geometric mean and the bars the 95% Confidence interval (CI). (C) the measured FIA versus (D) the FIA corrected to a serum ferritin (SF) concentration of 15 mg/L are shown. Differences betweenthe two variants were assessed by fitting linear mixed models with genotype as fixed effect, participants as the random effects, and FIA or FIA corrected for SF as the dependent variable CC ( ○, n=35) versus TT ( □, n=44).

due to differences in iron status and meal matrix, however, taken together, it has been estimated that iron status and food factors predict only ≈50% of the variance in iron absorption in a population.24 Cook et al.32 reported a striking positive correlation in body iron in iron-replete mothers and their young children and suggested this close correlation was due to a shared diet and/or possible genetic determinants of iron status such as shared iron-regulatory genes. In Mexican (n=18), and Senegalese mother-child pairs (n=17), non-heme-iron absorption measured with stable isotopes exhibited strong and intermediate correlations, respectively.25,26 A common polymorphism in the transferrin protein (G277S) has been associated with ID in American women,33 but a stable isotope study comparing 25 iron deficient, non-anemic women who had either a heterozygous G277S/G277G or wild-type G277G/G277G genotype did not find a significant difference in iron absorption.34 However, the G277S carriers did not show 2901

S. Buerkli et al.

Table 3. Predictors of iron absorption in healthy Taiwanese women (n=79).

Minimal Adequate Model†

Overall Model∗ Variables Intercept Variant (CC vs. TT)‡ Hemoglobin Serum Ferritin Transferrin Saturation Soluble Transferrin Receptor Plasma Hepcidin PBAC C-reactive Protein

b 0.14 -0.35 -0.13 -0.39 0.11 0.12 -0.31 0.07 -0.05

SE 0.10 0.16 0.07 0.08 0.06 0.07 0.06 0.08 0.07

P 0.159 0.032 0.085 <0.001 0.063 0.083 <0.001 0.389 0.434

b 0.13 -0.30 -0.11 -0.41 0.12 -0.29

SE 0.01 0.15 0.07 0.07 removed from the model§ 0.07 0.06 removed from the model§ removed from the model§

P 0.197 0.045 0.154 <0.001 0.105 <0.001

Analyzed by LMM using standardized variables, dependent variable: fractional iron absorption; fixed factors: potential continuous or categorical predictors; random effects: Subjects’ code. Shown are standardized b-coefficients standard errors (SE). Hemoglobin, menstrual blood loss scores (PBAC) were assessed on day (D) 1, inflammation, iron parameters, plasma hepcidin concentration, and fractional iron absorption (FIA) are based on data measured on D1 and D3. ∗Regression model fit: R2=0.498; R2adjusted=0.468;†assessed by backward linear regression; regression model fit: R2=0.486; R2adjusted=0.469; ‡nominal variable; 1=CC, 2=TT; §removed variable by the backward regression to assess the minimal adequate model.

Table 4. Potential predictors of iron absorption in variants of the TMPRSS6 rs855791 (nCC=35, nTT=44).

TT Variant†

CC Variant∗ Variables Intercept Hemoglobin Serum Ferritin Transferrin Saturation Soluble Transferrin Receptor Plasma Hepcidin PBAC C-reactive Protein

b 0.0003 0.05 -0.70 -0.02 -0.12 -0.16 -0.12 0.04

SE 0.08 0.09 0.010 0.07 0.08 0.08 0.09 0.08

P 0.997 0.559 <0.001 0.752 0.161 0.053 0.229 0.614

b -0.02 -0.25 -0.20 0.20 0.32 -0.35 0.14 -0.05

SE 0.10 0.11 0.11 0.08 0.10 0.08 0.11 0.10

P 0.851 0.023 0.060 0.011 0.004 <0.001 0.211 0.610

Analyzed by LMM using standardized variables, dependent variable: fractional iron absorption; fixed factors: potential continuous or categorical predictors; random effects: Subjects’ code. Shown are standardized b-coefficients with their standard errors. Hemoglobin, menstrual blood loss scores (PBAC) were assessed on day (D) 1, inflammation, iron parameters, plasma hepcidin concentration, and fractional iron absorption are based on data measured on D1 and D3. ∗Regression model fit of CC variant: R2=0.707; R2adjusted=0.669; †regression model fit of TT variant: R2=0.432; R2adjusted=0.375.

the typical inverse correlation between iron absorption and SF.34 Similarly, in our study with the TMPRSS6 rs855791 mutation, in the TT variant, the correlation between iron absorption and SF, and iron absorption and Hep, were only weak and moderate, respectively. In contrast to the CC variant, where both these correlations are strong. Also the models computed by variant show remarkable differences. In the CC variant SF alone is significantly associated with FIA, explaining 67% of the variability. In the TT variant, in contrast, several factors identify as being associated with FIA: Hep, sTfR, Hb, and TS, and their total contribution explain only 38% of the variability in iron absorption. Our hypothesis that the CC variant would have increased iron absorption was based on a regulatory model of TMPRSS6 acting as a negative regulator of the Hep activation pathway, and we hypothesized the largest effects would be seen in an iron replete population, where Hep expression would be activated. However, our findings indicate the effects of the genetic variant are likely most relevant at low iron status (Figure 3); at lower SF, women with the TT variant were less able to upregulate iron absorption, which could increase the risk for ID. Further, the overall model (Table 3) shows that iron status indices, Hep, and genotype, but not inflammation and menstrual blood loss are associated with fractional iron absorption. A recent large study in blood donors suggests an impaired capacity in the TT 2902

variant to replenish iron stores after repeated blood donations, even if the possibly protective CC variant was not enriched in high intensity donors.22 It is also possible that cellular mechanisms controlled by iron regulatory proteins are, especially at intermediate serum iron levels, able to compensate for the altered interplay of Hep and TS in the TT variant by inducing the translation of iron transporters (e.g., DMT1) and transcription factor HIF-2a.35 Such a compensatory mechanism was suggested in a recent study in women in whom an acute inflammatory stimulus increased Hep but did not affect iron absorption.36 Our findings suggest that, at low SF concentrations, women with the TT variant have lower iron absorption, whereas when iron stores are replete, they may be less able to downregulate iron absorption compared to the CC variant. Our variant-specific FIA correction to SF uses a similar approach as the original formula of Cook et al.30 used to correct dietary absorption measurements for the individual iron status; that formula employs a slope of -1 between log FIA and log SF. We propose adapted, regression formulas with slopes of -1.28 and -0.74 for CC and TT variants, respectively (Figure 3). In a case control study in Taiwanese women comparing women with IDA to non-anemic controls, the CC variant was less frequent in the IDA group compared to the control group (12% vs. 25%); this suggests the CC variant haematologica | 2021; 106(11)

A TMPRSS6 polymorphism affects iron absorption

may reduce risk of IDA.21 This effect is also suggested in our screening data: among the screened subjects, 60% of women with the TT variant had either Hb below 12 g/dL and/or SF below the study inclusion criteria, compared to 42% of women with the CC variant. Also, among women who received iron supplementation due to ID, 80% of the women with the CC variant replenished their iron stores, in contrast to only 59% of women with the TT variant. While this is consistent with the view that women with the TT variant are at a higher risk for ID and may have a blunted response to iron supplements when body iron stores are low, this hypothesis needs confirmation in larger prospective trials. Further mechanistic studies in monozygotic twins would be particularly informative as they may distinguish potential genetic and epigenetic sources of variability in iron absorption. Our findings are consistent with previous studies that have shown that Sfe and transferrin saturation are lower, haematologica | 2021; 106(11)

Figure 3. Correlations of fractional iron absorption. Correlations between fractional iron absorption (FIA) and serum ferritin (SF) (A and B), transferrin saturation (TS) (C and D) and hepcidin (Hep) (E and F) of the participants separated.by the variants in the TMPRSS6 rs855791, CC ( ○, n=35) and TT (□, n=44). Each point represents one participant and their mean of FIA, SF, TS, and Hep measured on study day 1 (D1) and D3. Pearson’s correlation factors r for FIA to SF correlation are: -0.79 (P<0.001) and -0.45 (P=0.002); for FIA to TS correlation: -0.45 (P=0.006) and -0.14 (P=0.360); for FIA to Hep correlation: -0.81 (P<0.001) and 0.45, (P=0.002), for the CC variants and TT variants respectively. **P<0.001, *P<0.05,

and TIBC is higher in the TT variant.18-20 The higher Hep/TS ratio reported in women with the TT variant in our study has been previously described in Italian18 and Dutch populations.20 In our study, the lack of association between variant and Hep suggests a different modulation of iron regulatory signals (transferrin bound iron and iron stores) in the regulation of Hep between the two different variants. Consistent with this interpretation, a recent study has suggested that the Hep/TS ratio may be a useful diagnostic marker to differentiate IRIDA patients from those with chronic ID.7 A strength of this study is that iron absorption was assessed from an isotopically labeled standardized labeled test meals in a relatively large number of subjects, using erythrocyte incorporation of stable iron isotope labels. Due to its precision, this approach allows, in combination with iron indices, to study regulatory aspects of iron metabolism in humans.36,37 We performed iron absorption 2903

S. Buerkli et al.

Table 5. The minimal adequate model and predictors of iron absorption in variants of the TMPRSS6 rs855791 (nCC=35, nTT=44).

TT Variant†

CC variant∗ Variables Intercept Hemoglobin Serum Ferritin Transferrin Saturation Soluble Transferrin Receptor Plasma Hepcidin PBAC C-reactive protein

b 0.0004 -0.67

-0.22 -0.11

SE 0.08 removed from the model‡ 0.09 removed from the model‡ removed from the model‡ 0.07 0.09 removed from the model

P 0.996 -0.23 <0.001

b -0.01 0.11

0.32 0.002 0.226

0.10 -0.41

SE 0.11 0.038 removed from the model‡ removed from the model‡ 0.003 0.08 removed from the model‡ removed from the model‡

P 0.911

<0.001

The minimal adequate model is assessed by backward linear regression using standardized variables. Parameters shown are analyzed by LMM, dependent variable: fractional iron absorption; fixed factors: potential continuous or categorical predictors; random effects: Subjects’ code. Shown are standardized b-coefficients with their standard errors (SE). Hemoglobin, menstrual blood loss scores (PBAC) were assessed on day (D) 1, inflammation, iron parameters, plasma hepcidin concentration, and fractional iron absorption are based on data measured on D1 and D3. ∗Regression model fit of CC variant: R2=0.702; R2adjusted=0.688; †regression model fit of TT variant: R2=0.378; R2adjusted=0.356; ‡removed variable by the backward regression to assess the minimal adequate model.

measurements twice in each subject; this increased statistical power and allowed us to make intra-individual comparisons. Our study also has limitations: our assessment of menstrual loss using PBAC is semi-quantitative, and while we found no association with iron absorption after correcting for iron status in the overall model, we cannot fully exclude a potential effect of menstrual blood loss on iron absorption. Our proposed genotype-specific slopes of SF and iron absorption are based on a relatively narrow range of iron status and should be studied in populations with broader iron status distribution. We focused our hypothesis on a single SNP, and we did not study the interplay with other SNP known to affect iron homeostasis. However, the HFE rs1800562 (C282Y) mutation is known to be rare in Taiwanese women.38 In contrast, the GNPAT rs11558492 has been associated with a high-iron phenotype,39 and a recent study in Taiwanese women has shown a minor allele frequency of 12%, and a significant higher serum iron response after a supplement.40 We did not study heterozygotes, despite the fact that effects on iron absorption are conceivable in this group. Furthermore, unknown SNP associated with rs855791 may explain the observed effects. However, we think this possibility is unlikely, as rs855791 has been repeatedly shown to be associated with iron status, as discussed above. To summarize, we have shown that in a fully adjusted model of iron absorption, women with the TT variant have lower iron absorption compared to women with the CC variant. This may be associated with higher Hep/TS and Hep/SFe ratios, suggesting impaired negative feedback on Hep synthesis by circulating iron. Furthermore, in the TT variant, regulation of iron absorption is less well

References 1. Zumerle S, Mathieu JR, Delga S, et al. Targeted disruption of hepcidin in the liver recapitulates the hemochromatotic phenotype. Blood. 2014;123(23):3646-3650. 2. 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. 3. Gao J, Chen J, Kramer M, Tsukamoto H, Zhang AS, Enns CA. Interaction of the hereditary hemochromatosis protein HFE

2904

predicted by iron stores. Thus, our findings suggest women with the TT variant are less able to upregulate iron absorption at low iron status, which may increase their risk of ID. Disclosures No conflicts of interest to disclose. Acknowledgments The authors thank all subjects who participated in the study and the nursing staff who essentially contributed to the conduction; MiaoChin Sun and Yu-Ching Chan for preparation of all the test meals, help in participant recruitment and study conduction; Min-Yi Tsai for sample handling and genetic variant analysis; Nicole Härter for careful preparation of the whole blood samples; Adam Krzystek, and Timo Christ for careful analysis of the samples on the MS-ICPMS.

Contributions MBZ, DM, SB, and SNP designed the study; SNP, SCH and CTL conducted the study and collected the samples; SB and CZ analyzed the samples and performed the statistical analyses, SB, DM, MBZ, and SNP participated in the data interpretation; SB wrote the first draft of the manuscript; all authors edited the manuscript and approved the final version.

Funding This study was supported by the Kaohsiung Chang-Gung Memorial Hospital, Kaohsiung, Taiwan, (grant CMRPG8F0721) and Laboratory of Human Nutrition, Institute of Food Nutrition and Health, Department of Health Science and Technology, Swiss Federal Institute of Technology (ETH Zurich), Zurich, Switzerland.

with transferrin receptor 2 is required for transferrin-induced hepcidin expression. Cell Metab. 2009;9(3):217-227. 4. Steinbicker AU, Bartnikas TB, Lohmeyer LK, et al. Perturbation of hepcidin expression by BMP type I receptor deletion induces iron overload in mice. Blood. 2011; 118(15):4224-4230. 5. Silvestri L, Pagani A, Nai A, De Domenico I, Kaplan J, Camaschella C. The serine protease matriptase-2 (TMPRSS6) inhibits hepcidin activation by cleaving membrane hemojuvelin. Cell Metab. 2008;8(6):502511.

6. Camaschella C. New insights into iron deficiency and iron deficiency anemia. Blood Rev. 2017;31(4):225-233. 7. Heeney MM, Guo D, De Falco L, et al. Normalizing hepcidin predicts TMPRSS6 mutation status in patients with chronic iron deficiency. Blood. 2018;132(4):448452. 8. McLachlan S, Giambartolomei C, White J, et al. Replication and characterization of association between ABO SNPs and red blood cell traits by meta-analysis in Europeans. PLoS One. 2016;11(6): e0156914.

haematologica | 2021; 106(11)

A TMPRSS6 polymorphism affects iron absorption

9. Ding K, Shameer K, Jouni H, et al. Genetic Loci implicated in erythroid differentiation and cell cycle regulation are associated with red blood cell traits. Mayo Clin Proc. 2012; 87(5):461-474. 10. Chambers JC, Zhang W, Li Y, et al. Genome-wide association study identifies variants in TMPRSS6 associated with hemoglobin levels. Nat Genet. 2009; 41(11):1170-1172. 11. Timmer T, Tanck MWT, Huis In 't Veld EMJ, et al. Associations between single nucleotide polymorphisms and erythrocyte parameters in humans: a systematic literature review. Mutat Res. 2019;779:58-67. 12. Benyamin B, Ferreira MA, Willemsen G, et al. Common variants in TMPRSS6 are associated with iron status and erythrocyte volume. Nat Genet. 2009;41(11):1173-1175. 13. Kamatani Y, Matsuda K, Okada Y, et al. Genome-wide association study of hematological and biochemical traits in a Japanese population. Nat Genet. 2010; 42(3):210-215. 14. Sorensen E, Rigas AS, Thorner LW, et al. Genetic factors influencing ferritin levels in 14,126 blood donors: results from the Danish Blood Donor Study. Transfusion. 2015;56(3):622-627. 15. Sorensen E, Rigas AS, Didriksen M, et al. Genetic factors influencing hemoglobin levels in 15,567 blood donors: results from the Danish Blood Donor Study. Transfusion. 2019;59(1):226-231. 16. Gan W, Guan Y, Wu Q, et al. Association of TMPRSS6 polymorphisms with ferritin, hemoglobin, and type 2 diabetes risk in a Chinese Han population. Am J Clin Nutr. 2012;95(3):626-632. 17. An P, Wu Q, Wang H, et al. TMPRSS6, but not TF, TFR2 or BMP2 variants are associated with increased risk of iron-deficiency anemia. Hum Mol Genet. 2012;21(9):21242131. 18. Nai A, Pagani A, Silvestri L, et al. TMPRSS6 rs855791 modulates hepcidin transcription in vitro and serum hepcidin levels in normal individuals. Blood. 2011;118(16):44594462. 19. Traglia M, Girelli D, Biino G, et al. Association of HFE and TMPRSS6 genetic variants with iron and erythrocyte parameters is only in part dependent on serum

haematologica | 2021; 106(11)

hepcidin concentrations. J Med Genet. 2011;48(9):629-634. 20. Galesloot TE, Geurts-Moespot AJ, den Heijer M, et al. Associations of common variants in HFE and TMPRSS6 with iron parameters are independent of serum hepcidin in a general population: a replication study. J Med Genet. 2013;50(9):593-598. 21. Pei SN, Ma MC, You HL, et al. TMPRSS6 rs855791 polymorphism influences the susceptibility to iron deficiency anemia in women at reproductive age. Int J Med Sci. 2014;11(6):614-619. 22. Mast AE, Langer JC, Guo Y, et al. Genetic and behavioral modification of hemoglobin and iron status among first-time and highintensity blood donors. Transfusion. 2020;60(4):747-758. 23. WHO. The global prevalence of anemia in 2011: Geneva: World Health Organization, 2015. 24. Reddy MB, Hurrell RF, Cook JD. Estimation of nonheme-iron bioavailability from meal composition. Am J Clin Nutr. 2000;71 (4):937-943. 25. Zimmermann MB, Harrington M, Villalpando S, Hurrell RF. Nonheme-iron absorption in first-degree relatives is highly correlated: a stable-isotope study in mother-child pairs. Am J Clin Nutr. 2010;91 (3):802-807. 26. Ndiaye NF, Idohou-Dossou N, Burkli S, et al. Polyphenol-rich tea decreases iron absorption from fortified wheat bread in Senegalese mother-child pairs and bioavailability of ferrous fumarate is sharply lower in children. Eur J Clin Nutr. 2020;74(8): 1221-1228. 27. Hotz K, Krayenbuehl PA, Walczyk T. Mobilization of storage iron is reflected in the iron isotopic composition of blood in humans. J Biol Inorg Chem. 2012;17(2):301309. 28. Walczyk T, Davidsson L, Zavaleta N, Hurrell RF. Stable isotope labels as a tool to determine the iron absorption by Peruvian school children from a breakfast meal. Fresen J Anal Chem. 1997;359(4-5):445449. 29. WHO. Serum ferritin concentrations for assessment of iron status and iron deficiency in populations. 2011. http://www.who.int/vmnis/indicators/ser

um_ferritin.pdf (accessed 17.04.2020). 30. Cook JD, Dassenko SA, Lynch SR. Assessment of the role of nonheme-iron availability in iron balance. Am J Clin Nutr. 1991;54(4):717-722. 31. Zimmermann MB, Troesch B, Biebinger R, Egli I, Zeder C, Hurrell RF. Plasma hepcidin is a modest predictor of dietary iron bioavailability in humans, whereas oral iron loading, measured by stable-isotope appearance curves, increases plasma hepcidin. Am J Clin Nutr. 2009;90(5):12801287. 32. Cook JD, Boy E, Flowers C, Daroca Mdel C. The influence of high-altitude living on body iron. Blood. 2005;106(4):1441-1446. 33. Lee PL, Halloran C, Trevino R, Felitti V, Beutler E. Human transferrin G277S mutation: a risk factor for iron deficiency anaemia. Br J Haematol. 2001;115(2):329333. 34. Sarria B, Navas-Carretero S, Lopez-Parra AM, et al. The G277S transferrin mutation does not affect iron absorption in iron deficient women. Eur J Nutr. 2007;46(1):57-60. 35. Wilkinson N, Pantopoulos K. The IRP/IRE system in vivo: insights from mouse models. Front Pharmacol. 2014;5:176. 36. Stoffel NU, Lazrak M, Bellitir S, et al. The opposing effects of acute inflammation and iron deficiency anemia on serum hepcidin and iron absorption in young women. Haematologica. 2019;104(6):1143-1149. 37. Moretti D, Goede JS, Zeder C, et al. Oral iron supplements increase hepcidin and decrease iron absorption from daily or twice-daily doses in iron-depleted young women. Blood. 2015;126(17):1981-1989. 38. Mah YH, Kao JH, Liu CJ, et al. Prevalence and clinical implications of HFE gene mutations (C282Y and H63D) in patients with chronic hepatitis B and C in Taiwan. Liver Int. 2005;25(2):214-219. 39. Bardou-Jacquet E, de Tayrac M, Mosser J, Deugnier Y. GNPAT variant associated with severe iron overload in HFE hemochromatosis. Hepatology. 2015; 62(6):1917-1918. 40. Hsiao SC, Lee CT, Pei SN. GNPAT variant is associated with iron phenotype in healthy Taiwanese women: a population without the HFE C282Y mutation. Hepatology. 2016;63(6):2057-2058.

2905

ARTICLE Ferrata Storti Foundation

Myelodysplastic Syndromes

Replication stress signaling is a therapeutic target in myelodysplastic syndromes with splicing factor mutations Johanna Flach,1 Johann-Christoph Jann,1 Antje Knaflic,1 Vladimir Riabov,1 Alexander Streuer,1 Eva Altrock,1 Qingyu Xu,1 Nanni Schmitt,1 Julia Obländer,1 Verena Nowak,1 Justine Danner,1 Arwin Mehralivand,1 Franziska Hofmann,1 Iris Palme,1 Ahmed Jawhar,2 Patrick Wuchter,3 Georgia Metzgeroth,1 Florian Nolte,1 Wolf-Karsten Hofmann1 and Daniel Nowak1

Haematologica 2021 Volume 106(11):2906-2917

1 Department of Hematology and Oncology, Medical Faculty Mannheim of Heidelberg University, Mannheim; 2Department of Orthopedic Surgery, Medical Faculty Mannheim of Heidelberg University, Mannheim and 3Institute of Transfusion Medicine and Immunology, Medical Faculty Mannheim of Heidelberg University, German Red Cross Blood Service Baden-Württemberg, Mannheim, Germany

ABSTRACT

Correspondence: DANIEL NOWAK daniel.nowak@medma.uni-heidelberg.de Received: March 31, 2020. Accepted: September 7, 2020. Pre-published: September 14, 2020. https://doi.org/10.3324/haematol.2020.254193

2906

omatic mutations in genes coding for splicing factors, e.g., SF3B1, U2AF1, SRSF2, and others are found in approximately 50% of patients with myelodysplastic syndromes (MDS). These mutations have been predicted to frequently occur early in the mutational hierarchy of the disease, therefore, making them particularly attractive potential therapeutic targets. Recent studies in cell lines engineered to carry splicing factor mutations have revealed a strong association with elevated levels of DNA:RNA intermediates (R-loops) and a dependency on proper ATR function. However, data confirming this hypothesis in a representative cohort of primary MDS patient samples have so far been missing. Using CD34+ cells isolated from MDS patients with and without splicing factor mutations as well as healthy controls we show that splicing factor mutation-associated R-loops lead to elevated levels of replication stress and ATR pathway activation. Moreover, splicing factor mutated CD34+ cells are more susceptible to pharmacological inhibition of ATR resulting in elevated levels of DNA damage, cell cycle blockade, and cell death. This can be enhanced by combination treatment with the low-dose splicing modulatory compound Pladienolide B. We further confirm the direct association between R-loops and ATR sensitivity and the presence of a splicing factor mutation using lentiviral overexpression of wild-type and mutant SRSF2 P95H in cord blood CD34+ cells. Collectively, our results from n=53 MDS patients identify replication stress and associated ATR signaling to be critical pathophysiological mechanisms in primary MDS CD34+ cells carrying splicing factor mutations, and provide a preclinical rationale for targeting ATR signaling in these patients.

Introduction With a frequency of approximately 50%, somatically acquired mutations in genes coding for splicing factors such as SF3B1, U2AF1, SRSF2, and others are the most recurrent mutations in patients with myelodysplastic syndromes (MDS).1-3 Apart from DNMT3A, TET2, and ASXL1 mutations, splicing factor (sf) mutations are also frequently determined to be founder mutations in MDS.4-7 Therefore, sf mutations are particularly attractive targets for therapeutic intervention in MDS. The precise mechanism of how sf mutations contribute to MDS pathophysiology is not fully understood. It has been shown that sf mutations perturb RNA splicing, e.g., by altering 3’-splice recognition sites, resulting in abnormal transcriptional programs in mutant cells.8-15 However, the overlap between the different sf mutations is limited16 raising the question how they can lead to a common myelodysplastic phenotype. Previous work has demonstrated that MDS-associated sf mutations induce increased formation of DNA:RNA intermediates (R-loops), suggesting ele-

haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

vated R-loops to be a novel common pathophysiological mechanism for sf mutated MDS.17,18 R-loops are transcription intermediates consisting of a DNA:RNA hybrids and a displaced single-stranded DNA (ssDNA).19 They are known to contribute to genome instability by exposing ssDNA and by blocking replication fork progression, causing replication stress and associated DNA damage.20 On the molecular level, perturbation in the dynamics of replication fork progression triggers phosphorylation of replication protein A (RPA) and activation of the ataxia telangiectasia and Rad3-related protein (ATR) pathway in order to repair replication stress-induced DNA damage.21 Chronically elevated R-loops as they occur in sf mutated cells make these cells dependent on proper ATR function in order to preserve the integrity of their genomes. This dependency represents a vulnerability towards pharmacological ATR inhibition that is specific to sf mutated MDS cells.22 A study on sf mutated cell lines has revealed that the R-loop-associated ATR activation seems to be independent of altered RNA processing.22 This opens the possibility for a potentially additive effect of the combination of ATR inhibition with splicing modulators, such as Pladienolide B (PladB), to selectively target sf mutated cells while sparing healthy non-mutated hematopoietic cells. PladB is a naturally occuring macrolide that targets the SF3b complex.23 PladB as well as its derivative E7107 have previously been shown to confer cytotoxic activity in several in vivo and in vitro models of sf mutated myeloid malignancies.24,2 A next-generation splicing modulator, H3B8800, a derivative of PladB, is currently under investigation in a phase I clinical trial in MDS and related myeloid malignancies (clinicaltrials gov. Identifier: NCT02841540).26 In this work we investigated the impact of R-loop-accumulation and inhibition on ATR signaling in primary CD34+ cells isolated from the bone marrow (BM) of MDS patients. We show that sf mutated cells display elevated R-loop-formation as compared to non-sf mutated MDS and healthy cells, which results in increased replication stress and associated activation of the ATR signaling pathway. Pharmacological inhibition of ATR preferentially kills sf mutated cells by increasing DNA damage. We further show that this effect can be increased by the splicing modulator PladB specifically in sf mutated cells of MDS patients. Our work provides preclinical evidence in primary patient cells for targeting ATR as a novel therapeutic strategy in MDS.

Methods

Tissue culture and cell-based assays Primary hematopoietic cells were cultured in StemSpan SFEM II supplemented with myeloid expansion supplement containing SCF, TPO, G-CSF, and GM-CSF (Stemcell Technologies). MOLM13 cells were obtained from the DSMZ (Germany) and cultured in RPMI-1614 supplemented with 20% heat-inactivated fetal calf serum.

Cell viability assay Cell viability in response to treatment was determined by CellTiter-Glo® assay (Promega) according to the manufacturer’s recommendations.

Cell apoptosis assay Apoptosis was determined by flow cytometry using Annexin V/ propidium iodide (PI) staining according to the manufacturer’s protocol (BD Bioscience) after treatment with AZD6738, PladB (alone and in combination), or vehicle for 24-48 hours.

Flow cytometry Differential BM fractions were isolated from MNC stained with fluorochrome-conjugated antibodies against CD33, CD19, CD3, CD34 and CD38 by flow cytometry-based cell sorting (FACS). Cell isolation was performed on a BD FACSMelody Flow Cytometer using double sorting to ensure maximum purity.

Immunofluorescence R-loops and RPA were quantified by immunofluorescence analysis using anti-S9.6 (Kerafast) and anti-RPA (pSer33; Novus Biologicals) antibodies. Quantification of mean fluorescence intensity (MFI) was performed using the following formula: mean fluorescence of selected cell – (area of selected cell x mean fluorescence of background readings). Values are displayed as arbitrary units (A.U.).

Single-molecule DNA replication analyses Replication track analyses were adapted from a published protocol.27 In brief, CD34+ cells in culture were pulsed with 25 mM 5-chloro-2′-deoxyuridine (CldU) followed by 250 mM 5-iodo-2′deoxyuridine (IdU), washed, and spotted onto SuperFrost® microscope slides for subsequent label detection.

Lentiviral transduction of CD34+ and MOLM13 cells Lentiviral constructs expressing SRSF2 wild-type (WT) and SRSF2 P95H were transiently transfected into 293T cells together with a third-generation lentiviral packaging mix (AMS.P904, Amsbio). Lentiviral supernatants were concentrated (Lenti-X concentrator, Clontech) and used to infect CD34+ or MOLM13 cells by spin-infection (2,000xg, 33°C, 90 min).

Patient and healthy donor material BM of MDS patients was obtained from residual diagnostic material. Healthy CD34+ cells were isolated from hip replacement surgery bone specimen (old CD34+ cells) or collected by iliac crest puncture of healthy volunteers (young CD34+ cells). For lentiviral transduction experiments, healthy CD34+ cells were isolated from cord blood. The study was approved by the Institutional Review Board of the Medical Faculty Mannheim, University of Heidelberg, Germany, and conducted in accordance with the declaration of Helsinki.

Sequencing Sf mutations were either assessed by diagnostic myeloid panel sequencing (n=9) or by germline controlled whole exome sequencing (WES) (n=44).

haematologica | 2021; 106(11)

RNA sequencing For RNA sequencing the Illumina TruSeq Stranded mRNA kit (San Diego, USA) with 500 ng RNA as input was used. Sequencing was performed with 100 bp paired end on an Illumina NovaSeq device.

Statistical analysis Data were analyzed using Prism 8 (GraphPad Software, La Jolla, CA) using t-test or one-way analysis of variance (ANOVA), or Mann-Whitney U test. P-value was considered significant at values less than 0.05 (n.s.: not significant, statistically significant *P<0.05, **P<0.01, ***P<0.001). For detailed protocols, please see the Online Supplementary Methods.

2907

J. Flach et al.

Figure 1. Splicing factor mutated myelodysplastic syndrome CD34+ cells exhibit elevated levels of R-loops. (A) Representative immunofluorescence images of R-loops using the antibody S9.6 (Kerafast, Boston) in CD34+ cells isolated from myelodysplastic syndrome (MDS) patient bone marrow with and without splicing factor (sf) mutations as well as young and old healthy controls. Scale bar, 5 mm. (B) Quantification of S9.6 mean fluorescence intensity (MFI); n=10 (658 cells, non-sf mutated), n=17 (809 cells, sf mutated), n=6 (363 cells, young CD34+), n=9 (301 cells, old CD34+). (C) Quantification of S9.6 MFI in fluorescence-activated cell sortingsorted hematopoietic subpopulations isolated from the bone marrow of three (n=1, SF3B1 K700 and n=2, SRSF2 P95) patients (CD34+: 103 cells; CD34+CD38–: 71 cells; CD33+: 81 cells; CD19+: 47 cells). (D) Quantification of S9.6 MFI in CD34+ cells carrying mutations in SRSF2, SF3B1, and ZRSR2; n=10 (658 cells, non-sf mutated), n=11 (484 cells, SRSF2 mutated), n=6 (325 cells, SF3B1 mutated); n=1 (88 cells, ZRSR2 mutated). Data are means +/-standard deviation. ***P≤0.001; n.s.: not significant; A.U: arbitrary units.

Results Splicing factor mutated CD34+ cells from myelodysplastic syndrome patients have elevated levels of R-loops Increased levels of R-loops have previously been associated with MDS-specific mutations in genes coding for splicing factors.17,22 Based on these cell-line data, we asked the question whether elevated R-loops would also represent a common vulnerability in sf mutated cells isolated from the BM of MDS patients. Patient characteristics are displayed in the Online Supplementary Table S1. 2908

Using an antibody against R-loops (S9.6, Kerafast), we evaluated levels of R-loops in CD34+ cells isolated from n=18 MDS patients carrying mutations in sf genes (SRSF2=10, SF3B1=7; ZRSR2=1) (sf mutated) and compared them to the levels of non-sf mutated CD34+ cells from MDS patients (n=10) (Figure 1A and B). We found a significant accumulation of R-loops in sf (SRSF2 and SF3B1) mutated cells (mean MFI [A.U.]: 83 vs. 163; P<0.001). In order to rule out that the accumulation of Rloops was solely an effect of aging we included CD34+ from healthy aged (>64 years of age; n=9) and young (<40 haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

Figure 2. Elevated levels of R-loops induce replication stress in splicing factor mutated myelodysplastic syndrome CD34+ cells. (A) Experimental procedure and representative images of CldU/IdU labeled DNA fibers. (B) Quantification of DNA fork velocity in CD34+ cells isolated from myelodysplastic syndrome (MDS) patient bone marrow with and without splicing factor (sf) mutations as well as healthy controls (young and old). n=441 (non-sf mutated), n=400 (sf mutated), n=279 (young CD34+), n=290 (old CD34+). (C) Schematic illustration of the DNA damage response pathway activated in response to replication stress and representative immunofluorescence images of pRPA (S33) in sf mutated and non-sf mutated MDS CD34+ cells. Scale bar, 5 mm. (D) Schematic illustration of the DNA damage response pathway activated in response to replication stress. (E) Representative immunofluorescence images of pRPA (S33) in healthy young, healthy old, MDS sf mutated, and MDS non-sf mutated CD34+ cells. Scale bar, 5 mm. (F) Quantification of pRPA (S33) mean fluorescence intensity (MFI). n=7 (373 cells, non-sf mutated), n= 13 (522 cells, sf mutated), n=4 (279 cells, young CD34+), n=6 (290 cells, old CD34+). Data are means +/- standard deviation. *P≤ 0.05; *** P≤0.001; n.s.: not significant; CldU: 5-chloro-2′-deoxyuridine; IdU: 5-iodo-2′-deoxyuridine (IdU); A.U: arbitrary units.

haematologica | 2021; 106(11)

2909

J. Flach et al.

Figure 3. Splicing factor mutated myelodysplastic syndrome CD34+ cells are hypersensitive to ATR inhibition. Dose-response experiments measuring cell viability with Celltiter-Glo® at 48 hours after continuous in vitro drug exposure with ATR inhibitors AZD6738 (n=8 [healthy old], n=6 [non-sf mutated], n=17 [sf mutated]) myelodysplastic syndrome (MDS) patients (A) and VE-821 (n=4 [non-sf mutated], n=3 [sf mutated MDS patients]) (B). (C) Sensitivity of MDS CD34+ cells carrying mutations in different sf genes towards AZD6738; n=6 (non-sf mutated patients), n=10 patients (SRSF2 mutated patients), n=5 (SF3B1 mutated patients) and n=2 (U1AF1 mutated patients). (D) Dose-response experiments measuring cell viability by CellTiter-Glo® 48 hours after continuous drug exposure with ATR inhibitor AZD6738 in an MDS patient cells carrying a ZRSR2 mutation; n=6 (non-sf mutated patients), n=15 (SRSF2, SF3B1, and U2AF1 mutated patients), n=1 (ZRSR2 mutated patient UPN18). IC50: 3.3 mM (non-sf mutated) versus 1.2 mM (sf mutated) versus 3.2 mM (ZRSR2 mutated). *P≤ 0.05; ***P≤0.001; n.s.: not significant; sf: splicing factor.

years of age, n=6) individuals as controls, which presented similar amounts of R-loops as non-mutated MDS cells (mean MFI [A.U.]: 91 [healthy young], 93 [healthy old]) (Figure 1A and B). It has previously been shown that the BM CD34+ fraction is enriched in MDS-initiating cells.4,28 In order to check whether certain cellular BM compartments were particularly prone to accumulate R-loops, we sorted several fractions, i.e., CD34+, CD34+CD38-, CD33+ (mature myeloid cells) and CD19– (B-lymphoid cells). As shown in Figure 1C, we detected similar levels of R-loops in CD34+ (mean MFI: 122), CD34+CD38– (mean MFI: 146), as well as in mature CD33+ (mean MFI: 126) myeloid BM cells. However, in accordance with the myeloid nature of MDS pathophysiology we did not find elevated levels of R-loops in CD19+ B cells (mean MFI: 68). We also confirmed the association of splicing pertur2910

bations induced by PladB with the accumulation of Rloops in non-sf mutated MOLM13 cell line cells (Online Supplementary Figure S1). Of note, the patients carrying a ZRSR2 mutation did not show increased levels of R-loops (Figure 1D).

Increased R-loops cause replication stress in splicing factor mutated myelodysplastic syndrome CD34+ cells In order to interrogate the hypothesis that elevated levels of R-loops cause replicative stress, we performed fiber assays. This assay evaluates alterations in DNA replication fork dynamics, a known complication of stalled or collapsed forks. As expected, the presence of sf mutations and associated elevated levels of R-loops reduced the rate of DNA replication from 0.404 (range, 0.385-0.423) kb/min to 0.282 (range, 0.272-0.293) kb/min (P<0.001), haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

Figure 4. Combinatorial effect of ATR inhibition and splicing modulation on splicing factor mutated and non-mutated myelodysplastic syndrome CD 34+ cells. (A) Dose-response experiment measuring cell viability at 48 hour after continuous drug exposure towards Pladienolide B; n=3 (healthy old), n=3 (nonsf mutated), n=3 (sf mutated) myelodysplastic syndrome (MDS) patients. (B) Cells were exposed to 500 nM AZD6738, 5 nM Pladienolide B, or a combination of both, and viability was determined by CellTiter-Glo® 48 hours after continuous drug exposure in vitro; n=5 (non-sf mutated patients), n=14 (sf mutated patients). (C) Combinatorial effect of ATR inhibition and splicing modulation on ZRSR2 mutated MDS CD34+ cells (UPN48) (n=3 technical replicates of one patient). Data are means +/- standard deviation. *P≤0.05; ** P≤0.01; *** P≤0.001; n.s.: not significant; sf: splicing factor; PladB: Pladienolide B; UPN: unique patient number.

strongly implying the presence of fork stalling and associated replication stress (Figure 2A and B). In comparison, healthy young CD34+ cells showed a median fork rate of 0.381 (range, 0.359-0.403) kb/min, and a median fork rate of 0.4206 (range, 0.388-0.453) was determined in healthy old (median age: 81 years, range, 76-88) CD34+ cells (Figure 2A to C).

Elevated R-loops cause ATR pathway activation As a consequence of replication stress and fork stalling, sf mutated myelodysplastic syndrome CD34+ cells showed increased activation of the ATR pathway as demonstrated by increased staining with pRPA (S33), a well-established downstream target of ATR kinase (Figure 2D to F) (mean MFI [A.U.]: 62 [non-sf mutated] vs. 110 [sf mutated]; P<0.001). This dependence on proper ATR function to repair replication stress-induced DNA damage represents a vulnerability presented primarily by sf mutated cells in a synthetic lethal fashion. ATR inhibitors, particularly AZD6738, are tested in early phase clinical trials, primarily for the treatment of solid cancers. For this reason, we next investigated whether inhibition of ATR in MDS patients could constitute a possible therapeutic target to selectively affect sf mutated CD34+ cells.

Splicing factor mutated myelodysplastic syndrome CD34+ cells are hypersensitive towards ATR inhibition In order to determine the sensitivity of MDS CD34+ haematologica | 2021; 106(11)

cells to ATR inhibitors we treated MDS CD34+ cells with two different ATR inhibitors (VE-821 and AZD6738, respectively) for 48 hours before cell viability assessment by CellTiter-Glo® assay. The presence of a sf mutation resulted in enhanced sensitivity associated with a reduced half maximal inhibitory concentration IC50 in MDS patient CD34+ cells (IC50=3.1 [healthy old] vs. 3.3 mM [MDS non-sf mutated] vs. 1.2 mM [MDS sf mutated]) (Figure 3A and B). We did not detect significant differences in IC50 values between the three most common sf hotspot mutations (IC50 SF3B1=1.2 [n=5] vs. SRSF2=1.2 [n=10], vs. U2AF1=0.8 mM [n=2]) (Figure 3C). Interestingly, the IC50 was unchanged in one MDS patient (UPN18) with a ZRSR2 mutation (3.3 [non-sf mutated] vs. 3.2 mM [ZRSR2]) (Figure 3D). This is consistent with the absence of increased levels of R-loops in this patient (Figure 1D).

Simultaneous ATR and splicing modulation show augmented efficacy in splicing factor mutant myelodysplastic syndrome CD34+ cells Splicing modulators, such as PladB or derivates such as H3B-8800 are under investigation for the treatment of sf mutated MDS.26,29 We confirmed that PladB as a single substance had increased efficacy in MDS cells as compared to age-matched healthy CD34+ cells and that there was a slight, but significant increase in sensitivity in sf mutated versus non-sf mutated MDS CD34+ cells (IC50= 17.5 [healthy old] vs. 6.9 [non-sf mutated] vs. 4.3 mM 2911

J. Flach et al.

Figure 5. Unresolved DNA damage leads to elevated cytotoxicity in splicing factor mutated myelodysplastic syndrome CD34+ cells treated with AZD6738 and Pladienolide B. (A to D) Representative immunofluorescence images of DNA repair kinetics following exposure to 500 nM AZD6738, alone or in combination with 5 nM Pladienolide B in cultured splicing factor (sf) not mutated (non-sf mutated) (A and B) or mutated (C and D) myelodysplastic syndrome (MDS) CD34+ cells (n=2 [non-sf mutated] and n=3 [sf mutated] independent experiments, respectively; 25-130 cells were analyzed per timepoint and experiment). For comparison reasons, the 48-hour timepoints are highlighted in red. Scale bar, 5 mm. 0, 1, 2, 3, 4, 5, >5 indicate the number of γH2AX/53BP1 foci. (E) Representative flow cytometry result to quantify apoptosis rates using AnnexinV/propidium iodide (PI). (F) Quantification of AnnexinV+ /PI+ cells in response to in vitro exposure to 500 nM AZD6738, 5 nM Pladienolide B (PladB), or a combination of both in sf mutated and non-sf mutated MDS CD34+ cells (n=2-7 independent experiments). Data are means +/- standard deviation. *P≤0.05; ***P≤0.001; n.s.= not significant; hr: hour.

2912

haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

Figure 6. Legend on the following page.

haematologica | 2021; 106(11)

2913

J. Flach et al.

Figure 6. Mechanistic validation that hypersensitivity towards ATR inhibition is driven by the presence of the splicing factor mutations and increased induction of aberrant splicing events. (A) Experimental procedure of lentivirus-mediated overexpression of SRSF2 wild-type (WT) and SRSF2 P95H in CD34+ isolated form cord blood. (B) Sequencing of the SRSF2 locus in transduced and sorted mCherry+ cells. (C) Representative immunofluorescence images and quantification of R-loops using antibody S9.6 (Kerafast) in mCherry+ cells (n=2 independent experiments; cells analyzed =150 [SRSF2 WT]/ 193 [SRSF2 P95]). (D) Cytotoxicity tests of AZD6738 alone and in combination with Pladienolide B (PladB) in transduced and sorted mCherry+ cells. Cell viability was determined by Celltiter-Glo® after 48 hours (n=3 independent experiments). (E) Experimental procedure of RNA sequencing experiment in order to determine aberrant splicing events in SRSF2 WT or P95H MOLM13 cells in the presence or absence of AZD6738. (F) Number of aberrantly spliced events decreased in SRSF2 WT MOLM13 cells treated with AZD6738. (G) Venn diagram demonstrating total numbers of splicing events in SRSF2 WT in comparison to SRSF2 P95H treated with dimethyl sulfoxide (DMSO), SRSF2 WT treated with AZD6738, and SRSF2 P95H treated with AZD6738. Data are means +/- standard deviation. *P≤0.05; ***P≤0.001; n.s.: not significant; A.U: arbitrary units.

[MDS sf mutated] [MDS sf mutated vs. non-sf mutated; P=0.0175], [healthy old vs. MDS; P<0.001]) (Figure 4A). Based on previous studies showing enhanced efficacy of splicing modulators in sf mutated cells of hematologic malignancies,26 we evaluated this concept, also in combination with ATR inhibition. We tested treatment with PladB in combination with AZD6738 in n=19 primary MDS samples (Figure 4B). Despite high inter-patient variability, we observed augmented activity of the combination of 500 nM AZD6738 and 5 nM PladB as compared to the single substances when analysing cytotoxicity in CellTiter-Glo® assays (Figure 4B). However, calculation of combination indexes by CompuSyn software (version 1.0; ComboSyn, Inc. Paramus, NJ, USA) did not result in significant additive or synergistic effects. Of note, combination of ATR inhibitor AZD6738 with an established standard therapy such as 5-Azacytitidine did not reveal additive or synergistic effects either. Again, a tendency for relatively lower cell viability was observed in the sf mutated group compared to the non-sf mutated group (P=0.20) (Online Supplementary Figure 2A to C). Interestingly, CD34+ cells of one available ZRSR2 mutated patient (UPN48) did neither respond to AZD6738 treatment alone nor to AZD6738 treatment in combination with PladB (Figure 4C).

ATR inhibition causes elevated levels of DNA damage associated with cell cycle arrest and cell death Our data demonstrated that increasing concentrations of ATR inhibitors led to a loss in cell viability likely due to the fact that replication stress-induced DNA damage is no longer efficiently repaired. In order to gain a deeper understanding of the underlying molecular mechanisms we treated non-sf mutated and sf mutated MDS CD34+ cells with 500 nM AZD6738 alone or in combination with 5 nM PladB and followed the DNA damage response dynamics over a period of 48 hours (h). As shown in Figure 5, treatment with AZD6738 led to the accumulation of γH2AX/53BP1 foci within 4-16 h, which were then largely repaired within 24-48 h post-exposure. However, in contrast to non-sf mutated CD34+ cells (Figure 5A), sf mutated MDS CD34+ not only showed delayed repair kinetics but they also retained levels of persistent DNA damage after 48 h (Figure 5B, 48-h bar graphs are highlighted in red). This was particularly evident in sf mutated cells when treated with both AZD6738 and PladB (Figure 5C and D). The inability to cope with increased levels of DNA damage upon both AZD6738 and PladB treatment was further reflected in increased apoptosis rates (Figure 5E and F) specifically in sf mutated cells. Results for vehicle and PladB only treated sf mutated and non-sf mutated cells are depicted in the Online Supplementary Figure S3A to D and show that even at baseline, sf mutated CD34+ cells have slightly elevated numbers of γH2AX/53BP1 foci as compared to non-sf mutated cells in line with previous publications.17,18 2914

Mechanistic validation that hypersensitivity towards ATR inhibition is driven by the presence of the splicing factor mutations and increased induction of aberrant splicing events The patient-derived CD34+ cells that we used in our study not only contained sf mutations but also other MDS typical recurrent mutations (Online Supplementary Table S1). In order to rule out confounding effects of other genetic alterations on ATR inhibitor sensitivity, we overexpressed WT and mutant SRSF2 in CD34+ cord blood cells, which we assumed not to carry any MDS-associated mutations (Figure 6A and B). As shown in Figure 6C we confirmed that increased levels of R-loops were directly associated with the presence of the sf mutation (mean MFI= 164.5. [SRSF2 WT] vs. 193.8 [SRSF2 P95]; P<0.001). In addition, SRSF2 P95H expressing CD34+ cells exhibited a reduced sensitivity to ATR inhibition alone and in combination with PladB (Figure 6D) similar to the levels seen in sf mutated MDS patient cells. This confirmed the direct association of the presence of sf mutations and elevated levels of R-loops and ATR hypersensitivity. As the presence of sf mutations has been shown to enhance alternative splicing,8-15 we evaluated whether ATR inhibition affects such alternative splicing events by performing RNA sequnecing. MOLM13 cells were transduced with and SRSF2 WT and P95H- expressing lentiviral particles and treated with AZD6738 or or dimethyl sulfoxide (DMSO) (Figure 6E; Online Supplementary Figure S4A). Of the n=4 experimental groups, SRSF2 WT (+ DMSO) was taken as reference and tested against SRSF2 WT + AZD6738, SRSF2 P95H + DMSO and SRSF2 P95H + AZD6738. The stability of the vector integration was confirmed within the RNA sequencing experiment (Online Supplemental Figure S4F). We found that n=1958 events in SRSF2 WT + AZD6738, n=2527 events in SRSF2 P95H, and n=3115 events in the SRSF2 P95H + AZD6738 group were differentially spliced (Figure 6F and G). As expected, the presence of SRSF2 P95H significantly increased differential splicing events. However, treatment of SRSF2 WT cells with ATR inhibitor AZD6738 also resulted in increased numbers of differential splicing events. SRSF2 P95H treated with AZD6738 showed the highest number of alternative splicing changes. Furthermore, we observed that the number of increased and decreased splicing events was approximately the same in all three experimental groups (Figure 6F) as were the patterns of splicing alterations (e.g., cassette exon, alternative 5’/3’ splice site) (Online Supplementary Figure S4G and I). Of note, only approximatly one third (45% and 33%, repectively) of events were shared between SRSF2 P95H and SRSF2 P95H + AZD6738 (Figure 6G), suggesting additional effects to aberrant splicing, beyond the presence of the sf mutation (Online Supplemental Figure S4I). Overall, the results suggest that ATR inhibition in combination with an sf mutation enhance differential splicing even further. haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

Discussion Somatic heterozygous mutations in SF genes such as U2AF1, SRSF2, and SF3B1 are the most frequently occurring mutations in MDS. Their high allele frequencies in clinical samples suggest that they often affect the founding or dominating clones of the disease and implicate RNA splicing alterations to be critical in disease initiation and progression. Due to their high frequency in MDS and dominating variant allele frequencies they are attractive targets for the development of new molecularly driven therapeutic strategies. Elevated levels of R-loops have emerged as a possible unifying pathomechanism common to U2AF1 S34 and SRFS2 P95 mutations in MDS.17 The association of sf mutations particularly in SRSF2 and U2AF1 with increased R-loop formation had previously been demonstrated in cell line models.17,22 Furthermore, a recent study has demonstrated increased R-loop accumulation in induced pluripotent stem cells and n=3 primary MDS patient samples carrying an SF3B1 mutation.18 Our study for the first time confirms this connection in a large representative clinical cohort of primary cells of MDS patients and therewith takes this finding into a pre-clinical setting. We screened a total of 28 MDS patient samples (ten non-sf mutated, 18 sf mutated [SRSF2=10, SF3B1=7; ZRSR2=1]) for their levels of Rloops in purified CD34+ BM cells, a cell population known to be enriched for MDS disease initiating cells. These results corroborate the hypothesis that accumulation of R-loops might indeed constitute a common pathophysiological mechanism in MDS patients with sf mutations. In line with previous studies on cell lines and as a consequence of elevated levels of R-loops we found highly activated replication stress signaling in sf mutated MDS patient CD34+ cells. This let us hypothesize that sf mutated BM cells from MDS patients may be dependent on proper ATR function in order to counteract replication stress-induced DNA damage. As a consequence, interfering with ATR-mediated DNA repair by ATR inhibitors VE-821 and AZD6738 causes mutated cells to no longer be able to tolerate R-loops driven replication stress causing them to cease their cell cycle and undergo apoptosis. This synthetic lethal connection between increased levels of R-loops and ATR inhibition makes therapeutic treatment possible without interfering too much with healthy hematopoiesis. While high doses of ATR inhibitors also impaired viability of non-sf mutated MDS CD34+ cells, careful titration allowed us to identify doses of ATR inhibitors that were still tolerated by CD34+ cells of nonsf mutated MDS patients and cells from healthy individuals. This suggests that at least in vitro there seems to be a therapeutic window for mutation-specific therapeutic targeting. Our in vitro experimental results support this hypothesis in an elaborate pre-clinical patient sample cohort of n=24, showing that hematopoietic cells with sf mutations become sensitive to even low doses of ATR inhibitor. It remains to be evaluated how this concept can be applied in the in vivo situation and ultimately in humans. We specifically chose AZD6738 as this substance has already moved into clinical trials. In addition, certain co-occuring mutations may affect response to ATR inhibition. Mutations interfering with the DNA damage response such as those affecting TP53, for instance, may predict resistance. Given the fact that ATR inhibition haematologica | 2021; 106(11)

affects the DNA damage response we investigated the influence of TP53 mutational status on the response towards AZD6738 (Online Supplemental Figure S5A). We did not detect statistically significant effects of the presence of mutated TP53 on the cells’ sensitivity towards AZD6738 in either the sf mutated or the non-sf mutated cohort. However, there was only one sample with a low level mutation of TP53 in the sf mutated group. Since the role of TP53 in the DNA damage response is primarily downstream of ATM signaling its exact impact on ATR inhibitor response needs to be carefully investigated in further studies. In order to assess any confounding effects of co-occurring mutations, our sample cohort was either analyzed by NGS panel sequencing (n=9) of recurrently mutated genes in myeloid diseases or by germline controlled WES (n=44). However, we did not discover molecular markers alleviating the sf mutation-mediated sensitivity against ATR inhibition. Mutations in ZRSR2 might be an exception among the sf mutations since the two patients carrying ZRSR2 mutations, which we were able to analyze, did neither display elevated levels of R-loops nor did they respond to ATR inhibition with the same sensitivity as the other sf mutations. The pathophysiological mechanism underlying ZRSR2 mutations in MDS may be different compared to common mutations in the other sf (SF3B1, U2AF1, and SRSF2) and possibly related to its aberrant retention of U12-type introns30 rather than the major U2-dependent spliceosome. Another confounding factor in our study could be due to an enrichment of higher risk IPSS-R categories in the non-sf mutated group mediating cell-intrinsic or, -extrinsic effects on ATR inhibitor responsiveness. The reason for this imbalance may be a natural bias in the distribution of sf mutations to be enriched in IPSS-R31 lower risk MDS categories. For SF3B1 this has been shown multiple times32,33 and is, last but not least, reflected in the WHO 2016 classification of myeloid neoplasms. Since our study includes a total of 12 patients with SF3B1 mutations, this per se accounts for an enrichment of lower risk categories in the sf mutated group. The absence of a sf mutation probably vice versa leads to an enrichment of higher risk categories in the non-sf mutated group.32 We did not have enough samples to artificially equalize this natural bias in differential mutational profiles among MDS risk groups. However, we have attempted to perform a subgroup analysis with matched IPSS-R risk status (n=4/4; Online Supplementary Figure S5B and C). In this analysis, the sf mediated hypersensitivity against ATR inhibition was retained. In a separate set of experiments we evaluated the effect of ATR inhibition on alternative splicing, which is known to be increased in sf mutated cells.8-15 Interestingly, treatment of non-sf mutated cells with AZD6738 resulted in elevated numbers of differentially spliced events to a similar extent as seen by ectopically expressed SRSF2 P95. The exact mechanism for this is unknown, however, splicing is inherently linked to transcription,34 a process known to be intertwined with DNA repair.35,36 As such, increased DNA damage induces alterations in RNA splicing of transcripts involved in genome stability maintenance.37,38 Furthermore, we found that the combination of ATR inhibitor treatment of SRSF2 mutated cells resulted in the highest number and unique differential splicing alterations. We hypothesize that the elevated levels of R-loops induced by the 2915

J. Flach et al.

mutation leads to further shifts in splicing factors and RNA binding proteins likely to prevent RNA-induced genome instability.39,40 Given the fact that splicing modulators have been shown to confer therapeutic activity on sf mutated cells,26 we hypothesized that combining low-dose AZD6738 with such a compound may further enhance the therapeutic window in sf mutated cells. We therefore performed experiments with PladB, which augmented the sensitivity of mutated cells towards ATR inhibitors. However, the combination failed to reach additive or synergistic effects. This could be a reflection of a possible lack of efficacy of splicesome modulators in MDS, which was also observed in the first clinical evaluation of H3B-8800 in myeloid malignancies.29 In conclusion, our study shows increased levels of R-loops in a large cohort of primary MDS CD34+ cells that carry sf mutations (SRSF2, SF3B1, and U2AF1), which results in high sensitivity towards pharmacologic ATR inhibition. Our data strengthen the preclinical rationale for this novel therapeutic strategy in MDS. A recently initiated single-arm phase Ib clinical trial is assessing safety and seeking preliminary evidence for efficacy for AZD6738 in patients with MDS and CMML, who have failed to respond to first-line treatments (clinicaltrials gov. Identifier: NCT03770429). R-loops could be potential novel biomarkers for response to this treatment.

References 1. Thol F, Kade S, Schlarmann C, et al. Frequency and prognostic impact of mutations in SRSF2, U2AF1, and ZRSR2 in patients with myelodysplastic syndromes. Blood. 2012;119(15):3578-3584. 2. Yoshida K, Sanada M, Shiraishi Y, et al. Frequent pathway mutations of splicing machinery in myelodysplasia. Nature. 2011;478(7367):64-69. 3. Haferlach T, Nagata Y, Grossmann V, et al. Landscape of genetic lesions in 944 patients with myelodysplastic syndromes. Leukemia. 2014;28(2):241-247. 4. Mian SA, Rouault-Pierre K, Smith AE, et al. SF3B1 mutant MDS-initiating cells may arise from the haematopoietic stem cell compartment. Nat Commun. 2015;6:10004. 5. Mossner M, Jann JC, Wittig J, et al. Mutational hierarchies in myelodysplastic syndromes dynamically adapt and evolve upon therapy response and failure. Blood. 2016;128(9):1246-1259. 6. Papaemmanuil E, Gerstung M, Malcovati L, et al. Clinical and biological implications of driver mutations in myelodysplastic syndromes. Blood. 2013;122(22):3616-3627; quiz 3699. 7. Pellagatti A, Roy S, Di Genua C, et al. Targeted resequencing analysis of 31 genes commonly mutated in myeloid disorders in serial samples from myelodysplastic syndrome patients showing disease progression. Leukemia. 2016;30(1):247-250. 8. Shiozawa Y, Malcovati L, Galli A, et al. Aberrant splicing and defective mRNA production induced by somatic spliceosome mutations in myelodysplasia. Nat Commun. 2018;9(1):3649. 9. Ilagan JO, Ramakrishnan A, Hayes B, et al. U2AF1 mutations alter splice site recognition in hematological malignancies. Genome Res. 2015;25(1):14-26. 10. Kim E, Ilagan JO, Liang Y, et al. SRSF2 mutations contribute to myelodysplasia by

2916

Future efforts should focus on combination strategies including ATR inhibition. Disclosures No conflicts of interest to disclose. Contributions JF, WKH and DN designed the study, analyzed data, and wrote the manuscript; JF, AK, and VR designed and performed all described cell biology experiments; JCJ and AS designed and performed bioinformatics analyses of NGS data and RNA-Seq experiments; VR, QX, AM, VN, JD, JO, FH and IP designed and performed wet lab and cell culture experiments; EA and NS helped with data analysis and interpretation; AJ and PW provided clinical samples; DN, JCJ, GM and FN performed clinical data analysis. Funding This work was supported by funds from the Deutsche Forschungsgemeinschaft (DFG, project FL1054/1-1 and NO817/5-2), funds from the “Deutsche Krebshilfe” (Project 70113953), funds from the Gutermuth Foundation, funds from the HW & J. Hector fund (Project M83), Baden Wuerttemberg, and the Dr. Rolf M. Schwiete Fund (Project 20/2016), Mannheim. DN is an endowed Professor of the German JoséCarreras-Stiftung (DJCLS H 03/01).

mutant-specific effects on exon recognition. Cancer Cell. 2015;27(5):617-630. 11. Komeno Y, Huang YJ, Qiu J, et al. SRSF2 is essential for hematopoiesis, and its myelodysplastic syndrome-related mutations dysregulate alternative pre-mRNA splicing. Mol Cell Biol. 2015;35(17):30713082. 12. Przychodzen B, Jerez A, Guinta K, et al. Patterns of missplicing due to somatic U2AF1 mutations in myeloid neoplasms. Blood. 2013;122(6):999-1006. 13. Zhang J, Lieu YK, Ali AM, et al. Diseaseassociated mutation in SRSF2 misregulates splicing by altering RNA-binding affinities. Proc Natl Acad Sci U S A. 2015; 112(34):E4726-4734. 14. Yip BH, Steeples V, Repapi E, et al. The U2AF1S34F mutation induces lineage-specific splicing alterations in myelodysplastic syndromes. J Clin Invest. 2017;127(9):3557. 15. Pellagatti A, Armstrong RN, Steeples V, et al. Impact of spliceosome mutations on RNA splicing in myelodysplasia: dysregulated genes/pathways and clinical associations. Blood. 2018;132(12):1225-1240. 16. Joshi P, Halene S, Abdel-Wahab O. How do messenger RNA splicing alterations drive myelodysplasia? Blood. 2017;129(18):24652470. 17. Chen L, Chen JY, Huang YJ, et al. The augmented R-loop is a unifying mechanism for myelodysplastic syndromes induced by high-risk splicing factor mutations. Mol Cell. 2018;69(3):412-425.e6. 18. Singh S, Ahmed D, Dolatshad H, et al. SF3B1 mutations induce R-loop accumulation and DNA damage in MDS and leukemia cells with therapeutic implications. Leukemia. 2020;34(9):2525-2530. 19. Santos-Pereira JM, Aguilera A. R loops: new modulators of genome dynamics and function. Nat Rev Genet. 2015;16(10):583-597. 20. Aguilera A, Garcia-Muse T. R loops: from transcription byproducts to threats to genome stability. Mol Cell. 2012;46(2):115-

124. 21. Zeman MK, Cimprich KA. Causes and consequences of replication stress. Nat Cell Biol. 2014;16(1):2-9. 22. Nguyen HD, Leong WY, Li W, et al. Spliceosome mutations induce R loop-associated sensitivity to ATR inhibition in myelodysplastic syndromes. Cancer Res. 2018;78(18):5363-5374. 23. Bonnal S, Vigevani L, Valcarcel J. The spliceosome as a target of novel antitumour drugs. Nat Rev Drug Discov. 2012;11(11): 847-859. 24. 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. 25. Obeng EA, Chappell RJ, Seiler M, et al. Physiologic expression of Sf3b1(K700E) causes impaired erythropoiesis, aberrant splicing, and sensitivity to therapeutic spliceosome modulation. Cancer Cell. 2016; 30(3):404-417. 26. Seiler M, Yoshimi A, Darman R, et al. H3B8800, an orally available small-molecule splicing modulator, induces lethality in spliceosome-mutant cancers. Nat Med. 2018;24(4):497-504. 27. Klusmann I, Rodewald S, Muller L, et al. p53 activity results in DNA replication fork processivity. Cell Rep. 2016;17(7):1845-1857. 28. Woll PS, Kjallquist U, Chowdhury O, et al. Myelodysplastic syndromes are propagated by rare and distinct human cancer stem cells in vivo. Cancer Cell. 2014;25(6):794-808. 29. Steensma DP, Wermke M, Klimek VM, et al. Results of a clinical trial of H3B-8800, a splicing modulator, in patients with myelodysplastic syndromes (MDS), acute myeloid leukemia (AML) or chronic myelomonocytic leukemia (CMML). Blood. 2019;134(Suppl 1):S673-673. 30. Madan V, Kanojia D, Li J, et al. Aberrant splicing of U12-type introns is the hallmark of ZRSR2 mutant myelodysplastic syn-

haematologica | 2021; 106(11)

MDS patients carrying splicing factor mutations exhibit increased levels of R-loops

drome. Nat Commun. 2015;6:6042. 31. Greenberg PL, Tuechler H, Schanz J, et al. Revised international prognostic scoring system for myelodysplastic syndromes. Blood. 2012;120(12):2454-2465. 32. Makishima H, Yoshizato T, Yoshida K, et al. Dynamics of clonal evolution in myelodysplastic syndromes. Nat Genet. 2017; 49(2):204-212. 33. Malcovati L, Karimi M, Papaemmanuil E, et al. SF3B1 mutation identifies a distinct subset of myelodysplastic syndrome with ring sideroblasts. Blood. 2015;126(2):233-241. 34. Kornblihtt AR, Schor IE, Allo M, Dujardin G,

haematologica | 2021; 106(11)

Petrillo E, Munoz MJ. Alternative splicing: a pivotal step between eukaryotic transcription and translation. Nat Rev Mol Cell Biol. 2013;14(3):153-165. 35. Fong YW, Cattoglio C, Tjian R. The intertwined roles of transcription and repair proteins. Mol Cell. 2013;52(3):291-302. 36. Svejstrup JQ. The interface between transcription and mechanisms maintaining genome integrity. Trends Biochem Sci. 2010;35(6):333-338. 37. Cloutier A, Shkreta L, Toutant J, Durand M, Thibault P, Chabot B. hnRNP A1/A2 and Sam68 collaborate with SRSF10 to control

the alternative splicing response to oxaliplatin-mediated DNA damage. Sci Rep. 2018;8(1):2206. 38. Shkreta L, Toutant J, Durand M, Manley JL, Chabot B. SRSF10 connects DNA damage to the alternative splicing of transcripts encoding apoptosis, cell-cycle control, and DNA repair factors. Cell Rep. 2016;17(8):19902003. 39. Cristini A, Groh M, Kristiansen MS, Gromak N. RNA/DNA hybrid interactome identifies DXH9 as a molecular player in transcriptional termination and R-loop-associated DNA damage. Cell Rep. 2018;23

2917

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2918-2926

Non-Hodgkin Lymphoma

CDKN2A deletion is a frequent event associated with poor outcome in patients with peripheral T-cell lymphoma not otherwise specified (PTCL-NOS) Francesco Maura,1-4 Anna Dodero,5 Cristiana Carniti,5 Niccolò Bolli,2,5 Martina Magni,5 Valentina Monti,6 Antonello Cabras,6 Daniel Leongamornlert,3 Federico Abascal,3 Benjamin Diamond,1 Bernardo Rodriguez-Martin,7 Jorge Zamora,7 Adam Butler,3 Inigo Martincorena,3 Jose M. C. Tubio,7 Peter J. Campbell,3 Annalisa Chiappella,8° Giancarlo Pruneri2,6 and Paolo Corradini2,5 Myeloma Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY, USA; 2Department of Oncology and Hemato-Oncology, University of Milan, Milan, Italy; 3The Cancer, Aging and Somatic Mutation Program, Wellcome Sanger Institute, Hinxton, Cambridgeshire, UK; 4Weill Cornell Medical College, New York, NY, USA; 5Department of Medical Oncology and Hematology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 6Department of Pathology and Laboratory Medicine, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy; 7CIMUS - Molecular Medicine and Chronic Diseases Research Center, University of Santiago de Compostela, Santiago de Compostela, Spain and 8Department of Hematology Azienda Ospedaliera Città della Salute e della Scienza, Turin, Italy. 1

°Current address: Department of Medical Oncology and Hematology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy.

ABSTRACT

N Correspondence: PAOLO CORRADINI paolo.corradini@unimi.it FRANCESCO MAURA francesco.maura85@gmail.com Received: June 12, 2020. Accepted: September 2, 2020. Pre-published: September 10, 2020. https://doi.org/10.3324/haematol.2020.262659

2918

odal peripheral T-cell lymphoma not otherwise specified (PTCLNOS) remains a diagnosis encompassing a heterogenous group of PTCL cases not fitting criteria for more homogeneous subtypes. They are characterized by a poor clinical outcome when treated with anthracycline-containing regimens. A better understanding of their biology could improve prognostic stratification and foster the development of novel therapeutic approaches. Recent targeted and whole exome sequencing studies have shown recurrent copy number abnormalities (CNA) with prognostic significance. Here, investigating five formalinfixed, paraffin embedded cases of PTCL-NOS by whole genome sequencing, we found a high prevalence of structural variants and complex events, such as chromothripsis likely responsible for the observed CNA. Among them, CDKN2A and PTEN deletions emerged as the most frequent aberration, as confirmed in a final cohort of 143 patients with nodal PTCL. The incidence of CDKN2A and PTEN deletions among PTCL-NOS was 46% and 26%, respectively. Furthermore, we found that co-occurrence of CDKN2A and PTEN deletions is an event associated with PTCLNOS with absolute specificity. In contrast, these deletions are rare and never co-occur in angioimmunoblastic and anaplastic lymphomas. CDKN2A deletion was associated with shorter overall survival in multivariate analysis corrected by age, International Prognostic Index, transplant eligibility and GATA3 expression (adjusted Hazard Ratio =2.53; 95% Confidence Interval: 1.006-6.3; P=0.048). These data suggest that CDKN2A deletions may be relevant for refining the prognosis of PTCLNOS and their significance should be evaluated in prospective trials.

Introduction Within the heterogeneous categories of nodal peripheral T-cell lymphoma (PTCL) subtypes currently recognized by the World Health Organization (WHO) classification system,1–4 many subtypes have defining molecular and phenotypic features. For example, within the anaplastic large cell lymphoma (ALCL) subtype, translocations

haematologica | 2021; 106(11)

CDKN2A and PTEN co-deletion in PTCL-NOS

of the ALK oncogene define a homogeneous subgroup, while ALK-negative cases carry mutations or translocations resulting in the activation of the JAK/STAT pathway.5 In angioimmunoblastic T-cell lymphoma (AITL), mutations in DNMT3A, TET2, IDH2 and RHOA have been described in a significant fraction of patients.6–8 These features are increasingly used for diagnostic and prognostic purposes, and may translate into targeted or rationale treatment approaches.9,10 A subgroup of PTCL not otherwise specified (PTCL-NOS) carries a T-cell follicular helper (TFH) phenotype and some phenotypic and genetic features of AITL including mutations affecting TET2, DNMT3A, and RHOA genes.6,11 Therefore, nodal PTCL with TFH phenotype have been included in the revised WHO classification as a provisional entity.4 On the contrary, the vast majority of PTCLNOS are a large yet ill-defined subgroup that lack defining genetic of phenotypic features.6,11–13 They are characterized by a poor clinical outcome when treated with anthracycline-containing regimens and some non-randomized studies suggest that high-dose chemotherapy can offer some survival benefit in young patients.14,15 However, a better understanding of their biology could improve prognostic stratification and foster the development of novel therapeutic approaches. The gene expression profile of PTCL-NOS suggests that two major subtypes can be identified, one characterized by GATA3 and the other by TBX21 overexpression, the former carrying worse prognosis.16–18 More recently, two groups identified recurrent copy-number alterations (CNA) in the tumor suppressors TP53 and CDKN2A with adverse prognostic significance, and generally associated with genomic instability.19,20 These efforts have been based on next-generation sequencing (NGS) of targeted gene panels and singlenucleotide polymorphism (SNP) arrays. Contrary to the above studies, whole genome sequencing (WGS) can interrogate the full repertoire of somatic mutations, CNA, structural variants (SV) and even mutational processes involved in cancer pathogenesis.21 However, WGS in PTCL have been hampered by the availability of high-quality, tumorrich DNA. On the other end, novel findings must be validated and applicable to everyday diagnostic routine before the benefits of additional information are translated into clinical improvements. With the hypothesis that WGS may reveal the genomic bases of recurrent genomic alterations in PTCL-NOS, here

we decided to test this approach in formalin-fixed, paraffin embedded (FFPE) samples. Furthermore, using an extended cohort of cases, we validate our and published results by fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC), and analyze prognostic correlates.

Methods Sample selection Eleven PTCL-NOS patients treated at the Department of Hematology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy were selected for the WGS analysis (Online Supplementary Table S1). The study protocol was approved by the Institutional Review Board (n. INT 19/13) and was conducted in accordance with the Declaration of Helsinki. The diagnosis was made according to the 2016 WHO classification4 and subtype assignment was independently reviewed by two expert hemato-pathologists. Only PTCL-NOS cases were selected. Nodal lymphomas with TFH cell origin, defined according to the 2017 WHO update4 by the association of a CD4+ phenotype with the expression of at least two TFH markers among ICOS, PD1, CXCL13, BCL-6 and CD10 by, were excluded on morphology and immunohistochemical bases. Additionally a comprehensive morphologic assessment was performed evaluating features commonly associated with nodal lymphomas with TFH cell origin such as the presence of large Reed-Sternberg-like, CD30+/CD15+ and PAX5+ cells, Eppstein Bar virus (EBV) positivity without a pabulum and with a growth pattern that mimics either a follicular lymphoma or progressively transformed germinal centers. Samples were chosen according to tumor cellularity and amount of extracted DNA (>500 ng). DNA was extracted from FFPE blocks in ten patients, as previously described.22 For one patient (PD30774a), DNA was extracted from a fresh frozen sample (peripheral blood mononuclear cells harvested during leukemic progression). For one patient, we sequenced two samples: one collected at diagnosis (PD30770a) and one at relapse (PD30770c). Therefore, all but two samples (PD30774a and PD30770c) were collected at diagnosis. DNA from buccal swabs was used as normal match for all samples.

Whole genome sequencing data analysis WGS libraries were prepared with the TruSeq DNA PCR-Free Library Preparation Kit (Illumina San Diego, CA, USA) from 500 ng of genomic DNA, aiming for an average target insert of 300

Table 1. Demographic and clinical characteristics of the peripheral T-cell lymphoma not otherwise specified cohort used for survival analysis.

Variable

All patients

CDKN2 deleted

CDKN2 wt

n/e

P-value

Age (range) Sex (female) Bone Marrow infiltration Extra-nodal disease IPI >2 SCT eligible Anthracycline induction CT Response to first line1 (CR) ASCT or AlloSCT GATA3 >50%

59.9 (22-85) 17/34 (50%) 9/30 (30%) 12/34 (35%) 14/32 (43%) 27/34 (79%) 31/34 (91%)

59.5 (22-84) 7/20 (35%) 4/18 (22%) 9/20 (45%) 7/18 (39%) 16/20 (80%) 18/20 (90%)

59.9 (34-85) 10/14 (71%) 5/12 (41%) 3/14 (21%) 8/14 (57%) 11/14 (78%) 13/14 (9%)

0 4 0 2 0 0

0.7 0.07 0.4 0.27 0.47 1 0.7

13/33 (39%) 14/34 (41%) 9/34 (26%)

5/20 (25%) 5/20 (25%) 6/20 (30%)

5/13 (38%) 9/14 (64%) 3/14 (21%)

1 0 0

0.06 0.035 0.7

SCT: stem cell transplantation; ASCT: autologous stem cell transplantation; AlloSCT: allogeneic stem cell transplantation; CR: complete remission; PT: partial remission; n/e: not evaluable; IPI: International Prognostic Index; wt: wild-type; CT: chemotherapy.

haematologica | 2021; 106(11)

2919

F. Maura et al.

basepairs (bp). Sequencing was performed on the Illumina X10 platform at the Wellcome Sanger Institute (WSI) on a 150 bppaired end protocol to a target depth of 40x for tumor samples and 30x for normal controls. The sequence data described here will be made available from the European Genome-phenome Archive (EGA) repository (EGAS00001002057). FASTQ files were aligned to the reference genome (GRCh37/hg19) using BWAmem and deduplicated aligned BAM files were analyzed using the following published tools available at the WSI: ASCAT and Battenberg to measure clonal and subclonal CNA and to estimate the tumor cell fraction of each sample,23 Caveman and Pindel for single nucleotide variants (SNV) and small insertion-deletions (indels);24,25 BRASS for SV (large inversions and deletions, translocations, internal tandem duplications).23 Complex events were defined, as previously described.26 TraFIC was used to described the somatic L1 retrotransposition landscape.27 The repertoire of mutational processes operative in PTCLNOS was analyzed by extracting the corresponding mutational signatures using the 96-class matrix of all possible substitutions in their 5’ and 3’ context with non-negative matrix factorization (NNMF), as previously described.28,29 Mutations were classified as drivers based on the COSMIC census catalogue of cancer genes (http://cancer.sanger.ac.uk/cosmic/).

Validation cohorts In order to expand our analysis on mutations in driver genes, we included published exomes data from 63 PTCL patients (30 PTCL-NOS ,15 AITL, 23 ALCL ALK neg and 16 EATL-II ).6,7,5,30,31,13 COSMIC census was used to create the catalogue of genes potentially involved by nonsynonymous mutations (https://cancer.sanger.ac.uk/census). CDKN2A and PTEN status was validated by dual-colour FISH on paraffin-embedded selected tumor areas using commercially available DNA probes: for p16 (CDKN2A) we used a spectrum orange-labeled locus-specific CDKN2A (9p21) probe and spectrum green-labeled chromosome 9 centromeric probe (LSI CDKN2A/CEP 9); for PTEN, a spectrum orange-labeled locusspecific PTEN (10q23) probe and a spectrum green-labeled chromosome 10 centromeric probe (LSI PTEN/CEP 10) – Vysis Inc., Downers Grove, IL, USA. A CDKN2A deletion was defined in the presence of either a homozygous deletion in >10% of cells or a hemizygous deletion in >40% of cells.27 A cutoff or 40% was applied to define the hemizygous PTEN loss. CDKN2A/p16 and GATA3 protein expression was tested by IHC, as previously described.16–18 In order to define GATA3 IHC positivity we used the recently proposed 50% cutoff.18 For further validation, we included published data from 20 AITL samples with copy number status defined using different NGS approach7,31 and 81 PTCL investigated by SNP array.32,33 For this last series, we downloaded all the available CEL files from GEO (accession numbers GSE15842 and GSE50252) and then LogR and BAF data were extracted using pennCNV.34 We then applied the ASCAT algorithm to perform segmentation and retrieve allele-specific absolute copy number alterations. The description of entire cohort used to validate CDKN2A and PTEN status can be found in the Online Supplementary Table S2. All contingency analyses were performed by Fisher’s exact test. Estimated progression-free survival (PFS) and overall survival (OS) were calculated with the Kaplan-Meier method and groups were compared with the log-rank test. Multivariate analysis was performed with Cox regression. All the analyses were performed using appropriate functions in the R 3.4.2 software (www.r-project.org).

2920

Results Peripheral T-cell lymphoma not otherwise specified show a complex genomic architecture We performed WGS on 12 tumors and 11 matched normal samples from 11 patients affected by PTCL-NOS, achieving an average depth of 27x (Online Supplementary Table S1). For one patient (PD30770) a diagnosis and a relapse tumor sample were available. As expected, analysis was severely hampered by the nature of samples analyzed, so that samples from six patients did not pass quality control. Causes ranged from insufficient cluster generation due to low quality DNA (one sample), low cancer cell fraction (CCF) (one sample), FFPE-induced artifacts (four samples). More data on quality control, FFPE artifacts and mutational signatures in PTCL-NOS WGS samples can be found in the Online Supplementary Appendix section. In the remaining five patients, no recurrent point mutations or indels in onco-driver genes were identified, in line with recent observations.6,11 The rare involvement of oncodriver gene in PTCL-NOS compared to other PTCL subtypes was confirmed in additional 84 published WES cases5– 7,30,31 (Online Supplementary Figure S1-2). We therefore focused the investigation on structural variants (SV: defined as inversions, translocations, internal tandem duplications [ITD] and deletions) and CNA, which could be investigated in depth in our five WGS samples. Three hundred and seventy-two SV were extracted with a median of 74 per sample (range, 56-86) (Figure 1A). Intriguingly, at least one complex event was observed in all but one patient, including five chromothripsis events in three patients (Figure 1B).35 Importantly, several cancer genes were affected either by SV directly or by CNA caused by SV and complex events (Figure 1C). For example, sample PD30771a was characterized by three different chromothripsis events on multiple chromosomes, including one causing a homozygous deletion of CDKN2A (Figure 1C). The genomic landscape was even more complex in relapse samples: PD30774a showed a whole genome duplication, and a large chromothripsis event was responsible for disruption of several known oncogenes resulting, for instance, in ARID1B and ARID2 losses (Figure 1C); analysis of serial diagnosis and relapse samples in patient PD30770 showed acquisition of numerous SV and CNA, and a complex event in chromosome 16 (Online Supplementary Figure S3). Overall, this confirms how driver events in PTCL-NOS may often go beyond coding gene mutations and involve complex structural events that can only be investigated by whole genome sequencing.

CDKN2A and PTEN are frequently co-deleted in peripheral T-cell lymphoma not otherwise specified, but not in other peripheral T-cell lymphoma Most of CNA and SV breakpoints were not recurrent, but we found the tumor suppressor CDKN2A deleted in four out of five patients. Collectively, we found a simple homozygous deletion (HD), and a second HD caused by chromothripsis. In two further cases, breakpoints around the copy-number loss could not be resolved by a distinct SV event, so the cause remains unknown (Figure 1C). Interestingly, PTEN loss was observed in two out of four CDKN2A-deleted patients. FISH analysis on sections from the five cases analyzed by WGS confirmed the CDKN2A and the PTEN deletion in all cases (Figure 2A), and the diploid status of the genes in the ones where no deletions

haematologica | 2021; 106(11)

CDKN2A and PTEN co-deletion in PTCL-NOS

were found by NGS. The high prevalence of these deletions in our small dataset confirmed recently published data and prompted us to extend our observations. To this end, we evaluated by FISH 37 PTCL archive cases from our institution (30 PTCL-NOS, five ALK negative ALCL and two AITL cases, Online Supplementary Table S2). The prevalence of

CDKN2A deletions among PTCL-NOS was 50%, and HD accounted for two thirds of cases. Overall, 31% of cases showed PTEN deletions. In order to compare the prevalence of CDKN2A and PTEN deletion across the main PTCL histologies, we included 101 cases from published datasets7,31–33 for a total of

Figure 1. Peripheral T-cell lymphoma not otherwise specified structural variants and complex events. Prevalence of structural variants (A) and complex events (B) among peripheral T-cell lymphoma not otherwise specified (PTCL-NOS). (C) Heatmap summarizing the oncogenes whose disruption was caused by at least one structural variants. The heatmap color highlighted the type of structural variant (SV) involved. Yellow: not determinable, i.e., the oncogene loss could not be mapped to a specific SV. For three events, the chromothripsis event responsible for the oncogene loss is drawn on the right. The horizontal black line indicates total copy number; the orange line minor copy number. Vertical lines represent SV breakpoints for deletion (red), inversion (blue), tandem-duplication (green) and translocation (black).

haematologica | 2021; 106(11)

2921

F. Maura et al.

143 PTCL of which 63 PTCL-NOS (Online Supplementary Table S2; Online Supplementary Figure S4). CDKN2A was deleted in 29 of 63 (46%) PTCL-NOS cases, including 19 of 29 (65%) HD. This was significantly more frequent than what was found in other histologies (Figure 2B and C). PTEN was deleted in 17 of 63 (26%) PTCL-NOS cases, 13 (76%) of which also carried a CDKN2A loss. Importantly, the co-occurrence of CDKN2A and PTEN was specific for PTCL-NOS (Figure 2B and C), representing the first genetic abnormality of such kind.

CDKN2A genetic status does not correlate with p16 nor with GATA3 expression We next evaluated CDKN2A expression (p16) by IHC in 31 PTCL-NOS samples for whose archival material was available. Overall, the average percentage of tumor cells expressing p16 in WT cases was 23.44% (range, 0-90%), while it was 6.21% (range, 0-30%) in deleted cases, either mono- or bi-allelic. This difference was significant (P=0.01, Student t-test) (Figure 3A and C). Using the previously described cutoff of 20% of tumor cells to define a categorical present/absent p16 expression status, only one case was p16 positive when CDKN2A was deleted, and no cases of

bi-allelic deletions showed p16 positivity. However, nine of 14 cases were p16 negative in spite of a wild-type (WT) CDKN2A locus (Figure 3A). This suggests that reduced expression of p16 can stem from non-genetic events, including epigenetic or post-transcriptional mechanisms as reported in other cancer types.36,37 Next, we correlated the genetic status of CDKN2A and GATA3 expression. Again, the picture was far from clear: the average percentage of GATA3-expressing cells was 23.12% in CDKN2A WT cases, and 33.72% in CDKN2A deleted cases. This difference was not significant (P=0.01, Student t-test) (Figure 3D). Even after setting the threshold for GATA3 expression at 50%,18 the number of GATA3 positive cases did not differ between CDKN2A WT and deleted cases. Finally, taking advantage of the combined WGS-SNP array cohort of 33 PTCL-NOS cases, we explored additional structural events in this subgroup. We observed 17p13 (TP53) deletions in 11 patients (33%), with partial overlap with CDKN2A deletions (Online Supplementary Figure S5).

CDKN2A deletions carry prognostic significance Clinical data were available for all PTCL-NOS patients

Figure 2. Prevalence of CDKN2A and PTEN deletions among different peripheral T-cell lymphoma subtypes. (A) Heatmap showing the prevalence of CDKN2A and PTEN deletions across the main peripheral T-cell lymphoma (PTCL) subtypes. (B) representative fluorescence in situ hybridization (FISH) pictures of PTEN (top) and CDKN2A (bottom) deletions. (C) Frequency table of CDKN2A deletions, mono- and bi-allelic, and co-occurrence with PTEN deletions across PTCL subtypes. All statistical comparisons were performed by Fisher’s exact test. Peripheral T-cell lymphoma not otherwise specified (PTCL-NOS) showed a significant enrichment for these three genomic aberrations compared to all the other PTCL histologies (P<0.01). ALCL: anaplastic large cell lymphoma; AITL: angioimmunoblastic T-cell lymphoma neg: negative; pos: positive.

2922

haematologica | 2021; 106(11)

CDKN2A and PTEN co-deletion in PTCL-NOS

from our WGS and FISH cohort except one that was not treated as a PTCL-NOS due to an initial diagnostic misclassification. The median age of the cohort was 59 years (range, 22-85), with 79% (27 of 34) transplant eligible and 43% (14 of 32) with an International Prognostic Index (IPI) >2. Most of the patients (n=31, 94.4%) received an induction therapy with an anthracycline. During the course of the disease, 14 patients (41%) received autologous stem cell transplantation (SCT) and/or allogeneic SCT (Table 1).38,14 After a median follow-up of 70 months, 11 patients (32%) were still alive. We thus evaluated whether CDKN2A and PTEN genomic status affected the response to therapy and outcome. PTEN deletion did not impact the clinical outcome in terms of both PFS and OS. In contrast, CDKN2A deleted cases were more frequent in primary refractory than WT cases (15 of 20 [75%] vs. five of 13 [38%]; P=0.06). Furthermore, 16 of 20 (80%) CDKN2Adeleted patients died in disease progression already within the first year of diagnosis. Consequently, PTCL-NOS patients carrying CDKN2A deletions had a significantly shorter PFS compared to WT patients (5-year PFS: 7.5% [95% Confidence Interval (CI): 1.3-42] vs. 24%, [95% CI: 963.5]; P=0.048) and OS (5-year OS: 22.5% [95% CI: 9.4-54] vs. 52% [95%CI: 30-90]; P=0.039) (Figure 4A and B). The

association between short survival and CDKN2A deletion retained its significance after multivariate correction by age, IPI, transplant eligibility and GATA3 IHC expression (Figure 4C and D).

Discussion In this paper we performed the first WGS experiment on PTCL-NOS samples. Despite the limited series, WGS showed ubiquitous instances of complex rearrangements and prevalent inactivation of classical tumor suppressor genes by SV and CNA. Chromothripsis in particular, a common mechanism underlying complex rearrangements, was never described in PTCL-NOS and frequently seen in our samples thanks to the comprehensive analysis of the genome allowed by WGS. This observation likely provides the mechanistic bases of the frequent CNA observed in PTCL-NOS cases in larger cohorts, including CDKN2A deletions.19,20 In our study, we could also observe how CDKN2A deletions were often focal and bi-allelic. This inactivation pattern is highly specific for driver tumor suppressors, and has been described for CDKN2A in other hematological malignancies.39–41 We went on to confirm this

Figure 3. CDKN2A protein expression in peripheral T-cell lymphoma not otherwise specified. (A) Expression of CDKN2A/p16 by immunohistochemistry (IHC). (B and C) Representative examples of CDKN2/p16 IHC results in cases with normal (B) and deleted (C) alleles. (D) GATA3 protein expression evaluated by IHC in CDKN2A deleted (del) and wild-type (wt) cases. P-value was estimated using a Wilcoxon test.

haematologica | 2021; 106(11)

2923

F. Maura et al.

Figure 4. CDKN2A deletion clinical impact. (A and B) Progression-free survival (PFS) and (C and D) overall survival (OS) according to CDKN2A status. (A and C) Kaplan-Meier plots of survival. P-values were estimated using a log-rank test. (B and D) multivariate analysis of survival, performed using Cox regression.

observation in a large validation cohort, showing how other PTCL subtypes did not share these abnormalities, pointing at a specific driver role of CDKN2A in PTCL-NOS pathogenesis. What was most striking though, was that PTEN and CDKN2A co-deletion had a 100% positive predictive value for the diagnosis of PTCL-NOS, providing the first genetic abnormality that provides absolute specificity for this PTCL subgroup. The integration of these genomic aberrations with point mutations recently reported in distinct subtypes might allow the generation of an accurate diagnostic flow based on NGS as described for other malignancies,42–44 potentially able to overcome historical classification difficulties in PTCL. Importantly, deletions of PTEN, a negative regulator of the PI3K pathway were also enriched in PTCL-NOS. From a mechanistic point of view, this suggests a possible epistatic relationship between the two pathways, and in fact the 2924

role of PTEN in T lymphomagenesis has been postulated in vivo.45 A distinction between PTCL-NOS cases has recently been proposed, as some cases would show a more complex genomic architecture with frequent CNA involving tumor suppressor genes and worse prognosis, while others would show fewer abnormalities and carry a better prognosis.19,20 A correlation between the former group and higher GATA3 expression has also been postulated by some groups,19 but not confirmed by others.20 High GATA3 expression is itself an adverse prognostic marker in PTCL-NOS.17 However, in our series, allelic status of CDKN2A did not correlate with GATA3 expression by IHC. Furthermore, we found that p16 expression by IHC was a poor predictor of CDKN2A genetic status, likely suggesting that, similarly to other cancers,37 non-genetic mechanisms converge towards p16 down-regulation as a haematologica | 2021; 106(11)

CDKN2A and PTEN co-deletion in PTCL-NOS

universal driver of PTCL-NOS. Given the high heterogeneity of the PTCL-NOS subgroup, and the variability of IHC staining of different proteins, further studies will be needed to validate these findings. However, our multivariate analysis of survival confirmed that CDKN2A allelic status was a significant predictor of inferior PFS and OS in PTCL-NOS. Importantly, this was independent of age, IPI, transplant eligibility and GATA3 expression by multivariate analysis. Future studies on larger number of patients will be required to confirm our findings. While the molecular categorization of PTCL-NOS still remains controversial, our study confirms the utility of WGS in the study of malignancies where recurrent gene mutations are lacking and the classification itself is often ambiguous. Moving from the complexities of WGS, we also show that an old-fashioned FISH test can be of relevance in PTCL-NOS. In the future, the diagnostic value of CDKN2A and PTEN co-deletion, and the prognostic value of CDKN2A deletions will need to be assessed in prospective studies.

References 1. de Leval L, Gaulard P. Cellular origin of T-cell lymphomas. Blood. 2014;123(19):29092910. 2. Inghirami G, Chan WC, Pileri S, the AIRC 5xMille consortium ‘Genetics-driven targeted management of lymphoid malignancies.’ Peripheral T-cell and NK cell lymphoproliferative disorders: cell of origin, clinical and pathological implications. Immunol Rev. 2015;263(1):124-159. 3. Maura F, Dodero A, Carniti C, Bolli N. Biology of peripheral T cell lymphomas – not otherwise specified: is something finally happening? Pathogenesis. 2016;3(1):9-18. 4. Swerdlow SH, Campo E, Harris NL, et al., editors. WHO classification of tumours of haematopoietic and lymphoid tissues. Revised 4th edition. Lyon: International Agency for Research on Cancer; 2017. 585 p. 5. Crescenzo R, Abate F, Lasorsa E, et al. Convergent mutations and kinase fusions lead to oncogenic STAT3 activation in anaplastic large cell lymphoma. Cancer Cell. 2015;27(4):516-532. 6. Palomero T, Couronné L, Khiabanian H, et al. Recurrent mutations in epigenetic regulators, RHOA and FYN kinase in peripheral T cell lymphomas. Nat Genet. 2014;46(2):166170. 7. Sakata-Yanagimoto M, Enami T, Yoshida K, et al. Somatic RHOA mutation in angioimmunoblastic T cell lymphoma. Nat Genet. 2014;46(2):171-175. 8. Yoo HY, Sung MK, Lee SH, et al. A recurrent inactivating mutation in RHOA GTPase in angioimmunoblastic T cell lymphoma. Nat Genet. 2014;46(4):371-375. 9. Allen PB, Pro B. Therapy of peripheral T cell lymphoma: focus on nodal subtypes. Curr Oncol Rep. 2020;22(5):44. 10. Lemonnier F, Dupuis J, Sujobert P, et al. Treatment with 5-azacytidine induces a sustained response in patients with angioimmunoblastic T-cell lymphoma. Blood. 2018;132(21):2305-2309. 11. Schatz JH, Horwitz SM, Teruya-Feldstein J, et al. Targeted mutational profiling of peripheral T-cell lymphoma not otherwise specified highlights new mechanisms in a heterogeneous pathogenesis. Leukemia. 2015;29(1):237-241. 12. Abate F, da Silva-Almeida AC, Zairis S, et al.

haematologica | 2021; 106(11)

Disclosures No conflicts of interest to disclose. Contributions PC, FM designed the study, collected and analyzed the data and wrote the paper; DL, AD, CC, NB analyzed the data and wrote the paper; BRM, JT, JZ, FA, AB, PJC, BD, and IM analyzed the data; AT, AC, MF and GP performed the IHC and FISH validations; GB, MM and AC collected the data. Funding This work is supported by AIRC (Associazione Italiana Ricerca sul Cancro) and AIL (Associazione Italiana Contro le LeucemieLinfomi e Mieloma ONLUS). This work is supported by the Memorial Sloan Kettering Cancer Center NCI core grant (P30 CA 008748). FM is supported by the American Society of Hematology, the International Myeloma Foundation and The Society of Memorial Sloan Kettering Cancer Center.

Activating mutations and translocations in the guanine exchange factor VAV1 in peripheral T-cell lymphomas. Proc Natl Acad Sci U S A. 2017;114(4):764-769. 13. Laginestra MA, Cascione L, Motta G, et al. Whole exome sequencing reveals mutations in FAT1 tumor suppressor gene clinically impacting on peripheral T-cell lymphoma not otherwise specified. Mod Pathol. 2020;33(2):179-187. 14. Corradini P, Vitolo U, Rambaldi A, et al. Intensified chemo-immunotherapy with or without stem cell transplantation in newly diagnosed patients with peripheral T-cell lymphoma. Leukemia. 2014;28(9):18851891. 15. d’Amore F, Leppä S, Silva MG da, et al. Final analysis of the front-line phase III randomized ACT-1 trial in younger patients with systemic peripheral T-cell lymphoma treated with CHOP chemotherapy with or without alemtuzumab and consolidated by autologous hematopoietic stem cell transplant. Blood. 2018;132(Suppl 1):S998-998. 16. Iqbal J, Wright G, Wang C, et al. Gene expression signatures delineate biological and prognostic subgroups in peripheral Tcell lymphoma. Blood. 2014;123(19):29152923. 17. Wang T, Feldman AL, Wada DA, et al. GATA-3 expression identifies a high-risk subset of PTCL, NOS with distinct molecular and clinical features. Blood. 2014;123 (19):3007-3015. 18. Amador C, Greiner TC, Heavican TB, et al. Reproducing the molecular subclassification of peripheral T-cell lymphoma-NOS by Immunohistochemistry. Blood. 2019;134(24):2159-2170. 19. Heavican TB, Bouska A, Yu J, et al. Genetic drivers of oncogenic pathways in molecular subgroups of peripheral T-cell lymphoma. Blood. 2019;133(15):1664-1676. 20. Watatani Y, Sato Y, Miyoshi H, et al. Molecular heterogeneity in peripheral T-cell lymphoma, not otherwise specified revealed by comprehensive genetic profiling. Leukemia. 2019;33(12):2867-2883. 21. The ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium. Pan-cancer analysis of whole genomes. Nature. 2020;578(7793):82-93. 22. Munchel S, Hoang Y, Zhao Y, et al. Targeted or whole genome sequencing of formalin

fixed tissue samples: potential applications in cancer genomics. Oncotarget. 2015;6(28): 25943-25961. 23. Maura F, Bolli N, Angelopoulos N, et al. Genomic landscape and chronological reconstruction of driver events in multiple myeloma. Nat Commun. 2019;10(1):3835. 24. Jones D, Raine KM, Davies H, et al. cgpCaVEManWrapper: simple execution of CaVEMan in order to detect somatic single nucleotide variants in NGS data. Curr Protoc Bioinformatics 2016;56:15.10.1-15.10.18. 25. Raine KM, Hinton J, Butler AP, et al. cgpPindel: identifying somatically acquired insertion and deletion events from paired end sequencing. Curr Protoc Bioinformatics. 2015;52:15.7.1-15.7.12. 26. PCAWG Structural Variation Working Group, PCAWG Consortium, Li Y, et al. Patterns of somatic structural variation in human cancer genomes. Nature. 2020;578 (7793):112-121. 27. Alvarez EG, Baez-Ortega A, Zamora J, et al. Pan-cancer analysis of whole genomes identifies driver rearrangements promoted by LINE-1 retrotransposition. Nat Genet. 2020;409(3):860-814. 28. Maura F, Degasperi A, Nadeu F, et al. A practical guide for mutational signature analysis in hematological malignancies. Nat Commun. 2019;10(1):2969. 29. PCAWG Mutational Signatures Working Group, PCAWG Consortium, Alexandrov LB, et al. The repertoire of mutational signatures in human cancer. Nature. 2020;578(7793):94-101. 30. Roberti A, Dobay MP, Bisig B, et al. Type II enteropathy-associated T-cell lymphoma features a unique genomic profile with highly recurrent SETD2 alterations. Nat Commun. 2016;7(1):12602. 31. Wang M, Zhang S, Chuang S-S, et al. Angioimmunoblastic T cell lymphoma: novel molecular insights by mutation profiling. Oncotarget. 2017;8(11):17763-17770. 32. Boi M, Rinaldi A, Kwee I, et al. PRDM1/BLIMP1 is commonly inactivated in anaplastic large T-cell lymphoma. Blood. 2013;122(15):2683-2693. 33. Hartmann S, Gesk S, Scholtysik R, et al. High resolution SNP array genomic profiling of peripheral T cell lymphomas, not otherwise specified, identifies a subgroup with chromosomal aberrations affecting the REL

2925

F. Maura et al. locus. Br J Haematol. 2010;148(3):402-412. 34. Wang K, Li M, Hadley D, et al. PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data. Genome Res. 2007;17(11): 1665-1674. 35. Korbel JO, Campbell PJ. Criteria for inference of chromothripsis in cancer genomes. Cell. 2013;152(6):1226-1236. 36. Alhejaily A, Day AG, Feilotter HE, Baetz T, LeBrun DP. Inactivation of the CDKN2A tumor-suppressor gene by deletion or methylation is common at diagnosis in follicular lymphoma and associated with poor clinical outcome. Clin Cancer Res. 2014;20(6):1676-1686. 37. Pasqualucci L, Khiabanian H, Fangazio M, et al. Genetics of follicular lymphoma transformation. Cell Rep. 2014;6(1):130-140.

2926

38. Dodero A, Spina F, Narni F, et al. Allogeneic transplantation following a reduced-intensity conditioning regimen in relapsed/refractory peripheral T-cell lymphomas: long-term remissions and response to donor lymphocyte infusions support the role of a graft-versus-lymphoma effect. Leukemia. 2012;26(3): 520-526. 39. Kataoka K, Nagata Y, Kitanaka A, et al. Integrated molecular analysis of adult T cell leukemia/lymphoma. Nat Genet. 2015;47 (11):1304-1315. 40. Reddy A, Zhang J, Davis NS, et al. Genetic and functional drivers of diffuse large B cell lymphoma. Cell. 2017;171(2):481-494. 41. Karube K, Enjuanes A, Dlouhy I, et al. Integrating genomic alterations in diffuse large B-cell lymphoma identifies new relevant pathways and potential therapeutic tar-

gets. Leukemia. 2018;32(3):675-684. 42. Bolli N, Li Y, Sathiaseelan V, et al. A DNA target-enrichment approach to detect mutations, copy number changes and immunoglobulin translocations in multiple myeloma. Blood Cancer J. 2016;6(9):e467. 43. Bolli N, Manes N, McKerrell T, et al. Characterization of gene mutations and copy number changes in acute myeloid leukemia using a rapid target enrichment protocol. Haematologica. 2015;100(2):214-222. 44. McKerrell T, Moreno T, Ponstingl H, et al. Development and validation of a comprehensive genomic diagnostic tool for myeloid malignancies. Blood. 2016;128(1):e1-9. 45. Liu X, Karnell JL, Yin B, et al. Distinct roles for PTEN in prevention of T cell lymphoma and autoimmunity in mice. J Clin Invest. 2010;120(7):2497-2507.

haematologica | 2021; 106(11)

ARTICLE

Non-Hodgkin Lymphoma

Transducin b-like protein 1 controls multiple oncogenic networks in diffuse large B-cell lymphoma Youssef Youssef,1 Vrajesh Karkhanis,1 Wing Keung Chan,1 Frankie Jeney,1 Alessandro Canella,1 Xiaoli Zhang,2 Shelby Sloan,1 Alexander Prouty,1 JoBeth Helmig-Mason,1 Liudmyla Tsyba,1 Walter Hanel,1 Xuguang Zheng,1 Pu Zhang,3 Ji-Hyun Chung,1 David M. Lucas,1 Zachary Kauffman,1 Karilyn Larkin,1 Anne M. Strohecker,4,5 Hatice G. Ozer,6 Rosa Lapalombella,1 Hui Zhou,7 Zijun Y. Xu-Monette,8 Ken H. Young,8 Ruolan Han,9 Elmar Nurmemmedov,10 Gerard Nuovo,11 Kami Maddocks,1 John C. Byrd,1 Robert A. Baiocchi,1 and Lapo Alinari1

Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2927-2939

Department of Internal Medicine, Division of Hematology, The Ohio State University, Columbus, OH, USA; 2Center for Biostatistics, Department of Biomedical Informatics, The Ohio State University, Columbus, OH, USA; 3Division of Pharmaceutics and Pharmaceutical Chemistry, College of Pharmacy, The Ohio State University, Columbus, OH, USA; 4Department of Cancer Biology and Genetics, The Ohio State University Columbus, OH, USA; 5Department of Surgery, Division of Surgical Oncology, The Ohio State University Columbus, OH, USA; 6Department of Biomedical Informatics, The Ohio State University, Columbus, OH, USA; 7Department of Lymphoma & Hematology, The Affiliated Tumor Hospital of Xiangya Medical School, Central South University, Changsha, Hunan, China. 8Department of Pathology, Division of Hematopathology, Duke University, Durham, NC, USA; 9Iterion Therapeutics, Huston, TX, USA; 10Department of Translational Neurosciences and Neurotherapeutics, John Wayne Cancer Institute, Providence Saint John's Health Center, Santa Monica, CA, USA and 11Discovery Life Sciences, Powell, OH, USA 1

ABSTRACT

iffuse large B-cell lymphoma (DLBCL) is the most common nonHodgkin lymphoma and is characterized by a remarkable heterogeneity with diverse variants that can be identified histologically and molecularly. Large-scale gene expression profiling studies have identified the germinal center B-cell (GCB-) and activated B-cell (ABC-) subtypes. Standard chemo-immunotherapy remains standard front-line therapy, curing approximately two thirds of patients. Patients with refractory disease or those who relapse after salvage treatment have an overall poor prognosis highlighting the need for novel therapeutic strategies. Transducin b-like protein 1 (TBL1) is an exchange adaptor protein encoded by the TBL1X gene and known to function as a master regulator of the Wnt signaling pathway by binding to b-CATENIN and promoting its downstream transcriptional program. Here, we show that, unlike normal B cells, DLBCL cells express abundant levels of TBL1 and its overexpression correlates with poor clinical outcome regardless of DLBCL molecular subtype. Genetic deletion of TBL1 and pharmacological approach using tegavivint, a first-in-class small molecule targeting TBL1 (Iterion Therapeutics), promotes DLBCL cell death in vitro and in vivo. Through an integrated genomic, biochemical, and pharmacologic analyses, we characterized a novel, b-CATENIN independent, post-transcriptional oncogenic function of TBL1 in DLBCL where TBL1 modulates the stability of key oncogenic proteins such as PLK1, MYC, and the autophagy regulatory protein BECLIN-1 through its interaction with a SKP1-CUL1-F-box (SCF) protein supercomplex. Collectively, our data provide the rationale for targeting TBL1 as a novel therapeutic strategy in DLBCL.

haematologica | 2021; 106(11)

Correspondence: LAPO ALINARI lapo.alinari@osumc.edu Received: July 27, 2020. Accepted: September 10, 2020. Pre-published: September 14, 2020. https://doi.org/10.3324/haematol.2020.268235

2927

Y. Youssef et al.

Introduction Diffuse large B-cell lymphoma (DLBCL) comprises the majority of adult non-Hodgkin lymphoma cases worldwide and is traditionally categorized, based on the molecular profile, into two major subtypes, germinal center B-cell (GCB) and activated B-cell (ABC) lymphoma, mirroring the cell of origin as well as reflecting the differences in clinical outcomes.1-3 DLBCL is curable in 40-60% of the cases following standard front-line immuno-chemotherapy.4 However, patients with refractory disease, those who relapse after salvage chemotherapy and autologous stem cell transplant or chimeric antigen receptor T-cell therapy have an overall poor prognosis highlighting the need for novel therapeutic approaches.4-6 Transducin b-like protein 1 (TBL1), encoded by the TBL1X gene, is an adaptor protein initially identified as a core component of the co-repressor silencing mediator for retinoid and thyroid hormone receptor (SMRT)– nuclear receptor co-repressor (N-COR) complex.7-9 SMRT/N-COR form a complex with TBL1, BCL6, and other proteins leading to transcriptional repression of target genes through HDAC3-mediated H3K9 deacetylation.7,10 Subsequently, TBL1 was found to be a key player in enhancing the canonical Wnt signaling pathway by directly binding to b-CATENIN and recruiting it to the promoter of Wnt target genes (MYC, BIRC5, CCND1) to promote uncontrolled cell proliferation and survival.9,11,12 Previous work also showed that TBL1 binds to a S-phase kinase-associated protein 1 (SKP1)/Cullin-1(CUL1)/F-box protein complex (SCF complex) and protects bCATENIN from proteasomal degradation.13,14 Tegavivint (Iterion Therapeutics) is a first-in-class small molecule compound that binds directly to the N-terminal domain of TBL1 and disrupts TBL1/b-CATENIN interaction leading to b-CATENIN degradation and subsequent downstream inhibition of the Wnt transcriptional program.15 Tegavivint has shown significant activity in preclinical models of acute myeloid leukemia (AML) and multiple myeloma (MM).15-17 Here, we report that, unlike normal B cells, DLBCL cells express abundant levels of TBL1 regardless of the molecular subtype and its overexpression correlates with poor clinical outcome. We demonstrate that both TBL1 genetic deletion and pharmacologic targeting with tegavivint induce DLBCL cell death in vitro and in vivo. More importantly, we show that b-CATENIN is dispensable for DLBCL cell survival and plays a minimal role in the tegavivint-mediated attenuation of known Wnt targets. Through an integrated genomic, biochemical, and pharmacologic approach we detail the crucial role played by TBL1 in controlling the stability of critical regulatory proteins involved in cell proliferation and autophagy induction. Our findings provide the rational for targeting TBL1 in DLBCL.

Methods Cell lines and primary samples DLBCL cell lines were purchased from ATCC and identity confirmed via short tandem repeats profiling (University of Arizona).18 Peripheral blood and tonsils from healthy donors and peripheral blood, bone marrow samples, and lymph nodes from

2928

DLBCL patients were obtained following written informed consent under a protocol approved by the Institutional Review Board of OSU in accordance with the Declaration of Helsinki.

Tissue microarray and immunohistochemistry Details on tissue microarray (TMA) preparation and immunohistochemistry (IHC) are presented in the Online Supplementary Appendix.

Gene knock-down using short hairpin RNA and CRISPRCas9 system TBL1X, CTNNB1, and CAND1 were genetically silenced using pLKO.1 short hairpin RNA (shRNA) vectors (Mission® shRNA, Sigma). Details are presented in the Online Supplementary Appendix.

Immunoblot and co-immunoprecipitation assay Details are presented in the Online Supplementary Appendix (see Online Supplementary Table S4).

RNA extraction and real-time polymerase chain recation Details are presented in the Online Supplementary Appendix (see Online Supplementary Table S5).

Chromatin-immunoprecipitation (ChIP)-sequencing and ChIP-polymerase chain reaction For chromatin-immunoprecipitation-sequencing (ChIP-Seq) assays, 5x106 Pfeiffer and Riva cells treated with either vehicle control or tegavivint for 12 hours were processed according to Active Motif protocols.19 All raw sequence data are available in the Gene Expression Omnibus (GSE148232). ChIP-polymerase chain reaction (ChIP-PCR) was performed according to standard procedures. Details are presented in the Online Supplementary Appendix.

Confocal microscopy, proximity ligation assay, cytotoxicity assay and transmission electron microscopy Details are presented in the Online Supplementary Appendix.

Animal studies Animal studies were approved by the OSU Institutional Animal Care and Use Committee. Experimental details for the animal studies with tegavivint are included in the result section. For the in vivo TBL1X knock-down experiment, luciferase and Cas9 positive Riva cells were transduced with either doxycycline-inducible single guide RNA (sgRNA) targeting TBL1X or vector control only using a lentiviral luciferase vector previously described.20 20×106 of engineered Riva cells were engrafted into NSG mice via tail vein injection. On day 23, the tumor burden was visualized after intraperitoneal administration of 200 mL of D-luciferin potassium salt (Caliper Life Sciences) in phospate buffered saline using in vivo live-imaging system (IVIS) (Perkin Elmer Inc.). Additional details are presented in the Online Supplementary Appendix.

Statistical analysis All the experiments include at least three independent replicates and results were expressed as the mean ± standard error of the mean (SEM). For correlated data such as patient cells treated with different conditions, paired t-tests for two group comparisons or linear mixed effects models for two or more group comparisons were used for analysis. Otherwise, for independent data, two-sample t-tests and analysis of variance (ANOVA) were used for comparisons between two groups and among more than two groups respectively. Survival analysis was performed using log-

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

rank test or Cox proportional hazard regression models. P-values <0.05 were considered significant after Holm’s procedure adjustment for multiple comparisons.

Results TBL1 is overexpressed and correlates with clinical outcome in diffuse large B-cell lymphoma regardless of its molecular subtype Immunoblot analysis showed abundant levels of TBL1 protein expression in seven DLBCL cell lines regardless of the molecular subtype (four GCB and three ABC, Figure 1A) and in primary DLBCL patient samples (n=4) (Figure 1B) (patient characteristics are listed in the Online Supplementary Table S1) compared to normal peripheral blood B cells and CD10+CD19+ sorted germinal center B cells from reactive tonsils (sorting strategy in the Online Supplementary Figure S1). In order to validate our findings, we performed IHC staining of TBL1 in tissue samples obtained from 83 primary de novo DLBCL patients uniformly treated with front-line R-CHOP (46 GCB- and 37 ABC-subtype, Figure 1C; Online Supplementary Figures S2 to S4; patient characteristics are listed in the Online Supplementary Tables S2 and S3). In our cohort, the percentage of DLBCL cells with strong (brown) TBL1 staining (similar to the intensity seen in the positive controls) ranged from 5% to 90% (GCB-DLBCL: mean percentage of positive cells 46%; range, 5-85%; ABC-DLBCL: mean percentage 49%; range, 5-90%) compared to 5% to 15% with a mean of 8% in tonsils with reactive lymphoid hyperplasia (n=5; P=0.0001). Normal GCB were identified by co-staining with CD10 (brown) and TBL1 (red) and coexpression showed in yellow (Figure 1C) or with TBL1 single staining (Online Supplementary Figure S2).21 No significant difference in the TBL1 expression patterns was seen between GCB- and ABC-DLBCL (dot plot in Figure 1C). Using the median percentage of TBL1 positive lymphoma cells (50% for GCB-DLBCL and 55% for ABC-DLBCL), patients were classified into high or low expressors. Importantly, a high percentage of TBL1 positive DLBCL cells was associated with significantly inferior progression free survival (PFS) and overall survival (OS) in both ABCand GCB-DLBCL (Figure 1D). However, the negative prognostic impact of TBL1 positivity on PFS/OS did not persist after adjusting for International Prognostic Index (IPI) and IPI factors in a Cox multivariate model.

TBL1 is critical for diffuse large B-cell lymphoma cell survival In vitro TBL1 knock-down using either a doxycyclineinducible CRISPR-Cas9 system or a TBL1X-specific shRNA was associated with significant DLBCL cell death regardless of the molecular subtype (Figure 2A) (P<0.05). Next, luciferase positive Riva cells were stably transduced with Cas9 and a doxycycline-inducible sgRNA targeting TBL1X exon 2 (guide 1). Prior to engraftment, efficient TBL1 knock-down after doxycycline induction was verified in vitro (Online Supplementary Figure S5). NSG mice (n=8 per group) received doxycycline (100 mg/kg) daily via oral gavage starting at day 24 after successful engraftment was verified by IVIS on day 23. TBL1 knock-down induced tumor regression after 5 days of doxycycline treatment as documented by in vivo imaging system (IVIS) and significantly prolonged the survival of these animals compared to conhaematologica | 2021; 106(11)

trol [median OS 48 days vs. 38 days, respectively, (P=0.0004)] (Figure 2B). Next, we tested the in vitro cytotoxicity of tegavivint. As shown in Figure 3A and B, 24 hours treatment with tegavivint (2-100 nM) induced significant DLBCL cell death in eight DLBCL cell lines (four GCB and four ABC) and primary DLBCL samples (n=8; patient characteristics are listed in the Online Supplementary Table S1) compared to the vehicle control (P<0.001). Importantly, cytotoxic studies on normal resting and activated immune cell subsets from the peripheral blood of healthy donors (n=3) incubated for 24 hours with the highest concentration of tegavivint (100 nM) showed no significant toxicity on B and natural killer (NK) cells while marginal toxicity (<10% decrease in viability) was noted in activated (P=0.0287) but not resting T cells, consistent with their low TBL1 expression (Online Supplementary Figure S6A and B). In support of a selective TBL1-mediated DLBCL cell death, tegavivint lost the majority of its cytotoxic activity in DLBCL cells when TBL1 was efficiently knocked-down via shRNA (Online Supplementary Figure S7). In order to evaluate the activity of tegavivint in vivo, we employed three murine models representative of disseminated ABC- and GCB-DLBCL. i) Seven-week-old NSG mice were engrafted with 20×106 Riva cells (ABC-DLBCL) via tail vein injection. Mice in groups of five were then randomized to receive either vehicle control or tegavivint at 25 mg/kg every 3 days via tail vein injection starting at day 7 post engraftment. Tegavivint significantly prolonged survival in the treated group (median OS of tegavivint vs. vehicle control mice 33 vs. 27 days, P=0.0031) (Figure 3C). Notably, 4 of 5 tegavivint-treated mice died of central nervous system (CNS) involvement with minimal systemic disease evaluated by the spleen size (Figure 3C) consistent with the known limited CNS penetration property of tegavivint. ii) Seven-week-old NSG mice were engrafted with 40×106 SU-DHL10 cells (GCB-DLBCL) via tail vein injection. Mice in groups of ten were then randomized to receive either vehicle control or intravenous tegavivint at 25 mg/kg starting at day 3 post engraftment on a twice weekly schedule (Monday-Thursday). Tegavivint significantly prolonged survival in the treated group (median OS of tegavivint vs. vehicle control mice 47 vs. 34.5 days, P=0.0002) (Figure 3D). ii) Seven-week-old NSG mice were tail vein engrafted with 1×106 of passage 3 DFBL-18689 cells from an ABC-DLBCL patient derived xenograft (PDX) mouse model obtained from the public repository of xenografts (Proxe).22 Mice in groups of five were then randomized to receive either vehicle control or tegavivint at 25 mg/kg via tail vein injection on a twice weekly schedule (Monday-Thursday) starting at day 3 post engraftment. Tegavivint significantly prolonged the survival of the treated animals (median OS of tegavivint vs. vehicle control mice 55 vs. 37 days, P=0.0026) (Figure 3E; Online Supplementary Figure S8). Collectively, these data demonstrate that TBL1 is a novel therapeutic target in DLBCL and that tegavivint induces the desired pharmacologic effect by exhibiting selective TBL1-mediated lymphoma cell death with very limited toxicity to normal immune cells.

TBL1 modulates Wnt targets in a novel, post-transcriptional, b-CATENIN independent manner Co-immunoprecipitation (Co-IP) experiments in Riva and Pfeiffer cells showed that treatment with tegavivint 2929

Y. Youssef et al.

Figure 1. Legend on following page.

2930

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

Figure 1. TBL1 is abundantly expressed and correlates with clinical outcome in diffuse large B-cell lymphoma. (A) Immunoblot showing abundant TBL1 expression in whole cell lysates from the indicated germinal center B-cell (GCB)- and activated B-cell (ABC)-diffuse large B-cell lymphoma (DLBCL) cell lines compared to normal resting and activated peripheral blood (PB) and germinal center (GC) B cells (n=3). (B) Immunoblot showing abundant TBL1 expression in whole cell lysates from four primary DLBCL patient (Pt) samples compared to normal B-cells and peripheral blood mononuclear cells (PBMC). ACTIN was used as loading control; mantle cell lymphoma cell line (Jeko) was used as positive control (+Ctl) for (A)-(B). (C) Immunohistochemical (IHC) staining of lymph nodes involved by ABC- and GCB-DLBCL (at 40x and 100x) showing strong (dark) nuclear and cytoplasmic positivity for TBL1. As comparison, tonsillar germinal center were co-stained with CD10 (brown) and TBL1 (red). TBL1-CD10 co-expression at 40x and 100x is shown in yellow (black arrows) using the Nuance Multispectral Imaging System (model 3.0.2) and colocalization tool which allows to select any color for co-localization evaluation. Positive controls (tonsillar epithelium [left] and breast ductal carcinoma [right]) are both shown at 40x and 100x. Dot plot showing percentage of TBL1 positive lymphoma cells by IHC staining in DLBCL patient samples (n=83 total with 46 GCBDLBCL and 37 (ABC-DLBCL) to reactive tonsillar germinal center controls (n=5). Data represent median percentage of TBL1 positive cells ± standard deviation. ***P<0.0005 by ANOVA. (D) Percentage of TBL1 positive lymphoma cells (≥55% for ABC-DLBCL and ≥50% for GCB-DLBCL) inversely correlated with progression free survival (PFS) and overall survival (OS) in de novo GCB- and ABC-DLBCL patients treated with front-line R-CHOP chemotherapy. P-values were determined using the log-rank tests.

for 12 hours led to the disruption of TBL1/b-CATENIN interaction (Figure 4A) which was further confirmed via PLA (Figure 4A; Online Supplementary Figure S9A and B). ChIP-seq analysis in the same cell lines revealed that genome-wide TBL1 promoter occupancy was not significantly affected by tegavivint treatment for 12 hours (Online Supplementry Figure S10A) while ChIP-PCR validation of selected Wnt/b-CATENIN target genes showed selective loss of b-CATENIN recruitment to the promoters of MYC and BIRC5 (Figure 4B) upon treatment (P<0.0005), TBL1 recruitment was unaffected. Treatment of four DLBCL cell lines with tegavivint for 24 hours

resulted in decreased protein levels of b-CATENIN, MYC and SURVIVIN with no effect on TBL1 expression when compared to the untreated control (Figure 4C). Similarly, shRNA-mediated TBL1 knock-down led to significant decrease in MYC, and SURVIVIN expression compared to controls (Figure 4D). Despite the loss of b-CATENIN recruitment to the promoter regions of MYC and BIRC5 following treatment with tegavivint, quantitative realtime PCR (qRT-PCR) showed significant increase in MYC transcript in Riva (P<0.005) and Pfeiffer (P<0.05) and no changes in OCI-Ly3 and WSU-NHL (Figure 4E). A similar pattern was observed with the transcript of two other

Figure 2. TBL1 genetic knock-down results in significant diffuse large B-cell lymphoma cell death in vitro and in vivo. (A) TBL1 was knocked-down (KD) using either a doxycycline-inducible CRISPR-Cas9 system (two different guides: one and two targeting exon 2 and 3, respectively) or a TBL1-specific short hairpin RNA (shRNA) construct (n=3). 72 hours after transduction, efficient TBL1 knock-down was verified via immunoblot and viability determined by annexin/propidium (Ann-/PI-) staining and flow cytometry. Data represent means ± standard error of the mean. ns: P>0.05, *P<0.05, **P<0.005, by paired t-test. Empty vector and scramble were used as controls. +Dox/-Dox: with (red) or without (black) doxycycline 0.75 mg/mL. Numbers below the immunoblots reflect normalized value of quantified protein bands relative to the controls. (B) Kaplan–Meier curve showing overall survival (OS) of NSG mice engrafted with Cas9+ Riva cells transduced with either a doxycycline-inducible TBL1-targeting single guide RNA (sgRNA) (n=8) or empty vector control (n=8). Daily doxycycline induction was started on day 24, in vivo imaging system (IVIS) images were obtained on day 23 and day 29 post engraftment. Median OS was 38 days for the empty vector controls (Ctl) versus 48 days for the TBL1 KD group (P=0.0004 using the log-rank test).

haematologica | 2021; 106(11)

2931

Y. Youssef et al.

well-known b-CATENIN target genes, WISP1 and AXIN2 (Online Supplementary Figure S10B) while treatment with tegavivint resulted in a minimal but significant decrease in BIRC5 transcript in three out of four DLBCL cell lines

tested (Figure 4E). Lastly, efficient shRNA knock-down of b-CATENIN did not significantly affect MYC or SURVIVIN protein expression and had minimal effect on cell viability in all the tested DLBCL cell lines (Figure 4F). In

Figure 3. Tegavivint shows significant activity in in vitro and in vivo models of diffuse large B-cell lymphoma. (A-B) Cytotoxicity assay in the indicated diffuse large B-cell lymphoma (DLBCL) cell lines (n=3) and DLBCL patient samples (n=8: peripheral blood [n=4], lymph node [n=3] and bone marrow [n=1]) treated with the indicated concentration of tegavivint (T) in nM for 24 hours. Viability was determined by annexinV/propidium (Ann-/PI-) staining and flow cytometry. Data represent means ± standard error of the mean. *P<0.05, **P<0.005, ***P<0.0005 by linear mixed effects models with adjustment for multiple dose comparisons in (A) and with two sample t-test in (B). (C) Kaplan–Meier curve showing overall survival (OS) of NSG mice engrafted with Riva (activated B-cell [ABC]-DLBCL) and randomized to receive either vehicle control (n=5) or tegavivint (n=5) given at 25 mg/kg via tail vein injection every 3 days starting at day 7 post engraftment. Median OS was 27 days for the controls (Ctl) versus 33 days for the treated group (P=0.0031 using the log-rank test). Representative picture of the spleen size (left image) and hematoxylin and eosin stain showing brain involvement by DLBCL cells (right image). (D) Kaplan–Meier curve showing OS of NSG mice engrafted with SU-DHL10 (germinal center B-cell [GCB]-DLBCL) and randomized to receive either vehicle control (n=10) or tegavivint (n=10) given at 25 mg/kg via tail vein injection twice weekly starting at day 3 post engraftment. Median OS was 34.5 days for the controls (Ctl) versus 47 days for the treated group (P=0.0002 using the log-rank test). (E) Using an adaptive transfer model of DFBL-18689, recipient NSG mice (n=5/group) were randomized to receive either vehicle control or tegavivint at 25 mg/kg via tail vein injection on a twice weekly schedule (Monday-Thursday) starting at day 3 post engraftment. Kaplan–Meier curve showing OS (P=0.0026 using the logrank test). Median OS: 55 days (tegavivint) and 37 days (control [Ctl]). Two mice in the treated group were censored due to engraftment failure assessed by flow cytometry for circulating lymphoma cells and magnetic resonance imaging for spleen volume.

2932

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

Figure 4. TBL1 modulates MYC levels in a post-transcriptional/b-CATENIN independent manner. (A) Diffuse large B-cell lymphoma (DLBCL) cell lines were treated with either dimethylsulphoxide (DMSO) control (NT) or tegavivint (T) for 12 hours. Lysates were immunoprecipitated with anti-TBL1 or anti-IgG and then immunoblotted with the indicated antibodies. Confocal images of proximity ligation assay (PLA) (60x) using anti-TBL1 and anti-b-CATENIN antibodies after treatment of DLBCL cells with either DMSO control (NT) or tegavivint (T) for 12 hours. Red indicates PLA signal, blue indicates cell nuclei (Dapi) (n=3). (B) Chromatin immunoprecipitation (ChIP) assay on crosslinked chromatin from DLBCL cells after treatment with either DMSO control (NT) or tegavivint (T) for 12 hours. Either preimmune (PI) or the indicated immune antibodies were used and the retained DNA was amplified using MYC or BIRC5 promoter-specific primers. Fold enrichment with each antibody was calculated relative to the PI. n=3; data represent means ± standard error of the mean (SEM). ns: P>0.05, *P<0.05, ***P<0.0005 by linear mixed effects models. (C) Immunoblot showing TBL1, b-CATENIN, MYC and SURVIVIN expression after treatment of DLBCL cells with either DMSO control (NT) or tegavivint (T) for 24 hours (n= 3). (D) Immunoblot showing MYC and SURVIVIN protein levels after TBL1 knock-down (KD) using two TBL1 specific shRNA constructs (n=3). Lysates are the same used in Figure 2A however probed for different proteins. (E) Quantitative real-time polymerase chain reaction (qRT-PCR) showing the mRNA fold changes of the indicated oncogenes after treating DLBCL cells with tegavivint (T) for 12 hours relative to the untreated control (NT). n=3, data represent means ± SEM. ns: P>0.05, *P<0.05, **P<0.005 by paired t-test. (F) Immunoblot showing MYC and SURVIVIN expressions after efficient b-CATENIN KD using two b-CATENIN specific short hairpin RNA (shRNA) constructs. Viability obtained by trypan blue exclusion 72 hours post-transduction. n=3, data represent means ± SEM. ns: P>0.05 by paired t-test (ns: not significant). Tegavivint (T): Riva, Pfeiffer: 70 nM; OCI-ly3: 50 nM; WSU-NHL: 15 nM (for all experiments).

haematologica | 2021; 106(11)

2933

Y. Youssef et al.

F G

Figure 5. Legend on following page.

2934

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

Figure 5. TBL1 is a critical component of a SCF complex in diffuse large B-cell lymphoma through which it modulates MYC stability. (A) The indicated diffuse large B-cell lymphoma (DLBCL) cell lines were treated with either dimethylsulfoxide (DMSO) control (NT) or tegavivint (T) for 12 hours, immunoprecipitated with either an anti-TBL1 or anti-SKP1, and then immunoblotted with the indicated antibodies. b-CATENIN and histone H3 were used as positive and negative controls, respectively. (B) Confocal images (60x) of proximity ligation assay (PLA) showing the CAND1-CUL1 interaction in the indicated DLBCL cells lines after treatment with either DMSO control (NT) or tegavivint (T) for 12 hours. Red indicates PLA signal, blue indicates cell nuclei (Dapi). Graph represents the mean value of PLA signal from no less than 300 cells (n=3). **P<0.005 by paired t-test. (C) Proposed model of the mechanism of tegavivint-induced activation of the SCF complex in DLBCL. (D) Immunoblot showing PLK1 and MYC expression levels after treatment with either DMSO control (NT) or tegavivint (T) at the indicated time points. (E) Immunoblot showing PLK1 level after treatment with either the translation inhibitor cycloheximide (CHX), tegavivint (T) or the combination (CHX+T) for the indicated time points. (F) Immunoblot showing PLK1 and MYC expression after treatment with either DMSO control (NT), tegavivint (T), the proteasome inhibitor MG132 or the combination (T+MG132) for 24 hours. (G) Cytotoxicity assay in DLBCL cell lines treated with either DMSO control (NT) or tegavivint (T), MG132 or the combination (T+MG132). Viability was determined by annexinV/propidium (Ann-/PI-) staining and flow cytometry at 24 hours. n= 3, data represent means ± SEM. *P<0.05, **P<0.005 by linear mixed effects models including interaction test between T and MG132. CHX: 70 mg/mL added for 1 hour and then washed before adding tegavivint treatment. MG132 (Riva, OCI-Ly3 and WSU-NHL: 0.5 uM and Pfeiffer: 0.3 uM) was added 6 hours after tegavivint treatment was started.

further support of a TBL1-mediated and b-CATENIN independent process, tegavivint maintained both its cytotoxic activity and its effect on MYC and SURVIVIN (protein and transcript) in the context of efficient b-CATENIN knock-down (Online Supplementary Figure S11). Collectively our data provide evidence that bCATENIN contributes minimally to the transcriptional regulation of critical Wnt targets and to DLBCL cell survival and that TBL1 regulates the stability of MYC in a bCATENIN independent fashion.

TBL1 is a critical component of the SCF complex in diffuse large B-cell lymphoma The SCF complex controls the ubiquitination and degradation of a large number of proteins through the specificity-conferring F-box proteins such as bTRCP and FBW7.23,24 Its activity is regulated by the interaction of CUL1 with two proteins: i) the cullin associated and neddylation dissociated 1 (CAND1) and ii) NEDD8. CAND1 functions as an inhibitor of the SCF complex while NEDD8 binding to CUL1 promotes the targeted substrate ubiquitination and degradation.23 We hypothesized that TBL1 modulates target protein stability through the SCF complex and that targeting TBL1 with tegavivint enhances SCF complex activity leading to increased downstream degradation of key regulatory proteins. CoIP and PLA in Riva and Pfeiffer cells showed TBL1-SKP1CUL1 association which was unaffected by tegavivint (Figure 5A; Online Supplementary Figure S12). Importantly, tegavivint treatment of Riva and Pfeiffer induced significant decrease of CAND1-CUL1 association by PLA (P<0.005) (Figure 5B; Online Supplementary Figure S13) with no effect on the expression of these two proteins (Online Supplementary Figure S14A). These results suggest a model in which TBL1, as a critical component of an SCF complex, modulates the stability of critical regulatory proteins (Figure 5C).

TBL1 modulates MYC stability through the SCF complex Polo-like kinase-1 (PLK1), an essential regulator of mitosis and a novel target in lymphoma, is degraded by SCFbTrCP.25 Importantly, PLK1 plays a critical protective role in MYC stability by preventing its SCFFbw7-mediated proteasomal degradation.26-28 We hypothesized that targeting TBL1 with tegavivint may enhance the proteasomalmediated degradation of MYC directly by SCFFbw7 and indirectly by enhancing SCFbTrCP–mediated PLK1 degradation. Time course analysis showed that treatment of four DLBCL cell lines with tegavivint led to PLK1 downregulation while leaving its transcript unaffected (Figure 5D; Online Supplementary Figure S14B and C), decreased MYC levels with concomitant increase in its phosphorylation haematologica | 2021; 106(11)

on Ser62 and Thr58 residues at 10 hours, consistent with enhanced MYC proteasomal degradation (Online Supplementary Figure S14D). Adding the translation inhibitor cycloheximide (CHX) to tegavivint significantly shortened the half-life of PLK1 protein compared to CHX alone (Figure 5E; Online Supplementary Figure S14E). Consistent with this, the proteasomal inhibitor MG132 partially restored PLK1 and MYC protein levels as well as cell viability when combined with tegavivint versus tegavivint alone (P<0.05) (Figure 5F and G). In further support of an SCF/TBL1-modulated process, efficient shRNA knock-down of CAND1 (and TBL1) in Riva and Pfeiffer cells led to significant decrease in viability as well as downregulation of PLK1 and MYC protein expression (Online Supplementary Figure 14F and G). Altogether, our data support a novel model in which TBL1 regulates MYC stability directly through SCFFbw7 and indirectly by enhancing SCFbTrCP-mediated PLK1 degradation.

TBL1 is involved in the modulation of cytoprotective autophagy Autophagy is a catabolic pathway that sustains metabolism by recycling cytoplasmic contents to support macromolecular synthesis during periods of cellular stress with accumulating evidence supporting its pro-survival role in lymphoma.29,30 As SCF complexes regulate the stability of key autophagy inducers such as BECLIN-1 and MYC,31-33 we hypothesized that tegavivint impairs the autophagy flux by affecting these key autophagy inducers. Immune blot analysis showed downregulation of BECLIN-1 protein level in four DLBCL cell lines at 12 and 24 hours (Figure 6A) whereas BECN1 transcript level was not affected by tegavivint (Online Supplementary Figure S15). Co-treatment with CHX and tegavivint significantly shortened the half-life of BECLIN-1 compared to either CHX or tegavivint alone (Figure 6B) and incubation with MG132 partially restored BECLIN-1 protein level when combined with tegavivint compared to tegavivint alone (Figure 6C). In order to evaluate the effect of tegavivint on the autophagy flux, four DLBCL cell lines were treated with either dimethylsulfoxide (DMSO) control, chloroquine (autophagy blocker), rapamycin (autophagy inducer), tegavivint or combinations. The increase in the amount of LC3-II observed in chloroquine-treated cells represents the amount of LC3 delivered to the lysosome for degradation and is a well-established measure of autophagic flux.34 As shown in Figure 6D, treatment with tegavivint led to an increase in the levels of LC3-II which could be due to autophagy induction or inhibition. Co-treatment with chloroquine led to further accumulation of LC3-II, suggesting autophagy induction in two of the four cell 2935

Y. Youssef et al.

Figure 6. Legend on following page.

2936

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

Figure 6. TBL1 modulates cytoprotective autophagy through the SCF complex. (A) Immunoblot showing BECLIN-1 expression in the indicated diffuse large B-cell lymphoma (DLBCL) cell lines after treatment with either dimethylsulfoxide (DMSO) control (NT) or tegavivint (T) for 12 hours and 24 hours (n=3). (B) Immunoblot showing BECLIN-1 expression in the indicated DLBCL cell lines after treatment with either cycloheximide (CHX), tegavivint (T) or the combination (CHX+T) (n=3). (C) Immunoblot showing BECLIN-1 expression in the indicated DLBCL cell lines after treatment with either DMSO control (NT), tegavivint (T), MG132 or the combination (T+MG132) at 24 hours (n=3). (D) Immunoblot showing BECLIN-1 and microtubule-associated protein light-chain 3 (LC3-I and LC3-II) expression in the indicated DLBCL cell lines after treatment with either DMSO control (NT), tegavivint (T), chloroquine (CQ), rapamycin (Rapa) or combinations at 12 hours and 24 hours (n=3). (E) Top: schematic depicting tandem LC3 reporter. Bottom: confocal microscopy images (120x) of the indicated DLBCL cell lines transduced with a lentivirus expressing a tandem LC3 plasmid (mCherry-eGFP-LC3) and subsequently treated with either DMSO control (NT), tegavivint (T), chloroquine (CQ) or rapamycin (Rapa) for 12 hours. (F) Histograms represent quantification of yellow (GFP/mCherry) puncta/total red (mCherry) puncta (%) of no less than 100 cells per condition. n=4, data represent means ± standard error of the mean (SEM). *P<0.05, **P<0.005 by a linear mixed effects model with Holm’s adjustment for multiple comparisons for each cell line. (G) Representative transmission electron microscopy images showing ultrastructural changes in the indicated DLBCL cell lines after treatment with either DMSO control (NT), tegavivint, chloroquine (CQ) or rapamycin (Rapa) at 12 hours. The arrows indicate accumulation of autophagic vacuoles containing cytoplasmic material after exposure to tegavivint or chloroquine (n=3). Chloroquine: 50uM and rapamycin: 10uM for all cell lines.

lines tested. However, treatment with rapamycin in combination with either tegavivint or chloroquine resulted in an increase of LC3-II compared with rapamycin alone across the four cell lines tested, suggesting that tegavivint may impair autophagic flux. In order to clarify this discordance, we transfected Riva and Pfeiffer cells with a GFP– mCherry-LC3 construct. This fluorescence reporter monitors LC3 flux by relying on both green-fluorescence protein (GFP) sensitivity and mCherry resistance to the acidic/proteolytic environment in the lysosomes. With autophagy blockade, both mCherry (red fluorescence protein) and GFP co-localize in a vesicular compartment that has not fused with lysosomes (autophagosomes) resulting in yellow puncta, whereas autophagy induction leads to fusion of these compartments (autolysosomes) and subsequent quenching of GFP signal resulting in singly mCherry red puncta.35 As shown in Figure 6E, treatment of these cells with either chloroquine or tegavivint increased yellow puncta assessed by confocal microscopy whereas rapamycin treatment was associated with increased red puncta solely. By calculating the ratio between yellow and red puncta, we found that the yellow signal was significantly increased in chloroquine and tegavivint-treated DLBCL cells (P<0.05) compared to controls, suggesting autophagosome accumulation (Figure 6E and F). Ultrastructurally, Riva and Pfeiffer treated with chloroquine or tegavivint for 12 hours showed accumulation of enlarged vesicles with characteristic enclosed cytoplasmic ultrastructures, supporting the notion that tegavivint treatment impairs the terminal stages of autophagy (Figure 6G). Collectively, these results indicate that tegavivint impairs cytoprotective autophagy by promoting SCFmediated degradation of key autophagy inducers such as MYC and BECLIN-1.

Discussion In the work presented here, we showed that TBL1 encoded by the TBL1X gene is abundantly expressed in DLBCL cells regardless of the molecular subtype. While further studies are required to validate TBL1 on DLBCL samples using IHC, and for this to predict relapse in a prospective manner, in this initial evaluation we show that TBL1 overexpression correlates with poor survival in de novo DLBCL patients uniformly treated with standard front-line immuno-chemotherapy. Genetic deletion of TBL1X or pharmacologic treatment with tegavivint significantly affected DLBCL cell viability in vitro while minimally toxic against normal immune cells and significantly prolonged the survival in four animal models of disseminated DLBCL, supporting the therhaematologica | 2021; 106(11)

apeutic potential of targeting TBL1 in this disease. While tegavivint has been shown to mediate AML and MM cell death by disrupting the TBL1/b-CATENIN interaction followed by proteasomal-mediated degradation of b-CATENIN and inhibition of the Wnt/transcriptional program,16,17 our findings are novel and provide insight into a previously undiscovered post-transcriptional, bCATENIN-independent oncogenic network modulated by TBL1 in DLBCL. Despite b-CATENIN expression in a subset of DLBCL cases,36,37 our data minimize the transcriptional relevance of b-CATENIN in promoting Wnt signaling and DLBCL cell survival while suggesting a possible role as a direct or indirect transcriptional repressor for some of the Wnt downstream targets (MYC, WISP1 and AXIN2) as previously shown in melanoma.38 However, this scenario may differ in a setting such as the tumor microenvironment where cytokine-induced Wnt signaling is active.39 The SCF-type of E3 ubiquitin ligase is a multi-protein complex composed of three static subunits RBX1, CUL1, SKP1 and variable F-box subunits which confer substrate selectivity to the complex.23,24,40-42 Here we showed that TBL1 modulates the stability of MYC, PLK1 and BECLIN1 through its direct interaction with the SCF complex. Mechanistically, our data support the hypothesis that tegavivint binding to TBL1 leads to conformational changes in the SCF complex resulting in CAND1-CUL1 dissociation followed by enhanced proteasomal degradation of these critical regulatory/pro-survival proteins. In further support of these findings, we showed that efficient shRNA knock-down of CAND1 led to significant decrease in DLBCL viability as well as downregulation of PLK1 and MYC protein expression. CAND1 functions as an inhibitor of the SCF complex by preventing the access of SKP1 and F-box proteins to the CUL1 catalytic core.23 In agreement with our work, C60, a small molecule inhibitor of CAND1, has been shown to destabilize CAND1-CUL1 interaction while increasing global ubiquitination in Epstein-Barr virus-positive lymphoma cell lines.43 Interestingly, TBL1 itself has an F-box motif and a WD40 domain for substrate recognition42 suggesting the possibility that TBL1 directly interacts with specific targets to modulate their stability. In support of this hypothesis which warrants further investigation, data produced in human embryonic kidney 293T cells showed that TBL1 directly interacts with and protects b-CATENIN from Siah-mediated ubiquitination and degradation as part of an SCFTBL1 complex.13,14 MYC rearrangement is identified in a subset of DLBCL cases (5-15%), MYC protein over-expression is seen more commonly (30–50% of DLBCL).44 Regardless of the genetic mechanism, MYC over-expression is associated 2937

Y. Youssef et al.

with a more aggressive clinical behavior and inferior outcome in DLBCL.45 In addition, it has been recently showed that PLK1 signaling promotes MYC protein stability, and in turn, MYC directly induces PLK1 transcription, establishing PLK1 as a therapeutic vulnerability in high grade B-cell lymphomas.28 Our findings are of particular importance because they indicate that tegavivint enhances the proteasomal-mediated degradation of MYC directly by SCFFbw7 and indirectly by enhancing SCFbTrCP– mediated PLK1 degradation supporting the notion that TBL1 holds promise as a therapeutic target particularly in MYC overexpressing DLBCL. Cytoprotective autophagy plays an important role in promoting lymphoma cell survival in adverse conditions and adequate levels of BECLIN-1 are necessary to maintain the autophagy flux as supported by the fact that BECLIN-1+/- mice have defective autophagy.29,30,33,46 Consistent with this published work and with that on MYC as a potent inducer of autophagy,31 our data indicate that tegavivint impairs pro-survival autophagy by promoting the SCF-mediated proteasomal degradation of both BECLIN-1 and MYC. Although we have identified TBL1 to play a central role in modulating the turnover of PLK1, MYC, and BECLIN1 through the SCF complex, there are a number of other potential SCF targets that will need to be explored to further define the role of TBL1 in the context of the SCF complex and to fully characterize the mechanism of action of tegavivint. For example, in addition to bCATENIN and PLK1, the TBL1 interacting SCF complex (SCFbTrCP) targets multiple proteins regulating mTOR signaling including the Ras homolog endriched in brain protein (RHEB), an upstream direct mTORC1 activator.47 Constitutive activation of mTOR signaling is critical for cellular growth and metabolism in both ABC- and GCBDLBCL48 and mTORC1 inhibition is known to lead to lymphoma cell cycle arrest and compensatory autophagy induction.49 While TBL1 targeting represents an attractive approach to simultaneously affect multiple prosurvival signaling pathways leading to a more profound lymphoma cell death than that achievable with agents targeting each of these pathways individually, this approach, given the complexity of the prosurvival pathways regulated by TBL1, may induce upregulation of compensatory mechanisms ultimately leading to drug resistance and providing the rational for combination strategies.

References 1. Teras LR, DeSantis CE, Cerhan JR, Morton LM, Jemal A, Flowers CR. 2016 US lymphoid malignancy statistics by World Health Organization subtypes. CA Cancer J Clin. 2016;66(6):443-459. 2. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature. 2000;403(6769):503-511. 3. Rosenwald A, Wright G, Chan WC, et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med. 2002;346(25):1937-1947. 4. Sehn LH, Donaldson J, Chhanabhai M, et al. Introduction of combined CHOP plus rituximab therapy dramatically improved

2938

While ongoing efforts are focused on studying the functional role of nuclear TBL1, the findings presented here demonstrate that TBL1 serves as a master regulator of a post-transcriptional oncogenic network and provide a rationale for TBL1 targeting in DLBCL with tegavivint which is currently in a phase-I clinical trial for patients with primary or recurrent desmoid tumors (clinicaltrials gov. Identifier: NCT03459469). Disclosure RH is an employee of Iterion therapeutics.The other authors have no conflicts of interest to declare. Contributions YY and LA conceptualized and designed research studies and wrote the manuscript; YY, VK, WC, FJ, AC, SS, AP, JHM, LT, WH, XZ, PZ, JC, ZK, EN, KL performed the experiments and analysed data; HGO provided bioinformatics support; AMS and RH provided technical and material support; KHY provided the DLBCL primary samples for IHC studies; KHY, HZ, ZYX, and GV performed and scored the IHC; XZ performed the statistical analysis; RL, KM, DML, JCB, and RAB provided conceptual advice and edited the manuscript; LA supervised the study and provided funding support. Acknowledgments We are grateful to the patients and healthy volunteers who provided tissue samples for these studies, to the OSU Comprehensive Cancer Center Leukemia Tissue Bank Shared Resource (supported by NCIP30 CA016058) for sample procurement, to the OSU Comparative Pathology and Mouse Phenotyping Shared Resource (CPMPSR,supported in part by NCIP30 CA016058) for excellent histotechnologic support (Ms. Brenda Wilson and Ms. Tessa VerStraete), to the OSU Campus Microscopy and Imaging Facility (CMIF supported in part by NIH Grant P30 CA016058) for excellent support with electron and confocal microscopy, to Somersault18:24 for the professional illustration of the proposed model of TBL1 posttranscriptional role within the SCF complex. The authors would also like to acknowledge Dr. Aharon Freud (OSU Department of Pathology) for providing normal pediatric tonsil controls. Funding This work was supported by the NCI (K08 CA226352 to LA; R35 CA198183 to JCB) and by a Research Scholar grant from the American Society of Hematology (LA)

outcome of diffuse large B-cell lymphoma in British Columbia. J Clin Oncol. 2005;23(22):5027-5033. 5. Chow VA, Shadman M, Gopal AK. Translating anti-CD19 CAR T-cell therapy into clinical practice for relapsed/refractory diffuse large B-cell lymphoma. Blood. 2018;132(8):777-781. 6. Sarkozy C, Sehn LH. Management of relapsed/refractory DLBCL. Best Pract Res Clin Haematol. 2018;31(3):209-216. 7. Yoon HG, Chan DW, Huang ZQ, et al. Purification and functional characterization of the human N-CoR complex: the roles of HDAC3, TBL1 and TBLR1. EMBO J. 2003;22(6):1336-1346. 8. Guenther MG, Lane WS, Fischle W, Verdin E, Lazar MA, Shiekhattar R. A core SMRT corepressor complex containing HDAC3

and TBL1, a WD40-repeat protein linked to deafness. Gene Dev. 2000;14(9):10481057. 9. Perissi V, Aggarwal A, Glass CK, Rose DW, Rosenfeld MG. A corepressor/coactivator exchange complex required for transcriptional activation by nuclear receptors and other regulated transcription factors. Cell. 2004;116(4):511-526. 10. Hatzi K, Jiang Y, Huang C, et al. A hybrid mechanism of action for BCL6 in B cells defined by formation of functionally distinct complexes at enhancers and promoters. Cell Rep. 2013;4(3):578-588. 11. Li J, Wang C-Y. TBL1–TBLR1 and β-catenin recruit each other to Wnt target-gene promoter for transcription activation and oncogenesis. Nat Cell Biol. 2008;10(2):160169.

haematologica | 2021; 106(11)

TBL1 modulates prosurvival proteins stability

12. Perissi V, Scafoglio C, Zhang J, et al. TBL1 and TBLR1 phosphorylation on regulated gene promoters overcomes dual CtBP and NCoR/SMRT transcriptional repression checkpoints. Mol Cell. 2008;29(6):755-766. 13. Dimitrova YN, Li J, Lee Y-T, et al. Direct ubiquitination of b-catenin by Siah-1 and regulation by the exchange factor TBL1. J Biol Chem. 2010;285(18):13507-13516. 14. Liu J, Stevens J, Rote CA, et al. Siah-1 mediates a novel b-catenin degradation pathway linking p53 to the adenomatous polyposis coli protein. Mol Cell. 2001;7(5):927-936. 15. Soldi R, Horrigan SK, Cholody MW, et al. Design, synthesis, and biological evaluation of a series of anthracene-9, 10-dione dioxime b-catenin pathway inhibitors. J Med Chem. 2015;58(15):5854-5862. 16. Fiskus W, Sharma S, Saha S, et al. Pre-clinical efficacy of combined therapy with novel b-catenin antagonist BC2059 and histone deacetylase inhibitor against AML cells. Leukemia. 2015;29(6):1267-1278. 17. Savvidou I, Khong T, Cuddihy A, McLean C, Horrigan S, Spencer A. b-catenin inhibitor BC2059 is efficacious as monotherapy or in combination with proteasome inhibitor bortezomib in multiple myeloma. Mol Cancer Ther. 2017;16(9): 1765-1778. 18. ANSI/ATCC ASN-0002-2011. Authentication of human cell lines: standardization of STR profiling. https://webstore.ansi.org/standards/atcc/ansiatccasn00022011. 19. de Almeida Nagata DE, Chiang EY, Jhunjhunwala S, et al. Regulation of tumorassociated myeloid cell activity by CBP/EP300 bromodomain modulation of H3K27 acetylation. Cell Rep. 2019;27(1): 269-281. 20. Aubrey BJ, Kelly GL, Kueh AJ, et al. An inducible lentiviral guide RNA platform enables the identification of tumor-essential genes and tumor-promoting mutations in vivo. Cell Rep. 2015;10(8):1422-1432. 21. Fiore C, Bailey D, Conlon N, et al. Utility of multispectral imaging in automated quantitative scoring of immunohistochemistry. J Clin Pathol. 2012;65(6):496-502. 22. Townsend EC, Murakami MA, Christodoulou A, et al. The public repository of xenografts enables discovery and randomized phase II-like trials in mice. Cancer Cell. 2016;29(4):574-586. 23. Goldenberg SJ, Cascio TC, Shumway SD, et al. Structure of the Cand1-Cul1-Roc1 complex reveals regulatory mechanisms for

haematologica | 2021; 106(11)

the assembly of the multisubunit cullindependent ubiquitin ligases. Cell. 2004;119 (4):517-528. 24. Cardozo T, Pagano M. The SCF ubiquitin ligase: insights into a molecular machine. Nat Rev Mol Cell Biol. 2004;5(9):739-751. 25. Giráldez S, Galindo-Moreno M, LimónMortés MC, et al. G1/S phase progression is regulated by PLK1 degradation through the CDK1/bTrCP axis. FASEB J. 2017;31(7): 2925-2936. 26. Popov N, Schülein C, Jaenicke LA, Eilers M. Ubiquitylation of the amino terminus of Myc by SCF b-TrCP antagonizes SCF Fbw7-mediated turnover. Nat Cell Biol. 2010;12(10):973-981. 27. Xiao D, Yue M, Su H, et al. Polo-like kinase-1 regulates Myc stabilization and activates a feedforward circuit promoting tumor cell survival. Mol Cell. 2016;64(3): 493-506. 28. Ren Y, Bi C, Zhao X, et al. PLK1 stabilizes a MYC-dependent kinase network in aggressive B cell lymphomas. J Clin Invest. 2018;128(12):5517-5530. 29. Alinari L. Toward autophagy-targeted therapy in lymphoma. Blood. 2017;129(13): 1740-1742. 30. Mathew R, Karantza-Wadsworth V, White E. Role of autophagy in cancer. Nat Rev Cancer. 2007;7(12):961-967. 31. Hart LS, Cunningham JT, Datta T, et al. ER stress–mediated autophagy promotes Mycdependent transformation and tumor growth. J Clin Invest. 2012;122(12):46214634. 32. Cui D, Xiong X, Zhao Y. Cullin-RING ligases in regulation of autophagy. Cell Div. 2016;11(1):1-14. 33. Kang R, Zeh H, Lotze M, Tang D. The Beclin 1 network regulates autophagy and apoptosis. Cell Death Differ. 2011;18(4): 571-580. 34. Mizushima N, Yoshimori T, Levine B. Methods in mammalian autophagy research. Cell. 2010;140(3):313-326. 35. Kimura S, Noda T, Yoshimori T. Dissection of the autophagosome maturation process by a novel reporter protein, tandem fluorescent-tagged LC3. Autophagy. 2007;3(5): 452-460. 36. Ge X, Lv X, Feng L, Liu X, Wang X. High expression and nuclear localization of bcatenin in diffuse large B-cell lymphoma. Mol Med Rep. 2012;5(6):1433-1437. 37. Bognar M, Vincendeau M, Erdmann T, et al. Oncogenic CARMA1 couples NF-κB and b-

catenin signaling in diffuse large B-cell lymphomas. Oncogene. 2016;35(32):4269-4281. 38. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic b-catenin signalling prevents anti-tumour immunity. Nature. 2015;523(7559):231-235. 39. Zhan T, Rindtorff N, Boutros M. Wnt signaling in cancer. Oncogene. 2017;36(11): 1461-1473. 40. Ciechanover A. The ubiquitin–proteasome pathway: on protein death and cell life. EMBO J. 1998;17(24):7151-7160. 41. Hussain M, Lu Y, Liu Y-Q, et al. Skp1: implications in cancer and SCF-oriented anti-cancer drug discovery. Pharmacol Res. 2016;111:34-42. 42. Wang Z, Liu P, Inuzuka H, Wei W. Roles of F-box proteins in cancer. Nat Rev Cancer. 2014;14(4):233-247. 43. Tikhmyanova N, Tutton S, Martin KA, et al. Small molecule perturbation of the CAND1-Cullin1-ubiquitin cycle stabilizes p53 and triggers Epstein-Barr virus reactivation. PLoS Pathog. 2017;13(7):e1006517. 44. Savage KJ, Johnson NA, Ben-Neriah S, et al. MYC gene rearrangements are associated with a poor prognosis in diffuse large B-cell lymphoma patients treated with R-CHOP chemotherapy. Blood. 2009;114(17):35333537. 45. Valera A, López-Guillermo A, CardesaSalzmann T, et al. MYC protein expression and genetic alterations have prognostic impact in patients with diffuse large B-cell lymphoma treated with immunochemotherapy. Haematologica. 2013;98(10):1554-1562. 46. Yue Z, Jin S, Yang C, Levine AJ, Heintz N. Beclin 1, an autophagy gene essential for early embryonic development, is a haploinsufficient tumor suppressor. Proc Natl Acad Sci U S A. 2003;100(25):15077-15082. 47. Harraz MM, Tyagi R, Cortés P, Snyder SH. Antidepressant action of ketamine via mTOR is mediated by inhibition of nitrergic Rheb degradation. Mol Psychiatry. 2016;21(3):313-319. 48. Sehn LH, Gascoyne RD. Diffuse large B-cell lymphoma: optimizing outcome in the context of clinical and biologic heterogeneity. Blood. 2015;125(1):22-32. 49. Wanner K, Hipp S, Oelsner M, et al. Mammalian target of rapamycin inhibition induces cell cycle arrest in diffuse large B cell lymphoma (DLBCL) cells and sensitises DLBCL cells to rituximab. Br J Haematol. 2006;134(5):475-484.

2939

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2940-2946

Non-Hodgkin Lymphoma

A prognostic index predicting survival in transformed Waldenström macroglobulinemia Eric Durot,1 Lukshe Kanagaratnam,2 Saurabh Zanwar,3 Efstathios Kastritis,4 Shirley D’Sa,5 Ramon Garcia-Sanz,6 Cécile Tomowiak,7 Bénédicte Hivert,8 Elise Toussaint,9 Caroline Protin,10 Jithma P. Abeykoon,3 Thomas GuerreroGarcia,11 Gilad Itchaki,12 Josephine M. Vos,13 Anne-Sophie Michallet,14 Sophie Godet,1 Jehan Dupuis,15 Stéphane Leprêtre,16 Joshua Bomsztyk,5 Pierre Morel,17 Véronique Leblond,18 Steven P. Treon,19 Meletios A. Dimopoulos,4 Prashant Kapoor,3 Alain Delmer1# and Jorge J. Castillo19# 1

Department of Hematology, University Hospital of Reims and UFR Médecine, Reims, France; 2Department of Research and Innovation, University Hospital of Reims, Reims, France; 3Division of Hematology, Department of Internal Medicine, Mayo Clinic, Rochester, MN, USA; 4Department of Clinical Therapeutics, National and Kapodistrian University of Athens, Athens, Greece; 5University College London Hospitals (UCLH) NHS Foundation Trust, London, UK; 6Department of Hematology, University Hospital of Salamanca, CIBERONC and Research Biomedical Institute of Salamanca (IBSAL), Salamanca, Spain; 7Department of Hematology and CIC U1402, University Hospital of Poitiers, Poitiers, France; 8Department of Hematology, Hospital of Lens, Lens, France; 9 Department of Hematology, University Hospital of Strasbourg, Strasbourg, France; 10 Department of Hematology, IUCT Oncopole, Toulouse, France; 11Department of Hematology/Oncology, Delbert Day Cancer Institute, Rolla, MO, USA; 12Institute of Hematology, Rabin Medical Center, Sackler Medical School, Tel-Aviv University, Tel-Aviv, Israel; 13Amsterdam University Medical Center (UMC) and LYMMCARE, Amsterdam, the Netherlands; 14Department of Hematology, Léon Bérard Center, Lyon, France; 15 Lymphoid Malignancies Unit, Henri Mondor Hospital, Créteil, France; 16 Inserm U1245 and Department of Hematology, Henri Becquerel Center and Normandie University UNIROUEN, Rouen, France; 17Department of Hematology, University Hospital of Amiens, Amiens, France; 18Department of Hematology, PitiéSalpêtrière Hospital and Sorbonne University, UPMC Paris, GRECHY, France and 19Bing Center for Waldenström Macroglobulinemia, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, USA #

AD and JJC contributed equally as co-senior authors.

ABSTRACT

Correspondence: ERIC DUROT edurot@chu-reims.fr Received: June 10, 2020. Accepted: October 23, 2020. Pre-published: November 12, 2020. https://doi.org/10.3324/haematol.2020.262899

2940

istological transformation into diffuse large B-cell lymphoma is a rare complication in patients with Waldenström macroglobulinemia (WM) and is usually associated with a poor prognosis. The objective of this study was to develop and validate a prognostic index for survival of patients with transformed WM. Through this multicenter, international collaborative effort, we developed a scoring system based on data from 133 patients with transformed WM who were evaluated between 1995 and 2016 (training cohort). Univariate and multivariate analyses were used to propose a prognostic index with 2-year survival after transformation as an endpoint. For external validation, a dataset of 67 patients was used to evaluate the performance of the model (validation cohort). By multivariate analysis, three adverse covariates were identified as independent predictors of 2-year survival after transformation: elevated serum lactate dehydrogenase (2 points), platelet count <100x109/L (1 point) and any previous treatment for WM (1 point). Three risk groups were defined: low-risk (0-1 point, 24% of patients), intermediate-risk (2-3 points, 59%; hazard ratio = 3.4) and high-risk (4 points, 17%; hazard ratio = 7.5). Two-year survival rates were 81%, 47%, and 21%, respectively (P<0.0001). This model appeared to be a better discriminant than either the International Prognostic Index or the revised International Prognostic Index. We validated this model in an independent cohort. This easy-to-compute scoring index is a robust tool that may allow identification of groups of transformed WM patients with different outcomes and could be used for improving the development of risk-adapted treatment strategies. haematologica | 2021; 106(11)

Prognostic index for transformed WM

Introduction Waldenström macroglobulinemia (WM) is a rare B-cell lymphoproliferative disorder characterized by lymphoplasmacytic bone marrow infiltration and production of an IgM monoclonal component.1 Histological transformation (HT) to diffuse large B-cell lymphoma (DLBCL) has been reported to occur in 2% to 10% of WM patients.2,3 Most transformed patients present with high-risk features such as extranodal disease, elevated serum lactate dehydrogenase (LDH) levels and high International Prognostic Index (IPI) scores.3,4 Patients who experience HT have an inferior survival compared to patients whose disease does not transform during its course.3,5 Patients with transformed WM are mainly treated with strategies used in de novo DLBCL but cure rates are low, with a median survival from the time of HT ranging from 16 to 32 months.3-5 However, as some patients experience prolonged survival, identifying those with high-risk features in order to select them for therapeutic intensification and/or novel agents is important. In a previous study of 77 patients, we identified elevated LDH and time to HT greater than 5 years as possible predictors of shorter survival4 but there is a need to develop an accurate predictive model for overall survival in a larger cohort of patients with transformed WM. Prognostic indices have been validated and are used routinely in aggressive non-Hodgkin lymphomas. The IPI was established in 1993 based on the clinical data of patients with de novo aggressive non-Hodgkin lymphoma treated with cyclophosphamide-doxorubicine-oncovin-prednisone (CHOP)-like chemotherapy.6 The revised IPI (R-IPI) was proposed in 2006 for a more accurate prediction of outcome in the era of treatment with CHOP plus rituximab (RCHOP).7 However, data pertaining to the prognostic value of these scores in the setting of transformed WM are sparse. Moreover, a majority of patients with transformed WM (65% to 76%) present with high IPI scores, probably limiting the accuracy of the IPI in this setting. The objectives of this large international collaborative study were to collect data on the characteristics of a large number of patients with transformed WM both at the time of the diagnosis of WM and at HT and to develop a prognostic index predicting survival following transformation, the transformed Waldenström International Prognostic Index (tWIPI). The final model was validated in an independent cohort of patients with transformed WM.

Methods Patients and data collection for development of the prognostic model Patients older than 18 years were included in the study if they had a diagnosis of WM and a concurrent or sequential histological diagnosis of DLBCL. The diagnosis of WM was based on criteria established in the Second International Workshop on WM.8 Patients with a diagnosis of indolent lymphoma other than WM were excluded. Histological assessment of transformation was mandatory for being considered in this study. We retrospectively identified 133 patients diagnosed with HT between January 1995 and December 2016 from French Innovative Leukemia Organization (FILO) centers (France), the Dana-Farber Cancer Institute (Boston, MA, USA), University College London Hospitals (UK) and Nieuwegein (the Netherlands) (details of the centers are provided in Online Supplementary Table S1). This retrospective study was conducted in accordance with the Declaration

haematologica | 2021; 106(11)

Table 1. Patients’ characteristics at diagnosis of Waldenström macroglobulinemia and at transformation in the training and validation sets.

Variable Entire cohort Median age, years Sex male/female, ratio Prior MGUS history

Training set

Validation set

n = 133 64 (range, 32-86) 75/58 (1.3/1) 20 (15%)

n = 67 63 (range, 27-82) 42/25 (1.7/1) 4 (17%)

WM characteristics n = 116 n = 64 (concurrent WM and HT excluded) Serum IgM level, g/L 17.7 (range, 1.4-66.7) 26 (range, 0.9-106) IPSS score n = 76 n = 43 0-1 31 (41%) 15 (40%) 2 30 (39%) 7 (18%) ≥3 15 (20%) 16 (42%) Median number of regimens 1 (range, 0-9) 1 (range, 0-9) prior to HT Therapies before HT n = 98 n = 54 Chlorambucil 43 (44%) 15 (28%) Fludarabine-based regimens 41 (42%) 16 (30%) Bendamustine +/- rituximab 19 (19%) 6 (11%) CHOP +/- rituximab 17 (17%) 11 (20%) Bortezomib-based regimens 15 (15%) 7 (13%) RCD 14 (14%) 12 (22%) Ibrutinib 5 (5%) 1 (2%) Autologous SCT 4 (3%) 0 (0%) Rituximab (alone or in combination) 67 (50%) 41 (76%) HT characteristics Median age, years PS (0-1/≥ 2) B symptoms Extranodal involvement Serum IgM level, g/L

n = 133 68 (range, 33-89) 59/48 (55%/45%) 56 (47%) 111 (86%) 6.9 (range, 0-66.6) Absolute neutrophils, x 109/L 4.1 (range, 0.2-20.2) Absolute lymphocytes, x 109/L 0.9 (range, 0.1-56) Hemoglobin, g/L 104 (range, 46-155) Platelets, x 109/L 172 (range, 9-610) Elevated LDH 85 (72%) Albumin level < 3.5 g/dL 62 (56%) Stage III or IV 96 (86%) Median number of lines 1 (range, 0-5) First-line therapies after HT n = 127 CHOP-like regimen +/- rituximab 102 (80%) DHAP +/- rituximab 10 (8%) GEMOX +/- rituximab 3 (2%) Rituximab-containing regimen 110 (87%) Autologous SCT 20 (16%) Allogeneic SCT 6 (5%)

n = 67 69 (range, 31-89) 30/18 (63%/37%) 30 (49%) 46 (69%) 6.3 (range, 0.3-43.9) 3.6 (range, 0.4-12.3) 1.2 (range, 0.2-30) 111 (range, 43-154) 186 (range, 8-576) 37 (55%) 30 (51%) 43 (83%) 2 (range, 0-5) n = 63 42 (67%) 3 (6%) 3 (6%) 44 (70%) 13 (21%) 2 (3%)

MGUS: monoclonal gammopathy of undetermined significance; WM: Waldenström macroglobulinemia; HT: histological transformation; IPSS: International Prognostic Scoring System; CHOP: cyclophosphamide-doxorubicin-oncovin-prednisone; RCD: rituximab-cyclophosphamide-dexamethasone; PS: performance status; LDH: lactate dehydrogenase; DHAP: dexamethasone-cytarabine-cisplatin; GEMOX: gemzar-oxaliplatin.

2941

E. Durot et al.

of Helsinki and was approved by the institutional review board at each participating institution. The clinical and disease characteristics considered as candidate prognostic factors were analyzed after reviewing medical records at the time of WM diagnosis and of HT. Variables considered as covariates for model building are detailed in the Online Supplementary Methods. In addition, the IPI and the R-IPI were assessed.6,7 The presence of MYD88L265P mutation was tested by allele-specific polymerase chain reaction on bone marrow samples at diagnosis of WM.9

Validation cohort The data from 96 patients diagnosed between 1988 and 2018 and treated at the Mayo Clinic (Rochester, MN, USA), or in Athens (Greece), Salamanca (Spain), Amsterdam (the Netherlands) and Toulouse (France) were analyzed (Online Supplementary Table S1). Information on the three parameters of the tWIPI was available for 67 patients.

Statistical methods The main endpoint of statistical analyses was 2-year overall survival calculated from the diagnosis of HT to the date of death or last follow-up. The survival curves were plotted using the Kaplan-Meier method and compared using the log-rank test for categorical variables. Univariate and multivariate analyses were performed using the Cox proportional hazards model. For continuous variables, the cutoffs were defined on the basis of published thresholds, for ease of clinical use. The multivariate Cox proportional hazards model included all variables with a P-value <0.10 by univariate analysis. A manual backward selection of covariates was used. The results were presented as hazard ratio (HR) and 95% confidence intervals (95% CI). A weighted risk score was assigned to each factor included in the final multivariable model. The prognostic score was then defined as the sum of single-risk parameters. Risk subgroups were pooled according to the number of patients within each category and the relative risk of death. The discriminatory value of the prognostic model and the score was assessed using concordance probability estimates by the Harrell concordance index (C-index).10 Calibration was assessed using the May and Hosmer test for goodness-of-fit. An internal validation of both the model and score was performed using the bootstrap resampling method11 (replication on 2,000 different samples drawn with replacement). External validation was performed in a second dataset of subjects. All tests of statistical significance were two-sided, and a P-value <0.05 was considered statistically significant. Statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA).

Results Patients’ characteristics The patients’ characteristics are summarized in Table 1. Of the 133 patients, 17 (13%) were diagnosed at the time of initial diagnosis of WM. Fifty-six percent of patients were male and the median age at WM diagnosis was 64 years (range, 32-86 years). According to the International Prognostic Scoring System (IPSSWM)12 when available, 31 patients (41%) were classified as low risk, 30 (39%) as intermediate risk and 15 (20%) as high risk. Following the diagnosis of WM, treatment was not initiated in 35 patients (26%) until the diagnosis of HT. The median number of lines of therapy for WM was one (range, 0-9). Half of the patients (n=67) had received rituximab alone or in combination for WM before HT. 2942

The median time from WM diagnosis to HT was 4.3 years (range, 0-25 years). The median age at HT was 68 years (range, 33-89 years). Extranodal involvement by the DLBCL component was noted in 86% of patients. Serum LDH was elevated in 85 patients (72%). The first-line regTable 2. Results of the univariate analysis of prognostic factors.

Characteristic

N. of patients (%) 2-year OS, %

Sex Male Female Previous treatment for WM No Yes Prior rituximab exposure No Yes Time to transformation Less than 5 years 5 years or more Age at transformation 65 years or less More than 65 years Performance status (ECOG) 0-1 More than 1 B symptoms Absent Present Extranodal involvement Absent Present Ann-Arbor stage I-II III-IV Leukocyte count 4 x 109/L or more Less than 4 x 109/L Hemoglobin level 100 g/L or more Less than 100 g/L Platelet count 100 x 109/L or more Less than 100 x 109/L Serum albumin 35 g/L or more Less than 35 g/L Serum LDH Less than or equal to ULN Greater than ULN Serum b2-microglobulin Less than 3 mg/L 3 mg/L or more

75 (56) 58 (44)

54.7 48.3

0.64

35 (26) 98 (74)

65.7 46.9

0.02

66 (50) 67 (50)

57.6 46.3

0.01

77 (58) 56 (42)

59.7 41.1

0.006

44 (33) 89 (67)

45.5 55.1

0.61

59 (55) 48 (45)

50.9 45.8

0.22

62 (53) 56 (47)

58.1 41.1

0.02

18 (14) 111 (86)

61.1 51.4

0.78

16 (14) 96 (86)

43.8 49

0.88

56 (48) 61 (52)

46.4 55.6

0.73

68 (57) 52 (43)

52.9 46.2

0.78

88 (75) 29 (25)

56.8 27.6

0.006

48 (44) 62 (56)

54.2 50

0.80

33 (28) 85 (72)

78.8 42.4

0.001

16 (28) 41 (71)

50 53.7

0.37

OS: overall survival; WM: Waldenström macroglobulinemia; ECOG: Eastern Cooperative Oncology Group; LDH, lactate dehydrogenase; ULN, upper limit of normal.

haematologica | 2021; 106(11)

Prognostic index for transformed WM

Figure 1. Kaplan-Meier curves for survival after transformation according to subgroups defined by the transformed Waldenström International Prognostic Index. (A) The model was built using three variables: previous treatment for Waldenström macroglobulinemia, lactate dehydrogenase at transformation and platelet count at transformation. It divided the cohort into three risk groups: low-, intermediate-, and high-risk with 2-year survival rates after transformation of 80.8%, 46.9% and 21.1%, respectively. (B) Validation cohort: 2-year survival after transformation of 71.4%, 39.4% and 0%, for the low-, intermediate-, and high-risk groups, respectively.

imens given at HT are listed in Table 1. The median number of lines of therapy given for HT was one (range, 0-5). The majority of patients (80%) were treated with first-line regimens used in de novo DLBCL (CHOP +/- rituximab). Rituximab was part of the first-line treatment for HT in 110 patients (87%). The median follow-up for all patients was 6.4 years (range, 0.1-33.7 years) after the diagnosis of WM and 2.3 years (range, 0-16.6 years) after HT. The median overall survival after HT was 19 months (95% CI: 12-31 months) (Online Supplementary Figure S1A). When we divided the cohort into two groups on the basis of the date of diagnosis of HT (using three cutoffs: 2004, 2008 and 2012), we did not observe any significant differences in terms of survival (data not shown). At the date of last follow-up, 83 patients (62%) had died. The majority of deaths were related to progressive disease (75%) or infections (14%).

Prognostic factors In univariate analysis, six variables that were associated with lower 2-year survival after HT were identified for inclusion in multivariate analyses: previous treatment for WM (P=0.02), prior exposure to rituximab (P=0.01), time to transformation more than or equal to 5 years (P=0.006), elevated LDH (P=0.001), B symptoms (P=0.02) and platelet count less than 100x109/L (P=0.006) (Online Supplementary Figure S2). Age and Eastern Cooperative Oncology Group performance status at HT were of no significant prognostic value. Among other variables included in the IPI, Ann Arbor stage III and IV and extranodal involvement were not only very common (86% for both) but also not associated with worse outcome. Serum IgM level at transformation was of no significant prognostic value (P=0.51). The prognostic values of the clinical and biological characteristics for survival at transformation are reported in Table 2. haematologica | 2021; 106(11)

Development of the prognostic model and the scoring system In multivariate analysis, independent factors for 2-year survival after HT were elevated serum LDH (P=0.003; HR=3.6; 95% CI: 1.53-8.50), platelet count less than 100x109/L (P=0.03; HR=1.8; 95% CI: 1.04-3.19) and previous treatment for WM (P=0.04; HR=2; 95% CI: 1.04-3.94) (Table 3). Bootstrapping of the multivariable model showed good internal validity. The May and Hosmer goodness-of-fit test did not identify any calibrations issues (P>0.6 for each stratum) and the model’s Harrell C-index was 0.75 (CI 95%: 0.65-0.84). The prognostic model comprised these three variables all available for 109 patients. Based on the relative hazard ratios, platelet count <100x109/L and previous treatment for WM were scored with 1 point and elevated serum LDH with 2 points. As a result, there were groups of patients with scores ranking from 0 to 4. Patients with score 0 were combined with those with score 1 because they were too few to constitute a separate risk group. Patients with scores of 2 and 3 were combined because they both corresponded to a group with an intermediate prognosis. The tWIPI was thus created and comprised three risk categories: low (0-1 point, 24% of patients), intermediate (2-3 points, 59%) and high (4 points, 17%). The 2-year survival rates were 81%, 47% and 21%, respectively (P<0.0001). The distribution of patients into these three groups and hazard ratios are shown in Table 4. The survival curves are shown in Figure 1A. The prognostic index displayed high model performances, as assessed by concordance probability estimates. The Harrell C-index was 0.75 (95% CI: 0.66-0.85). The May and Hosmer goodness-of-fit test did not identify any calibrations issues (P>0.7 for each stratum). Excluding patients with concurrent disease (WM and DLBCL), the model also identified three risk groups with significant different 2-year survivals and displayed good discrimination 2943

E. Durot et al.

Figure 2. Kaplan-Meier curves for survival after transformation according to risk group as defined by (A) the International Prognostic Index (IPI) and (B) the revised IPI (R-IPI).

Table 3. Results of the Cox regression analysis: final prognostic model.

Variable Previous treatment for WM Platelet count at HT LDH at HT

Adverse factor

Hazard ratio

95% CI

≥1 < 100 x 109/L > ULN

2 1.8 3.6

1.04-3.94 1.04-3.19 1.53-8.50

0.04 0.03 0.003

95% CI: confidence interval; WM: Waldenström macroglobulinemia; HT: histological transformation; LDH: lactate dehydrogenase; ULN: upper limit of normal.

Table 4. The transformed Waldenström International Prognostic Index: outcome and relative risk of death according to risk group.

Risk group Low Intermediate High

Score

N. of patients (%)

2-year OS %

Median survival, months

95% CI

0-1 2-3 4

26 (24) 64 (59) 19 (17)

80.8 46.9 21.1

NR 16.8 4.8

1.0 3.4 7.5

NA 1.3-8.7 2.7-20.7

OS: overall survival; HR: hazard ratio; 95% CI: 95% confidence interval; NR: not reached; NA: not applicable.

and calibration properties (Online Supplementary Figure S3A and Online Supplementary Table S2A).

Comparison with the International Prognostic Index and its revised form Complete information for the parameters of the IPI (age, serum LDH level, performance status, Ann Arbor stage and number of extranodal sites of disease) was available for 99 of the 109 patients used to build the tWIPI. The distribution of patients into the four IPI and the three R-IPI risk groups is shown in Online Supplementary Table S3. Neither the IPI nor the R-IPI was able to discriminate subgroups of patients with significantly different survival outcomes (P=0.33 and 0.24, respectively) (Figure 2).

External validation We applied the tWIPI to 67 other patients with transformed WM. The median follow-up from WM diagnosis and from HT was 8.8 (range, 0.2-20.8) and 3.1 years (range, 0-13.4) respectively. The main clinical characteristics of this 2944

cohort are shown in Table 1. The median survival after HT was 18 months (95% CI: 13 months – not reached) (Online Supplementary Figure S1B). The model successfully divided the cohort into three groups with 2-year survival rates of 71%, 39% and 0% for the low-, intermediate- and high-risk groups, respectively (P=0.0001) (Figure 1B). The prognostic significance of the tWIPI in the external cohort demonstrated good performance for discrimination. The Harrell Cindex was 0.79 (95% CI: 0.64-0.92). The May and Hosmer goodness-of-fit test did not identify any calibrations issues (P>0.8 for each stratum). In the same way as for the training cohort, the results were similar when patients with concurrent disease were excluded (Online Supplementary Figure S3B and Online Supplementary Table S2B).

Impact of MYD88 mutation status on survival after transformation By combining the training and the validation cohorts, we were able to analyze 64 patients with available data on MYD88 mutation status at the time of WM. Fortyhaematologica | 2021; 106(11)

Prognostic index for transformed WM

three patients (67%) carried a MYD88L265P mutation and 21 (33%) had wild-type (WT) alleles. The characteristics of the subset of patients for whom MYD88 mutation results were available and those for whom the status was not known (n=136) were comparable, except for a shorter time to transformation in the cohort with known MYD88L265P mutation status. The 2-year survival rates after HT were 67% and 49% in patients with MYD88WT and MYD88L265P, respectively (P=0.018) (Online Supplementary Figure S4).

Discussion The prognosis of transformed indolent lymphomas is historically poor despite combination chemoimmunotherapy, especially in chronic lymphocytic leukemia (Richter syndrome) and WM.4,13 Characterization of adverse prognostic factors in this setting is important for identifying specific risk groups and comparing different therapeutic approaches. There is no specific prognostic score for transformed WM and the existing scores such as the IPI appear not to discern prognosis appropriately. We developed an easy-to-use prognostic index relying on a model with three risk groups defined by the presence, or not, of one or more of the following parameters: previous treatment for WM, serum LDH level and platelet count at the time of HT. Previous treatment for WM is typically associated with prior exposure to rituximab and a prolonged time to transformation. This parameter could reflect chemo-resistance and/or immunological impairment related to the disease and its previous treatment. Serum LDH level is a well-established prognostic factor both in hematologic malignancies and solid tumors.14-17 Its prognostic role has been validated in both WM and DLBCL, being one of the variables included in the revised IPSSWM and the IPI, respectively.6,18 Low platelet count, also part of the IPSSWM, is usually associated with a poor prognosis in hematologic malignancies12,19 and could reflect a critical level of bone marrow involvement. For development of the prognostic score, only pretreatment characteristics were considered. Nevertheless, despite the retrospective nature of the study, first-line treatments at HT were quite uniform with a majority of patients being treated with R-CHOP chemoimmunotherapy, similarly to de novo DLBCL. This is unlikely to have influenced the analysis. Using this index, we were able to separate patients with transformed WM into three risk groups. In patients with a good prognosis (score 0-1), the 2-year survival rate was 81%. This indicates that the standard R-CHOP regimen could lead to prolonged control of the high-grade component in a majority of these patients. In the intermediate-risk group (score 2-3), less than half of the patients were alive after 2 years. In this group, the role of consolidative therapies such as high-dose therapy with stem cell transplantation in younger patients or association with targeted therapies would be interesting to investigate. For patients in the high-risk group (score 4), the outcome was very poor with a 2-year survival of 21%. Innovative therapies are required and these patients should be directed to clinical trials with new agents. Chimeric antigen receptor (CAR) T-cell therapies have been shown to be effective and to lead to durable responses in relapsed/refractory haematologica | 2021; 106(11)

DLBCL including transformed follicular lymphomas.20-22 The potential effectiveness of CAR T-cell therapy in transformed WM has recently been suggested based on one case report with complete response maintained at 1 year.23 Clinical trials are needed to evaluate the place of CAR T-cell therapy in WM and transformed WM. An important finding of our study is that the IPI and the R-IPI do not seem appropriate to identify patients with significantly different outcomes in the particular setting of transformed WM. Application of the IPI in our cohort was not able to separate the intermediate-risk and highrisk groups, most patients with transformed WM falling in the high-intermediate or the high-risk category. In addition, of the IPI risk factors, only serum LDH level showed prognostic relevance in univariate analyses. The IPI and the R-IPI have been studied in other transformed lymphomas such as transformed follicular lymphoma and marginal zone lymphoma and could predict survival.24,25 A Richter syndrome prognostic score has been proposed and is based on five adverse risk factors.13 Interestingly, the three variables of our score are part of the Richter synrome score. We performed internal validation by bootstrap11 and confirmed marked stability of the model developed. Despite the rarity of the disease, we were able to validate the prognostic index in an independent cohort of patients with transformed WM. Our model displayed good discrimination properties in the validation cohort, identifying three risk groups with similar 2-year survival after transformation to the ones in the training set. This external validation confirms the robustness and the reproducibility of the tWIPI. Advances in the biology of WM have demonstrated the role of mutation status in outcome prediction. The MYD88L265P mutation is found in 95% of WM patients and represents an important diagnostic marker.26 MYD88WT WM patients seem to have a worse outcome and a higher incidence of DLBCL events.5,27 In our study, molecular parameters were available only for one-third of the patients and so could not be included in the initial analysis. By combining the two cohorts, we could analyze 64 patients and found that patients with MYD88L265P mutation had a significantly lower 2-year survival rate after transformation compared to patients with MYD88WT disease. Although this finding should be confirmed by multivariate analysis in a larger cohort of patients, it is in line with previous studies showing that MYD88 mutations are associated with worse survival in de novo DLBCL.28-30 Our study has some limitations. First, the majority of patients were exposed to chlorambucil and/or fludarabine-based regimen as therapy for WM. Half of the patients received rituximab alone or in combination and very few patients were treated with Bruton tyrosine kinase inhibitors such as ibrutinib. The tWIPI warrants further validation in a cohort of patients with transformed WM treated with more contemporary regimens at the time of WM. Secondly, in the present study, we were not able to assess the clonal relationship between the original WM and DLBCL. It is known that the occurrence of DLBCL in WM can result from HT or arise as a de novo, not clonally related lymphoma.31 This phenomenon has been widely described in Richter syndrome in which de novo DLBCL usually carries a better prognosis (median survival of 5 years vs. 8-16 months for clonally 2945

E. Durot et al.

related DLBCL transformation).32 Nevertheless, one strength of our study was the strict and homogeneous definition of transformation by restricting inclusion to histologically documented transformation. In conclusion, through this large multicenter study aimed at identifying factors predicting survival in transformed WM, we developed a prognostic model and validated it in an independent series of patients. Retrospective and prospective international multicenter studies are needed to define the optimal therapeutic strategies for transformed WM. Our prognostic score could help physicians individualize treatment strategy and improve the management of patients with transformed WM by selecting the most appropriate treatment.

Contributions ED, LK, AD, and JJC designed the study; ED, SZ, EK, SD, RG, CT, BH, ET, CP, JPA, TG, GI, JMV, SG, JD, SL, and JB collected patients’ clinical data; ED and LK performed the statistical analysis; ED, LK, AD and JJC wrote the manuscript, all authors contributed to analyzing and interpreting the data and provided final approval of the manuscript. Acknowledgments The authors would like to acknowledge the following people who participated in the study: Fatiha Merabet (Versailles), Eric Van Den Neste (Bruxelles), Sarah Ivanoff (Amiens), Xavier Roussel (Besançon), Jean-Marc Zini (Paris, St-Louis), Caroline Regny (Grenoble), Richard Lemal (Clermont-Ferrand), Laurent Sutton (Argenteuil), Aurore Perrot (Nancy) and Katell Le Dû (Le Mans).

Disclosures No conflicts of interest to disclose.

References 1. Swerdlow SH, Berger F, Pileri SA, et al. Lymphoplasmacytic lymphoma. In: Swerdlow SH, Campo E, Harris NL, et al., editors. WHO Classification of Tumours of Haematopoietic and Lymphoïd Tissues. Lyon, Franc. IARC 2008. pp. 194-195. 2. Leblond V, Johnson S, Chevret S, et al. Results of a randomized trial of chlorambucil versus fludarabine for patients with untreated Waldenström macroglobulinemia, marginal zone lymphoma, or lymphoplasmacytic lymphoma. J Clin Oncol. 2013; 31(3):301-307. 3. Castillo JJ, Gustine J, Meid K, et al. Histological transformation to diffuse large B-cell lymphoma in patients with Waldenström macroglobulinemia. Am J Hematol. 2016;91(10):1032-1035. 4. Durot E, Tomowiak C, Michallet AS, et al. Transformed Waldenström macroglobulinaemia: clinical presentation and outcome. A multi-institutional retrospective study of 77 cases of the French Innovative Leukemia Organization (FILO). Br J Haematol. 2017; 179(3):439-448. 5. Zanwar S, Abeykoon JP, Durot E, et al. Impact of MYD88 L265P mutation status on histological transformation of Waldenström macroglobulinemia. Am J Hematol. 2020;95(3):274-281. 6. The International Non-Hodgkin’s Lymphoma Prognostic Factors Project. A predictive model for aggressive nonHodgkin’s lymphoma. N Engl J Med. 1993;329(14):987-994. 7. Sehn LH, Berry B, Chhanabhai M, et al. The revised International Prognostic Index (R-IPI) is a better predictor of outcome than the standard IPI for patients with diffuse large B-cell lymphoma treated with RCHOP. Blood. 2007;109(5):1857-1861. 8. Owen RG, Treon SP, Al-Katib A, et al. Clinicopathological definition of Waldenstrom’s macroglobulinemia: consensus panel recommendations from the Second International Workshop on Waldenstrom’s Macroglobulinemia. Semin Oncol. 2003;30(2):110-115. 9. Xu L, Hunter ZR, Yang G, et al. MYD88 L265P in Waldenstrom macroglobulinemia, immunoglobulin M monoclonal gammopathy, and other B cell lymphoproliferative disorders using conventional and quantitative allele-specific polymerase chain reac-

2946

tion. Blood. 2013;121(11):2051-2058. 10. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361-387. 11. Efron B, Tibshirani RJ. An Introduction to the Boostrap. Vol. 57: Chapman and Hall/CRC. 1994. 12. Morel P, Duhamel A, Gobbi P, et al. International prognostic scoring system for Waldenstrom macroglobulinemia. Blood. 2009;113(18):4163-4170. 13. Tsimberidou AM, O’Brien S, Khouri I, et al. Clinical outcomes and prognostic factors in patients with Richter’s syndrome treated with chemotherapy or chemoimmunotherapy with or without stem-cell transplantation. J Clin Oncol. 2006;24(15):2343-2351. 14. Solal-Céligny P, Roy P, Colombat P, et al. Follicular lymphoma international prognostic index. Blood. 2004;104(5):1258-1265. 15. Hoster E, Dreyling M, Klapper W, et al. A new prognostic index (MIPI) for patients with advanced-stage mantle cell lymphoma. Blood. 2008;111(2):558-565. 16. Palumbo A, Avet-Loiseau H, Oliva S, et al. Revised International Staging System for Multiple Myeloma: a report from International Myeloma Working Group. J Clin Oncol. 2015;33(26):2863-2869. 17. Manola J, Atkins M, Ibrahim J, et al. Prognostic factors in metastatic melanoma: a pooled analysis of Eastern Cooperative Oncology Group trials. J Clin Oncol. 2000; 18(22):3782-3793. 18. Kastritis E, Morel P, Duhamel A, et al. A revised international prognostic score system for Waldenström's macroglobulinemia. Leukemia. 2019;33(11):2654-2661. 19. Ochi Y, Kazuma Y, Hiramoto N, et al. Utility of a simple prognostic stratification based on platelet counts and serum albumin levels in elderly patients with diffuse large B cell lymphoma. Ann Hematol. 2017; 96(1):1-8. 20. 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. 21. Locke FL, Ghobadi A, Jacobson CA, et al. Long-term safety and activity of axicabtagene ciloleucel in refractory large B-cell lymphoma (ZUMA-1): a single-arm, multicenter, phase 1-2 trial. Lancet Oncol. 2019; 20(1):31-42.

22. 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. 23. Bansal R, Jurcic JG, Sawas A, et al. Chimeric antigen receptor T cells for treatment of transformed Waldenström macroglobulinemia. Leuk Lymphoma. 2020;61(2):465-468. 24. Alonso-Alvarez S, Magnano L, Alcoceba M, et al. Risk of, and survival following, histological transformation in follicular lymphoma in the rituximab era. A retrospective multicentre study by the Spanish GELTAMO group. Br J Haematol. 2017; 178(5):699-708. 25. Alderuccio JP, Zhao W, Desai A, et al. Risk factors for transformation to higher-grade lymphoma and its impact on survival in a large cohort of patients with marginal zone lymphoma from a single institution. J Clin Oncol. 2018 Oct 12. [Epub ahead of print] 26. Treon SP, Xu L, Yang G, et al. MYD88 L265P somatic mutation in Waldenström's macroglobulinemia. N Engl J Med. 2012; 367(9):826-833. 27. Treon SP, Gustine J, Xu L, et al. MYD88 wild-type Waldenstrom macroglobulinaemia: differential diagnosis, risk of histological transformation, and overall survival. Br J Haematol. 2018;180(3):374-380. 28. Rovira J, Karube K, Valera A, et al. MYD88 L265P mutations, but no other variants, identify a subpopulation of DLBCL patients of activated B-cell origin, extranodal involvement, and poor outcome. Clin Cancer Res. 2016;22(11):2755-2764. 29. Dubois S, Viailly PJ, Bohers E, et al. Biological and clinical relevance of associated genomic alterations in MYD88 L265P and non-L265P-mutated diffuse large B-cell lymphoma: analysis of 361 cases. Clin Cancer Res. 2017;23(9):2232-2244. 30. Schmitz R, Wright GW, Huang DW, et al. Genetics and pathogenesis of diffuse large B-cell lymphoma. N Engl J Med. 2018; 378(15):1396-1407. 31. Talaulikar D, Biscoe A, Lim JH, et al. Genetic analysis of diffuse large B-cell lymphoma occurring in cases with antecedent Waldenström macroglobulinaemia reveals different patterns of clonal evolution. Br J Haematol. 2019;185(4):767-770. 32. Rossi D, Spina V, Gaidano G. Biology and treatment of Richter syndrome. Blood. 2018;131(25):2761-2772.

haematologica | 2021; 106(11)

ARTICLE

Platelet Biology & its Disorders

Platelet proteome and function in X−linked thrombocytopenia with thalassemia and in silico comparisons with gray platelet syndrome

Ferrata Storti Foundation

Daniel Bergemalm,1 Sofia Ramström,2,3 Caroline Kardeby,3 Kjell Hultenby,4 Anna Göthlin Eremo,5 Carina Sihlbom,6 Jörgen Bergström,6 Jan Palmblad7 and Maria Åström1 Department of Medicine, Faculty of Medicine and Health, Örebro University, Örebro; Department of Clinical Chemistry, and Department of Biomedical and Clinical Sciences, Linköping University, Linköping; 3Cardiovascular Research Center, School of Medical Sciences, Örebro University, Örebro; 4Department of Laboratory Medicine, Karolinska Institute, Karolinska University Hospital Huddinge, Stockholm; 5Department of Clinical Research Laboratory, Faculty of Medicine and Health, Örebro University, Örebro; 6Proteomics Core Facility, University of Gothenburg, Gothenburg and 7 Departments of Medicine and Hematology, Karolinska Institutet, Karolinska University Hospital Huddinge, Stockholm, Sweden 1 2

Haematologica 2021 Volume 106(11):2947-2959

ABSTRACT

n X-linked thrombocytopenia with thalassemia (XLTT; OMIM 314050), caused by the mutation p.R216Q in exon 4 of the GATA1 gene, male hemizygous patients display macrothrombocytopenia, bleeding diathesis and a b-thalassemia trait. Herein, we describe findings in two unrelated Swedish XLTT families with a bleeding tendency exceeding what is expected from the thrombocytopenia. Blood tests revealed low P-PAI-1 and P-factor 5, and elevated S-thrombopoietin levels. Transmission electron microscopy showed diminished numbers of platelet a- and dense granules. The proteomes of isolated blood platelets from five male XLTT patients, compared to five sex- and agematched controls, were explored. Quantitative mass spectrometry showed alterations of 83 proteins (fold change ≥±1.2, q<0.05). Of 46 downregulated proteins, 39 were previously reported to be associated with platelet granules. Reduced protein levels of PTGS1 and SLC35D3 were validated in megakaryocytes of XLTT bone marrow biopsies by immunohistochemistry. Platelet function testing by flow cytometry revealed low dense- and a-granule release and fibrinogen binding in response to ligation of receptors for ADP, the thrombin receptor PAR4 and the collagen receptor GPVI. Significant reductions of a number of agranule proteins overlapped with a previous platelet proteomics investigation in the inherited macrothrombocytopenia gray platelet syndrome. In contrast, Ca2+ transporter proteins that facilitate dense granule release were downregulated in XLTT but upregulated in gray platelet syndrome. Ingenuity pathway analysis showed altered coagulation system and protein ubiquitination pathways in the XLTT platelets. Collectively, the results revealed protein and functional alterations affecting platelet a- and dense granules in XLTT, probably contributing to bleeding.

Introduction The inherited platelet disorder, X-linked thrombocytopenia with thalassemia (XLTT; OMIM 314050) was first described in 1977 in a family where three men presented with macrothrombocytopenia, bleeding diathesis, splenomegaly and mild hemolysis of the b-thalassemia type.1 Three additional families were reported2-5 prior to our description of two Swedish XLTT families exhibiting a previously not reported grade 1−2/3 myelofibrosis.6 Recently, a Danish−Swedish whole-exome sequencing study of 156 patients with bleeding tendency identified two additional

haematologica | 2021; 106(11)

Correspondence: MARIA ÅSTRÖM maria.astrom@regionorebrolan.se Received: February 19, 2020. Accepted: September 15, 2020. Pre-published: September 28, 2020. https://doi.org/10.3324/haematol.2020.249805

2947

D. Bergemalm et al.

families (three individuals) with the disease.7 All reported XLTT patients carried the same exon 4 GATA1 p.R216Q mutation.2-4,6-8 Similarities with the autosomal inherited disorder gray platelet syndrome (GPS)9 have been noted regarding deficiency of platelet a-granules,2,4,6,10 splenomegaly,1,2,6,11 and, more recently, myelofibrosis.6,11 Bleeding diathesis beyond what could be expected from blood platelet counts was observed in several male XLTT patients, with occasional severe bleeds requiring platelet and/or erythrocyte transfusions1-4,6,7 We, therefore, set out to evaluate hemostasis and platelet functions in members of our two Swedish XLTT families, as a complement to earlier investigations. Our approach was to map the XLTT platelet proteome in order to disclose anticipated platelet granule deficiencies and other abnormalities. Subsequently, we sought to validate alterations of selected proteins by immunohistochemistry (IHC) in bone marrow (BM) megakaryocytes and platelet functional reactions related to granule deficiencies by flow cytometry. Finally, we compared our platelet proteomic findings to those of a published dataset from a NBEAL2 mutated patient diagnosed with GPS (by courtesy of Dr. Meral Gunay-Aygun, Johns Hopkins University, e-mail, 24 October 2018), aiming to disclose similarities and differences.12

were investigated using flow cytometry (see the Online Supplementary Methods and Online Supplementary Figure S1 for details). In summary, diluted whole blood was incubated for 10 minutes with specific receptor agonists, and platelet activation was detected as follows: (i) a conformational change in the platelet fibrinogen receptor GPIIb/IIIa was detected as binding of a chicken anti-human fibrinogen antibody; (ii) exocytosis of platelet agranules was detected as binding of a mouse anti-human Pselectin (CD62P) antibody; (iii) exocytosis of platelet lysosomes was detected as binding of an antibody towards human LAMP1 (CD107a); and (iv) exposure of the procoagulant phospholipid phosphatidylserine (PS) was detected as binding of annexin V. Capacity of platelet dense granule release of ADP was detected by an indirect method, where the effect of addition of apyrase to samples to degrade released ADP was investigated as previously described.14,15 Compared to e.g., CD63 or LAMP2 exposure, which only proves granule exocytosis but does not give information on granule contents, this alternative approach was used as it illustrates the functional consequences of released dense granule contents and not only the possibility to exocytose the granules. For additional methods regarding hematological and other blood tests, transmission electron microscopy (TEM), platelet proteomics, bioinformatics, immunohistochemistry, flow cytometry, and statistical analyses, see the Online Supplementary Methods.

Data sharing statement Methods Patients and healthy controls Five male XLTT patients from two unrelated Swedish families (A and B) were recruited for the study. Subjects gave written informed consent in accordance with institutional guidelines and the Declaration of Helsinki. The Regional Ethical Review Board (Uppsala, Sweden) approved the studies (reference 2010/294). Pyrosequencing confirmed hemizygosity of the p.R216Q GATA1 mutation, corresponding to the amino acid change Arg216Gln, in the five investigated males.2,4 The phenotypical aspects, including bleedings, of the four adult patients were described previously.6,13 Additionally, an 8-year old boy (at sampling) from family B,6 was now included for proteomics and routine investigations. The control material for proteomics consisted of platelets from five sexand age-matched healthy blood donors.

Platelet proteomics For proteomic analysis, platelets from the five male XLTT patients and the five sex- and age-matched healthy volunteers were isolated. Whole blood samples were collected into EDTA Vacutainer® tubes and processed as described in the Online Supplementary Methods in order to obtain lysed platelet pellets. Quantitative mass spectrometry (QMS) was performed at the Proteomics Core Facility (PCF), Sahlgrenska Academy, University of Gothenburg, Sweden. Equal amounts of total protein from each sample were trypsin digested, alkylated, and peptides subjected to the isobaric mass tagging reagent TMT® and further acidified. Peptides were purified, fractionated and analyzed on Q Exactive™ or Orbitrap Fusion Tribrid mass spectrometers (Thermo Scientific™, Waltham, MA, USA). For the identification of proteins, a database search was performed using the Mascot search engine (Matrix Science, Boston, MA, USA) followed by protein quantification based on TMT reporter ion intensities (see the Online Supplementary Methods for details).

Flow cytometry Platelet activation responses to stimulation of the platelet receptors for ADP, thrombin (PAR1 and PAR4) and collagen (GPVI)

2948

The complete set of quantified platelet proteins that was uploaded to Ingenuity Pathway Analysis (IPA)16 is listed in the Online Supplementary Appendix.

Results Bleeding and laboratory tests The five male patients (including one child) included in the proteomics investigation had bleeding diathesis with recurring nose bleeds and spontaneous hematomas. One of the adult patients (who had stable platelet counts between 50−90×109/L) had on one occasion, after a minor trauma during sports activity, severe thigh muscle bleeding with compartment syndrome, necessitating transfusions. Pedigrees, case reports including bleeding information, hematological indices and BM fibrosis grades of members of the A and B XLTT families were given previously.6 Platelet and hemoglobin values as well as hemostasis related laboratory characteristics are shown in Table 1. The patients displayed mild hemolytic anemia and moderate macrothrombocytopenia (Online Supplementary Figure S2A and B), the latter despite increased numbers of CD61-positive megakaryocytes in the BM (Online Supplementary Figure S2C and D).6 Among routine coagulation tests, APTT and PK-INR were slightly elevated in three and two adults, respectively, from family B. Plasma levels of plasminogen activator inhibitor 1 (PPAI-1) were consistently low. P-Factor 8 levels were normal, but P-Factor 5 (P-F5) levels were below the normal range in the three adults from family B, whereas two patients had P-F5 levels in the low normal range (Table 1). Thus, the previously noted low P-F5 value in one XLTT patient6 was also found in other family members, probably contributing to prolonged APTT and bleeding tendency. No mutation was detected in the F5 gene from the individual with the lowest P-F5 using a TruSight One Expanded sequencing panel (Illumina, Inc., San Diego, CA, USA). haematologica | 2021; 106(11)

Platelet proteome in XLTT

Table 1. Age at samplig for proteomics and laboratory characteristics of the five investigated male X-linked thrombocytopenia with thalassemia patients (with normal range for individual biomarkers in parentheses).

ID* Age, years Hemoglobin† (134-170 g/L) Platelet count† (145-387×109/L) MPV (7-9 fL) APTT (29-42 s) PK-INR (<1.2 INR) P-Fibrinogen (2.0-4.0 g/L) P-Factor 5 (0.6-1.5 kIU/L) P-Factor 8 (0.5-1.8 kIU/L P-PAI-1 (<15 kIU/L) P-VWF (0.5-1.5 kIU/L) S-TPO (14-75 pg/mL)

III

59 115-131 25-67 11.6 47 1.0 2.4-2.7 0.37 1.21 <2 0.78 156

37 130-137 44-64 11.3 43 (ref 28-40) 1.2 1.6 0.55 0.90 <2 0.86 121

33 125-141 57-87 11.1 48 1.2-1.3 2.2 0.52 1.89 <2 0.89 99

8 116-126 96-115 11.3 29 (ref 26-33) 1.1 2.5 0.86 1.19 3.5 0.96 ND

39 116-147 22-101 11.8 38 1.0 2.3 0.70 0.88 <2 1.98 (ref 0.5-2.0) ND

*Patients were from two unrelated families, where I, II, III and IV belong to family B and V belongs to family A.6 †Range from several sampling occasions. The maximum platelet count of patient V (101×109/L) was sampled at an infectious episode. P: plasma; S: serum; MPV: mean platelet volume; APTT: activated partial thromboplastin time; PK-INR: prothrombin complex – international normalized ratio; PAI-1: plasminogen activator inhibitor 1; VWF: Von Willebrand Factor; TPO: thrombopoietin; ND: not determined.

We investigated whether or not the thrombocytopenia could be explained by defective regulation of thrombopoietin (TPO) turnover. Serum TPO levels were higher in all three sampled XLTT males (range, 99−156 pg/mL), and in a female carrier (95 pg/mL), compared to age- and sex-matched controls (n=10) showing a mean value of 47 pg/mL (range, 14−75) pg/mL (P<0.001) (Table 1; Online Supplementary Figure S3)

Transmission electron microscopy TEM of platelets from three males with XLTT, representing both families, showed the presence of abnormally large platelets and deficiencies in the numbers and contents of a-granules compared to controls investigated in parallel (Figure 1A to I). Empty looking vacuoles were abundant, probably representing “ghost a-granules”.12 The dense tubular system and open canalicular system were well represented. Dense granules were not observed in XLTT but were found in platelets from healthy controls, although whole mounts were not used. Thus, we corroborated previous reports of a-granule2,4,10 and one report of dense granule10 deficiencies in XLTT, with probable significance for the bleeding diathesis. Overall, the ultrastructural alterations in XLTT platelets were largely similar to those described in an earlier patient.10

Platelet proteomics results In order to explore potential alterations in the platelet proteome, we used QMS. In isolated platelets from the five XLTT patients and five age-matched male healthy controls, >3,100 proteins were identified, similar to previous reports.17 Out of these, >2,200 proteins could be quantified in both patients and controls and further analyzed by statistical comparison of groups (Online Supplementary Methods). Eighty-three proteins were shown to be significantly altered (fold change [FC] ≥±1.2, q<0.05); 46 showing reduced and 37 elevated levels (Tables 2 and 3). From two previously reported datasets of >800 proteins predicted to be of granule origin in healthy individuals,18,19 47 proteins were here identified to be differentially regulated. Congruent to findings from TEM of sparse numbers of granules, 39 of 47 (83%) predicted granule proteins identified by QMS with FC ≥±1.2 and q<0.05 were haematologica | 2021; 106(11)

downregulated in XLTT (Table 2). SLC35D3, a protein involved in the biogenesis of platelet dense granules,20 was downregulated in all XLTT patients, mean 3.4-times compared to healthy controls. Similar to the finding in plasma (Table 1), SERPINE1/PAI-1 (stored in a-granules) showed significantly reduced levels in XLTT platelets. In accordance with the slightly hemolytic phenotype, haptoglobin (HP) was four-times downregulated (Table 2). In contrast, the antioxidative enzyme carbonic anhydrase 2 (CA2) was almost three-times more abundant in XLTT compared to healthy controls. Also the seventh among the most upregulated significant proteins in XLTT platelets, peroxiredoxin 1 (PRDX1), has antioxidant effects. Some other top upregulated proteins including tubulin-tyrosine ligase-like protein 12 (TTLL12), spectrin a chain, non-erythrocytic 1 (SPTAN1) and nexilin (NEXN) were cytoskeletal components (Table 3). Notably, the protein level of NBEAL2 (mutated in GPS) was not significantly altered in XLTT compared to control platelets (FC =1.05, q=0.24). The 83 significantly altered proteins were predicted in IPA16 to originate from different subcellular compartments, the majority from the cytoplasm (not shown).

Pathway and network analyses In a core analysis in IPA16 (13/05/2019) of the 83 platelet proteins with FC ≥±1.2 and q<0.05 compared to the controls, coagulation system was the most significant pathway, with reductions of the a-granule proteins F13A1, SERPINE1/PAI-1 and von Willebrand factor (VWF). The second most significant pathway was protein ubiquitination, with five upregulated proteins (HSPA1A/HSPA1B, PSMA4, PSMB2, PSMB4, PSMC4) and one downregulated (UBE2O). Protein−protein interaction network analysis using STRING21 (28/07/2020) suggested two clusters with altered granule and vesicle domain proteins, and one cluster with altered proteasomal proteins (Figure 2).

X-linked thrombocytopenia with thalassemia versus gray platelet syndrome Comparison with a published GPS platelet a-granule fraction sub-proteome,12 Online Supplementary Table S2 containing 230 proteins with FC ≥±1.2 from one GPS 2949

D. Bergemalm et al.

patient compared to one control, revealed that nine proteins, all of them known to be present in a-granules, were downregulated in both XLTT and GPS (for XLTT with criteria FC ≥±1.2 and q<0.05): LTBP1, PPBP, THBS1, SELP/Pselectin, MMRN1, APP, F13A1, HSD17B4 and ANO6. In addition, there was one jointly upregulated granule protein, SACM1L, and six granule proteins that were downregulated in XLTT but upregulated in GPS: FHL1,

YWHAH, ATP2A3, WDR1, MLEC and ATP2A2 (Tables 2 and 3). Using equal criteria as for the GPS study, among 729 XLTT platelet proteins with FC ≥±1.2 regardless of statistical significance, six were found to be commonly upregulated and 30 commonly downregulated whereas 24 were contraregulated in XLTT in comparison to GPS (Online Supplementary Figure S4; Online Supplementary Table S1). The three Ca2+ transporting proteins ATP2A3,

Figure 1. Occurrence of giant platelets and granule deficiency in X-linked thrombocytopenia with thalassemia. Transmission electron microscopy (TEM) graphs of platelets (A and B) from a healthy control and (C to I) from three X-linked thrombocytopenia with thalassemia (XLTT) patients. Bars in (A and D−I) =2 mm, in (B and C) =1 mm. (A) Platelets from controls show a normal morphology with few mitochondria, some open canalicular system components and several rounded a-granules. (B) A few dense granules (arrow) were also observed in the controls. (C) Detail of a platelet from XLTT patient I (Table 1) shows reduced amounts of a-granules and mostly empty vacuoles, some containing small remnants (arrowheads) of a-granules. (D) Overview of platelets from XLTT patient V (Table 1) showing reduced amounts of αgranules and some platelets with increased dilated open canicular system (arrowheads). A platelet containing a secondary platelet was also observed (arrow). (E) Overview of platelets from XLTT patient II (Table 1) showing reduced amounts of a-granules and platelets with increased dilated open canicular system (arrowheads). (F) Overview of platelets from XLTT patient I, showing reduced amounts of a-granules and increased dilated open canicular system (arrowhead). (G) An agranular macrothrombocyte from XLTT patient II, with microtubuli in the periphery (arrowheads), vacuoles and some mitochondria. (H) An agranular macrothrombocyte from XLTT patient II, with a tubular inclusion (arrow) and some vacuoles containing electron dense cell debris (arrowheads). (I) An agranular macrothrombocyte from XLTT patient II, containing some mitochondria (arrowheads) and elements of the dense tubular system (arrow).

2950

haematologica | 2021; 106(11)

Platelet proteome in XLTT

Table 2. Significantly downregulated proteins in X-linked thrombocytopenia with thalassemia platelets and overlap in gray platelet syndrome.

Uniprot ID

IPA ID

P00738 Q5M8T2 P05121 Q14766

HP SLC35D3 SERPINE1 LTBP1

P0C7M8 P02775 Q12912 Q16799 P07996 Q13642 P16109 P09486 Q13201 Q13576 P04275 Q8WXF7 Q9UIB8 Q9NRW1 Q16643 P05067 Q8WXE9 P27338 Q8WWA1 Q8TDZ2 Q5VWC8

CLEC2L PPBP LRMP RTN1 THBS1 FHL1 SELP SPARC MMRN1 IQGAP2 VWF ATL1 CD84 RAB6B DBN1 APP STON2 MAOB TMEM40 MICAL1 HACD4

Q5SQ64

LY6G6F

O95219 O43488 Q7L9L4 P31146 P00390 P47755 P00488 Q04917 P43304

SNX4 AKR7A2 MOB1B CORO1A GSR CAPZA2 F13A1 YWHAH GPD2

P51659 Q9NR12 Q4KMQ2 Q93084

HSD17B4 PDLIM7 ANO6 ATP2A3

O75083 Q14165 P23219 Q96AX2 P16615

WDR1 MLEC PTGS1 RAB37 ATP2A2

Q9C0C9 Q16851

UBE2O UGP2

Protein name

XLTT FC

q-values

Gran*

GPS†

Haptoglobin Solute carrier family 35 member D3 Plasminogen activator inhibitor 1 Latent-transforming growth factor betabinding protein 1 C-type lectin domain family 2 member L Platelet basic protein Lymphoid-restricted membrane protein Reticulon-1 Thrombospondin-1 Four and a half LIM domains protein 1 P-selectin Secreted protein acidic and rich in cystein Multimerin-1 Ras GTPase-activating-like protein IQGAP2 von Willebrand factor Atlastin-1 SLAM family member 5 Ras-related protein Rab-6B Drebrin Amyloid beta A4 protein Stonin-2 Amine oxidase [flavin-containing] B Transmembrane protein 40 Protein-methionine sulfoxide oxidase Very-long-chain (3R)-3-hydroxyacyl-[acylcarrier protein] dehydratase 4 Lymphocyte antigen 6 complex locus protein G6f Sorting nexin-4 Aflatoxin B1 aldehyde reductase member 2 MOB kinase activator 1B Coronin-1A Glutathione reductase, mitochondrial F-actin-capping protein subunit alpha-2 Coagulation factor XIII A chain 14-3-3 protein eta Glycerol-3-phosphate dehydrogenase, mitochondrial Peroxisomal multifunctional enzyme type 2 PDZ and LIM domain protein 7 Anoctamin-6 Sarcoplasmic/endoplasmic reticulum calcium ATPase 3 WD repeat-containing protein 1 Malectin Prostaglandin G/H synthase 1 Ras-related protein Rab-37 Sarcoplasmic/endoplasmic reticulum calcium ATPase 2 Ubiquitin-conjugating enzyme E2 O UTP--glucose-1-phosphate uridylyltransferase

-4.03 -3.36 -2.17 -2.11

0.0319 0.0156 0.0041 0.0028

x x

Down

-2.10 -2.08 -2.01 -2.00 -1.78 -1.77 -1.74 -1.74 -1.69 -1.67 -1.67 -1.63 -1.62 -1.60 -1.59 -1.58 -1.57 -1.57 -1.55 -1.48 -1.45

0.0319 0.0103 0.0312 0.0202 0.0088 0.0285 0.0041 0.0242 0.0220 0.0009 0.0092 0.0202 0.0110 0.0243 0.0364 0.0194 0.0259 0.0088 0.0259 0.0074 0.0335

Down

x x x x x x x x x x x x x x x x x

Down Up Down

-1.45

0.0103

-1.45 -1.44 -1.43 -1.43 -1.41 -1.40 -1.39 -1.37 -1.35

0.0006 0.0333 0.0333 0.0497 0.0220 0.0223 0.0220 0.0061 0.0074

-1.32 -1.31 -1.31 -1.30

0.0220 0.0393 0.0043 0.0103

x x x x

-1.30 -1.29 -1.28 -1.24 -1.24

0.0331 0.0413 0.0460 0.0471 0.0333

x x x x x

-1.23 -1.23

0.0219 0.0241

x x

Down

x x x x x x x

Down Up

Down Down Up Up Up

*Proteins potentially associated with platelet granules, as published before.17,18 †Protein levels elevated or reduced, here defined as fold change [FC] ≥±1.2, in a published gray platelet syndrome (GPS) platelet a-granule fraction proteome.12(Online Supplementary Table S2) XLTT: X-linked thrombocytopenia with thalassemia.

haematologica | 2021; 106(11)

2951

D. Bergemalm et al.

Table 3. Significantly upregulated proteins in X-linked thrombocytopenia with thalassemia platelets and overlap in gray platelet syndrome.

Uniprot ID

IPA ID

P00918 P61247 P49247 Q14166

CA2 RPS3A RPIA TTLL12

P51692

STAT5B

Q8N3F0 Q06830 Q13813

MTURN PRDX1 SPTAN1

Q0ZGT2 Q99447

NEXN PCYT2

Q9UK76

JPT1

P04080 P54727

CSTB RAD23B

P34949

MPI

Q6YHK3 P46783 P48147 Q5T0N5 Q6UX71 P05023

CD109 RPS10 PREP FNBP1L PLXDC2 ATP1A1

Q99733

NAP1L4

P28070 P98082 P49721 Q14392

PSMB4 DAB2 PSMB2 LRRC32

P30085 P26640 P43686 P07814

CMPK1 VARS PSMC4 EPRS

P50454 Q9NTJ5

SERPINH1 SACM1L

Q15257

PTPA

P49327 P08107 Q9NQC3 P25789 P61201

FASN HSPA1A/HSPA1B RTN4 PSMA4 COPS2

Protein name

XLTT FC

q-values

Gran*

Carbonic anhydrase 2 40S ribosomal protein S3a Ribose-5-phosphate isomerase Tubulin--tyrosine ligase-like protein 12 Signal transducer and activator of transcription 5B Maturin Peroxiredoxin-1 Spectrin alpha chain, nonerythrocytic 1 Nexilin Ethanolamine-phosphate cytidylyltransferase Hematological and neurological expressed 1 protein Cystatin-B UV excision repair protein RAD23 homolog B Mannose-6-phosphate isomerase CD109 antigen 40S ribosomal protein S10 Prolyl endopeptidase Formin-binding protein 1-like Plexin domain-containing protein 2 Sodium/potassium-transporting ATPase subunit alpha-1 Nucleosome assembly protein 1-like 4 Proteasome subunit beta type-4 Disabled homolog 2 Proteasome subunit beta type-2 Leucine-rich repeat-containing protein 32 UMP-CMP kinase Valine--tRNA ligase 26S protease regulatory subunit 6B Bifunctional glutamate/proline-tRNA ligase Serpin H1 Phosphatidylinositide phosphatase SAC1 Serine/threonine-protein phosphatase 2A activator Fatty acid synthase Heat shock 70 kDa protein 1A/1B Reticulon-4 Proteasome subunit alpha type-4 COP9 signalosome complex subunit 2

2.91 2.59 2.17 1.94

0.0006 0.0092 0.0202 0.0156

1.86

0.0202

1.79 1.70 1.68

0.0022 0.0335 0.0082

1.60 1.58

0.0317 0.0061

1.55

0.0202

1.54 1.51

0.0202 0.0076

1.48

0.0227

1.47 1.46 1.41 1.41 1.41 1.35

0.0092 0.0223 0.0103 0.0373 0.0103 0.0028

1.35

0.0333

1.35 1.34 1.33 1.33

0.0468 0.0487 0.0227 0.0317

1.32 1.31 1.31 1.29

0.0220 0.0010 0.0357 0.0471

1.29 1.29

0.0497 0.0330

1.27

0.0202

1.26 1.26 1.25 1.21 1.21

0.0103 0.0220 0.0312 0.0333 0.0317

GPS†

*Proteins potentially associated with platelet granules, as published before.17,18 †Protein levels elevated or reduced, here defined as fold change [FC] ≥±1.2, in a published gray platelet syndrome (GPS) platelet a-granule fraction proteome.12(Online SupplementaryTable S2) XLTT: X-linked thrombocytopenia with thalassemia.

2952

haematologica | 2021; 106(11)

Platelet proteome in XLTT

ATP2A2 and ATP2C1 were downregulated in XLTT but upregulated in GPS. One of the jointly upregulated proteins was PRDX2 (FC =2.56, P=0.03 but q=0.15 in XLTT, FC =1.35 in GPS),12 Online Supplementary Table S2 with similar antioxidant functions as PRDX1 (that was only found in XLTT). Upstream regulators, predicted from the dataset using IPA, were then compared between the XLTT and GPS datasets. All proteins with FC ≥±1.2 (compared to the respective controls regardless of statistical significance) were included. The upstream regulators with predicted inhibition and activation, respectively (Z-score ≥±2.0; P<0.05), in XLTT and GPS are presented in Figure 3.

RPTOR independent companion of mTOR complex 2 (aka RICTOR) showed the strongest predicted altered activity in XLTT, with its inhibition predicted mainly by elevated expression of downstream proteasome proteins (Online Supplementary Figure S5A) participating in the protein ubiquitination pathway. In GPS, the X-box binding protein 1 (XBP1, which responds to unfolded protein increases) had the highest absolute Z-score (activated with Z-score 4.02) among upstream regulators (Figure 3; Online Supplementary Figure S5B). VIPAS39, a protein that regulates platelet granule biogenesis,22 had predicted inhibited activity in both XLTT and GPS (Figure 3; Online Supplementary Figure S5A and B).

Figure 2. Analysis of possible protein−protein interaction network using STRING tool. The list of 83 differentially expressed proteins in X-linked thrombocytopenia with thalassemia (XLTT) platelets was subjected to STRING analysis, which found significantly more associations between proteins than would have occurred by chance (P=7.85e-12). Three clusters detected by k-means algorithm were colored as follows: cluster 1, 60 proteins, red (those of the 83 proteins of Table 2 and 3 not included in cluster 2 and 3); cluster 2, 15 proteins, green; cluster 3, eight proteins, cyan. The top three gene ontology (GO) enrichment terms regarding Biological Process given in STRING were for cluster 1: regulated exocytosis, vesicle-mediated transport, and secretion; for cluster 2: platelet degranulation, regulated exocytosis, and vesicle-mediated transport; for cluster 3: protein modification by small protein removal, cellular macromolecular catabolic process, and protein deubiquitination.

haematologica | 2021; 106(11)

2953

D. Bergemalm et al.

Validation of proteomics data Immunohistochemistry We assumed that protein alterations in the platelet proteome could reflect regulations taking place in precursor megakaryocytes in the BM, and performed IHC on two downregulated proteins (see the Online Supplementary Methods). The established semi-quantitative H-score (“histo-score”) method was used for evaluation of megakaryocyte staining intensities. The H-score is obtained by the formula: three-times the percentage of strongly staining cells + twice the percentage of moder-

ately staining cells + the percentage of weakly staining cells, giving a range of 0-300.23 Prostaglandin G/H synthase 1 (PTGS1/COX1, FC =1.28) plays a role in production of the autocrine platelet activator thromboxane A2, important for hemostasis.24 Although megakaryocytes from both controls and XLTT stained for the presence of PTGS1 in the perinuclear region and cytoplasm, there was an almost 50% reduction in median megakaryocyte H-score for cytoplasmic staining intensity in XLTT compared to controls (P=0.012) (Figure 4A). The perinuclear PTGS1 staining

Figure 3. Analysis of upstream regulators in X-linked thrombocytopenia with thalassemia and gray platelet syndrome. Comparison of Ingenuity Pathway Analysis (IPA) core analyses of X-linked thrombocytopenia with thalassemia (XLTT) and gray platelet syndrome (GPS), where fold change (FC) ≥±1.2 compared to controls was used as the only criterion for inclusion of dysregulated proteins. The upstream regulators are sorted by ascending Z-scores in XLTT. The results are filtered to show genes, mRNA and proteins with upstream regulator Benjamini-Hochberg adjusted P<0.05 and absolute Z-score ≥±2. However, when present as a result of comparison with the other group, predicted regulators with Z-score <±2 are marked by dots in the heatmap. Several known/suggested fibrosis regulators including RICTOR, SMAD2, CTGF, TGFB1, FLI1, ERK1/2, PI3K, TGFBR1 and TGFBR2 were predicted to be inhibited in XLTT, and less so or activated in GPS. Contrarily, EGFR and PDGFDD (also involved in fibrosis) were predicted to be activated in XLTT. NBEAL2 (mutated in GPS) showed predicted inhibited activity in XLTT. GATA1 (mutated in XLTT) was predicted to be activated in GPS. IL1, OSM, VIPAS39, RICTOR, HIF1A and F2 were significant predicted inhibited upstream regulators (Z-score ≤ -2) in XLTT when only the 83 proteins with FC ≥±1.2 and q<0.05 were included for the underlying core analysis; no upstream regulator was then predicted to be activated with Z-score ≥2 (not shown).

2954

haematologica | 2021; 106(11)

Platelet proteome in XLTT

was slightly but not significantly reduced in XLTT (numerical data not shown). When BM biopsies from controls and XLTT patients were stained for solute carrier family 35 member D3 (SLC35D3), the second most downregulated protein (FC =-3.40), important for dense granule formation,20 we found a greater than 50% reduction in median megakaryocyte cytoplasmic H-score in XLTT compared to controls (P=0.006) (Figure 4B).

Platelet function testing by flow cytometry As flow cytometry can be used for platelet function testing even at low platelet counts,15,25,26 this method was chosen for testing platelet activation responses ex vivo. Blood samples from XLTT patients II and V from Table 1 were analyzed by flow cytometry for evaluation of platelet function (Figure 5). The common surface markers CD41 (GPIIb) and CD42b (GPIba) were used to identify the platelets. These levels were not quantified, but no marked differences were noticed between healthy donors and patients (not shown). Platelets from patient II showed a markedly low activation response to ADP, thrombin receptor, PAR4-activating peptide (AP) and CRP-XL (cross-linked collagen-related peptide, activates collagen receptor GPVI), both considering the binding of fibrinogen to its receptor and the exposure of the a-granule protein P-selectin upon activation (Figure 5A and B). Only the response to PAR1-AP (platelet thrombin receptor PAR1activating peptide, also known as TRAP) was close to normal. Patient V also showed a low activation response, especially to PAR4-AP and CRP-XL.

For both patients, the decrease of activation in the presence of apyrase was very low (Figure 5C and D), even for the agonists where the primary response was closer to normal. This indicates that the release and contribution to activation by ADP from platelet dense granules was low.14,15 Control experiments were performed with blood from a normal donor diluted to the same platelet count as patient V. In these control experiments, apyrase decreased the platelet activation response considerably for all agonists, showing that released ADP normally can contribute to platelet activation even at these low platelet counts (not shown). No pronounced abnormalities in the capacity for induction of pro-coagulant platelet features (exposure of phosphatidylserine on the cell surface, detected as annexin V binding) upon strong stimulation by a combination of CRP-XL and PAR-activating peptides were observed for any of the patients (Online Supplementary Table S2), although patient V showed results in the lower part of the reference range for normal donors. The platelets from the two patients showed a capacity to expose the lysosomal protein LAMP1 on their surface upon activation, indicating the presence of lysosomes (Online Supplementary Table S2).27 As for P-selectin and fibrinogen receptor activation, the LAMP1 exposure for patient V was relatively low as compared to results in normal donors. This could suggest the presence of fewer than normal lysosomes in XLTT platelets, but no conclusions should be drawn as the platelets showed a generally low activation potential, leading to lower potential of lysosome granule release even if normal numbers of lysosomes were present.

Figure 4. PTGS1/COX1 and SLC35D3 megakaryocyte staining in X-linked thrombocytopenia with thalassemia and controls. (A) Representative immunohistochemical staining of PTGS1/COX1 in bone marrow (BM) megakaryocytes from healthy control (left) and X-linked thrombocytopenia with thalassemia (XLTT) patient (center). Original magnification 60x. The graph (right) shows H-score distribution between controls (n=7) and XLTT (n=4) where the lines represent median H-scores (132 in controls, 72 in XLTT). (B) Representative immunohistochemical staining of SLC35D3 in BM megakaryocytes from healthy control (left) and XLTT patient (center). Original magnification 60x. The graph (right) shows H-score distribution between controls (n=7) and XLTT (n=4) where the lines represent median H-scores (206 in controls, 92 in XLTT).

haematologica | 2021; 106(11)

2955

D. Bergemalm et al.

Figure 5. Examination of platelet activation by flow cytometry. Platelet activation responses were measured as binding of an antibody towards fibrinogen and binding of an antibody towards the a-granule protein P-selectin in response to exposure to platelet agonists specifically activating receptors for ADP, thrombin receptor PAR1 (PAR1-AP), PAR4 (PAR4-AP) and collagen receptor GPVI (CRP-XL). The scatter plots show results for all normal donors (n=26-30) as black circles, while results for patient II are displayed as open squares and for patient V as open circles. (A) Fibrinogen receptor activation. Percentage of platelets binding fibrinogen upon activation. (B) a-granule exocytosis. Percentage of platelets exposing P-selectin upon activation. (C) Dense granule exocytosis: contribution to fibrinogen binding. Percentage decrease in median fluorescence intensity (MFI) for the anti-fibrinogen antibody in the presence of apyrase to degrade ADP released from dense granules. (D) Dense granule exocytosis: contribution to a-granule exocytosis. Percentage decrease in MFI for the anti-P-selectin antibody in the presence of apyrase to degrade ADP released from dense granules.

Discussion Hemostatic platelet functions are largely mediated by soluble factors released from membrane-bound storage organelles including a-granules, dense granules and lysosomes.28 In the present study of XLTT patients, platelet TEM suggested diminished numbers of both dense- and a-granules (Figure 1), in congruence with earlier studies.2,4,10 For an improved understanding of the molecular mechanisms behind the bleeding diathesis in XLTT, we compared the platelet proteome of five patients with five matched controls. In addition to findings of granule content deficiencies, altered protein ubiquitination was thereby suggested in XLTT. Exploration of the cellular effects of the XLTT causing GATA1 mutation in platelets is hampered by their lack of nuclear DNA and transcriptional regulation. Protein translation, however, is continuous throughout the lifespan of platelets and regulated by external and internal signaling.29 Several causes of platelet protein alterations 2956

might exist. Some might be the result of GATA1 mutation p.R216Q induced transcriptional changes, in similarity to the transcriptional dysregulation of the erythropoiesis in XLTT due to defect GATA1, leading to low expression of b-globin chains and thereby the b-thalassemia-like trait.30 Other mechanisms could include altered trafficking of vesicles/granules from megakaryocytes to proplatelets, and changes in vesicle/granule release in vivo (including such occurring after platelet/megakaryocyte activation).31 Some dysregulations of the platelet proteome could be due to absorption/endocytosis of proteins from plasma into circulating platelets, thus reflecting plasma concentrations.32 One example of the latter is haptoglobin (FC =-4.03, q=0.03), which was generally low in XLTT plasma, probably largely due to the continuous thalassemia-like hemolysis. Notably, the altered protein ubiquitination pathway in our data (mostly elevated levels of proteasomal proteins) could imply protein degradation as an important cause and/or effect of granule deficiencies. Increased haematologica | 2021; 106(11)

Platelet proteome in XLTT

amounts of proteasomal proteins have recently been reported also in GFI1B mutated macrothrombocytopenia.31 Some blood analyses may aid in the differential diagnosis of XLTT patients. For example, the high S-TPO in XLTT noted here differs from the reported normal or mildly elevated level in ITP.33 Increased P-TPO levels were found in mice with the hypomorphic Gata1low mutation.34 If the high S-TPO level reflects deficient uptake of TPO by the TPO-receptor MPL in platelets and megakaryocytes of the thrombocytopenic XLTT patients should be further evaluated. MPL is transcriptionally regulated by GATA1,35 and the TPO-MPL axis has crucial effects on platelet production and life span.36 The MPL protein seemed possibly downregulated (FC =-1.84, q=0.06) in XLTT platelets. P-PAI-1 (aka SERPINE1, an inhibitor of fibrinolysis) was low in all XLTT patients, which could contribute to bleeding. P-F5 values were below normal in three of the patients, and in the lower normal range in two. Deficiency of F5 has previously been associated with a modest bleeding diathesis.37 Interestingly, a partially cleaved form of F5 (comprising approximately 20% of the total F5 in blood) resides in the a-granules of platelets, in complex with the protein multimerin (MMRN1).38 Multimerin was downregulated in our proteomics assay (FC =-1.68, q=0.02) whereas the platelet F5 level change was not statistically significant. Multimerin deficiency (found also in the platelet proteome of GPS)12 might be of interest for future evaluations regarding bleeding diathesis.38 The platelet QMS revealed significant (FC ≥±1.2, q<0.05) reductions of 39 proteins associated with granules. These included SERPINE1/PAI-1, vWF, SELP/Pselectin and PTGS1/COX1 which are all important for hemostasis (Table 2). Latent transforming growth factor bbinding protein 1 (LTBP1), the fourth most downregulated protein and found in a-granules, has not previously been associated with bleeding diathesis, but its interactions with transforming growth factor b (TGFB), found in platelets in high concentrations, might be of significance for BM fibrosis development. This has also been discussed for GPS.12 Notably, TGFB1 was identified as a predicted inhibited upstream regulator in XLTT, but appeared somewhat activated in GPS (Figure 3), with possible implications for the respective myelofibrosis developments.39 In XLTT, both the present study (Figure 3 and LTBP1) and our former IHC investigation on BM expression of CTGF and VEGF6 showed low TGFB stimulated protein expressions in XLTT megakaryocytes/platelets. Thrombospondin-1 (THBS1) was downregulated in XLTT platelets, and downregulated also in GPS. THBS1 is a matricellular glycoprotein first discovered in activated platelets. It interacts with a number of ligands and is of significance for, inter alia, inhibition of angiogenesis:40 XLTT BM fibrosis is characterized by increased angiogenesis.6 Though the exact role of THBS1 in hemostasis is unclear, it may interact with coagulation factor 13/F13A141 which was also found in reduced amounts in the XLTT and GPS platelet proteomes. The identification of several jointly downregulated agranule proteins in XLTT and GPS is congruent with ultrastructural similarities regarding deficiencies of agranules. However, NBEAL2 mutated GPS platelets have shown normal morphology and numbers of dense bodhaematologica | 2021; 106(11)

ies/granules,42,43 whereas our XLTT results suggested ultrastructural and functional dense granule deficiency, consistent with an earlier ultrastructural study.10 Possible proteomic correlates to a suggested functional difference between XLTT and GPS regarding dense granule release included contraregulations of the three Ca2+ transporting ATPases ATP2A3, ATP2A244 and ATP2C1,45 downregulated in XLTT (although q=0.057 for ATP2C1) but upregulated in GPS (Online Supplementary Table S1). Deficient dense granule ADP release could be a consequence in XLTT (Figure 5C and D).44,45 The most upregulated protein of our study, CA2 (FC =2.91, q=0.0006), was one of 26 dysregulated proteins included in the “response to stress” gene ontology found significantly enriched in STRING analysis (not shown).21 CA2 participates in several biological processes, including regulations of ion transport and cytosol acidity. CA2 was recently found to predict aspirin resistance in platelet aggregation tests with arachidonic acid.46 Elevated expression of CA2 mRNA was found in Down syndrome-associated acute megakaryoblastic leukemia (AMKL-DS), which harbors GATA1 exon 2 mutations, compared to other AMKL (NCBI GEO2 Accession: GSE4119).47 A recent investigation based on patients with GPS (NBEAL2 mutations) and the GATA1 mutations p.D218G and p.D218Y suggested that GATA1 enhances NBEAL2 expression via interaction with the GATA1 co-activator friend of GATA1 (FOG1),48 possibly explaining the agranule deficiency in the GATA1 mutated patients. However, GATA1 interaction with FOG1 should not be affected in XLTT due to the different mutation localization of p.R216Q which does not alter the FOG1 binding site.8,30 In addition, NBEAL2 protein expression was not altered in XLTT platelets in our study. Thus, the mechanisms for the platelet defects in XLTT must be investigated and evaluated on their own terms. Flow cytometry is the only method available for reliable studies of platelet functional responses at low platelet counts. Both investigated XLTT patients’ platelets showed low reactivity to several platelet agonists (Figure 5), as has similarly been described in GPS.49 We observed a release of a-granules and lysosomes, but to a lower extent compared to normal donors. If this was just a consequence of the low primary reactivity or due to lower numbers or contents of the granules is difficult to ascertain with this method. However, the TEM and proteomics data also strongly suggested a decrease in both a- and dense granules in XLTT platelets, and activation was low even at very high agonist concentrations (not shown). Taken together, this suggests that a lower amount of dense- and a-granules indeed caused the low P-selectin exposure in XLTT platelets. In addition, as fibrinogen is normally stored in platelet a-granules and released upon activation to aid aggregation,29 reduced levels of platelet fibrinogen in XLTT (non-significant by QMS) might contribute to the low fibrinogen binding observed upon platelet activation. For both patients, the dense granule ADP release seemed very deficient, indicating reduced dense granule contents and/or a dense granule release defect.25 The TEM picture in combination with low levels of proteins affecting dense granule biogenesis (SLC35D3)20 and function (ATP2A3, ATP2A244 and ATP2C1)45 could support both mechanisms. 2957

D. Bergemalm et al.

Although investigation of a rare disease such as XLTT implies limited statistical power, 83 dysregulated proteins (FC ≥±1.2, q<0.05) were identified in our proteomics study. Combining the proteomic results analyzed by IPA and STRING with flow cytometry and information from electron microscopy and IHC, several pieces of evidence pointed to dense- and a-granule deficiencies as contributors to platelet functional defects. Impaired dense granule biogenesis and function might differentiate XLTT from GPS, but further studies in additional families are needed. Novel findings suggesting altered protein ubiquitination and degradation should be investigated further in relation to the pathogenesis of platelet granule disorders. Disclosure No conflicts of interest to disclose. Contribution MÅ and JP included patients; DB, CK, CS and JB worked with proteomics; DB and MÅ performed bioinformatics analyses; SR performed flow cytometry and analyzed the results; JP, AGE and MÅ performed and evaluated immunohistochemistry;

References 1. Thompson AR, Wood WG, Stamatoyannopoulos G. X-linked syndrome of platelet dysfunction, thrombocytopenia, and imbalanced globin chain synthesis with hemolysis. Blood. 1977;50(2):303-316. 2. Balduini CL, Pecci A, Loffredo G, et al. Effects of the R216Q mutation of GATA-1 on erythropoiesis and megakaryocytopoiesis. Thromb Haemost. 2004;91(1): 129-140. 3. Hughan SC, Senis Y, Best D, et al. Selective impairment of platelet activation to collagen in the absence of GATA1. Blood. 2005;105(11):4369-4376. 4. Tubman VN, Levine JE, Campagna DR, et al. X-linked gray platelet syndrome due to a GATA1 Arg216Gln mutation. Blood. 2007;109(8);3297-3299. 5. Balduini CL, De Candia E, Savoia A. Why the disorder induced by GATA1 Arg216Gln mutation should be called "X-linked thrombocytopenia with thalassemia" rather than "X-linked gray platelet syndrome". Blood. 2007;110(7):2770-2771; author reply 2771. 6. Åström M, Hahn-Strömberg V, Zetterberg E, Vedin I, Merup M, Palmblad J. X-linked thrombocytopenia with thalassemia displays bone marrow reticulin fibrosis and enhanced angiogenesis: comparisons with primary myelofibrosis. Am J Hematol. 2015;90(3):E44-48. 7. LeinØe E, Zetterberg E, Kinalis S, et al. Application of whole-exome sequencing to direct the specific functional testing and diagnosis of rare inherited bleeding disorders in patients from the Öresund Region, Scandinavia. Br J Haematol. 2017;179(2): 308-322. 8. Ciovacco WA, Raskind WH, Kacena MA. Human phenotypes associated with GATA1 mutations. Gene. 2008;427(1-2):1-6. 9. Gunay-Aygun M, Falik-Zaccai TC, Vilboux T, et al. NBEAL2 is mutated in gray platelet syndrome and is required for biogenesis of platelet alpha-granules. Nat Genet. 2011;43 (8):732-734.

2958

KH performed electron microscopy; DB, MÅ, SR and JP cowrote the manuscript, which was revised and approved by all authors. Acknowledgements The authors would like to thank all participating individuals with XLTT as well as the healthy volunteers. Doctor Anders Fritzberg at the Department of Pediatrics, Falun Hospital, Sweden, included the pediatric patient. Professor Tomas Lindahl at the Department of Clinical Chemistry, and bioinformatics scientist Rada Ellegard at the Department of Clinical Genetics, both at Linköping University, Sweden, carried out the TruSight panel genetic investigation. Inger Vedin (PhD) at the Departments of Medicine and Hematology, Karolinska Institute, Karolinska University Hospital Huddinge, Sweden, performed TPO measurements and immunohistochemistry. Doctor Ian Jones at the Department of Laboratory Medicine, Falun Hospital, Sweden, revised the language. Funding Örebro County Council Research Committee, grant numbers OLL-158661, OLL-164431, OLL-239301, OLL-255251, OLL-268261 and OLL-347141.

10. White JG, Thomas A. Platelet structural pathology in a patient with the X-linked GATA-1, R216Q mutation. Platelets. 2009;20(1):41-49. 11. Yang W, Xiaofan Z, Xiaojuan C, et al. Splenectomy in a child of X-linked thrombocytopenia with thalassemia and bone marrow fibrosis: hemoglobin and platelet count were improved. Blood. 2014;124:1367. http://www.bloodjournal.org/content/124/2 1/1367 12. Maynard DM, Heijnen HF, Gahl WA, Gunay-Aygun M. The alpha-granule proteome: novel proteins in normal and ghost granules in gray platelet syndrome. J Thromb Haemost. 2010;8(8):1786-1796. 13. Danielsson S, Merup M, Olsson L, Palmblad J, Åström M. [X-linked thrombocytopenia with thalassemia in two families in Sweden. Consider hereditary causes of thrombocytopenia and bone marrow fibrosis]. Läkartidningen. 2012;109(34-35):1474-1477. 14. Boknäs N, Ramström S, Faxälv L, Lindahl TL. Flow cytometry-based platelet function testing is predictive of symptom burden in a cohort of bleeders. Platelets. 2018;29(5):512519. 15. Connolly-Andersen AM, Sundberg E, Ahlm C, et al. Increased thrombopoiesis and platelet activation in Hantavirus-infected patients. J Infect Dis. 2015;212(7):1061-1069. 16. Data were analyzed through the use of QIAGEN’s Ingenuity Pathway Analysis (IPA®, QIAGEN Redwood City, CA, USA). https://digitalinsights.qiagen.com 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. Maynard DM, Heijnen HF, Horne MK, White JG, Gahl WA. Proteomic analysis of platelet alpha-granules using mass spectrometry. J Thromb Haemost. 2007;5(9): 1945-1955. 19. Zufferey A, Schvartz D, Nolli S, Reny JL, Sanchez JC, Fontana P. Characterization of the platelet granule proteome: evidence of

the presence of MHC1 in alpha-granules. J Proteomics. 2014;101:130-140. 20. Meng R, Wang Y, Yao Y, et al. SLC35D3 delivery from megakaryocyte early endosomes is required for platelet dense granule biogenesis and is differentially defective in Hermansky-Pudlak syndrome models. Blood. 2012;120(2):404-414. 21. Szklarczyk D, Gable AL, Lyon D, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47(D1):D607-D613. 22. Nurden AT, Nurden P. Should any genetic defect affecting alpha-granules in platelets be classified as gray platelet syndrome? Am J Hematol. 2016;91(7):714-718. 23. Specht E, Kaemmerer D, Sanger J, Wirtz RM, Schulz S, Lupp A. Comparison of immunoreactive score, HER2/neu score and H score for the immunohistochemical evaluation of somatostatin receptors in bronchopulmonary neuroendocrine neoplasms. Histopathology. 2015;67(3):368-377. 24. Santos MT, Valles J, Lago A, et al. Residual platelet thromboxane A2 and prothrombotic effects of erythrocytes are important determinants of aspirin resistance in patients with vascular disease. J Thromb Haemost. 2008;6(4):615-621. 25. Frelinger AL, 3rd, Grace RF, Gerrits AJ, et al. Platelet function tests, independent of platelet count, are associated with bleeding severity in ITP. Blood. 2015;126(7):873-879. 26. Boknäs N, Macwan AS, Södergren AL, Ramström S. Platelet function testing at low platelet counts: when can you trust your analysis? Res Pract Thromb Haemost. 2019;3:(2);285-290. 27. Södergren AL, Svensson Holm AC, Ramström S, Lindström EG, Grenegård M, Ollinger K. Thrombin-induced lysosomal exocytosis in human platelets is dependent on secondary activation by ADP and regulated by endothelial-derived substances. Platelets. 2016;27(1):86-92. 28. Heijnen H, van der Sluijs P. Platelet secretory behaviour: as diverse as the granules ... or

haematologica | 2021; 106(11)

Platelet proteome in XLTT

not? J Thromb Haemost. 2015;13(12):21412151. 29. Weyrich AS, Lindemann S, Tolley ND, et al. Change in protein phenotype without a nucleus: translational control in platelets. Semin Thromb Hemost. 2004;30(4):491-498. 30. Yu C, Niakan KK, Matsushita M, Stamatoyannopoulos G, Orkin SH, Raskind WH. X-linked thrombocytopenia with thalassemia from a mutation in the amino finger of GATA-1 affecting DNA binding rather than FOG-1 interaction. Blood. 2002;100(6): 2040-2045. 31. van Oorschot R, Hansen M, Koornneef JM, et al. Molecular mechanisms of bleeding disorderassociated GFI1B(Q287*) mutation and its affected pathways in megakaryocytes and platelets. Haematologica. 2019;104(7): 1460-1472. 32. Palmblad J. To give and take - life of a platelet. Blood. 2009;113(12):2617. 33. Makar RS, Zhukov OS, Sahud MA, Kuter DJ. Thrombopoietin levels in patients with disorders of platelet production: diagnostic potential and utility in predicting response to TPO receptor agonists. Am J Hematol. 2013;88(12):1041-1044. 34. Zingariello M, Sancillo L, Martelli F, et al. The thrombopoietin/MPL axis is activated in the Gata1(low) mouse model of myelofibrosis and is associated with a defective RPS14 signature. Blood Cancer J. 2017;7(6): e572. 35. Sunohara M, Morikawa S, Fuse A, Sato I. GATA-dependent regulation of TPO-

haematologica | 2021; 106(11)

induced c-mpl gene expression during megakaryopoiesis. Okajimas Folia Anat Jpn. 2014;90(4):101-106. 36. Saur SJ, Sangkhae V, Geddis AE, Kaushansky K, Hitchcock IS. Ubiquitination and degradation of the thrombopoietin receptor c-Mpl. Blood. 2010;115(6):1254-1263. 37. Dahlbäck B. Pro- and anticoagulant properties of factor V in pathogenesis of thrombosis and bleeding disorders. Int J Lab Hematol. 2016;38(Suppl 1):S4-11. 38. Jeimy SB, Fuller N, Tasneem S, et al. Multimerin 1 binds factor V and activated factor V with high affinity and inhibits thrombin generation. Thromb Haemost. 2008;100(6):1058-1067. 39. Malara A, Abbonante V, Zingariello M, Migliaccio A, Balduini A. Megakaryocyte contribution to bone marrow fibrosis: many arrows in the quiver. Mediterr J Hematol Infect Dis. 2018;10(1):e2018068. 40. Huang T, Sun L, Yuan X, Qiu H. Thrombospondin-1 is a multifaceted player in tumor progression. Oncotarget. 2017;8 (48):84546-84558. 41. Dardik R, Solomon A, Loscalzo J, et al. Novel proangiogenic effect of factor XIII associated with suppression of thrombospondin 1 expression. Arterioscler Thromb Vasc Biol. 2003;23(8):1472-1477. 42. Gunay-Aygun M, Falik-Zaccai TC, Vilboux T, et al. NBEAL2 is mutated in gray platelet syndrome and is required for biogenesis of platelet a-granules. Nat Genet. 2011;43(8): 732-734.

43. Bottega R, Pecci A, De Candia E, et al. Correlation between platelet phenotype and NBEAL2 genotype in patients with congenital thrombocytopenia and a-granule deficiency. Haematologica. 2013;98(6):868-874. 44. Feng M, Eliab Z, Borgel D, et al. NAADP/SERCA3-dependent Ca2+ stores pathway specifically controls early autocrine ADP secretion potentiating platelet activation. Circ Res. 2020;127(7): e166-e183. 45. Unsworth AJ, Bombik I, Pinto-Fernandez A, et al. Human platelet protein ubiquitylation and changes following GPVI activation. Thromb Haemost. 2019;119(1):104-116. 46. Jakubowski M, Debski J, SzahidewiczKrupska E, et al. Platelet carbonic anhydrase II, a forgotten enzyme, may be responsible for aspirin resistance. Oxid Med Cell Longev. 2017;2017:3132063. 47. Bourquin JP, Subramanian A, Langebrake C, et al. Identification of distinct molecular phenotypes in acute megakaryoblastic leukemia by gene expression profiling. Proc Natl Acad Sci U S A. 2006;103(9):3339-3344. 48. Wijgaerts A, Wittevrongel C, Thys C, et al. The transcription factor GATA1 regulates NBEAL2 expression through a long-distance enhancer. Haematologica. 2017;102(4):695706. 49. Larocca LM, Heller PG, Podda G, et al. Megakaryocytic emperipolesis and platelet function abnormalities in five patients with gray platelet syndrome. Platelets. 2015;26(8): 751-757.

2959

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(11):2960-2970

Red Cell Biology & its Disorders

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I (CDA-I) identifies defects in differentiation and nucleolar abnormalities Caroline Scott,1 Damien J. Downes,1 Jill M. Brown,1 Robert A. Beagrie,1 Aude-Anais Olijnik,1 Matthew Gosden,1 Ron Schwessinger,1 Christopher A. Fisher,1 Anna Rose,1 David J.P Ferguson,2 Errin Johnson,3 Quentin A. Hill,4 Steven Okoli,5 Raffaele Renella,6 Kate Ryan,7 Marjorie Brand,8 Jim Hughes,1 Noemi B.A. Roy,9,10 Douglas R. Higgs,10 Christian Babbs1 and Veronica J. Buckle1 1

MRC Weatherall Institute of Molecular Medicine, Oxford University, Oxford, UK; Ultrastructural Morphology Group, NDCLS, John Radcliffe Hospital, Oxford, UK; 3Sir William Dunn School of Pathology, Oxford University, Oxford, UK; 4Leeds Teaching Hospital NHS Trust, Leeds, UK; 5Imperial College, The Commonwealth Building, The Hammersmith Hospital, Du Cane Rd, London, UK; 6Pediatric Hematology-Oncology Research Laboratory, CHUV-UNIL Lausanne, Lausanne, Switzerland; 7Department of Hematology, Manchester Royal Infirmary, Manchester, UK; 8Sprott Center for Stem Cell Research, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada; 9Department of Hematology, Oxford University Hospitals NHS Trust, Churchill Hospital, Headington, and NIHR Biomedical Research Centre, Oxford, UK and 10National Institute of Health Research Oxford Biomedical Research Center, Oxford, UK 2

ABSTRACT

T Correspondence: CAROLINE SCOTT caroline.scott@imm.ox.ac.uk VERONICA BUCKLE veronica.buckle@imm.ox.ac.uk Received: May 20, 2020. Accepted: September 17, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.260158

he investigation of inherited disorders of erythropoiesis has elucidated many of the principles underlying the production of normal red blood cells and how this is perturbed in human disease. Congenital dyserythropoietic anemia type 1 (CDA-I) is a rare form of anemia caused by mutations in two genes of unknown function: CDAN1 and CDIN1 (previously called C15orf41), whilst in some cases, the underlying genetic abnormality is completely unknown. Consequently, the pathways affected in CDA-I remain to be discovered. In order to enable detailed analysis of this rare disorder we have validated a culture system which recapitulates all of the cardinal hematological features of CDA-I, including the formation of the pathognomonic ‘spongy’ heterochromatin seen by electron microscopy. Using a variety of cell and molecular biological approaches we discovered that erythroid cells in this condition show a delay during terminal erythroid differentiation, associated with increased proliferation and widespread changes in chromatin accessibility. We also show that the proteins encoded by CDAN1 and CDIN1 are enriched in nucleoli which are structurally and functionally abnormal in CDA-I. Together these findings provide important pointers to the pathways affected in CDA-I which for the first time can now be pursued in the tractable culture system utilized here.

2960

Introduction Many key discoveries in the process of erythropoiesis have come from identification and analysis of individuals with forms of inherited anemia.1-4 In some cases, because of the rarity of the disease and limited access to primary erythroid progenitors and precursors, progress can only be made by developing appropriate models of the disease or, ideally, methods that only require access to peripheral blood. Here we have utilised such an approach to study congenital dyserythropoietic anemia type 1 (CDA-I). CDA-I is a rare autosomal recessive disease associated with ineffective erythropoiesis and macrocytic anemia. Disease severity is commonly mild (not necessitating treatment) to moderate (requiring Interferon or occasional blood transfusion) but patients with severe disease can be transfusion-dependent from birth or even

haematologica | 2021; 106(11)

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I

present as hydrops fetalis.5,6 Light microscopy reveals abnormalities in erythroblast nuclei, including binucleate cells and inter-nuclear bridging.5 A diagnostic feature of CDA-I is the “Swiss-cheese” or spongy pattern of abnormal chromatin in up to 50% of erythroblasts obtained from bone marrow aspirates, visualised using transmission electron microscopy (TEM).6 In ~90% of patients, biallelic mutations in CDAN1 (encoding Codanin-1) or CDIN1 (previously C15orf41) (encoding CDIN1) are causative,7,8 with the genetic basis of the remaining ~10% of patients yet to be determined. Both Codanin-1 and CDIN1 are widely expressed and appear to be essential to life5 but their precise functions are unclear. CDIN1 comprises a helix-turn-helix binding domain and a predicted nuclease domain8 whilst Codanin-1 has sequence similarity with a scaffold protein, CNOT1, involved in mRNA stability and translational control.9 The two proteins form a complex where Codanin-1 is required for stability of CDIN19-11 and both directly interact with the histone chaperone ASF1.12-14 CDA-I is predominantly an erythroid-restricted disease but most of the structural and functional assessments of Codanin-1 and CDIN1 have been performed in non-erythroid cells and some characteristics described for the proteins are not recapitulated in patient-derived erythroblasts.9,11 Here we analyze the distribution and role of these proteins in erythroid cells. Differentiation of CD34+ hematopoietic stem and progenitor cells (HSPC) from peripheral blood has been used to study normal erythropoiesis15-19 and to elucidate disease mechanisms in a number of hematological disorders including Diamond Blackfan anemia (DBA),2,3 hereditary spherocytosis,20 congenital dyserythropoietic anemia type II21 and myelodysplastic syndrome (MDS).16 In order to fully understand the defects that arise in patients who do not generate sufficient mature cells, any culture system must recapitulate terminal erythropoiesis through to enucleation and erythroblasts from controls and patients should be stage matched.22 Here we use an ex vivo threephase culture system15 (broadly expansion, differentiation and enucleation) whereby CD34+ HSPC, obtained from peripheral blood of healthy donors and patients with TEM-confirmed CDA-I, are successfully differentiated into reticulocytes. We use immunophenotyping by fluorescence activated cell sorting (FACS), morphological assessment, single cell proteomics using mass cytometry time of flight (CyTOF) and assay for transposase-accessible chromatin using sequencing (ATAC-seq) to enable direct comparisons of cell populations. This approach successfully recapitulates the pathognomonic feature of ‘spongy’ heterochromatin in patient erythroblasts. We find a delay in terminal erythroid differentiation and increased proliferation of CDA-I erythroblasts, associated with widespread changes in chromatin accessibility. We demonstrate that CDIN1 and Codanin-1 are enriched in nucleoli, which are structurally and functionally abnormal in CDA-I. These findings provide important indicators to the pathways affected in CDA-I, which can now be pursued in the tractable model of erythropoiesis utilised here.

Methods Patient recruitment Subjects were referred for next-generation sequencing

haematologica | 2021; 106(11)

through the Oxford Molecular Diagnostic Center. If a molecular diagnosis of CDA-I was made, patient consent was obtained for entry into this research study approved by the Wales Research Ethics Committee (REC5) (13/WA/0371), with written consent and compliance with the Declaration of Helsinki.

Isolation and differentiation of CD34+ hematopoietic stem and progenitor cells Peripheral blood mononuclear cells were isolated from 50 mL of EDTA anti-coagulated peripheral blood from three healthy donors and ten CDA-I patients using Histopaque. The CD34+ HSPC were extracted with the Human CD34 Microbead Kit (Miltenyi Biotec), according to the manufacturer’s instructions. 1x105 frozen CD34+ HSPC were recovered into Phase I media of a three-phase protocol15 (see Online Supplementary Figure S2A and the Online Supplementary Appendix) and monitored by cytospin (see the Online Supplementary Appendix) and FACS (Online Supplementary Table S1; Online Supplemental Figure S2).

Transmission electron microscopy 5x106 staged intermediate erythroblasts were prepared for TEM as previously described.4,23

Iso-electric focusing 1x106 cultured erythroblasts were analysed by iso-electric focusing (IEF) (see the Online Supplementary Appendix).

Real-time quantitative polymerase chain reaction for globins RNA was extracted using a Tri-reagent protocol and real-time quantitative polymerase chain reaction (RT-qPCR) conducted with commercial TaqMan assays (Online Supplementary Table S2).

Chromatin accessibility and NF-E2 chromatin immunoprecipitation sequencing ATAC-seq was performed as previously described.24,25 NF-E2 chromatin immunoprecipitation sequencing (ChIP-seq) was conducted on 5x106 day 10 erythroblasts with previously described modifications26 using rabbit anti-NFE2 (4 mg, sc-22827X; Santa Cruz discontinued). For ATAC-seq library preparation and analysis see the Online Supplementary Appendix.

Antibody labeling, barcoding and mass cytometry for mass cytometry time of flight Samples were prepared and analysed for CyTOF as previously described27 (Online Supplementary Table S6). Subsampled events were concatenated for uniform manifold approximation and projection (UMAP).28 See the Online Supplementary Appendix.

Data availability Sequencing data generated for this work is available on the Gene Expression Omnibus (GSE125175).

Immunofluorescence Cells were washed and processed essentially as described previously.29 See the Online Supplementary Appendix.

Fluorescence in situ hybridization Fluorescence in situ hybridization (FISH) probes used were p7.1 (covering most of the rDNA array) and BAC CT476834 (demarcating perinucleolar heterochromatin) and were kindly gifted by Prof B. McStay.30 Probes were labeled with Cy3-dUTP (GE Healthcare) or indirectly with digoxygenin-11-dUTP (Roche).31 See the Online Supplementary Appendix.

2961

C. Scott et al.

Figure 1. Differentiating erythroblasts from healthy donors and CDA-I patients are broadly equivalent by immunophenotyping. (A) Representative fluorescence activated cell sorting profiles of cultured erythroblasts from a healthy donor, CDA-I patient with a mutation in CDIN1 and CDA-I patient with a mutation in CDAN1. (B) Uniform manifold approximation and projection (UMAP) plots showing mass cytometry time of flight (CyTOF) data from healthy donors (n=3) and CDIN1-patient derived erythroblasts (n=3) at day 11 of differentiation for the erythroid markers CD235 (glycophorin A), CD36 (Scavenger receptor), CD71 (Transferrin receptor) and transcription factor Gata1. There is no difference in the clustering patterns observed between healthy donors and CDIN1 patients for any of the 25 markers tested (Online Supplementary Table S6).

Ethynyl uridine labeling and analysis RNA transcripts were labeled by 5ethynyl uridine (EU) incorporation (1 mM for 30 minutes or 2 hours) and detected by click chemistry with Alexa488 azide using Click-iT RNA imaging Kit (ThermoFisher Scientific). Quantitation of EU was performed using FIJI.32 All images were acquired using standardized settings and maximum-intensity projected. Mean EU intensities were quantitated using a nuclear mask demarcated by DAPI.

cell distribution width (RDW) and a reduced red cell count (RBC) compared to healthy donors (Online Supplementary Figure S1B). In one patient (UPID6) with CDA-I, confirmed by TEM, a potentially pathogenic homozygous variant was identified in CDAN1 although the allele frequency for this mutation is >1% in specific populations. Data from this patient was included in the CDAN1 mutation group.

CDA-I patients

Establishing a suitable model system using peripheral blood-derived CD34+ hematopoietic stem and progenitor cells

To date, approximately 60 mutations have been reported in CDAN1 and CDIN1,5,7,8,33 and six mutational hotspots have been identified in the CDAN1 gene.10 There are often differences in the severity of the disease between individuals, even for those with identical mutations.34 We examined erythropoiesis in ten CDA-I patients (Online Supplementary Figure S1A; Online Supplementary Table S8). These patients (excluding those receiving regular blood transfusion or venesections) have hemoglobin (Hb) levels and mean cell volumes (MCV) within the normal range (Online Supplementary Figure 1B), consistent with ~30% of clinical cases6 but tend to have higher mean cell hemoglobin (MCH), an increased red

We initially validated a three-phase ex vivo culture system15 for differentiation of CD34+ HSPC obtained from the peripheral blood of healthy donors (Online Supplementary Figures S2 and S3). In addition to the morphological assessment of cultured erythroblasts (Online Supplementary Figure S2B), we characterized their chromatin landscape, globin gene expression profile and the expression of erythroid proteins and transcription factors, to comprehensively evaluate differentiation status (Online Supplementary Figure S4). Immunophenotyping revealed the expected gain of glycophorin A (CD235a) and transferrin receptor (CD71), which typically occurred by day 7 (Online Supplementary Figure S2C). As maturation pro-

Results

2962

haematologica | 2021; 106(11)

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I

Figure 2. Erythroblasts from CDA-I patients display delayed differentiation and increased proliferation. (A) Cell morphology counts from cytospins on day 10 (n=101348), day 13 (n=92-607) and day 17 (n=63-329) of erythroblasts divided into proerythroblast (Pro), basophilic (Baso), polychromatic (Poly), orthochromatic (Ortho) and enucleated (Enuc). Data are shown as mean ± standard deviation. Statistical significance was tested using Mann-Whitney with a Benjamani-Hochberg adjustment where Q=0.05. CDIN1 patients and CDAN1 patients were tested separately against healthy donors for each timepoint. CDIN1 *P=0.0095 at day 17 for both polychromatic and enucleated erythroblasts. CDAN1 *P=0.0087 for day 10 basophilic erythroblasts and **P=0.0043 for day 10 polychromatic erythroblasts. (B) Representative cytospin images stained with modified Wright’s stain (magnification 40x) for healthy donors, CDIN1 and CDAN1 patients at days 10, 13 and 17, with marked examples of cell types scored for (A). (C) Proliferation of cultured erythroblasts from healthy donors (n=6), patients with mutations in CDIN1 (n=3) and patients with mutations in CDAN1 (n=5) showing increased proliferation in both patient cohorts. All scores are for viable cells only (see Online Supplementary Figure S10), normalised to 100,000 cells at day 5. Data are shown as mean ± standard error of the mean. Dashed vertical lines denote the three culture phases. Statistical significance was tested using Mann-Whitney with a Benjamani-Hochberg adjustment where Q=0.05. CDIN1 patients and CDAN1 patients were tested separately against healthy donors for each timepoint. CDIN1 **P=0.0043 at day 7 and at day 17. CDAN1 *P=0.0087 at day 6 and **P=0.0043 at day 17. Cell counts from two patients were not scored in a comparable manner for this analysis.

gressed, cells visibly hemoglobinized by day 10 (Online Supplementary Figure S2D) and this coincided with increasing expression of the adult globins (Online Supplementary Figure S4A), with the a- to b-globin ratios remaining around one throughout differentiation (Online Supplementary Figure S4B). Iso-electric focusing (IEF) confirmed that predominantly adult globin was produced (Online Supplementary Figure S4C), and ATAC-seq showed open chromatin at the HBA1/2, and HBB genes and their associated locus control regions, again indicative of definitive erythropoiesis (Online Supplementary Figure S4D to E). During the third phase of culture, erythroblasts underwent the normal final stages of differentiation, with increased levels of Band 3 (CD233) and loss of the adhesion protein a-4 integrin (CD49d) (Online Supplementary Figure S2C). Enucleated cells were visible on cytospins at this stage (Online Supplementary Figure S2B).

haematologica | 2021; 106(11)

Differentiating erythroblasts from healthy donors and CDA-I patients are broadly equivalent by immunophenotyping We studied the differentiation of erythroblasts derived from CD34+ HSPC from CDA-I patients with a variety of mutations in CDIN1 and CDAN1 (Online Supplementary Figure S1A). Flow cytometry bulk population analysis showed that differentiation of CDA-I patient HSPC appeared to be equivalent to that of the healthy donors with loss of CD34+ and gain of erythroid markers CD71 and CD235 from day 7, and expected changes in CD36, CD49d and Band 3 occurring from day 10. (Figure 1A, and for gating strategy see Online Supplementary Figure S3). We also analysed the erythroblast immunophenotype by CyTOF, a next-generation flow cytometry platform that allows functional and phenotypic characterisation of cell populations.35 We examined the levels of 25 erythroid

2963

C. Scott et al.

Figure 3. Evidence of defective differentiation in patient erythroblasts, most severe for CDIN1 mutations. (A) Transmission electron microscopy at day 11 of healthy and CDA-I cultured erythroblasts showing the diagnostic phenotype of abnormal chromatin in patients. Inset shows enlarged area to illustrate the pattern of euchromatin and heterochromatin and how this is disrupted in CDA-I patients. (B) Percentage of nuclei ± standard deviation with spongy heterochromatin at day 11, determined from large field images. Numbers of nuclei scored were 96-436 per individual. *P=0.0189 with Kruskal-Wallis test. (C) Fluorescence activated cell sorting histograms (gated on all viable single cells in the CD235a+ population (nucleated and enucleated)) of Band 3 intensity at day 17 (healthy donors, n=6; CDIN1 patients, n=3; and CDAN1 patients, n=4). (D) Median fluorescence intensity (MFI) ± standard deviation of Band 3-FITC at day 17 (**P=0.0088 and *P=0.0127 with Kruskal-Wallis test).

transcription factors and cell surface markers (Online Supplementary Table S6) at day 11 in healthy donor and CDIN1 patient-derived erythroblasts. Visualization by UMAP, widely used to identify distinct cell populations in cytometry data,28 revealed that erythroblasts from both groups follow a continuous trajectory during differentiation with a clear separation of cells expressing, for example, low levels of CD235 (denoted in blue) from those in the tail which have high expression of CD235 (Figure 1B). Cells from healthy donors and CDIN1 patients appear similarly distributed along the trajectory of differentiation at day 11 for all 25 markers analysed by CyTOF (Figure 1B; Online Supplementary Figure S5) and no significant differences in the clustering patterns were identified by Kmeans clustering analysis.36 Together with the FACS data (Figure 1A), this indicates that the patient and healthy donor samples cannot be distinguished by day 10/11 on the basis of immunophenotype.

from healthy donors than CDIN1 patients, whose erythroblasts were still delayed at the polychromatic stage, with CDAN1 erythroblasts having progressed a little further. Therefore CDA-I patients exhibit clear delay in their erythroid differentiation. In order to firmly establish whether this represents a delay or a block, cultures would need to be continued beyond day 17 to look for persistence of precursor forms. We also observed greater expansion in patient erythroblast numbers compared with healthy donors (Figure 2C) and this was significant even at the end of the expansion phase by day 6/7 (CDAN1, P=0.0087; CDIN1, P=0.0043). The increase in viable cell counts for patient erythroblasts became especially marked in the later phase of culture, reaching significance at day 17 of P=0.0043 for both patient groups.

Erythroblasts from CDA-I patients display delayed differentiation with increased proliferation

TEM revealed that the pattern of chromatin abnormalities characteristic of CDA-I was present in patient erythroblasts by day 11 of culture. This feature was observed in all our patient samples, averaging 29% (±7.7 standard deviation) of nuclei affected (Figures 3A and B; Online Supplementary Figure S7A). Furthermore, elevated expression of growth differentiation factor 15 (GDF15), a marker of ineffective erythropoiesis known to be increased in CDA-I patients,37 was detected at day 10 by immunofluorescence (IF) (Online Supplementary Figure S7B). Therefore, ex vivo differentiation of CDA-I erythroblasts successfully

Despite bulk population immunophenotyping indicating that patient and healthy donor cells are grossly stage matched in expansion and into differentiation, analysis of the proportions of morphologically-classed erythroid cells identified differences in the progression through differentiation, detectable from day 10 (Figure 2A and B). At this time point, significantly more erythroblasts from healthy donors than CDAN1 patients had reached the polychromatic stage. By day 17, there were more enucleated cells 2964

Evidence of defective differentiation in patient erythroblasts, most severe for CDIN1 mutations

haematologica | 2021; 106(11)

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I

Figure 4. ATAC-seq analysis reveals the emergence of an altered regulatory landscape in patient-derived erythroblasts. (A) Principal component analysis (PCA) of assay for transposase-accessible chromatin using sequencing (ATAC-seq) from healthy donors (n=6) CDIN1 patients (n=3) and CDAN1 patients (n=4). The distribution of cells along PC1 follows differentiation stage and PC2 distinguishes patients and healthy donors. (B) MA plot for differential expression analysis (DESeq2) comparison of ATAC-seq from healthy donors and CDA-I patients at day 10 and 13 of ex vivo differentiation with significantly different peaks (q<0.01) highlighted as either more accessible in patients (red - up) or less accessible in patients (blue - down). (C) Comparison of chromatin state annotations26 for differentially accessible peaks shows enrichment for enhancers in less accessible peaks. Strong and weak refers to the level of H3K27ac signal. (D) MEME motif discovery identified a motif matching that of NF-E248,60 as being significantly enriched (E-value <10-50) in ATAC-seq peaks that were less accessible in patients.

recapitulates abnormal cellular phenotypes observed in bone marrow derived erythroblasts and these features are already apparent midway through terminal differentiation. We next assessed the effects of patient mutations on the enucleation stage of differentiation. Firstly, analysis of cellular morphology indicated a persistence of erythroid precursors in CDA-I cultures, particularly in those from patients with CDIN1 mutations, together with a significant reduction in the percentage of enucleated cells (Figure 2A). Secondly, immunophenotyping of cultured erythroblasts in the enucleation phase revealed changes in Band 3 expression in CDA-I patients. In normal erythropoiesis Band 3 shows a marked increase from the pro-erythrobhaematologica | 2021; 106(11)

lasts to late erythroblasts,16 and while CDA-I patient erythroblasts did progressively gain Band 3, the level of protein was significantly less at day 17 than in healthy donors (Figure 3C and D). This supports other indications of a delay in the progression of differentiation in patientderived erythroblasts. Notably the Band 3 reduction was more severe in patients with CDIN1 mutations (n=3, P=0.0088) than in the CDAN1 mutant cells (n=4, P=0.0127) (Figure 3D).

Differentiating erythroblasts from CDA-I patients display an altered regulatory landscape Analysis of open chromatin regions can be used to distinguish cell types and deconvolve mixed cell popula2965

C. Scott et al.

Figure 5. Nucleolar structure is disrupted in CDA-I patients. (A) Nucleoli detected with probe BAC-CTD (green) surrounding rDNA probe p7.1 (red) in healthy nuclei. In example patients UPID20 (CDAN1 mutation) and UPID25 (CDIN1 mutation) this order is disrupted. (B) Nucleolar output, judged by 5 ethynyl uridine (EU) labeling, is significantly reduced in two CDA-I patients with CDAN1 mutations, each compared with a healthy donor in the same experiment. (C) Despite the disrupted nucleolar structure, CDIN1 and Codanin-1 proteins (green) continue to associate with nucleolar proteins UBF and Fibrillarin (red) respectively, in two example CDA-I patients with CDAN1 mutations. All analyses are on day 10/11 cultured erythroblasts, using DAPI counterstain (blue).

2966

haematologica | 2021; 106(11)

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I

tions;18,38 therefore to further compare our differentiating healthy and patient cells we assayed chromatin accessibility at day 10 and day 13 using ATAC-seq.24 We mapped healthy and patient cell populations against a differentiation trajectory of sorted erythropoietic cell populations and 136,698 nucleosome-depleted regions,18,38 using principal component analysis (PCA) of all open chromatin regions (Online Supplementary Figure S6). This mapping could not distinguish patient material from healthy donors on the differentiation trajectory. Next we undertook PCA of our cultured erythroblast data alone (without plotting against other erythroid populations), where a distinction in the accessibility profile between healthy donor and patient samples became apparent (Figure 4A), and this was more marked by day 13. We looked at pooled day 10 and day 13 ATAC-seq data by differential expression analysis (DESeq2) for differences in DNA accessibility peaks not attributable to differentiation status (Figure 4B). There were 61 peaks displaying increased accessibility in patients and 531 less accessible sites. The latter showed a marked enrichment for enhancers (strong and weak) (65%) (Figure 4C) when assessed for chromatin state annotations.26 Furthermore, 40% of the 531 less accessible sites in CDA-I patients have a binding motif for the NF-E2 family of transcription factors (Figure 4D). Consistent with motif distribution, the mean level of NF-E2 binding in ChIP-seq from healthy donor erythroblasts and K562 erythroleukemic cells was significantly higher than background at the peaks with patient-specific decreased accessibility (CDA-I down) (Online Supplementary Figure S8A), indicating that NF-E2 normally binds these sites. Any difference in NF-E2 binding could not be attributed to altered protein abundance as similar levels of NF-E2 were detected in healthy donors and CDIN1 patients by CyTOF mass cytometry (Online Supplementary Figure S8B). Similarly, no differences were seen in abundance for either MAFG, which dimerizes with NF-E2,39 or BACH1 which competitively binds with the NF-E2 motif.40,41 Of note, the Band 3 encoding gene Slc4a1 has erythroidspecific enhancer elements (Online Supplementary Figure S8C). The 5' enhancer has a binding motif for NF-E2 and is bound by NF-E2 in ChIP analyses of three different erythroblast cultures so that reduced accessibility at this site could account for reduced levels of Band 3 observed in CDA-I patients. The decrease in accessibility at this specific site did not quite reach significance in patient erythroblasts however the ATAC was performed at day 10 and day 13 which is possibly too early to observe an effect for this gene.

The structure of nucleoli is disrupted in CDA-I patients With multiple strands of evidence for an altered pattern of differentiation in CDA-I patients, we looked for abnormal features that might be linked to the affected proteins. We have previously shown both Codanin-1 and CDIN1 endogenous proteins to be enriched in erythroblast nucleoli.10 Both proteins also show nucleolar enrichment in HEK293T (human embryonic kidney cell line), G-292 (human osteocarcinoma line), mES E14 (mouse embryonic stem cells) and B16F10 (mouse melanoma cell line) (data not shown), indicating this is a common feature across a range of cell types. We therefore examined nucleolar structure in day 10/11 erythroblasts by FISH using probes haematologica | 2021; 106(11)

detecting the heterochromatic region adjacent to rDNA arrays (BAC-CTD), and the rDNA arrays themselves (p7.1),30 where the heterochromatin normally surrounds the rDNA signal within nucleoli. In all four patients examined (UPID10, 20, 22 and 25), nucleoli in a proportion of nuclei appeared more numerous, less ordered and less regular in shape, and the rDNA arrays appeared less open (Figure 5A). This was observed at a timepoint when 19-29% of nuclei in these patients exhibited abnormal chromatin distribution by TEM. Such disrupted organization might be expected to impact on the synthesis of ribosomal RNA, which accounts for a major portion of RNA synthesis in the cell.41,42 We therefore assessed RNA synthesis in two patients by measuring incorporation of the uridine analogue 5 ethynyl uridine (EU) into newly synthesised RNA42-44 at day 10/11 of differentiation. In both cases there is a significant reduction in nuclear EU labeling (Figure 5B) despite only a percentage of nuclei apparently affected and, particularly in UPID15, there is a distinct cell population with low EU signal. We have previously shown for four patients that the two mutated proteins are not destabilized by missense and in-frame mutations and remain detectable.10 Therefore, we investigated whether the normal enrichment of Codanin-1 and CDIN1 in nucleoli is also disrupted, in patients with predicted non-destabilizing mutations. Using IF with day 10 erythroblasts derived from CDAN1 patients UPID 6, 16, 20 and 22, in combination with nucleolar proteins UBF and Fibrillarin, we observed that the disrupted appearance of nucleoli recapitulated that observed by FISH (Figure 5C). Further, we were able to detect that both CDIN1 and the mutant Codanin-1 remained associated with nucleolar proteins in patient erythroblasts (Figure 5C).

Using the culture system to validate novel variants in CDA-I patients The pathogenicity of novel CDIN1 or CDAN1 variants identified by sequencing requires further evidence, such as chromatin abnormalities in bone marrow biopsies by TEM. TEM is not only relatively inaccessible but often requires a second bone marrow biopsy.5 By contrast, peripheral blood is usually accessible and was used here in the ex vivo culture system to confirm the diagnosis of CDA-I in an infant with two novel mutations in CDAN1 (UPID33) (Online Supplementary Figure S1).10 Genetic analysis was conducted on the patient, who presented with unexplained anemia, using the Oxford Red Cell Panel (ORCP)45 (Online Supplementary Figure S9). Following identification of two novel variants, CD34+ HSPC from UPID33 were extracted from peripheral blood and after 11 days in ex vivo culture, TEM on the resulting intermediate erythroblasts revealed 39% of erythroblasts with abnormal chromatin morphology, thus confirming diagnosis of CDA-I.

Discussion Although understanding the cellular and molecular basis of CDA-I has the potential to elucidate new insights into the process of erythropoiesis, research is constrained by the limited studies that can be conducted using primary erythroid progenitors and precursors derived from patients with this condition. Here, using a modified exvivo culture system, we demonstrate that healthy control 2967

C. Scott et al.

erythroblasts pass through the expected stages of differentiation with appropriate expression of erythroid cell surface markers16 and are able to enucleate. Further, we recapitulate the cardinal hematological features of CDA-I and show by TEM that up to 40% of patient-derived erythroblasts have spongy heterochromatin, indicating that ex vivo culture can be used to elucidate mechanisms underlying this disease. We staged the cultures using an array of methods including FACS and CyTOF, which rely on immunophenotyping cell populations. While such methods showed healthy and diseased erythroblasts were immunophenotypically similar during the expansion phase and then into differentiation (day 10-13), aspects of disordered erythropoiesis were already evident at this mid-differentiation timepoint. In particular at this stage we noted a delay in progression through differentiation together with an increased proliferation of CDA-I erythroid precursors, producing increased amounts of GDF15, characteristic of dyserythropoiesis.37,46 Many of these cells exhibited the abnormal chromatin structure associated with CDA-I. These findings show that the effects of the mutant proteins start to operate early in terminal differentiation, indeed the viable cell counts would suggest that there may already be effects by the end of the expansion phase at day 6/7. Chromatin accessibility has become a superior approach for cell type classification, including hematopoietic lineages.38 The accessibility of transcription factor motifs within chromatin changes as subsets of regulatory elements are systematically activated and repressed during commitment to different lineages.47 Therefore, we used ATAC–seq as a genome-wide method to stage cell populations. When healthy donor and patient material from day 10 and day 13 cultures were aligned with a defined ATAC-seq erythroid trajectory,38 both map as expected with the intermediate and later stages of erythropoiesis. A more detailed PCA revealed a distinction between healthy donors and patients, more marked by day 13, identified a reduced accessibility in patient erythroblasts at gene enhancers containing the erythroid-specific NF-E2 motif. NF-E2 (comprising NFE2-p45 and MAFG) and BACH1 (which binds the same motif) are important transcription factors for erythropoiesis and the oxidative stress pathway respectively.39-41,48 CyTOF analysis indicates that levels of these three proteins appear to be normal in patient cells at day 11. NF-E2 motifs normally increase accessibility towards late erythropoiesis,18 in parallel with the level of the protein complex,49,50 and we show that in normal erythroblasts, NF-E2 does bind those sites. Together these facts suggest that reduced accessibility of this motif at enhancers could affect the later stages of erythroid differentiation. It is possible that reduced accessibility of this motif reflects a generally altered regulatory landscape due to delayed differentiation, however motifs for other erythroid-specific transcription factors such as Gata1 did not reach significance in terms of altered accessibility. The observation of disrupted nucleoli raises interesting ideas about the roles the two proteins may play in erythropoiesis and could explain the erythroid-specific nature of the disease. Mutations to the ribosomal proteins themselves can underlie tissue restricted disorders, including erythroid-specific disorders such as Diamond-Blackfan syndrome, Schwachmann-Diamond syndrome, 2968

Dyskeratosis Congenita and MDS.43,51 Impaired function in the nucleolus could affect the number of available ribosomes and have similar effects to these other conditions in producing anemia. Beyond that, the nucleolus appears to have other, regulatory roles.43,52,53 Of interest, given the importance of cell cycle described above, is the proposed role for the nucleolus in cell cycle regulation.52 Another possibility is that CDIN1, with its sequence similarity to the Holliday junction resolvase family of proteins, could function in a repair pathway. The high transcription rate within nucleoli can lead to topological stress and double strand breaks53 whilst partial deletion of rDNA arrays has been shown to cause disordered nucleolar structure.54 Further work is required to test these possibilities. The final stages of erythropoiesis involve nuclear condensation prior to expulsion of the pyknotic nuclei by enucleation55,56 and this process is highly organized50 with chromatin condensation playing an important role.57 The abnormal spongy heterochromatin observed in CDA-I could have a significant impact on the usual processes that precede enucleation, such as the selective loss of histones.55 Remarkably, a substantial number of erythroblasts progress to enucleation without developing the catastrophic changes in chromatin compaction and organization apparent in spongy nuclei. This implies that the effects of the aberrant proteins must reach a threshold within individual cells to produce the pathognomonic phenotype and could be related to the balance between euchromatin and heterochromatin under nucleolar regulation.54,58 Two distinct types of CDA-I have been reported (CDAIa MIM 224120 and CDA-Ib MIM 615631)59 based on the levels of Hb and MCV, with the CDA-Ib patients (caused by CDIN1 mutations) thought to be more severely affected. In our patient cohort (excluding those regularly transfused or venesected), there is overlap between the blood indices irrespective of the mutation (Online Supplementary Figure S1). However, we observe a more pronounced delay in differentiation, increased proliferation and significantly reduced levels of Band 3 expression in erythroblasts cultured from CDIN1 patients, as compared to those with mutations in CDAN1. This implies that there may indeed be a distinction based on patient genotype where the CDA-I phenotype is more severe when arising from CDIN1 mutations. In this study we provide a detailed characterization of CDA-I erythroblasts. We recapitulate aspects of the disease pathology seen in CDA-I, including high levels of cells with spongy heterochromatin and increased GDF15 expression. We report that CDA-I patient erythroblasts have elevated levels of proliferation, together with delay in the differentiation process and reduced levels of enucleation. There are difficulties in identifying and quantifying abnormalities in this disorder since only a proportion of erythroblasts exhibit defects whilst the majority differentiate and manage to function as red cells in many patients. Further, nurturing culture conditions may diminish the abnormal phenotypes observed.21 Nevertheless, ATACseq analysis provides clear evidence of an altered regulatory landscape during terminal differentiation. This, together with the observations of aberrant nucleolar structure and transcriptional output, gives insight into the underlying disease mechanism and highlights several new avenues for further investigation of the functional role of the two proteins in erythroid differentiation. haematologica | 2021; 106(11)

Recapitulation of erythropoiesis in congenital dyserythropoietic anemia type I

Disclosures No conflicts of interest to disclose. Contributions CS and DJD extracted CD34+ cells from CDA-I patient blood and normal donors respectively; CS, DJD, MG performed the experiments with help from AAO, RB, DJD and RS undertook the ATAC analysis; DJPF and EJ performed the electron microscopy; JMB conducted the immunofluorescence, FISH and EU labeling and analyzed the images; CF and AR conducted the IEF; MB conducted the CyTOF and the data was analyzed by RB, QAH, SO, RR, KR and NR were the clinicians responsible for the care of the of the CDA-I patients; CS, DJD, JH, CB and VJB conceived and designed experiments; DRH provided conceptual advice and clinical oversight; DJD and VJB created the figures; CS, DRH and VJB wrote the paper. All authors reviewed and critically edited the manuscript. Acknowledgments We thank the CDA-I patients for providing blood samples. We acknowledge the flow cytometry facility at the WIMM for providing cell analysis services and technical expertise, supported by the MRC HIU; MRC MHU (MC_UU_12009); NIHR Oxford BRC; Kay Kendall Leukemia Fund (KKL1057), John Fell Fund

References 1. Migliaccio AR, Varricchio L. Concise Review: advanced cell culture models for Diamond Blackfan anemia and other erythroid disorders. Stem Cells. 2018; 36(2):172-179. 2. Moniz H, Gastou M, Leblanc T, et al. Primary hematopoietic cells from DBA patients with mutations in RPL11 and RPS19 genes exhibit distinct erythroid phenotype in vitro. Cell Death Dis. 2012 3(7):e356. 3. O'Brien KA, Farrar JE, Vlachos A, et al. Molecular convergence in ex vivo models of Diamond-Blackfan anemia. Blood. 2017; 129(23):3111-3120. 4. Renella R, Roberts NA, Brown JM, et al. Codanin-1 mutations in congenital dyserythropoietic anemia type 1 affect HP1alpha localization in erythroblasts. Blood. 2011; 117(25):6928-6938. 5. Roy NBA, Babbs C. The pathogenesis, diagnosis and management of congenital dyserythropoietic anaemia type I. Br J Haematol. 2019;185(3):436-449. 6. Wickramasinghe SN. Congenital dyserythropoietic anaemias: clinical features, haematological morphology and new biochemical data. Blood Rev. 1998;12(3):178200. 7. Dgany O, Avidan N, Delaunay J, et al. Congenital dyserythropoietic anemia type I is caused by mutations in codanin-1. Am J Hum Genet. 2002;71(6):1467-1474. 8. Babbs C, Roberts NA, Sanchez-Pulido L, et al. Homozygous mutations in a predicted endonuclease are a novel cause of congenital dyserythropoietic anemia type I. Haematologica. 2013;98(9):1383-1387. 9. Swickley G, Bloch Y, Malka L, et al. Characterization of the interactions between Codanin-1 and C15Orf41, two proteins implicated in congenital dyserythropoietic anemia type I disease. BMC Mol Cell Biol. 2020;21(1):18. 10. Olijnik AA, Roy NBA, Scott C, et al.

haematologica | 2021; 106(11)

(131/030 and 101/517), the EPA fund (CF182 and CF170) and by the WIMM Strategic Alliance awards G0902418 and MC_UU_12025. We also acknowledge the Electron Microscopy Facility at the Sir William Dunn School of Pathology for conducting the majority of the TEM and Raman Dhaliwal for help with sample preparation and imaging. Funding Further support came from grants to the Wolfson Imaging Center Oxford (Wolfson Foundation 18272, joint MRC/BBSRC/EPSRC MR/K015777X/1, Wellcome Trust Multi-User Equipment 104924/Z/14/Z). We would like to acknowledge Giorgio Napolitani and Michalina Mazurczyk for help in the mass cytometry facility at the WIMM, providing technical expertise and cell analysis services. The facility is supported by the MRC HIU core funded project MC_UU_00008 and the Oxford Single Cell Biology Consortium (OSCBC). This work was supported by the charity Congenital Anemia Network (CAN)(UK charity no. 1176864), Blood Buddies (UK charity no. 1108692), the Medical Research Council MC_UU_00016/1, a Wellcome Trust Strategic Award (106130/Z/14/Z) and the National Institute for Health research (NIHR) Oxford Biomedical Research Center Hematology Theme at Oxford University Hospitals NHS Trust and Oxford University.

Genetic and functional insights into CDA-I prevalence and pathogenesis. J Med Genet. 2021;58(3):185-195. 11. Shroff M, Knebel A, Toth R, Rouse J. A complex comprising C15ORF41 and Codanin-1- the products of two genes mutated in congenital dyserythropoietic anemia type I (CDA-I). Biochem J. 2020; 477(10):1893-1905. 12. Noy-Lotan S, Dgany O, Lahmi R, et al. Codanin-1, the protein encoded by the gene mutated in congenital dyserythropoietic anemia type I (CDAN1), is cell cycleregulated. Haematologica. 2009;94(5):629637. 13. Ask K, Jasencakova Z, Menard P, Feng Y, Almouzni G, Groth A. Codanin-1, mutated in the anaemic disease CDAI, regulates Asf1 function in S-phase histone supply. EMBO J. 2012;31(8):2013-2023. 14. Ewing RM, Chu P, Elisma F, et al. Largescale mapping of human protein-protein interactions by mass spectrometry. Mol Syst Biol. 2007;3:89. 15. Griffiths RE, Kupzig S, Cogan N, et al. Maturing reticulocytes internalize plasma membrane in glycophorin A-containing vesicles that fuse with autophagosomes before exocytosis. Blood. 2012; 119(26):6296-6306. 16. Hu J, Liu J, Xue F, et al. Isolation and functional characterization of human erythroblasts at distinct stages: implications for understanding of normal and disordered erythropoiesis in vivo. Blood. 2013; 121(16):3246-3253. 17. Mettananda S, Clark K, Fisher CA, SloaneStanley JA, Gibbons RJ, Higgs DR. Phenotypic and molecular characterization of a serum-free miniature erythroid differentiation system suitable for high-throughput screening and single-cell assays. Exp Hematol. 2018;60:10-20. 18. Ludwig LS, Lareau CA, Bao EL, et al. Transcriptional states and chromatin accessibility underlying human erythropoiesis. Cell Rep. 2019;27(11):3228-3240.

19. Satchwell TJ, Hawley BR, Bell AJ, Ribeiro ML, Toye AM. The cytoskeletal binding domain of band 3 is required for multiprotein complex formation and retention during erythropoiesis. Haematologica. 2015; 100(1):133-142. 20. Satchwell TJ, Bell AJ, Hawley BR, et al. Severe Ankyrin-R deficiency results in impaired surface retention and lysosomal degradation of RhAG in human erythroblasts. Haematologica. 2016;101(9):10181027. 21. Satchwell TJ, Pellegrin S, Bianchi P, et al. Characteristic phenotypes associated with congenital dyserythropoietic anemia (type II) manifest at different stages of erythropoiesis. Haematologica. 2013;98(11):17881796. 22. Ulirsch JC, Lareau C, Ludwig LS, Mohandas N, Nathan DG, Sankaran VG. Confounding in ex vivo models of Diamond-Blackfan anemia. Blood. 2017; 130(9):1165-1168. 23. Riffelmacher T, Clarke A, Richter FC, et al. Autophagy-dependent generation of free fatty acids is critical for normal neutrophil differentiation. Immunity. 2017;47(3):466480. 24. Buenrostro JD, Wu B, Litzenburger UM, et al. Single-cell chromatin accessibility reveals principles of regulatory variation. Nature. 2015;523(7561):486-490. 25. Hay D, Hughes JR, Babbs C, et al. Genetic dissection of the alpha-globin superenhancer in vivo. Nat Genet. 2016;48(8): 895-903. 26. Downes DJ, Schwessinger R, Hill SJ, et al. An integrated platform to systematically identify causal variants and genes for polygenic human traits. bioRxiv. 2020;813618. 27. Palii CG, Cheng Q, Gillespie MA, et al. Single-Cell Proteomics reveal that quantitative changes in co-expressed lineage-specific transcription factors determine cell fate. Cell Stem Cell. 2019;24(5):812-820. 28. Becht E, McInnes L, Healy J, et al. Dimensionality reduction for visualizing

2969

C. Scott et al. single-cell data using UMAP. Nat Biotechnol. 2019;37:38–44. 29. Brown JM, Leach J, Reittie JE, et al. Coregulated human globin genes are frequently in spatial proximity when active. J Cell Biol. 2006;172(2):177-187. 30. Floutsakou I, Agrawal S, Nguyen TT, Seoighe C, Ganley AR, McStay B. The shared genomic architecture of human nucleolar organizer regions. Genome Res. 2013;23(12):2003-2012. 31. Brown J, Saracoglu K, Uhrig S, Speicher MR, Eils R, Kearney L. Subtelomeric chromosome rearrangements are detected using an innovative 12-color FISH assay (M-TEL). Nat Med. 2001;7(4):497-501. 32. Schindelin J, Arganda-Carreras I, Frise E, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods. 2012; 9(7):676-682. 33. Heimpel H, Schwarz K, Ebnother M, et al. Congenital dyserythropoietic anemia type I (CDA I): molecular genetics, clinical appearance, and prognosis based on longterm observation. Blood. 2006;107(1):334340. 34. Heimpel H, Kellermann K, Neuschwander N, Hogel J, Schwarz K. The morphological diagnosis of congenital dyserythropoietic anemia: results of a quantitative analysis of peripheral blood and bone marrow cells. Haematologica. 2010;95(6):1034-1036. 35. Kay AW, Strauss-Albee DM, Blish CA. Application of mass cytometry (CyTOF) for functional and phenotypic analysis of natural killer cells. Methods Mol Biol. 2016;1441:13-26. 36. MacQueen J. Some methods for classification and analysis of multivariate observations. Proc Fifth Berkley Sympon Math Statist and Prob. 1967;1:281-297. 37. Tamary H, Shalev H, Perez-Avraham G, et al. Elevated growth differentiation factor 15 expression in patients with congenital dyserythropoietic anemia type I. Blood. 2008;112(13):5241-5244. 38. Corces MR, Buenrostro JD, Wu B, et al. Lineage-specific and single-cell chromatin accessibility charts human hematopoiesis

2970

and leukemia evolution. Nat Genet. 2016;48(10):1193-1203. 39. Blank V, Kim MJ, Andrews NC. Human MafG is a functional partner for p45 NF-E2 in activating globin gene expression. Blood. 1997;89(11):3925-3935. 40. Oyake T, Itoh K, Motohashi H, et al. Bach proteins belong to a novel family of BTBbasic leucine zipper transcription factors that interact with MafK and regulate transcription through the NF-E2 site. Mol Cell Biol. 1996;16(11):6083-6095. 41. Sun J, Brand M, Zenke Y, Tashiro S, Groudine M, Igarashi K. Heme regulates the dynamic exchange of Bach1 and NF-E2related factors in the Maf transcription factor network. Proc Natl Acad Sci U S A. 2004;101(6):1461-1466. 42. Jao CY, Salic A. Exploring RNA transcription and turnover in vivo by using click chemistry. Proc Natl Acad Sci U S A. 2008; 105(41):15779-15784. 43.Bohnsack KE, Bohnsack MT. Uncovering the assembly pathway of human ribosomes and its emerging links to disease. EMBO J. 2019;38(13):e100278. 44. Warner JR. The economics of ribosome biosynthesis in yeast. Trends Biochem Sci. 1999;24(11):437-440. 45. Roy NB, Wilson EA, Henderson S, et al. A novel 33-Gene targeted resequencing panel provides accurate, clinical-grade diagnosis and improves patient management for rare inherited anaemias. Br J Haematol. 2016; 175(2):318-330. 46. Tanno T, Noel P, Miller JL. Growth differentiation factor 15 in erythroid health and disease. Curr Opin Hematol. 2010; 17(3):184-190. 47. Heuston EF, Keller CA, Lichtenberg J, et al. Establishment of regulatory elements during erythro-megakaryopoiesis identifies hematopoietic lineage-commitment points. Epigenetics Chromatin. 2018;11(1):22. 48. Andrews NC, Kotkow KJ, Ney PA, Erdjument-Bromage H, Tempst P, Orkin SH. The ubiquitous subunit of erythroid transcription factor NF-E2 is a small basicleucine zipper protein related to the v-maf

oncogene. Proc Natl Acad Sci U S A. 1993;90(24):11488-11492. 49. Gillespie MA, Palii CG, Sanchez-Taltavull D, et al. Absolute quantification of transcription factors reveals principles of gene regulation in erythropoiesis. Mol Cell. 2020;78(5):960-974. 50. Gautier EF, Ducamp S, Leduc M, et al. Comprehensive proteomic analysis of human erythropoiesis. Cell Rep. 2016; 16(5):1470-1484. 51. Nakhoul H, Ke J, Zhou X, Liao W, Zeng SX, Lu H. Ribosomopathies: mechanisms of disease. Clin Med Insights Blood Disord. 2014;7:7-16. 52. Andersen JS, Lyon CE, Fox AH, et al. Directed proteomic analysis of the human nucleolus. Curr Biol. 2002;12(1):1-11. 53. Tsekrekou M, Stratigi K, Chatzinikolaou G. The nucleolus: in genome maintenance and repair. Int J Mol Sci. 2017;18(7):1411. 54. Paredes S, Maggert KA. Ribosomal DNA contributes to global chromatin regulation. Proc Natl Acad Sci U S A. 2009; 106(42): 17829-17834. 55. Zhao B, Mei Y, Schipma MJ, et al. Nuclear condensation during mouse erythropoiesis requires Caspase-3-mediated nuclear opening. Dev Cell. 2016;36(5):498-510. 56. Baron MH, Barminko J. Chromatin condensation and enucleation in red blood cells: an open question. Dev Cell. 2016;36(5):481482. 57. Ney PA. Normal and disordered reticulocyte maturation. Curr Opin Hematol. 2011; 18(3):152-157. 58. Guetg C, Lienemann P, Sirri V, et al. The NoRC complex mediates the heterochromatin formation and stability of silent rRNA genes and centromeric repeats. EMBO J. 2010;29(13):2135-2146. 59. Gambale A, Iolascon A, Andolfo I, Russo R. Diagnosis and management of congenital dyserythropoietic anemias. Expert Rev Hematol. 2016;9(3):283-296. 60. Daftari P, Gavva NR, Shen CK. Distinction between AP1 and NF-E2 factor-binding at specific chromatin regions in mammalian cells. Oncogene. 1999;18(39):5482-5486.

haematologica | 2021; 106(11)

ARTICLE

Red Cell Biology & its Disorders

The interactome of the N-terminus of band 3 regulates red blood cell metabolism and storage quality

Ferrata Storti Foundation

Aaron Issaian,1 Ariel Hay,2 Monika Dzieciatkowska,1 Domenico Roberti,3 Silverio Perrotta,3 Zsuzsanna Darula,4 Jasmina Redzic,1 Micheal P. Busch,5 Grier P. Page,6 Stephen C. Rogers,7 Allan Doctor,7 Kirk C. Hansen,1 Elan Z. Eisenmesser,1 James C. Zimring2 and Angelo D’Alessandro1 Department of Biochemistry and Molecular Genetics, University of Colorado Denver – Anschutz Medical Campus, Aurora, CO, USA; 2University of Virginia, Charlottesville, VA, USA; 3Università della Campania "L. Vanvitelli", Naples, Italy; 4Laboratory of Proteomics Research, Biological Research Center, Szeged, Hungary; 5Vitalant Research Institute, San Francisco, CA, USA; 6RTI International, Pittsburgh, PA, USA and 7University of Maryland, Baltimore, MD, USA 1

Haematologica 2021 Volume 106(11):2971-2985

ABSTRACT

and 3 (anion exchanger 1; AE1) is the most abundant membrane protein in red blood cells, which in turn are the most abundant cells in the human body. A compelling model posits that, at high oxygen saturation, the N-terminal cytosolic domain of AE1 binds to and inhibits glycolytic enzymes, thus diverting metabolic fluxes to the pentose phosphate pathway to generate reducing equivalents. Dysfunction of this mechanism occurs during red blood cell aging or storage under blood bank conditions, suggesting a role for AE1 in the regulation of the quality of stored blood and efficacy of transfusion, a life-saving intervention for millions of recipients worldwide. Here we leveraged two murine models carrying genetic ablations of AE1 to provide mechanistic evidence of the role of this protein in the regulation of erythrocyte metabolism and storage quality. Metabolic observations in mice recapitulated those in a human subject lacking expression of AE1 (band 3 Neapolis), while common polymorphisms in the region coding for AE1 correlate with increased susceptibility to osmotic hemolysis in healthy blood donors. Through thermal proteome profiling and crosslinking proteomics, we provide a map of the red blood cell interactome, with a focus on AE1 and validate recombinant AE1 interactions with glyceraldehyde 3-phosphate dehydrogenase. As a proof-of-principle and to provide further mechanistic evidence of the role of AE1 in the regulation of redox homeostasis of stored red blood cells, we show that incubation with a cell-penetrating AE1 peptide can rescue the metabolic defect in glutathione recycling and boost post-transfusion recovery of stored red blood cells from healthy human donors and genetically ablated mice. 1-11

1-56

Correspondence: ANGELO D’ALESSANDRO angelo.dalessandro@ucdenver.edu JAMES C ZIMRING jcz2k@virginia.edu Received: December 24, 2020. Accepted: February 26, 2021. Pre-published: May 13, 2021. https://doi.org/10.3324/haematol.2020.278252

Introduction Band 3, also known as anion exchanger 1 (AE1) because of its role in the exchange of chloride for bicarbonate anions,1 is one of the main targets of oxidant stress during red blood cell (RBC) aging in vivo and in the blood bank.2,3 AE1 also regulates RBC metabolism to allow adaptation to hypoxia and oxidant stress.4 The N-terminal cytosolic domain of AE1 serves as a docking site for deoxygenated hemoglobin, with significantly higher affinity than for oxygenated hemoglobin.5,6 Based on this background, it has been proposed that hemoglobin, in addition to its central role of carrying oxygen from the lungs to peripheral tissues, may serve as the “oxygen sensor” that directs AE1 to regulate metabolism appropriately. Indeed, studies had provided compelling, albeit indirect evidence that, when not encumbered by deoxy-hemoglobin, the N-terminus of AE1 can bind to and, in so doing, inhibit the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH).7-11 Notably, RBC host approximately 106 copies of both AE1 and GAPDH.4 A wealth of indirect evidence supports

haematologica | 2021; 106(11)

2971

A. Issaian et al.

the likelihood and biological relevance of AE1-GAPDH interactions, based on data from immunofluorescencebased experiments,7,8 enzymatic activity assays, metabolic flux analysis via nuclear magnetic resonance (NMR),9 analytical ultracentrifugation,10 in silico prediction,11 and genetic ablation of the N-terminus of band 3 in mice.12 However, early crosslinking studies had hitherto failed to produce direct evidence of an interaction of GAPDH with the N-terminal residues.7 Nevertheless, a widely accepted model was proposed, according to which AE1 serves as a “railway switch” diverting glucose down different tracks (glycolysis vs. the pentose phosphate pathway [PPP]) depending on cellular needs.8,13,14 This model explains why, when oxidant stress is high and GAPDH is bound to AE1 and thereby inhibited, RBC favor glucose oxidation via the PPP, to generate the reducing cofactor NADPH and fuel related antioxidant systems;15 on the other hand, RBC would rely on glycolysis when oxidant stress is low (e.g., at high altitude16), hemoglobin is deoxygenated and binds to the N-terminus of AE1, which in turn favors the release of glycolytic enzymes from the membrane to promote the generation of energy in the form of ATP and NADH. Regulation of the balance between glycolysis and PPP is essential in order for RBC to be able to respond to their particular metabolic needs (membrane pump function, cytoskeleton and lipid homeostasis), which are a function of whether the RBC are exposed to high or low oxygen tensions in the lungs or peripheral capillaries.17 We and others have proposed that this balance may be dysregulated as RBC age and/or become damaged as a result of pathology (including SARS-CoV-2 infection18) or iatrogenic intervention (e.g., blood storage),19,20 depriving RBC of critical metabolic plasticity and leading to their demise.3 So far, the role of AE1 in the quality of stored RBC has only been hypothesized, based on a body of evidence that includes: (i) an increase in AE1 fragmentation as a function of oxidant stress in stored RBC;21-25 (ii) a progressive loss of the capacity to activate the PPP in stored RBC,26-28 despite the gradual inhibition of GAPDH as a function of its oxidation at the active site Cys152 and 156, His179;29 (iii) an increase in band 3 phosphorylation at tyrosine (Y) residues as a function of storage,30 which correlates with the release of microparticles from stored RBC; (iv) disruption of GAPDH localization in the membrane upon phosphorylation of Y8 and 21 of AE1;1,11,31 (v) a poorer storage quality and post-transfusion recovery of RBC from donors who are incapable of activating the PPP because of deficient glucose 6-phosphate dehydrogenase (G6PD) activity;32,33 and (vi) changes in AE1 oligomeric structure correlate with the loss of phospholipid asymmetry in stored RBC, which could affect survival upon transfusion.3,34 Despite this evidence, the role of AE1 in the quality of stored blood has not yet been elucidated mechanistically.

Methods The methods are detailed extensively in the Online Supplementary Methods.

later), originally generated by Chu et al.,35 were acquired from the National Institutes of Health mouse embryo repository and were bred with C57BL/6 females. The use of Ubi-GFP and HOD mice has been described in prior publications by our group.36 Whole blood was drawn by cardiac puncture as a terminal procedure for the mice.

Band 3 Neapolis RBC (100 mL) were obtained from a subject carrying a mutation resulting in the lack of AE1 amino acids 1-11, as extensively described by Perrotta et al.31

REDS-III RBC Omics Details of the Recipient Epidemiology and Donor assessment Study (REDS)-III RBC Omics, including the patients’ enrollment, blood processing, osmotic hemolysis and development of the transfusion medicine genome-wide genotyping array have been described extensively in prior studies on the background of the project.33,37,38

Methylene blue treatment, red blood cell storage and tracing experiments RBC from wild-type (WT) and band 3 knockout (KO) mice were incubated with methylene blue (100 mM, Sigma Aldrich) at 37°C for 1 h, as described elsewhere.32 RBC from the four main mouse strains investigated in this study, along with 13 different mouse strains described before,39 were collected, processed, stored, and transfused and post-transfusion recovery was determined as previously described.39 Tracing experiments were conducted with labeled glucose, citrate, glutamine, arachidonic acid and methionine. RBC (100 mL) from all the mouse strains investigated in this study or RBC lysates from healthy donor volunteers (n=3) or the individual carrying band 3 Neapolis RBC were incubated at 37ºC for 1 h in the presence of stable isotope-labeled substrates, as described previously.23

Metabolomics and proteomics Metabolomics analyses were performed using a Vanquish ultrahigh performance liquid chromatograph (UHPLC) coupled online to a Q Exactive mass spectrometer (Thermo Fisher, Bremen, Germany), as described previously.40,41 Proteomics analyses, thermal proteome profiling and crosslinking experiments were performed via FASP digestion and nanoUHPLC tandem mass spectrometry identification (Thermo Fisher), as previously described.42,43 and extensively detailed in the Online Supplementary Methods.

Isothermal titration calorimetry All isothermal titration calorimetry binding experiments were performed with a MicroCal iTC (Cytiva) set at 25°C. GAPDH and band 3 proteins were dialyzed into matching buffer (20 mM Bis-Tris pH 6.5, 50 mM NaCl, 2 mM DTT). The cell contained GAPDH at 0.1 mM while the syringe contained band 3 at 1 mM. Reference power was set to 10 µcal/s with a constant stirring speed of 1,000 rpm. In total, 20 injections were performed. The first injection was excluded from the data analysis. Experiments were performed in duplicate and the results were analyzed with the Origin ITC module. 200

Nuclear magnetic resonance and structural models Animal studies with mice All the animal studies described in this manuscript were reviewed and approved by the University of Virginia Institutional Animal Care and Use Committee (protocol n: 4269). Band 3 mouse founders, including huB3, HA Del and BS KO mice (see

2972

15 N-heteronuclear single quantum coherence (HSQC) spectra were collected at 25ºC for the recombinantly expressed band 3 peptide 1-56 in the presence of 100 and 200 mM GAPDH. Data were collected on a Varian 900 using a standard 15N-HSQC sequence.

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

Results Genetic ablation of the N-terminus of band 3 affects metabolism and glycolysis to pentose phosphate ratios in stored red blood cells Despite similar function, the amino acid sequence of AE1 at the N-terminus is poorly conserved between mice and humans6 (Figure 1A). Low’s group generated a knockin mouse by inserting the human binding site (amino acids 135) into the mouse AE1, resulting in a human-mouse hybrid AE1 (HuB3) that could be used for a reductionist analysis of the human binding motif.5 Two additional mouse models were subsequently made: (i) the first one carries a deletion of the first 11 amino acid residues - hereon referred to as high affinity deletion or HA Del, since deletion of these residues negatively affects binding of glycolytic enzymes, as suggested by immunofluorescence assays,7,8 enzymatic activity assays, metabolic flux analysis via NMR,9 analytical ultracentrifugation10, and in silico modeling;11 (ii) the second model is characterized by a deletion of amino acid residues 12-23, which results in loss of affinity for the binding of deoxygenated hemoglobin (hereon referred to as binding site knockout or BS KO)35 (Figure 1B). Based on this model (Online Supplementary Figure S1A-C), we hypothesized that blood from HA Del and BS KO mice would store poorly, because of the incapacity to activate the PPP as a result of the lack of residues 1-11 or 12-23, respectively, of the N-terminus of AE1 (Figure 1A, B). Importantly, genetic ablation of the AE1 N-terminus in these mice recapitulates the storage-induced fragmentation of the AE1 N-terminus at multiple residues in human RBC (Online Supplementary Figure S1D). RBC from WT, HuB3, HA Del and BS KO mice were stored under refrigerated conditions for 12 days, prior to metabolomics and unsupervised multivariate analyses (Figure 1C, D; Online Supplementary Table S1). The latter highlighted a significant impact of genotype on the end-ofstorage levels of metabolites involved in glycolysis, the PPP and glutathione homeostasis (Figure 1E; Online Supplementary Figure S2A, B). To further investigate the impact on glycolysis and the PPP, RBC from these four mouse strains were incubated with 1,2,3-13C3-glucose upon stimulation with methylene blue (MB) (Figure 1F). MB is metabolized into leukomethylene blue in a reaction that consumes NADPH and thus promotes PPP activation by law of mass action.32 The levels of lactate isotopologues 2,313 C and 1,2,3-13C – and their ratios – are indicative of activation of the PPP and glycolysis, respectively, since the first carbon atom of 1,2,3-13C -glucose is lost in the form of 13CO in the oxidative phase of the PPP (Figure G-I, Online Supplementary Figure S2C). Activation of the PPP following MB stimulation was comparable in WT, HuB3 and BS KO mouse RBC, but ablated in HA Del mice; the defect in PPP activation of HA Del mice was rescued, while BS KO mouse RBC had higher basal levels and further increases in PPP activation by MB when incubated with the AE1 peptide (Figure 1I; Online Supplementary Figure S2C). 2

1-56

Red blood cells from mice lacking the N-terminus of band 3 have poorer post-transfusion recovery Fresh and stored RBC from HA Del or BS KO mice also showed increased levels of carboxylic acids (succinate, fumarate, malate), acyl-carnitines (hexanoyl-, decanoyl-, dodecenoyl-, tetradecenoyl- and hexadecanoyl-carnitine) and lipid peroxidation products compared to WT and HuB3 mouse RBC, especially with regard to metabolites of the haematologica | 2021; 106(11)

arachidonic and linoleic acid pathways (Figure 2A-C). Tracing experiments with labeled glutamine, citrate and methionine confirmed the steady-state aberrations in AE1 KO mice following MB challenge, while incubation with a recombinant AE1 peptide only rescued glutaminolysis in HA Del mice and malate generation in BS KO mice (Figure 2A-C). Expanding on previous targeted metabolomics studies,36,39 untargeted metabolomics analyses on fresh and stored RBC from 13 different mouse strains identified metabolic correlates of post-transfusion recovery (Figure 2D-G), a Food and Drug Administration gold standard for storage quality. Pathway analyses and top ranked correlates are shown in Figure 2E-G, including previously reported oxylipins,36,39 but also purine oxidation products (xanthine), short chain and long chain saturated fatty acids, as well as carboylic acids as top negative correlates (Figure 2F, G; Online Supplementary Table S1). All of these metabolites accumulated at higher levels in stored RBC from HA Del and BS KO mice compared to WT and HuB3 mice. Consistently, determination of end-of-storage post-transfusion recovery of RBC from the four mouse strains showed significant decreases in posttransfusion recovery in the HA Del and BS KO mice (63.2±3.1% and 35.4±5.8%, respectively compared to 88.9±5.6% for WT and 85.5±3.6% for HuB3 mouse RBC) (Figure 2H). 1-56

Lack of the 11 N-terminal amino acids of AE1 in humans results in red blood cell metabolic alterations comparable to those in AE1 knockout mice In 2005, Perrotta and colleagues31 identified a son of a consanguineous marriage with severe anemia, which was associated with a deficiency of AE1 expression (~10% of normal levels). Characterization of the gene coding for AE1 (SLC4A1) revealed a single base substitution (T-->C) at position +2 in the donor splice site of intron 2, resulting in the generation of a novel mutant protein that lacked residues 1-11 (band 3 Neapolis) (Figure 3A). The HA Del mice in this study recapitulate, at least in part, band 3 Neapolis RBC in humans. Consistent with observations in HA Del mice, “omics” analyses of band 3 Neapolis RBC compared to control RBC showed significant alterations of glycolysis, glutathione homeostasis, decreases in methylgroup donors for isoaspartyl damage repair,23 increases in carboxylic acid, and altered arginine and polyamine metabolism (Figure 3B; Online Supplementary Figure S3A, B). Incubation of band 3 Neapolis RBC lysates with 1,2,3-13C glucose confirmed a deficit in the capacity to activate the PPP following stimulation with MB, a defect that was corrected by rescue with a recombinant AE1 peptide (Figure 2C), similar to HA Del RBC. Proteomics analyses showed higher levels of hexokinase, but lower levels of all the remaining glycolytic enzymes downstream to it, as well as of enzymes of the PPP and glutathione synthesis in the RBC from the band 3 Neapolis individual (Online Supplementary Figure S4C), despite comparable mean corpuscular value, mean corpuscular hemoglobin and hematocrit (Online Supplementary Table S1). Alterations in RBC metabolism in band 3 Neapolis RBC comparable to those observed in AE1 KO mice were confirmed not just at steady state (e.g., higher levels of lipid oxidation products) (Figure 3D), but also by tracing experiments with 13C15N-glutamine; 13C-citrate, 13C15N-methionine, and deuterium-labeled arachidonate (Online Supplementary Figure S3D-G). 3

1-56

2973

A. Issaian et al.

Single nucleotide polymorphisms in the region coding for AE1 correlate with increased osmotic fragility of stored human red blood cells 1-56

Mutations targeting the N-terminus of AE1 (e.g., AE1 , band 3 Neapolis) in humans result in significant hemolysis, requiring blood transfusion.31 Therefore, humans carrying equivalent mutations to those used in our mechanistic mouse models would not be eligible to donate blood. The region coding for AE1 (gene name SLC4A1) is highly polymorphic in humans. AE1 polymorphisms were captured in the genetic array of over 879,000 single nucleotide polymorphisms designed for the Recipient Epidemiology 1-11

and Donor assessment Study (REDS) III-RBC Omics.37 The REDS III RBC Omics enrolled a large cohort of 13,806 healthy donor volunteers in four different blood centers across the USA.33 Units from these donors were stored for up to 42 days prior to determination of the RBC susceptibility to hemolysis following osmotic insults.38 Genomewide association studies revealed that the highly polymorphic region on chromosome 17 coding for AE1 ranked among the top correlates with osmotic fragility (Figure 3E). While the most significant single nucleotide polymorphisms associated with greater osmotic hemolysis were intronic variants, three polymorphisms were identified in

Figure 1. Continued on the following page.

2974

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

Figure 1. The N-terminus of band 3 controls red blood cell metabolism during storage under blood bank conditions. (A) Human and mouse N-terminal band 3 sequences are slightly different. (B) We leveraged humanized AE1 mice,12 and compared them to wild-type (WT) mice (C57BL6), or humanized mice lacking residues 1-11 (deletion of the high affinity binding region for glycolytic enzyme:12 HA Del) or 12-23 (hemoglobin binding site knockout: BS KO). (C) Red blood cells (RBC) from these mice were stored under refrigerated conditions for 12 days, prior to metabolomics analyses. (D, E) Storage and genotypes had a significant impact on RBC metabolism, as determined by unsupervised principal component analysis (D) and hierarchical clustering analysis of significant metabolites by analysis of variance (E). Specifically, strain-specific differences in fresh and stored RBC were noted in glycolysis, the pentose phosphate pathway (PPP) and glutathione homeostasis. (F) To further investigate the impact on glycolysis and the PPP, RBC from the four mouse strains were incubated with 1,2,3-13C -glucose upon stimulation with methylene blue (MB). (G, H) Rescue experiments were also performed by incubating RBC with a recombinantly expressed N-terminus AE1 peptide (residues 1-56), prior to determination of the ratios of lactate isotopologues +3 and +2, deriving from glycolysis and the PPP, respectively. (I) MB activated the PPP in all mouse strains except for the HA Del mice. Supplementation of the AE1 peptide increased PPP activation and decreased glycolysis in HA Del mice and further exacerbated responses in all the other strains.± 3

1-56

Figure 2. Continued on the following page.

haematologica | 2021; 106(11)

2975

A. Issaian et al.

Figure 2. Metabolic alterations in band 3 knockout mice correlate with poor post-transfusion recovery. (A-C) Analyses of fresh and stored red blood cells (RBC) from wild-type (WT) (C57BL6) mice [green], humanized band 3 mice (HuB3) [white] and mice lacking amino acids 1-11 (HA Del) [yellow] or 12-23 (BS KO) [orange] of band 3 showed strain-specific differences in several pathways in fresh and stored RBC. Above all, stored RBC from the KO mice were characterized by increased levels of carboxylic acid and lipid peroxidation products (A), especially metabolites of the arachidonate metabolism (B), including prostaglandins, eicosanoids, hydroxyeicosatetraenoates (HETE) and hydroxyoctadecenoates (HODE) (C); the full, vectorial version of this list is provided in Online Supplementary Table S1). (D-F) Metabolomics analyses on fresh and stored RBC from 13 different mouse strains (D) highlight these metabolites as significant correlates with poor post-transfusion recovery (E), as highlighted by the metabolite set enrichment analysis in (F) and the ranked top positive and negative correlates with post-transfusion recovery reported in (G). (H) Determination of end-of-storage post-transfusion recovery of RBC from the four mouse strains showed significant decreases in recovery in the HA Del and BS KO mice.

the blood donor population as D38A, E40K and K56E. These single nucleotide polymorphisms had been previously reported to be associated with hereditary spherocytosis (band 3 Foggia and Neapolis II),44,45 a condition associated with mild to moderate hemolysis in eligible blood donors,46 especially when heterozygous for these traits. The D38A allele was rare in Europeans (4% of the donor population) and relatively more common in east Asians (8%), while the K56E allele was common in donors of African ancestry (10%). Although low, these allele frequencies appear to be relevant to the field of transfusion medicine when considering the over 100 million units donated every year around the world.

2976

Proteomics and structural studies on the interactome of band 3 After confirming the lack of the N-terminal portion of AE1 in the KO mice (Online Supplementary Figure S4E-G), we characterized the proteomes of RBC from the four mouse strains, with a focus on protein levels that were similarly altered in both HA Del and BS KO mice, or uniquely altered in either KO strain (Online Supplementary Figure S4HK). The list of proteins whose levels were decreased in band 3 KO mice included numerous proposed interactors of AE1 (in red in Online Supplementary Figure 4H-K, Online Supplementary Table S1). This observation could be explained by altered protein expression in erythroid cells

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

Figure 3. Metabolic impact of AE1 deletion in human red blood cells and alterations of osmotic fragility of stored red blood cells from donors with polymorphisms in AE1 . (A, B) Lack of amino acid residues 1-11 of AE1 in humans, known as band 3 Neapolis, (A) results in increased activation of glycolysis and decreases in glutathione pools (B). (C) Tracing experiments with 1,2,3-13C -glucose showed impaired responses to methylene blue (MB)-induced activation of the pentose phosphate pathway (PPP) compared to glycolysis in band 3 Neapolis, a phenotype that is partially rescued in vitro by supplementation of a recombinant AE1 peptide. (D) However, band 3 Neapolis RBC were also characterized by higher levels of oxylipins, comparable to those observed in stored RBC from mice lacking AE1 amino acids 12-23 (BS KO). (E) Genome-wide association studies on 13,806 healthy donor volunteers revealed increased osmotic fragility in subjects carrying a polymorphism in the region coding for band 3 (gene name SLC4A1) residues 1-56 and neighboring introns. CTRL: control; GSH: reducd glutathione; GSSG: oxidized glutathione; HETE: hydroxyeicosatetraenoates; HODE: hydroxyoctadecenoates; GWAS: genome-wide association studies, SNP: single nucleotide polymorphisms. 1-11

1-56

during maturation to RBC, or could be indicative of a protective role from proteolytic degradation by interaction of these proteins with the N-terminus of AE1, as proposed in other studies of the RBC degradome.47 To test the latter hypothesis, we used thermal proteome profiling coupled to tandem mass tag 10 (TMT10) (Figure 4A) to determine candidate interacting partners to AE1 . Candidate interactors were identified as those with the most extreme alterations in the temperature at which their solubility decreased/increased and precipitation was observed (DTm) (Figure 4B, Online Supplementary Table S1). In other words, an increase/decrease in the DTm of a pro1-56

haematologica | 2021; 106(11)

tein indicates a stabilizing/destabilizing effect on the protein (or the complex it is part of) through direct/indirect interaction with AE1 . A few representative melting curves for the top hits for proteins stabilized or destabilized by the presence of AE1 (red) compared to untreated controls (blue) are provided in Figure 4C. Hits include several glycolytic enzymes GAPDH, aldolase A, phosphoglycerate kinase and lactate dehydrogenase A; hemoglobin a and beta; enzymes involved in redox homeostasis and oxidant damage-repair such as protein L-isoaspartate O-methyltransferase, peroxiredoxin 2, γ-glutamyl cysteine ligase, glyoxalase 1, aldehyde dehydrogenase 1, glutathione peroxi1-56

1-56

2977

A. Issaian et al.

dase 4, and phosphogluconate dehydrogenase; or other metabolic enzymes including acetyl-CoA lyase, arginase 1, adenosyl homocysteine hydrolase, glutamate oxalacetate transaminase, and adenylate kinase (Figure 4C).

Immunoprecipitation and crosslinking proteomics studies with recombinantly expressed AE1

1-56

To confirm and expand on the thermal proteome profiling-TMT10 analyses, we recombinantly expressed AE1 with a His-SUMO-tag at either the N- or C-terminus, prior to incubation with human plasma, RBC cytosols and membrane from human RBC lysates and purification/pull-down of interacting partners (Online Supplementary Figure S5A). Pathway analyses of the hits from these experiments confirmed most of the hits from thermal proteome profiling (Online Supplementary Table S1), revealing a widespread network of interactors of AE1, including up to 63 proteins involved in metabolic regulation, as mapped against the KEGG pathway map of human metabolism (Online 1-56

Supplementary Figure S5B, C). One limitation of the thermal proteome profiling and immunoprecipitation approaches is that they identify both direct and indirect interactors to AE1 that can be pulled down because of their complexing with direct AE1 interactors. To distinguish between direct and indirect interactors of AE1 , we repeated the precipitation experiments by also introducing crosslinking agents, disuccinimidyl sulfoxide (DSSO) or 4-(4,6-dimetoxy-1,3,5-triazin-2-yl)-4-metylmorpholinium (DMTMM) (Online Supplementary Figure S6A, B; Online Supplementary Table S1). As proof of principle, we first performed these analyses on RBC lysates without the addition of recombinant AE1, providing an experimental report of the RBC interactome (shown in the form of a circos plot or network view in Online Supplementary Figure S6C, D, respectively; Online Supplementary Table S1), significantly expanding on in silico predictions reported in the literature.48 This analysis recapitulated decades of structural studies on RBC hemoglobin (Online Supplementary Figure S7) and the RBC membrane

Figure 4. Continued on the following page.

2978

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

Figure 4. Thermal proteome profiling and crosslinking proteomics of recombinant peptide 1-56 of band 3 in red blood cell lysates. (A) Thermal proteome profiling experiments were performed by incubating red blood cell (RBC) lysates with a recombinantly expressed peptide coding for the amino acids 1-56 in the N-terminus of band 3 at a gradient of temperatures from 37°C to 67ºC, labeling with ten different tandem mass tags (TMT10), pooling and analysis via nano-ultrahigh performance liquid chromatography tandem mass spectrometry (nanoUHPLC-MS/MS). (B) Proteins were ranked as a function of their alterations in the temperature at which their solubility decreases and precipitation is observed (ΔTm). (C) A few representative melting curves for the top hits for proteins stabilized or destabilized by the presence of the band 3 peptide 1-56 (red) compared to untreated controls (blue) are provided. (D) A peptide coding for amino acids 1-56 of the N-terminus of band 3 was recombinantly expressed with a SUMO-tag and/or a His-Flag tag at either the N- or C-terminus of the peptide, prior to incubation with plasma, RBC cytosols and membrane in independent experiments, enrichment in nickel columns, pull-down against the SUMO tag, and crosslinking (XLINK) with disuccinimidyl sulfoxide (DSSO) or 4-(4,6-dimetoxy-1,3,5-triazin-2-yl)-4-metylmorpholinium (DMTMM), prior to protein digestion, fractionation of crosslinked peptides and nanoUHPLC-MS/MSbased identification of band 3 interacting partners via MS2/MS3 analyses. (E, F) Top interactors for RBC membrane and cytosol interactors are listed in (E) and (F), respectively, divided by pathway.

interactome (Online Supplementary Figure S8A-C), relevant to the regulation of the RBC cytoskeleton, and the role AE1 plays as a lynchpin for RBC membrane structural proteins (Online Supplementary Figure S8C) and AE1 multimerization in the intra- and extra-cellular compartments (Online Supplementary Figure S8D, E). A detailed description and discussion of these results is provided in the Online Supplementary File. To further focus on the AE1 N-terminus, crosslinking proteomics studies were repeated with the addition of recombinant AE1 . This approach enabled determination of direct AE1 protein interactors by identifying crosslinked free ε-amines of lysine (K) with adjacent identical residues (DSSO) or carbon atoms in the carboxylic groups of aspartic acid (D) or glutamic acid (E) (DMTMM) (Figure 5), provided these residues face each other within a range spanning from 11.4 to 24 Å.49 The list (Online Supplementary Table S1) included several structural proteins, proteasomal components, and enzymes involved in glycolysis, redox homeostasis, and purine, glutamine or sulfur metabolism, both in RBC membranes and cytosols (Figure 4E, F, respectively). An overview of the AE1 interactome is shown in Figure 5. This approach revealed three hot spots of crosslinks, at D or E residues in amino acid sequences 1-7, 13-25 and 40-45 of the AE1 N-terminus (Figure 5). Intramolecular and intermolecular interactions were observed for AE1 itself at the N-terminus (positions 45-6; 45-38; 45-56; 45-1; 25-1, 38-1, etc.) and C-terminus (833-840; 837-765; 771-849, etc.) (Online Supplementary Table S1). We report direct interactions of AE1 with hemoglobin B (1-2; 25-2; 23-2; 10-2; 10-134; 197; 38-2; 6-2; 1-100) (Online Supplementary Figure S9) and 1-56

1-56

haematologica | 2021; 106(11)

hemoglobin A (38-7, 38-86; 1-97: less frequently and with lower scores than the crosslinks with hemoglobin B).5 In addition, direct crosslinks were identified for AE1 with spectrin b (residues 23 to 1664), ankyrin (residues 10-381) (Online Supplementary Figure S8A-C), phosphoglucomutase 2 (residues 23-88), enolase b (residues 38-198). glyoxylase 2 (residues 10-299), Rab GDP dissociation inhibitor b (position 2-2) (Online Supplementary Table S1), Golgi integral membrane protein 4 (residues 23-570 – consistent with work on endosomes),50 carbonic anhydrase 2 (residues 1-81) and biliverdin reductase B (residues 1-40). Pull-down of AE1 followed by crosslinking studies identified further candidate indirect interactors that are part of macromolecular complexes with direct AE1 interactors, including peroxiredoxin 2,51 peroxiredoxin 6, catalase, phosphofructokinase isoforms,12 proteins 4.1 and 4.2, glucose phosphate isomerase, stomatin, adducin b, carbonic anhydrase 1, and 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase η-2 (full list in Online Supplementary Table S1). Other enzymes were pulled-down, but not directly crosslinked (suggestive of an indirect interaction with AE1, although direct interactions with other residues beyond residue 56 cannot be ruled out), such as acetyl-CoA lyase, bleomycin hydrolase, serine/threonine-protein kinase OSR1, desmoplakin, calpastatin, and glycophorin A. 1-56

1-56

Biophysical studies of the GAPDH and AE1 interaction

1-56

The interaction between recombinant GAPDH tetramers (Figure 6A) and AE1 (also including K56) was confirmed via isothermal titration calorimetry with a Kd of 2.56 mM 1-56

2979

A. Issaian et al.

(Figure 6B), and two-dimensional NMR upon titration of GAPDH (0, 100 and 200 mM) (Figure 6C). Of note, recombinant expression of an AE1 peptide spanning from residues 1-30 resulted in weaker interactions with GAPDH (Kd: 44.8 mM) (Online Supplementary Figure S10A, B). Calculations of chemical shift perturbations induced by GAPDH titration to the 15N-labeled AE1 (Figure 6D) provided clues on the structural organization of the otherwise intrinsically disordered N-terminus of AE1. Specifically, residues that comprise the AE1 binding site exhibit severe line-broadening due to exchange with the much larger GAPDH. For example, the terminal residues are the primary binding site as illustrated by the near complete disappearance of the central Y8 amide, compared to the Y46 amide, which exhibits no chemical shift perturbations (Figure 6E). DSSO and DMTMM crosslinks of GAPDH and AE1 (a representative spectrum between AE1 K56 and GAPDH D203 is shown in Figure 6F) are reported in Figure 6G and Online Supplementary Table S1. Cross-linked residues are mapped against the GAPDH monomeric structure in Figure 6H. Similar results were obtained when probing (Online interactions between GAPDH and AE1 Supplementary Figure S10C-E), with the exception of crosslinks between E40, a hotspot for interaction with several GAPDH residues in the longer AE1 peptide (3, 5, 84, 86, 194, 219, 251, 259, 260, 263) (Figure 6G). These experiments provide further proof of a direct interaction of the Nterminus of AE1 with GAPDH and indicate that the N-terminal region comprising residues 1-20 is the primary binding site with additional contacts that also contribute to the AE1/GAPDH interaction. Isothermal titration calorimetry,

NMR and crosslinking proteomics data were used to build a model of the AE1 interaction with GAPDH with the Rosetta software, indicating that the active site pocket of GAPDH is exposed to residues 1-15 of AE1 (Figure 6I, K; distances between K and D/E residues are shown in Online Supplementary Figure S10F). 1-56

1-56

1-30

Cell membrane-permeable AE peptides reverse the defect in glutathione recycling capacity of red blood cells from band 3 knockout mice 11-56

To rescue the genetic defect resulting from ablation of the AE1 N-terminus, we designed three membrane permeable versions52 of AE1 through addition of a polyarginine (polyArg), internalization sequence or TAT sequence at the C-terminus of the peptide (Figure 7A). Human RBC were thus incubated with a control (scramble), an AE1 peptide (non-penetrating) or the three cell-penetrating peptides in the presence of 1,2,3-13C -glucose and MB stimulation to activate the PPP (Figure 7A). Tracing experiments showed that the PPP was activated following MB in all cases, but ATP synthesis was only preserved in the RBC treated with the cell-penetrating peptides (Figure 7B), consistent with the schematic in Figure 7C. We thus hypothesized that incubation with polyArg-AE1 would at least partially rescue the defect in PPP activation upon MB stimulation of RBC from AE1 KO mice, especially HA Del (Figure 7D). The results showed increased glutathione oxidation (GSSG/GSH ratios) in response to MB treatment in all cases, but a significant rescue by the AE1 peptide in both AE1 KO. In the presence of polyArg-AE1 post transfusion recovery of stored murine RBC was restored in WT and HA 1-56

1-56

Figure 5. The interactome of AE1 , as gleaned from crosslinking mass spectrometry. A network view of the interactome of AE1 , as determined by merging the data from all the crosslinking proteomics studies. The network shows direct and indirect interactors with AE1 and the residues on band 3 with which the proteins were identified to crosslink (XLINK). 1-56

1-56

2980

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

Del mice back to levels observed in HuB3, but not in BS KO mice (Figure 7E)

calorimetry, (v) NMR, (vi) crosslinking with two different agents (DSSO and DMTMM) of recombinantly expressed GAPDH and AE1 or AE1 - the former showing a stronger interaction at least in part explained by a hotspot of crosslinks between E40 on AE1 and GAPDH. Of note, crosslinking studies have hitherto failed to show direct proximity of the (many) acidic residues of the AE1 N-terminus with GAPDH K and/or D/E residues, with only one cross-link reported between the two proteins bridging peptides spanning residues 356-384 of AE1.7 In this view, it should be noted that results from prior studies suggest that the AE1 N-terminus might not necessarily be the only region of this protein interacting with GAPDH. Finally, (vii) we modeled GAPDH-AE1 structural interactions, as constrained by all the data above. We also provide evidence of direct and indirect interactions of the N-terminus of AE1 with several other glycolytic enzymes and show that the metabolic defect in glutathione recycling of oxidatively challenged or stored RBC can be rescued in WT human and genetically ablated mouse RBC through incubation with a 1-56

Discussion Here we provide mechanistic evidence of a role for the Nterminus of AE1 in regulating the quality of stored RBC. Using multiple state-of-the-art direct and indirect structural approaches, we provide an experimental overview of the RBC interactome, with a focus on the N-terminus of band 3. A further discussion of the results is provided in the Online Supplementary File. While the structure of the N-terminus of AE1 beyond residue K56 has been solved through crystallography,53 AE1 is an intrinsically disordered domain that has hitherto eluded structural studies. In the present study, we provide conclusive, direct evidence of the interaction between GAPDH and band 3 by: (i) thermal proteome profiling; (ii) pull-down and (iii) crosslinking of recombinantly expressed AE1 ; (iv) isothermal titration 1-56

1-56

1-30

1-56

Figure 6. Continued on the following page.

haematologica | 2021; 106(11)

2981

A. Issaian et al. F

K J

Figure 6. Structural studies of recombinantly expressed GAPDH and its interaction with the N-terminus of band 3 (residues 1-56). (A) Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was recombinantly expressed in E.coli prior to purification and interaction studies with the recombinantly expressed band 3 peptide (residues 1-56). (B-E) These studies included isothermal titration calorimetry (B), nuclear magnetic resonance of band 3 (AE1 ) with 100 or 200 μM GAPDH (C) and derived calculation of chemical shift perturbations (CSP) (D), and in silico modeling of band 3 (1-56) structure based on NMR data (E). Following these studies, crosslinking proteomics analyses were performed in vitro by co-incubating GAPDH and AE1 in the presence of disuccinimidyl sulfoxide (DSSO) or 4-(4,6-dimetoxy1,3,5-triazin-2-yl)-4-metylmorpholinium (DMTMM). (F, G) A representative spectrum (F) from one of the most abundant crosslinks, comprehensively mapped in the circos plot (light blue for intramolecular crosslinks, red for intermolecular ones) (G). (H) Results were thus mapped against the GAPDH monomeric structure (blue: all lysine residue on GAPDH that were experimentally found to crosslink to band 3 aspartyl or glutamyl side chains; yellow: the lysines that were available for crosslinking but were not found to face band 3 acidic residues within the reach of the crosslinker of ~20 Å). (H-K) Nuclear magnetic resonance and crosslinking proteomics data were used to build a model of the band 3 (1-56) interaction with GAPDH with the software Rosetta, resulting in the active site pocket of GAPDH being exposed to residues 1-15 of band 3. 1-56

1-56

cell-penetrating AE1 peptide. Clearly, these proof-of-principle mechanistic experiments do not suggest the potential use of such peptides as a supplement to current blood bags, since the approach would be logistically unfeasible because of the prohibitive costs. An additional concern about such an approach would be the potential immunogenicity of such a peptide upon infusion in vivo. Notably, activation of the PPP and post-transfusion recovery were normalized in band 3 KO RBC lacking AE1 residues 1-11 (HA Del) upon incubation with a cell-penetrating AE1 peptide. It is worth stressing that the band 3 KO mice lacking residues 1-11 (but not those lacking residues 12-23) phenocopied defects in PPP activation previously described in G6PD-deficient RBC32,33 with respect to the decreased capacity to activate 1-56

1-56

2982

the PPP and ultimately poor post-transfusion recovery. The use of multiple approaches to investigate AE1 interacting partners is justified by the intrinsic limitations of each independent method. For example, false hits may result from nonspecific pull-down in the immunoprecipitation studies or indirect effects of AE1 on the stabilization of multimeric complexes with proteins identified as candidate interacting partners in thermal proteome profiling studies. To mitigate these problems, we performed additional crosslinking proteomics studies, alone or in combination with immunoprecitation strategies. Acknowledging these caveats, here we validated well-established AE1 interactors (e.g., hemoglobin, ankyrin, spectrin, carbonic anhydrase, glycophorin A), and identified other likely interactors haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

including several components of glutathione synthesis (e.g., γ-glutamyl cysteine ligase, glutathione synthetase, glutathione recycling (6PGD), glutathione-dependent lipid peroxidation pathways (glutathione peroxidase 4), as well as other antioxidant enzymes (peroxiredoxins 2 and 6, catalase). While peroxiredoxin 2 had been previously suggested to interact with the N-terminus of AE1,51 here we provide

direct evidence of this interaction. Tracing experiments suggested that RBC lacking the N-terminus of AE1 suffer from an altered degree of de novo glutathione synthesis, at least in part explained by altered glutaminolysis and transamination. Indeed, glutamate oxalacetate transaminase, postulated to be present42 but never before identified through proteomics in RBC, was identified as a potential interactor

Figure 7. Cell membrane-permeable band 3 peptides reverse the defect in the glutathione recycling capacity of red blood cells from band 3 knockout mice. (A) Three membrane permeable versions of the band 3 (AE1 ) peptide were generated through addition of a poly arginine (polyArg), internalization sequence or TAT sequence at the C-terminus of the peptide. Human red blood cells (RBC) were thus incubated with a control (scramble), a band 3 (B3) 1-56 peptide (non-penetrating) or the three penetrating peptides in the presence of 1,2,3-13C -glucose and methylene blue (MB) stimulation to activate the pentose phosphate pathway (PPP). (B) 13C isotopologues of glycolysis and the PPP are reported for all groups at baseline and following MB stimulation of healthy human RBC, showing increases in PPP activation following MB in all cases, even though ATP levels were only preserved in the RBC treated with the cell-penetrating peptides, suggesting a metabolic reprogramming consistent with the schematic in (C). (D) RBC from wild-type (WT), humanized band 3 (HuB3) or band 3 knockout mice lacking residues 1-11 (HA Del) or 12-23 (BS KO) were stimulated with MB in the presence of a cell-penetrating version of the band 3 peptide (polyArg). Results show increased glutathione oxidation in response to MB treatment in all cases, but significant rescue by the band 3 peptide, especially in the HA Del and BS KO groups. (E) Similarly, storage of murine RBC from the WT and band 3 KO mouse strains in the presence of the polyArg cell-penetrating AE1 peptide promoted increases in post-transfusion recovery in WT and HA Del mice, but not in BS KO mice. 1-56

1-56

haematologica | 2021; 106(11)

2983

A. Issaian et al.

with AE1 . Still, the caveats remain that these interactors would have to compete with a finite number of AE1 N-terminal domains per RBC (~106 copies/cell). Of note, in the plasma samples we also immunoprecipitated high mobility group nucleosome binding protein and histone H2A type 2, very basic proteins, likely with affinity for the very acidic N-terminus of AE1. While speculative at this stage, if confirmed, this interaction would suggest a role for RBC hemolysis in mitigating the endogenous danger signaling associated with the release of these damageassociated molecular pattern (DAMP) proteins in response to organ damage or traumatic injury or other DAMP-stimulating events.54 In conclusion, here we leveraged mouse models carrying a humanized AE1 N-terminus, either canonical or with genetic ablation of sequences coding for amino acid residues 1-11 or 12-23. RBC from these mice, which recapitulate N-terminus AE1 fragmentation observed in human packed RBC during storage in the blood bank or the genetic defect observed in humans carrying band 3 Neapolis, are characterized by increased markers of the storage lesion and poor post-transfusion recovery. Phenotypes recapitulate those observed in humans carrying G6PD-deficient traits in mice lacking AE1 residues 1-11 (which are required for binding to GAPDH in proximity to its active side) and in band 3 Neapolis human RBC, because of defects in the capacity to activate the PPP – a defect that could be mechanistically rescued by a cell membrane-penetrating AE1 peptide in proof-of-principle studies. While humans carrying mutations of AE1 N-terminus are not eligible for blood donation, polymorphisms on the chromosome 17 region coding for AE1 are common and associate with an increased susceptibility to lysis following osmotic insult, especially in subjects with single nucleotide polymorphisms targeting residues that we have identified as hotspots for the AE1 N-term interactome (D38, E40 and K56). Finally, here we provide the first experimental map of the RBC interactome (discussed extensively in the Online Supplementary File), with a specific focus on the N-terminus of AE1. Through a combination of proteomics approaches (co-immunoprecipitation, thermal proteome profiling and crosslinking proteomics) we validated previously described interactors (e.g., GAPDH) and identified potential new candidate binding partners for AE1 . Some of the interactions reported here may also be nonspecific (e.g., those driven by electrostatic interactions with the negatively charged acidic 1-56

1-56

References 1. Pantaleo A, Ferru E, Pau MC, et al. Band 3 rrythrocyte membrane protein acts as redox stress sensor leading to its phosphorylation by p (72) Syk. Oxid Med Cell Longev. 2016;2016:6051093. 2. Yoshida T, Prudent M, D'Alessandro A. Red blood cell storage lesion: causes and potential clinical consequences. Blood Transfus. 2019;17(1):27-52. 3. Bosman GJ, Stappers M, Novotny VM. Changes in band 3 structure as determinants of erythrocyte integrity during storage and survival after transfusion. Blood Transfus. 2010;8(Suppl 3):s48-52. 4. Bryk AH, Wisniewski JR. Quantitative analysis of human red blood cell proteome. J Proteome Res. 2017;16(8):2752-2761.

2984

residue-rich N-terminus of AE1) or affected by environmental conditions (e.g., low vs. high oxygen saturation). Further studies will address whether a membrane-penetrating version of the AE1 N-terminus peptide could rescue oxidant stress-induced lesions to RBC AE1 in the context of those pathologies that have been reported to target the N-terminus of AE1, including COVID-19.18 Disclosures ADA and KCH are founders of Omix Technologies Inc. and ADA is a founder of Altis Biosciences LLC. AD and JCZ are consultants for Rubius Therapeutics. ADA is an advisory board member for Hemanext Inc. and FORMA Therapeutics Inc. None of the other authors has any conflicts of interest relevant to this study to disclose. Contributions ADA, JCZ, SCR and ADO designed the study. AI, JR and EZE performed structural studies. AI, ZD and KCH performed crosslinking proteomics studies. AH and JCZ performed mouse studies. ADA performed metabolomics analyses. AI, MD and KCH performed proteomics analyses; DR and SP enrolled and characterized the band 3 Neapolis patient investigated in the human studies. GP and MPB directed the REDS-III RBC Omics study and related analysis of genome-wide association studies. AI, JCZ and ADA analyzed data. ADA prepared the figures, wrote the first draft of the manuscript and revised it. All the authors contributed to the finalization of the manuscript. Acknowledgments The authors are grateful to Devin P. Champagne for his contribution in the early stages of this study. Funding This research was supported by funds from the Boettcher WebbWaring Investigator Award (to ADA), RM1GM131968 (to ADA) and R01GM113838 (to ADO) from the National Institute of General and Medical Sciences, and R01HL146442 (to ADA), R01HL149714 (to ADA), R01HL148151 (to ADA and JCZ), R21HL150032 (to ADA), from the National Heart, Lung, and Blood Institute. Data sharing statement Source data for all the figures and analyses in this study are included within Online Supplementary Table S1, divided by figures and experiments.

5. Chu H, Breite A, Ciraolo P, Franco RS, Low PS. Characterization of the deoxyhemoglobin binding site on human erythrocyte band 3: implications for O2 regulation of erythrocyte properties. Blood. 2008;111(2):932-938. 6. Sega MF, Chu H, Christian J, Low PS. Interaction of deoxyhemoglobin with the cytoplasmic domain of murine erythrocyte band 3. Biochemistry. 2012;51(15):32643272. 7. Puchulu-Campanella E, Chu H, Anstee DJ, Galan JA, Tao WA, Low PS. Identification of the components of a glycolytic enzyme metabolon on the human red blood cell membrane. J Biol Chem. 2013;288(2):848858. 8. Sun K, Zhang Y, D'Alessandro A, et al. Sphingosine-1-phosphate promotes erythrocyte glycolysis and oxygen release for adap-

tation to high-altitude hypoxia. Nat Commun. 2016;7:12086. 9. Lewis IA, Campanella ME, Markley JL, Low PS. Role of band 3 in regulating metabolic flux of red blood cells. Proc Natl Acad Sci U S A. 2009;106(44):18515-18520. 10. von Ruckmann B, Schubert D. The complex of band 3 protein of the human erythrocyte membrane and glyceraldehyde-3-phosphate dehydrogenase: stoichiometry and competition by aldolase. Biochim Biophys Acta. 2002;1559(1):43-55. 11. Eisenmesser EZ, Post CB. Insights into tyrosine phosphorylation control of protein-protein association from the NMR structure of a band 3 peptide inhibitor bound to glyceraldehyde-3-phosphate dehydrogenase. Biochemistry. 1998;37(3):867-877. 12. Chu H, Low PS. Mapping of glycolytic

haematologica | 2021; 106(11)

AE1 regulates RBC metabolism and storage

enzyme-binding sites on human erythrocyte band 3. Biochem J. 2006;400(1):143-151. 13. Castagnola M, Messana I, Sanna MT, Giardina B. Oxygen-linked modulation of erythrocyte metabolism: state of the art. Blood Transfus. 2010;8 Suppl 3(Suppl 3):s5358. 14. Rogers SC, Said A, Corcuera D, McLaughlin D, Kell P, Doctor A. Hypoxia limits antioxidant capacity in red blood cells by altering glycolytic pathway dominance. FASEB J. 2009;23(9):3159-3170. 15. D'Alessandro A, Hansen KC, Eisenmesser EZ, Zimring JC. Protect, repair, destroy or sacrifice: a role of oxidative stress biology in inter-donor variability of blood storage? Blood Transfus. 2019;17(4):281-288. 16. D'Alessandro A, Nemkov T, Sun K, et al. AltitudeOmics: red blood cell metabolic adaptation to high altitude hypoxia. J Proteome Res. 2016;15(10):3883-3895. 17. Messana I, Orlando M, Cassiano L, et al. Human erythrocyte metabolism is modulated by the O2-linked transition of hemoglobin. FEBS Lett. 1996;390(1):25-28. 18. Thomas T, Stefanoni D, Dzieciatkowska M, et al. Evidence of structural protein damage and membrane lipid remodeling in red blood cells from COVID-19 patients. J Proteome Res. 2020;19(11):4455-4469. 19. Chapman RG, Schaumburg L. Glycolysis and glycolytic enzyme activity of aging red cells in man. Changes in hexokinase, aldolase, glyceraldehyde-3-phosphate dehydrogenase, pyruvate kinase and glutamicoxalacetic transaminase. Br J Haematol. 1967;13(5):665-678. 20. Low PS, Waugh SM, Zinke K, Drenckhahn D. The role of hemoglobin denaturation and band 3 clustering in red blood cell aging. Science. 1985;227(4686):531-533. 21. Rinalducci S, Ferru E, Blasi B, Turrini F, Zolla L. Oxidative stress and caspase-mediated fragmentation of cytoplasmic domain of erythrocyte band 3 during blood storage. Blood Transfus. 2012;10(Suppl 2):s55-62. 22. D'Alessandro A, D'Amici GM, Vaglio S, Zolla L. Time-course investigation of SAGMstored leukocyte-filtered red bood cell concentrates: from metabolism to proteomics. Haematologica. 2012;97(1):107-115. 23. Reisz JA, Nemkov T, Dzieciatkowska M, et al. Methylation of protein aspartates and deamidated asparagines as a function of blood bank storage and oxidative stress in human red blood cells. Transfusion. 2018;58(12):2978-2991. 24. Kriebardis AG, Antonelou MH, Stamoulis KE, Economou-Petersen E, Margaritis LH, Papassideri IS. Progressive oxidation of cytoskeletal proteins and accumulation of denatured hemoglobin in stored red cells. J Cell Mol Med. 2007;11(1):148-155. 25. Prudent M, Delobel J, Hubner A, Benay C, Lion N, Tissot JD. Proteomics of stored red blood cell membrane and storage-induced microvesicles reveals the association of flotillin-2 with band 3 complexes. Front Physiol. 2018;9:421. 26. Messana I, Ferroni L, Misiti F, et al. Blood

haematologica | 2021; 106(11)

bank conditions and RBCs: the progressive loss of metabolic modulation. Transfusion. 2000;40(3):353-360. 27. Tzounakas VL, Kriebardis AG, Georgatzakou HT, et al. Glucose 6-phosphate dehydrogenase deficient subjects may be better "storers" than donors of red blood cells. Free Radic Biol Med. 2016;96:152-165. 28. Roch A, Magon NJ, Maire J, et al. Transition to 37 degrees C reveals importance of NADPH in mitigating oxidative stress in stored RBCs. JCI Insight. 2019;4(21): e126376. 29. Reisz JA, Wither MJ, Dzieciatkowska M, et al. Oxidative modifications of glyceraldehyde 3-phosphate dehydrogenase regulate metabolic reprogramming of stored red blood cells. Blood. 2016;128(12):e32-42. 30. Azouzi S, Romana M, Arashiki N, et al. Band 3 phosphorylation induces irreversible alterations of stored red blood cells. Am J Hematol. 2018;93(5):E110-E112. 31. Perrotta S, Borriello A, Scaloni A, et al. The N-terminal 11 amino acids of human erythrocyte band 3 are critical for aldolase binding and protein phosphorylation: implications for band 3 function. Blood. 2005;106(13):4359-4366. 32. Francis RO, D'Alessandro A, Eisenberger A, et al. Donor glucose-6-phosphate dehydrogenase deficiency decreases blood quality for transfusion. J Clin Invest. 2020;130(5): 2270-2285. 33. D'Alessandro A, Fu X, Kanias T, et al. Donor sex, age and ethnicity impact stored red blood cell antioxidant metabolism through mechanisms in part explained by glucose 6phosphate dehydrogenase levels and activity. Haematologica. 2021;106(5):1290-1302. 34. Karon BS, Hoyer JD, Stubbs JR, Thomas DD. Changes in band 3 oligomeric state precede cell membrane phospholipid loss during blood bank storage of red blood cells. Transfusion. 2009;49(7):1435-1442. 35. Chu H, McKenna MM, Krump NA, et al. Reversible binding of hemoglobin to band 3 constitutes the molecular switch that mediates O2 regulation of erythrocyte properties. Blood. 2016;128(23):2708-2716. 36. Howie HL, Hay AM, de Wolski K, et al. Differences in Steap3 expression are a mechanism of genetic variation of RBC storage and oxidative damage in mice. Blood Adv. 2019;3(15):2272-2285. 37. Guo Y, Busch MP, Seielstad M, et al. Development and evaluation of a transfusion medicine genome wide genotyping array. Transfusion. 2019;59(1):101-111. 38. Kanias T, Lanteri MC, Page GP, et al. Ethnicity, sex, and age are determinants of red blood cell storage and stress hemolysis: results of the REDS-III RBC-Omics study. Blood Adv. 2017;1(15):1132-1141. 39. de Wolski K, Fu X, Dumont LJ, et al. Metabolic pathways that correlate with post-transfusion circulation of stored murine red blood cells. Haematologica. 2016;101(5): 578-586. 40. Nemkov T, Reisz JA, Gehrke S, Hansen KC, D'Alessandro A. High-throughput

metabolomics: isocratic and gradient mass spectrometry-based methods. Methods Mol Biol. 2019;1978:13-26. 41. Reisz JA, Zheng C, D'Alessandro A, Nemkov T. Untargeted and semi-targeted lipid analysis of biological samples using mass spectrometry-based metabolomics. Methods Mol Biol. 2019;1978:121-135. 42. D'Alessandro A, Dzieciatkowska M, Nemkov T, Hansen KC. Red blood cell proteomics update: is there more to discover? Blood Transfus. 2017;15(2):182-187. 43. Franken H, Mathieson T, Childs D, et al. Thermal proteome profiling for unbiased identification of direct and indirect drug targets using multiplexed quantitative mass spectrometry. Nat Protoc. 2015;10(10):15671593. 44. Miraglia del Giudice E, Vallier A, Maillet P, et al. Novel band 3 variants (bands 3 Foggia, Napoli I and Napoli II) associated with hereditary spherocytosis and band 3 deficiency: status of the D38A polymorphism within the EPB3 locus. Br J Haematol. 1997;96(1):70-76. 45. Ma SY, Liao L, He BJ, Lin FQ. [Application of high resolution melting curve analysis in detection of SLC4A1 gene mutation in patients with hereditary spherocytosis]. Zhongguo Shi Yan Xue Ye Xue Za Zhi. 2018;26(6):1826-1830. 46. Weinstein R, Martinez R, Hassoun H, Palek J. Does a patient with hereditary spherocytosis qualify for preoperative autologous blood donation? Transfusion. 1997;37(1112):1179-1183. 47. Lange PF, Huesgen PF, Nguyen K, Overall CM. Annotating N termini for the human proteome project: N termini and Nalphaacetylation status differentiate stable cleaved protein species from degradation remnants in the human erythrocyte proteome. J Proteome Res. 2014;13(4):20282044. 48. D'Alessandro A, Righetti PG, Zolla L. The red blood cell proteome and interactome: an update. J Proteome Res. 2010;9(1):144-163. 49. Li X, Liu S, Jiang J, et al. CryoEM structure of Saccharomyces cerevisiae U1 snRNP offers insight into alternative splicing. Nat Commun. 2017;8(1):1035. 50. Beck KA, Nelson WJ. A spectrin membrane skeleton of the Golgi complex. Biochim Biophys Acta. 1998;1404(1-2):153-160. 51. Matte A, Bertoldi M, Mohandas N, et al. Membrane association of peroxiredoxin-2 in red cells is mediated by the N-terminal cytoplasmic domain of band 3. Free Radic Biol Med. 2013;55:27-35. 52. Ruseska I, Zimmer A. Internalization mechanisms of cell-penetrating peptides. Beilstein J Nanotechnol. 2020;11:101-123. 53. Zhang D, Kiyatkin A, Bolin JT, Low PS. Crystallographic structure and functional interpretation of the cytoplasmic domain of erythrocyte membrane band 3. Blood. 2000;96(9):2925-2933. 54. Silk E, Zhao H, Weng H, Ma D. The role of extracellular histone in organ injury. Cell Death Dis. 2017;8(5):e2812.

2985

LETTERS TO THE EDITOR Cluster of differentiation 33 single nucleotide polymorphism rs12459419 is a predictive factor in patients with nucleophosmin1-mutated acute myeloid leukemia receiving gemtuzumab ozogamicin Gemtuzumab ozogamicin (GO) is a humanized anticluster of differentiation (CD) 33 monoclonal antibody linked to the cytotoxic agent calicheamicin.1-4 GO binds the CD33 antigen and is internalized; calicheamicin is released inside the blasts leading to DNA damage and death of leukemic cells.2,5,6 GO was approved for the treatment of patients with acute myeloid leukemia (AML) with expression of CD33 on blasts by the European Medicines Agency (here only for newly diagnosed patients in combination with intensive therapy) and the Food and Drug Administration based on positive results from the ALFA-0701 trial.1,7 However, a number of clinical trials with GO yielded partially conflicting results emphasizing the importance of a better understanding of factors influencing the clinical response to GO.6,7 Several factors can influence the response to GO. A meta-analysis by Hills et al. demonstrated a benefit of GO only in patients with Medical Research Council (MRC) favorable and intermediate cytogenetic risk, while GO did not improve outcome of patients with adverse cytogenetic risk.7 Another factor that might have an impact on the response to GO is the CD33-coding single nucleotide polymorphism (SNP) rs12459419 (NM_001772.4:c.41 C>T; Ala14Val in exon 2) as GO needs to bind to CD33. CD33 consists of an amino-terminal V-set immunoglobulin (Ig)-like domain, coded by exon 2, and a C2-set Ig-like domain in its extracellular component. The presence of rs12459419 in exon 2 affects alternative splicing of CD33 resulting in the loss of exon 2 for the T allele which leads to a shorter isoform lacking the GO binding site. The C allele leads to the CD33 full-length isoform including the GO binding site.3,6 The genotype frequencies in the European population for the SNP are c.41C/C: 46.9%, c.41C/T: 44.1%, and c.41T/T: 8.9%.

Given the strong clinical need to clarify who could benefit from GO and the mechanisms responsible for a differential response to GO we studied the effect of rs12459419 in the AMLSG 09-09 phase III study (NCT00893399). This trial is a large, randomized study in adult AML patients who were eligible for intensive therapy and had a mutation in nucleophosmin1 (NPM1). According to MRC definitions,8 99.5% of the patients in both treatment arms had intermediate-risk cytogenetics (Online Supplementary Table S1). Since the benefit of GO was only observed in favorable- and intermediate-risk groups,7 our cohort is well suited for analyzing the effect of the CD33 SNP. Of 588 patients included in this study, 545 patients had samples available for SNP analysis. Patients were randomly assigned (1:1) to receive 3 mg/m2 of GO (n=273, GO arm) versus no GO (n=272, standard arm) in combination with two cycles of induction chemotherapy.9 Samples were obtained from peripheral blood (n=459) and bone marrow (n=86); 507 samples were obtained at the time of diagnosis and 38 during follow-up. The SNP genotype was determined using the TaqMan SNP Genotyping Assay rs12459419 (Thermo Fisher Scientific, Waltham, MA, USA) on a 7500 Fast Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). The allelic discrimination protocol was modified from that described by Easton et al.10 The Genotyping Assay revealed robust results for 540 out of 545 samples (call rate >99%). DNA from the remaining five patients was genotyped by Sanger sequencing as previously reported.11 In order to confirm the genotype we additionally analyzed three samples with a genotype known from the Genotyping Assay. Results from the Genotyping Assay and Sanger sequencing were concordant. Genotype proportions were 45.2% c.41 C/C, 44.1% c.41 C/T and 10.7% c.41 T/T in the standard arm and 46.1% c.41 C/C, 45.3% c.41 C/T and 8.6% c.41 T/T in the GO arm. We then analyzed whether rs12459419 genotypes influenced hematologic recovery times, response to therapy and clinical outcome in patients treated with or without GO.

Figure 1. Analysis comparing gemtuzumab ozogamicin versus standard treatment in the three subgroups defined by the genotypes with regard to complete remission/complete remission with incomplete hematologic recovery. HR: hazard ratio; 95% CI: 95% confidence interval; CR: complete remission; CRi: complete remission with incomplete hematologic recovery; OR: odds ratio; GO: gemtuzumab ozogamicin.

2986

haematologica | 2021; 106(11)

Letters to the Editor

Figure 2. Association of rs12459419 genotypes with clinical outcome by treatment arm. (A) Event-free survival. (B) Relapse-free survival. (C) Cumulative incidence of relapse. (D) Cumulative incidence of death in remission. EFS: event-free survival; GO: gemtuzumab ozogamicin; RFS: relapse-free survival; CR: complete remission; CRi: complete remission with incomplete hematologic recovery; CIR: cumulative incidence of relapse; CID: cumulative incidence of death in remission.

haematologica | 2021; 106(11)

2987

Letters to the Editor

Figure 3. Analysis comparing gemtuzumab ozogamicin versus standard treatment in the three subgroups defined by the genotypes with regard to clinical endpoints. Effect estimates are taken from multivariate Cox models and are therefore adjusted for age, gender, European LeukemiaNet 2010 risk class and white blood cell count. HR: hazard ratio; 95% CI: 95% confidence interval; EFS: eventfree survival; RFS: relapsefree survival; CR: complete remission; CIR: cumulative incidence of relapse; CID: cumulative incidence of death in remission: GO: gemtuzumab ozogamicin.

Time to hematologic recovery from the start of the first and second induction cycles was not influenced by the SNP genotype in the GO arm (Online Supplementary Figure S1). With respect to the clinical endpoints complete remission (CR) or CR with incomplete hematologic recovery (CRi) as well as event-free survival, relapse-free survival, cumulative incidence of relapse and cumulative incidence of death in remission (defined according to European LeukemiaNet 2017 criteria12), multivariate regression models (logistic regression and Cox proportional hazards models) were used to evaluate whether the effect of GO varies with the SNP genotype. More explicitly, the likelihood ratio test was used to test for the presence of an interaction between treatment with GO and SNP genotype while other prognostic factors (age, gender, European LeukemiaNet 2010 risk class13 and white blood cell count) were accounted for by including them as additional covariates. Of 485 patients who achieved CR/CRi, 157 patients experienced a relapse and 46 patients died in CR. No significant differences for CR/CRi could be observed between the genotypes in the treatment arms (Figure 1). For event-free survival, no significant difference in the treatment effect of GO compared to the standard arm was observed across the SNP genotypes (Figures 2A and 3). Specifically, patients in the GO arm with the c.41 C/C genotype did not have a superior event-free survival compared to those with other genotypes. However, relapsefree survival was significantly improved for patients with the c.41 C/C genotype in the GO arm (P=0.03) compared to those in the standard arm (Figures 2B and 3). It is important to state that the early death rate during induction was significantly higher (10.3%) in the GO arm than in the standard arm (5.7%) (P=0.05) but not influenced by the different SNP genotypes (data not shown). However, the higher early death rate in the GO arm may explain why we see an effect of the CD33 SNP on relapse-free survival 2988

but not on event-free survival. Furthermore, patients with c.41 C/C genotype had a significantly reduced cumulative incidence of relapse when treated with GO (P=0.023) (Figures 2C and 3). Patients with c.41 C/T or c.41 T/T genotype did not benefit from GO with regard to relapse-free survival and cumulative incidence of relapse (Figures 2B, C and 3). There were no significant differences in cumulative incidence of death in remission in relation to genotypes between the treatment arms or between the genotypes within a treatment arm (Figures 2D and 3). Because subgroup analysis of the AMLSG 09-09 trial revealed a significant prognostic advantage in female patients, patients <70 years and FLT3-ITD-negative patients when receiving GO, we analyzed whether these observations could be explained by different distributions of the SNP genotypes. We found no significant association of the SNP with mutations in FLT3-ITD, FLT3-TKD or DNMT3A (Online Supplementary Table S1). For separate analysis of female and male patients see Online Supplementary Figure S2A, B. Additionally, we separately analyzed the influence of rs12459419 on the effect of GO in younger patients (≤40 years) versus older ones (>40 years). Due to low event numbers (especially in the group of younger patients) additional covariate adjustment was omitted for these subgroup analyses. Here, younger patients with the c.41 C/C genotype had no extra benefit from GO compared to younger patients with other SNP genotypes (data not shown). Published data about whether CD33 SNP rs12459419 can influence the response to GO have been conflicting. The first study to analyze the effect of the rs12459419 on the response to GO was the AAML0531 trial, a randomized phase III study in pediatric and adolescent AML patients (aged 0-29 years).3 In that study, Lamba et al. showed that the T allele was significantly associated with haematologica | 2021; 106(11)

Letters to the Editor

higher levels of the CD33 isoform lacking exon 2. Patients harboring the c.41 C/C genotype treated with GO had a significantly reduced cumulative incidence of relapse and higher relapse-free survival than patients in the non-GO arm. Importantly, this positive effect was restricted to patients with favorable and intermediate cytogenetic risk. The clinical benefit of GO was not visible in patients with the c.41 C/T or c.41 T/T genotype of SNP rs12459419.3 In contrast, in the UK MRC/NCRI AML 15 (ISRCTN17161961) and AML 17 (ISRCTN55675535) trials, GO was given to adult AML patients aged 13 to 69 years. Here, the SNP genotype had no impact on outcome, even if only patients with favorable-risk cytogenetics were considered.14 Of note, the trial included multiple randomization steps with different chemotherapy regimens.14 Moreover, Short et al. could not show a significant impact of the SNP on overall survival or relapse-free survival in patients (n=113) treated with decitabine plus GO.15 However, this study was performed in a population of patients unlikely to benefit from GO as it included patients with high-risk myelodysplastic syndrome and AML patients with unfavorable risk features.15 Our results show a similar signal as the data from Lamba et al. with regard to an improved relapse-free survival and reduced cumulative incidence of relapse for patients with the c.41 C/C genotype treated with GO. It is unlikely that the discrepancies between the results of the studies by Gale et al. and Lamba et al. are only related to the difference in age of the patients as we showed an impact of the SNP in adult patients. In summary, our study suggests that CD33 SNP rs12459419 is one of the predictive factors that affects the rate of relapse in patients with NPM1-mutated AML receiving GO. As testing for the SNP is technically simple, it can be easily included in routine diagnostics. In order to support our findings a prospective study in a larger cohort is desirable. Katrin Teich,1 Julia Krzykalla,2 Silke Kapp-Schwoerer,3 Verena I. Gaidzik,3 Richard F. Schlenk,4,5 Peter Paschka,3 Daniela Weber,3 Walter Fiedler,6 Michael W. M. Kühn,7 Thomas Schroeder,8 Karin Mayer,9 Michael Lübbert,10 Dhanya Ramachandran,11 Axel Benner,2 Arnold Ganser,1 Hartmut Döhner,3 Michael Heuser,1 Konstanze Döhner3 and Felicitas Thol1 1 Department of Hematology, Hemostaseology, Oncology, and Stem Cell Transplantation, Hannover Medical School, Hannover; 2 Division of Biostatistics, German Cancer Research Center Heidelberg, Heidelberg; 3Department of Internal Medicine III, University Hospital of Ulm, Ulm; 4Nationales Centrum für Tumorerkrankungen Trial Center, National Center of Tumor Diseases, German Cancer Research Center, Heidelberg; 5Department of Internal Medicine V, Heidelberg University Hospital, Heidelberg; 6Department of Internal Medicine II, University Medical Center Hamburg-Eppendorf, Hamburg; 7Department of Hematology, Medical Oncology and Pneumology, University Medical Center Mainz, Mainz; 8Department of Hematology, Oncology, and Clinical Immunology, University of Düsseldorf, Medical Faculty, Düsseldorf; 9Internal Medicine III, University Hospital of Bonn, Bonn; 10Klinik für Innere Medizin I, Universitätsklinikum Freiburg, Faculty of Medicine, Freiburg and 11 Department Molecular Gynecology, Hannover Medical School, Hannover, Germany Correspondence: KATRIN TEICH - teich.katrin@mh-hannover.de doi:10.3324/haematol.2021.278894 Received: April 1, 2021.

haematologica | 2021; 106(11)

Accepted: May 17, 2021. Pre-published: May 27, 2021. Disclosures: no conflicts of interest to disclose. Contributions: conception and design of the study: FT and KT; financial support: FT; provision of study materials or patients: JK, SKS, VIG, RFS, PP, DW, WF, MWMK, TS, KM, KL, AB, AG, HD, MH and FT; collection and assembly of data: JK, SKS, VIG, RFS, PP, DW, WF, MWMK, TS, KM, KL, AB, AG, HD, MH and FT; data analysis and interpretation: JK, FT, DR and KT; manuscript writing: FT and KT; final approval of the manuscript: all authors. Each author is accountable for all aspects of the work. Funding: this study was supported by grant DJCLS R13/14 from the Deutsche José Carreras Leukämie-Stiftung e.V., grant 01GM1909A from the German Federal Ministry of Education and Research and by Sonderforschungsbereich SFB 1074, project B3 (to KD) from the Deutsche Forschungsgemeinschaft. Data sharing statement: all data are available from the corresponding author upon request.

References 1. Lambert J, Pautas C, Terré C, et al. Gemtuzumab ozogamicin for de novo acute myeloid leukemia: final efficacy and safety updates from the open-label, phase III ALFA-0701 trial. Haematologica. 2019;104(1):113-119. 2. Ricart AD. Antibody-drug conjugates of calicheamicin derivative: gemtuzumab ozogamicin and inotuzumab ozogamicin. Clin Cancer Res. 2011;17(20):6417-6427. 3. Lamba JK, Chauhan L, Shin M, et al. CD33 splicing polymorphism determines gemtuzumab ozogamicin response in de novo acute myeloid leukemia: report from randomized phase III Children’s Oncology Group trial AAML0531. J Clin Oncol. 2017;35(23):26742682. 4. Thol F, Schlenk RF. Gemtuzumab ozogamicin in acute myeloid leukemia revisited. Expert Opin Biol Ther. 2014;14(8):1185-1195. 5. Sievers EL. Efficacy and safety of gemtuzumab ozogamicin in patients with CD33-positive acute myeloid leukaemia in first relapse. Expert Opin Biol Ther. 2001;1(5):893-901. 6. Godwin CD, Gale RP, Walter RB. Gemtuzumab ozogamicin in acute myeloid leukemia. Leukemia. 2017;31(9):1855-1868. 7. Hills RK, Castaigne S, Appelbaum FR, et al. The addition of gemtuzumab ozogamicin to induction chemotherapy in acute myeloid leukaemia: an individual patient data meta-analysis of randomised trials in adults. Lancet Oncol. 2014;15(9):986-996. 8. Kühnl A, Grimwade D. Molecular markers in acute myeloid leukaemia. Int J Hematol. 2012;96(2):153-163. 9. Schlenk RF, Paschka P, Krzykalla J, et al. Gemtuzumab ozogamicin in NPM1 -mutated acute myeloid leukemia: early results from the prospective randomized AMLSG 09-09 phase III study. J Clin Oncol. 2020;38(6):623-632. 10. Easton EF, Pooley KA, Dunning AM, et al. Genome-wide association study identifies novel breast cancer susceptibility loci. Nature. 2007;447(7148):1087-1093. 11. Thol F, Damm F, Lüdeking A, et al. Incidence and prognostic influence of DNMT3A mutations in acute myeloid leukemia. J Clin Oncol. 2011;29(21):2889-2896. 12. 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. 13. Döhner H, Estey EH, Amadori S, et al. Diagnosis and management of acute myeloid leukemia in adults: recommendations from an international expert panel, on behalf of the European LeukemiaNet. Blood. 2010;115(3):453-474. 14. Gale RE, Popa T, Wright M, et al. No evidence that CD33 splicing SNP impacts the response to GO in younger adults with AML treated on UK MRC/NCRI trials. Blood. 2018;131(4):468-471. 15. Short NJ, Richard-Carpentier G, Kanagal-Shamanna R, et al. Impact of CD33 and ABCB1 single nucleotide polymorphisms in patients with acute myeloid leukemia and advanced myeloid malignancies treated with decitabine plus gemtuzumab ozogamicin. Am J Hematol. 2020;95(9):E225-E228.

2989

Letters to the Editor

Aurora A kinase as a target for therapy in TCF3-HLF rearranged acute lymphoblastic leukemia B-cell acute lymphoblastic leukemia (B-ALL) harboring the t(17;19)(q22;p13) is a rare subtype of leukemia with a dismal prognosis.1 This translocation produces an aberrant TCF3-HLF fusion with distinct gene expression profiles and drug sensitivity. Recent studies have shown that this subtype of B-ALL may be amenable to therapies inhibiting BCL2 and the pre-B cell receptor through inhibition of SRC family kinases.2-4 Using RNA sequencing (RNA-seq) in combination with ex vivo drug sensitivity analyses we identified an overexpression pattern of Aurora A kinase (AURKA) in five of five TCF3-HLF ALL samples. This finding further translated to enhanced sensitivity to inhibition by one of the AURKA inhibitors alisertib. Our studies suggest a molecular susceptibility of TCF3-HLF ALL to alisertib providing support to pursue further clinical testing within this rare and lethal subtype of ALL. The t(17;19)(q22;p13) rearrangement encodes for a chimeric transcription factor consisting of type 1 rearrangements with a fusion of exon 16 of TCF3 with exon 4 of HLF; and type 2 rearrangements aligning with a fusion of exon 15 of TCF3 with exon 4 of HLF.5 Because of its rarity and poor prognosis, there are few established recommended therapies. Patients who achieve remission with no evidence of disease historically recur and die of disease even with stem cell transplant as consolidation therapy.6 Recent advances using immunologic therapies has shown some promise with durable responses targeting expression of CD19 by B-cell leukemia,7 however not all patients remained in remission and not all may be eligible for immunologic therapies. Therefore, novel therapies remain essential to improve outcomes. We developed an in vitro assay using a panel of small-molecule inhibitors to identify patient specific targeted therapies and employed this assay on the diagnostic marrow sample of a patient who presented to our institution with TCF3-HLF ALL.2 We observed in vitro sensitivity to the ABL/SRC inhibitor dasatinib with some clinical benefit.2 Meanwhile, other studies identified targeting the BCL2 family of proteins with venetoclax as a therapeutic potential.3,4 We thus sought to validate and identify other novel targets across an independent set of samples. We partnered with the Children’s Oncology Group (COG) and St. Jude Children’s Research Hospital to obtain four separate patient samples combined with our sample which were xenografted and expanded for in vitro as well as in vivo experiments. All clinical samples were obtained with informed consent with approval by the Institutional Review Board of Oregon Health & Science University, the Children’s Oncology Group, and St. Jude Children’s Research Hospital. All murine studies were approved by the OHSU Institutional Animal Care and Use Committee (IACUC, protocol #2358). Standard reverse transcription polymerase chain reaction (RT-PCR), Sanger sequencing and immunoblot analysis of the TCF3-HLF fusion identified three samples with a type 1 translocation (10-199, 3.316, and 3.332) with variable insertions corresponding to different protein length, while two others carried a type 2 translocation (3310, 3324) (Online Supplementary Figure S1A). We then undertook next-generation sequencing studies with these samples compared to 35 B-ALL samples (Online Supplementary Figure S1B and C). Using whole-exome sequencing within our ALL cohort, and specifically focusing on variations found in pathways in B-cell ALL,8 we 2990

found that three of the five samples carried an IKZF1 variant of unknown clinical significance (IKZF1 I125V; Online Supplementary Figure S1B). Further, two of the five samples carried KRAS mutations (KRAS G12V/S) as previously described.9 Finally, RNA-seq analysis of the top 1.000 most variably expressed genes across the cohort revealed gene expression signatures that clustered strongly with TCF3-HLF ALL (Online Supplementary Figure S1C), suggesting relatively uniform gene expression within these five Type 1 and Type 2 TCF3-HLF leukemias. Implementation of our small-molecule kinase inhibitor panel10 on 69 ALL samples revealed some consistent patterns of responses in TCF3-HLF ALL including sensitivity to the class of aurora kinase inhibitors (Online Supplementary Figures S1D and S2). Harnessing the RNAseq data for expression identified significantly higher RNA levels for AURKA, slightly higher levels of AURKB with no difference in AURKC levels in TCF3-HLF ALL as compared to the other B-ALL samples (Figure 1A). Increased AURKA expression was further validated by protein expression as compared to three separate B-ALL xenograft samples (Figure 1B). Gene set enrichment analysis (GSEA) of the top 15 up-regulated Reactome pathways also emphasized cell cycle pathways (Online Supplementary Figures S3A) as well as pathways involving Aurora kinases in TCF3-HLF ALL (Online Supplementary Figures S3B). Within the class of Aurora kinase inhibitors, the AURKA selective alisertib appeared to be the most advanced in clinical testing, including pediatric dosing.11 Utilizing our functional assay, the population of B-ALL appear by and large resistant to alisertib with a median half maximal inhibitory concentration (IC50) >10 mM while the majority of TCF3-HLF ALL samples showed some sensitivity (Figure 1D; Online Supplementary Figures S1D and S2). As AURKA levels can be dependent on the cell cycle we compared the levels of BIRC5 and INCENP, two proteins critical for the chromosome passenger complex during mitosis.12 Interestingly, both BIRC5 and INCENP levels were increased compared to B-ALL suggesting that these samples may have a higher population of cells progressing through the cell cycle (Figure 1C). These findings correlated with an increase in the number of cells in S phase amongst TCF3-HLF ALL and considerable sensitivity to the BIRC5 inhibitor YM155 (Figure 1D). In order to further interrogate the effects of alisertib on TCF3-HLF ALL, HAL01 cells that carry a type 1 fusion was used as a model (Figure 2). We found that HAL01 displayed a low IC50 of 1 nM for alisertib when assessing viability, increased apoptosis by annexin V staining and a significant G2/M arrest (Figure 2A to C). Treatment of these cells with alisertib appeared to disrupt the normal assembly of the spindle poles with more cells having only one spindle pole, and some cells with more than two spindle poles (Figure 2D and E). These phenotypes are similar to RNAi depletion of AURKA consistent with alisertib’s effects on Aurora A kinase inhibition.13 Each TCF3-HLF ALL sample was then tested for therapeutic response in vivo in cohorts of immunodeficient NSG mice. One week after tail vein injection, cohorts were treated with vehicle, the SRC inhibitor dasatinib (40 mg/kg/day), the BCL2 inhibitor venetoclax (25 mg/kg/day), combination of dasatinib and venetoclax (n=5)14 or alisertib (30 mg/kg/day)15 by oral gavage for five days each week (Figure 3). All control animals engrafted rapidly within 2 months after injection. Only sample 3.332 and 3.310 showed significant response to dasatinib, venetoclax, or combination. Interestingly, haematologica | 2021; 106(11)

Letters to the Editor

Figure 1. Sensitivity of TCF3-HLF ALL to AURKA inhibition. (A) Expression levels of AUK kinases within TCF3-HLF acute lymphoblastic leukemia (ALL) samples from RNA sequencing (RNA-seq) analysis. RPKM (reads per kilo base per million mapped reads) levels of AURKA, AURKB and AURKC were compared between t(17;19) and remaining B-cell acute lymphoblastic leukemia (B-ALL) samples by two-tailed student’s t-test. (B) Protein expression of AURKA in ALL. Protein extract from three primary B-ALL xenograft samples (11-064, TCF3-PBX1; 11-504, KMT2A-rearranged; 12-225, BCR-ABL1) were compared to three t(17;19) TCF3-HLF ALL xenograft samples by immunoblot (green denotes type 1 fusions, while red denotes type 2 fusions). Cell lysate was extracted and separated by standard SDS-PAGE. Protein was then transferred to PVDF membranes and immunoblotted using anti-AURKA or anti-actin. Three independent blots for each sample were quantified and AURK level for each sample was normalized to actin. Comparison of the level of AURK to each sample was normalized to the level of AURK in 11-064. Green column indicates type 1 samples, red column indicates type 2. ANOVA ***P=0.0033. (C) Expression from RNA-seq analysis comparing RPKM levels of BIRC5 and INCENP between t(17;19) and remaining ALL samples by two-tailed student’s t-test. Graphic representation of asynchronous populations within the cell cycle. Cells were fixed in 70% methanol, then stained with propidium iodide (PI), and analyzed for DNA content by flow cytometry. Comparison of the percentage of cells in G1, S and G2/M were performed from three B-ALL xenografted samples compared to three TCF3-HLF ALL samples using two-tailed student’s t-test. (D) Half maximal inhibitory concentration (IC50) comparison of alisertib and YM155 between t(17;19) and remaining ALL samples by two-tailed student’s t-test.

haematologica | 2021; 106(11)

2991

Letters to the Editor

these two samples do not carry the IKZF1 variant. In contrast, all animals treated with alisertib did show a significant survival advantage compared to their control counterparts. Our results continue to confirm that TCF3-HLF ALL is a unique subset of ALL with varying degrees of response to prior published targets for therapy. In vivo validation in the murine model with our five samples suggests hetero-

geneity of response to current pursued targets such as BCL2 and SRC. In contrast, we found significant in vivo response to the AURKA inhibitor alisertib in all of our samples tested. It is interesting that prior studies using a different set of samples showed significant in vivo response to venetoclax3 compared to our studies. This difference may be due to differences in the methodology used in the in vivo studies and/or the biology of the sam-

Figure 2. Alisertib effects on TCF3-HLF HAL01 acute lymphoblastic leukemia cells. (A) Dose-response to varying concentrations of alisertib. Cells were exposed to alisertib for 3 days then assessed for viability with MTS. Viability was normalized to untreated cells. All data points were performed in triplicate. (B) Assessment of apoptosis with annexin V staining. Cells were exposed to 100 nM alisertib (IC90) for 24 hours, then assessed for annexin V staining using Guava Nexin within the population of cells. All data points were performed in triplicate and compared by unpaired student’s t-test (****P<0.0001). (C) G2/M cell cycle arrest after exposure to alisertib. HAL01 cells were incubated with 100 nM alisertib for 24 hours, fixed with 70% methanol and stained with propidium iodide (PI), then analyzed by flow cytometry for PI intensity. (D) Immunocytochemical fluorescence of AURKA within an asynchronous population of cells. HAL01 cells were exposed to 100 nM alisertib for 24 hours in culture, then fixed with 4% paraformaldehyde, permeabilized with triton X-100 and stained with rabbit antiAURKA, mouse anti-a-tubulin, secondary anti-rabbit FITC, secondary anti-mouse Texas Red, and DAPI. (E) Quantification of cells with spindle poles. Within an asynchronous population, the number of nuclei with 0 (none), 1 (single), 2 (double), or more (>2) spindle poles were counted. Each population was compared by an unpaired student’s t-test (*P<0.05).

2992

haematologica | 2021; 106(11)

Letters to the Editor

Figure 3. In vivo response of TCF3-HLF acute lymphoblastic leukemia xenografts. Kaplan-Meier plots for event-free-survival (EFS) comparing control to drug treated cohorts of five separate TCF3-HLF acute lymphoblastic leukemia (ALL) samples. Cohorts of five animals were injected with 1x106 leukemia cells per animal by tail vein. Within the first week there was some attrition due to animal loss in samples 3.316, 3.332, and 3.310 (where n= the number of animals per cohort). One week after injection cohorts of four to five individuals began treatment with vehicle (control), dasatinib (40 mg/kg/day), venetoclax (25 mg/kg/day), combination of dasatinib (40 mg/kg/day) and venetoclax (25 mg/kg/day) (das/ven), or alisertib (30 mg/kg/day) by oral gavage for 5 days each week. Animals were monitored weekly for peripheral blood chimerism for human CD45 and murine CD45 by flow cytometry and daily well-being. An event was defined as peripheral blood chimerism ≥10% or if the animal became moribund. The top row represents type 1 TCF3-HLF rearrangements, while the bottom row represents type 2 rearrangements. Statistical comparison performed by log-rank test (*P<0.05; **P<0.01).

ples within each study. Further, earlier studies did not identify AURKA as a potential target, although there appear to be similar trends with their studies of other AURK inhibitors.3,4 One very interesting observation was the sensitivity of TCF3-HLF ALL to YM155 both in our studies as well as Fischer et al.3 Unfortunately, this compound is not currently under significant clinical evaluation making the future of this drug unknown. There are also novel immunotherapeutic targets that may be pursued in future therapies including the use of the CD19 target using blinatumomab for therapeutic response followed by stem cell transplantation.7 Our studies suggest that TCF3-HLF ALL may have clinical benefit from alisertib. This compound has been and is currently under investigation for multiple different disease types. The Children's Oncology Group (COG) phase II study (clinicaltrials gov. Identifier: ADVL0921) achieved their target plasma level, but minimal objective responses were seen with significant toxicities. Their studies concurrently ran xenograft experiments with alisertib using twice daily dosing to show similar therapeutic effects, with their concern that the dose tolerated in the murine model did not reflect what is tolerated in the patient. In contrast we used 30 mg/kg/dose daily in our xenograft studies as described in Manfredi et al.15 this dose was not only well-tolerated for 21 days but had similar to improved response compared to 20 mg/kg given twice daily. These findings would suggest the dosing needed for therapeutic effect for TCF3-HLF ALL may be less in patients. Clearly our studies do not describe complete disease control with the single agent alisertib, rather a delay in disease progression as all xenograft animals

haematologica | 2021; 106(11)

eventually engrafted and succumbed to their disease. It remains unknown as to whether addition of this drug in combination therapies (e.g., conventional chemotherapy or other targeted agents such as venetoclax) may provide a beneficial response to select tumors that are dependent on AURKA. Because of the rarity of this tumor, it will take collaborative studies to test this compound in patients with TCF3-HLF ALL. Jessica Leonard,1 Joelle S.J. Wolf,2 Michelle Degnin,1 Christopher A. Eide,1 Dorian LaTocha,1 Kyle Lenz,2 Beth Wilmot,3 Charles G. Mullighan,4 Mignon Loh,5 Stephen P, Hunger,6 Brian J. Druker,1,7 Marc M. Loriaux,8 Jeffrey W. Tyner1,9 and Bill H. Chang2 1 Division of Hematology and Medical Oncology, Department of Medicine and OHSU Knight Cancer Institute, OHSU, Portland, OR; 2Division of Hematology and Oncology, Department of Pediatrics and OHSU Knight Cancer Institute, OHSU, Portland, OR; 3Division of Bioinformatics and Computational Biology, Department of Medical Informatics and Clinical Epidemiology, and OHSU Knight Cancer Institute, OHSU, Portland, OR; 4Department of Pathology, St Jude Children's Research Hospital and University of Tennessee Health Science Center, Memphis, TN; 5Department of Pediatrics, Benioff Children's Hospital, Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, San Francisco, CA; 6 Department of Pediatrics and Center for Childhood Cancer Research, Children’s Hospital of Philadelphia and Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; 7Howard Hughes Medical Institute, Chevy Chase, MD; 8Department of Pathology and OHSU Knight Cancer Institute, OHSU, Portland, OR and 9 Department of Cell, Molecular, and Cancer Biology, and OHSU Knight Cancer Institute, OHSU, Portland, OR, USA 2993

Letters to the Editor

Correspondence: BILL H. CHANG - changb@ohsu.edu doi:10.3324/haematol.2021.278692 Received: March 2, 2021. Accepted: June 21, 2021. Pre-published: July 8, 2021. Disclosures: CGM received research funding from AbbVie and Pfizer, speaking fees and stock from Amgen; Advisory Board Illumina; JWT received research support from Agios, Aptose, Array, AstraZeneca, Constellation, Genentech, Gilead, Incyte, Janssen, Petra, Seattle Genetics, Syros, Takeda, and Tolero; SPH has received consulting fees from Novartis, honoraria from Jazz, Amgen and Servier, and owns common stock in Amgen; BJD’s potential competing interests include: SAB: Aileron Therapeutics, Therapy Architects (ALLCRON), Cepheid, Vivid Biosciences, Celgene, RUNX1 Research Program, Nemucore Medical Innovations, Novartis, Gilead Sciences (inactive), Monojul (inactive); SAB and stock: Aptose Biosciences, Blueprint Medicines, EnLiven Therapeutics, Iterion Therapeutics, GRAIL; scientific founder: MolecularMD (inactive, acquired by ICON); board of directors and stock: Amgen, Vincerx Pharma; board of directors: Burroughs Wellcome Fund, CureOne; joint steering committee: Beat AML LLS; founder: VB Therapeutics; sponsored research agreement: EnLiven Therapeutics; clinical trial funding: Novartis, Bristol-Myers Squibb, Pfizer; royalties from Patent 6958335 (Novartis exclusive license) and OHSU and Dana-Farber Cancer Institute (one Merck exclusive license and one CytoImage, Inc. exclusive license). Contributions: JL performed the research and wrote the manuscript; JSW performed the research and wrote the paper; MD analyzed the data and wrote the manuscript; CAE analyzed the data and wrote the manuscript; DL performed the flow cytometry and wrote the paper; KL performed the research and wrote the paper; BW contributed analytical tools, analyzed the data and wrote the manuscript; CGM contributed vital reagents and wrote the manuscript; ML contributed vital reagents and wrote the manuscript; SPH contributed vital reagents and wrote the manuscript; BJD contributed vital reagents and wrote the manuscript; MML contributed vital reagents and wrote the manuscript; JWT designed the research, contributed vital reagents and wrote the manuscript; BHC designed the research, performed the research, analyzed the data and wrote the manuscript. Acknowledgments: we wish to express our deepest gratitude to the patients and families that agreed to submit these precious samples to their respective biorepositories. Funding: these studies were partially funded by the Hyundai Hope on Wheels, Newman’s Own Foundation, and Tucker’s Toy box, American Lebanese Syrian Associated Charities of St. Jude Children’s Research Hospital, NCI P30 CA021765 and R35 CA197695.

2994

Data sharing statement: for original data, please contact changb@ohsu.edu.

References 1. Gu Z, Churchman ML, Roberts KG, et al. PAX5-driven subtypes of B-progenitor acute lymphoblastic leukemia. Nat Genet. 2019; 51(2):296-307. 2. Glover JM, Loriaux M, Tyner JW, Druker BJ, Chang BH. In vitro sensitivity to dasatinib in lymphoblasts from a patient with t(17;19)(q22;p13) gene rearrangement pre-B acute lymphoblastic leukemia. Pediatr Blood Cancer. 2012;59(3):576-579. 3. Fischer U, Forster M, Rinaldi A, et al. Genomics and drug profiling of fatal TCF3-HLF-positive acute lymphoblastic leukemia identifies recurrent mutation patterns and therapeutic options. Nat Genet. 2015;47(9):1020-1029. 4. Frismantas V, Dobay MP, Rinaldi A, et al. Ex vivo drug response profiling detects recurrent sensitivity patterns in drug-resistant acute lymphoblastic leukemia. Blood. 2017;129(11):e26-e37. 5. Panagopoulos I, Micci F, Thorsen J, et al. A novel TCF3-HLF fusion transcript in acute lymphoblastic leukemia with a t(17;19)(q22;p13). Cancer Genet. 2012;205(12):669-672. 6. Minson K, Vear S, Prasad P, Domm J, Borinstein S, Frangoul H. The prognostic significance of t(17;19) on outcome of children with preB acute lymphoblastic leukemia (ALL). Pediatr Blood Cancer. 2011; 56(6):927. 7. Mouttet B, Vinti L, Ancliff P, et al. Durable remissions in TCF3-HLF positive acute lymphoblastic leukemia with blinatumomab and stem cell transplantation. Haematologica. 2019;104(6):e244-e247. 8. Ma X, Edmonson M, Yergeau D, et al. Rise and fall of subclones from diagnosis to relapse in pediatric B-acute lymphoblastic leukaemia. Nat Commun. 2015;6:6604. 9. Watanabe A, Inukai T, Kagami K, et al. Resistance of t(17;19)-acute lymphoblastic leukemia cell lines to multiagents in induction therapy. Cancer Med. 2019;8(11):5274-5288. 10. Tyner JW, Tognon CE, Bottomly D, et al. Functional genomic landscape of acute myeloid leukaemia. Nature. 2018;562(7728):526-531. 11. Mosse YP, Fox E, Teachey DT, et al. A phase II study of alisertib in children with recurrent/refractory solid tumors or leukemia: Children's Oncology Group Phase I and Pilot Consortium (ADVL0921). Clin Cancer Res. 2019;25(11):3229-3238. 12. Adams R, Maiato H, Earnshaw W, Carmena M. Essential roles of Drosophila inner centromere protein (INCENP) and aurora B in histone H3 phosphorylation, metaphase chromosome alignment, kinetochore dysjunction, and chromosome segregation. J Cell Biol. 2001; 153(4):856-880. 13. Zhou H, Kuang J, Zhong L, et al. Tumour amplified kinase STK15/BTAK induces centrosome amplification, aneuploidy and transformation. Nat Genet. 1998;20(2):189-193. 14. Leonard JT, Rowley JS, Eide CA, et al. Targeting BCL-2 and ABL/LYN in Philadelphia chromosome-positive acute lymphoblastic leukemia. Sci Transl Med. 2016;8(354):354ra114. 15. Manfredi MG, Ecsedy JA, Chakravarty A, et al. Characterization of Alisertib (MLN8237), an investigational small-molecule inhibitor of aurora A kinase using novel in vivo pharmacodynamic assays. Clin Cancer Res. 2011;17(24):7614-7624.

haematologica | 2021; 106(11)

Letters to the Editor

IL4-STAT6 signaling induces CD20 in chronic lymphocytic leukemia and this axis is repressed by PI3Kδ inhibitor idelalisib Efforts to combine anti-CD20 antibodies (such as rituximab or obinutuzumab) with BCR inhibitors or venetoclax lead to the necessity to better understand the largely unclear mechanisms of CD20 regulation and its function(s) (reviewed in Pavlasova and Mraz1). This is underscored by the observation that in chronic lymphocytic leukemia (CLL) the combination of ibrutinib with rituximab fails to provide a clinical benefit in comparison to ibrutinib alone2 likely as ibrutinib downmodulates CD20 levels.1,3-5 PI3Kδ inhibitor idelalisib has been approved in

combination with rituximab or ofatumumab;6 however, it remains unclear if idelalisib affects CD20 levels or function(s). Here we show for the first time that single-agent idelalisib therapy in CLL leads to CD20 downmodulation in vivo by interfering with a previously unknown mechanism of CD20 transcriptional regulation via the IL4STAT6 axis. We describe a novel mechanism of CD20 regulation in CLL B cells, which has implications for combinatorial therapy with PI3K inhibitors. We have recently shown a concurrent upregulation of CD20 and cell surface immunoglobulin M (IgM) in CLL cells from immune niches and demonstrated that this functionally serves to increase BCR signaling propensity.3,7 The SDF1 (CXCL12) chemokine induces approximately 30-50% upregulation of CD20,3 however, this

Figure 1. IL4 upregulates CD20 expression via STAT6. (A) Cell surface level of CD20 after interleukin 4 (IL4) treatment (20 ng/mL, PeproTech) for 24 hours (hrs) (n=25), 48 hrs (n=23), or 72 hrs (n=22) in comparison to untreated control cells (ctrl). (B) Representative immunoblot of CD20 and pSTAT6 (Tyr 641) protein levels in chronic lymphocytic leukemia (CLL) cells after IL4 stimulation (24-72 hrs). (C) Densitometric quantification of CD20 protein levels for independent replicates of the experiment described in (B) (n=17; 24-72 hrs). Untreated control (ctrl) without IL4 was set as 1 and compared to the other samples. (D) Normalized cell surface CD20 levels in primary CLL cells treated by SDF1a (100 ng/mL, PeproTech), IL4 (20 ng/mL), or their combination (SDF1a + IL4) for 24-48 hrs (n=7). The untreated ctrl was set as 1. (E) Peripheral blood CLL cells were electroporated with small interfering RNA (siRNA) against STAT6 (siSTAT6, 500 nM; Dharmacon) or negative control (Neg.Ctrl). IL4 (20 ng/ml) was added 48 hrs after transfection and cells were cultured for another 24 hrs. For the immunoblot, β-actin was used as a loading control and pSTAT6 (Tyr 641)/tSTAT6 levels were assessed. (F) Densitometric quantification of CD20 protein levels for independent replicates (n=4) of the experiment described in (E). Neg.Ctrl without IL4 was set as 1 and compared to the other samples. (G) Peripheral blood CLL cells were pretreated with pSTAT6 inhibitor (AS1517499, 1 mM, Selleckchem) for 12 hrs. Subsequently, IL4 (20 ng/mL) was added to the media, and cells were cultured for another 24 hrs. CD20 expression was determined by real-time quantitative polymerase chain reaction (TaqMan, ABI), and the expression of CD20 was normalized to endogenous control HPRT (n=6). (H) Chromatin immunoprecipitation analysis of samples that were immunoprecipitated with anti-STAT6 antibody in comparison with immunoglobulin G (IgG) antibody before (-) and after (+) IL4 stimulation (n=5; 40 ng/mL, 30 minutes). IgG antibody was used as an isotype control. For all in vitro experiments in Figure 1 and 2 CLL cells were purified by RosetteSep Human B Cell Enrichment Cocktail (Stemcell Technologies) to obtain purity ≥95% of CD5+CD19+ cells. For immunoblots, b-actin was used as a loading controls. In all experiments, the statistical difference was tested using a paired t-test, and the error bars indicate standard error of the mean.

haematologica | 2021; 106(11)

2995

Letters to the Editor

Eii

Fii

Eiii

Eiv

Hii

Figure 2. CD20 is downmodulated by idelalisib and embedded in the IL4 pathway. (A) Cell-surface CD20 levels and (B) relative expression of CD20 mRNA in paired samples before (Pre) and after 5 weeks (idelalisib-week 5) and 9 weeks (idelalisib-week 9) of idelalisib therapy in vivo (cell surface CD20: n=1 week 4, n=6 week 5, n=6 week 9; CD20 mRNA: n=6 week 5 and week 9). Chronic lymphocytic leukemia (CLL) cells were isolated by density centrifugation (Ficoll-Paque) followed by magnetic anti-CD3 MicroBeads separation (Miltenyi Biotec) or in some cases negative selection with RosetteSep Human B Cell Enrichment Cocktail (Stemcell Technologies) was used to obtain purity of ≥95% of CD5+19+ cells. (C) Representative examples (n=3) of CD20 protein levels in CLL cells obtained before (Pre) and during idelalisib therapy in vivo (week 4/5 and 9). (D) CLL cells (purity ≥95%) were pretreated with idelalisib (2 mM, Selleckchem) for 4 hours (hrs) or plerixafor (5 mg/mL, Selleckchem) for 4 hrs and then SDF1 (100 ng/mL, CXCR4 ligand) or interleukin 4 (IL4) (20 ng/mL) were added into the media for 24 hrs (for the result of the 48 hrs treatment with SDF1/IL4 see the Online Supplementary Figure S3A). Cell surface CD20 levels were measured and the results are visualized as a fold-change to untreated control (ctrl) (n=12). Viable CLL cells were gated for assessment of cell surface CD20 levels. The pretreatment of CLL cells by idelalisib or plerixafor (CXCR4 inhibitor) for 4 hrs was performed to ensure a full inhibition of the pathways before exposure to the receptor ligands. (Ei) Representative immunoblot of CLL cells treated in vitro with idelalisib (2 mM; 48 hrs) and subsequently stimulated by IL4 (40 ng/mL; 3 minutes [min]). (Eii) Densitometric quantification of pSTAT6 (Tyr 641) protein levels for independent replicates (n=7) of the experiment described in (Ei). (Eiii) Representative immunoblot of CLL cells treated in vitro with idelalisib (2 mM; 48 hrs) then washed twice with clean culture media and stimulated by IL4 in full media (20 ng/mL; 24 hrs). (Eiv) Densitometric quantification of CD20 protein levels for independent replicates of the experiment described in (Eiii) (n=7). (Fi) Representative immunoblot of pSTAT6 (Tyr 641) downmodulation after pretreatment of CLL cells with idelalisib (2 mM; 4 hrs) followed by IL4 stimulation (40 ng/mL; 3 and 5

Legend continued on the following page

2996

haematologica | 2021; 106(11)

Letters to the Editor

min). (Fii) Densitometric quantification of pSTAT6 (Tyr 641) protein level for independent replicates (n=5) of the experiment described in (Fi). (G) Representative immunoblot of CLL cells transfected with small interfering RNA (siRNA) against the PI3Kδ isoform (siPI3Kδ, 500 nM, Dharmacon) or a negative control (Neg.Ctrl). Seventy-two hrs after transfection, cells were stimulated by IL4 (40 ng/mL; 3 min) and STAT6 phosphorylation (Tyr 641) was assessed. Phosphorylation of AKT (Ser 473) was used as a surrogated marker for PI3Kδ downmodulation after siPI3Kδ since we were not able to detect PI3Kδ protein due to issues with antiPI3Kδ primary antibody. (Hi) Representative immunoblot of MEC1 cells transfected with siRNA against CD20 (siCD20, 500 nM, Thermo Fisher Scientific) or negative control (Neg.Ctrl). Forty-eight hrs after the transfection, cells were stimulated by IL4 (40 ng/mL; 3 min). (Hii) Densitometric quantification of pSTAT6 (Tyr 641) protein level for independent replicates (n=4) of the experiment described in (Hi). Neg.Ctrl without IL4 was set as 1 and compared to the other samples. (I) Representative immunoblot of primary CLL cells transfected with siRNA against CD20 (siCD20) or negative control (Neg.Ctrl), cultured for 48 hrs, and then stimulated by IL4 (40 ng/mL; 3 min). For immunoblots, β-actin or GAPDH were used as a loading controls. In all experiments, the statistical difference was tested using a paired t-test, and the error bars indicate standard error of the mean.

cannot fully explain the ~2-fold higher CD20 levels in CLL cells from immune niches.3,7 Here we hypothesize that the same factor inducing cell surface IgM in the CLL microenvironment, namely interleukin 4 (IL4) produced by T cells,8 might also be inducing CD20. Indeed, stimulating primary CLL cells with IL4 led to a significant upregulation of CD20 (mean fold-change [FC] =1.6-3.1 [time span, 24-72 hrs], n=25; Figure 1A to C) and IgM on the cell-surface (Online Supplementary Figure S1A). IL4 also induced cell surface CD20 in normal B cells (~2-fold induction; Online Supplementary Figure S1B). This is in line with previous anecdotal observations suggesting a role for IL4 in CD20 regulation.9 We next compared the effects of SDF1 and IL4 on CD20 levels and noted that each factor independently induces CD20, and their combination has a more potent effect (Figure 1D). IgM induction was used as a control in this experiment since it is known to only be induced by IL48 and not SDF1 (Online Supplementary Figure S1C). The effect of IL4 on CD20 was transcriptional since CD20 mRNA (gene MS4A1) was induced similarly to its cell surface levels (Online Supplementary Figure S1D). The IL4 effect was observed irrespective of the immunoglobulin heavy-chain variable region (IGHV) status or the presence of chromosomal aberrations (Online Supplementary Figure S1E to J and data not shown). It is known that IL4 supports CLL cell viability, and to avoid any potential confounding effects from the different viability on CD20 levels, we gated on viable cells and we also separately analyzed CLL samples with comparable viability in control and IL4-treated cells (Online Supplementary Figure S1K to O). Altogether, the data show that IL4 (produced by T cells) induces CD20 in CLL cells. We next hypothesized that CD20 induction by IL4 might be mediated by STAT6 as it is a well-known key IL4 signaling effector. Indeed, STAT6 silencing by small interfering RNA (siRNA) or a specific STAT6 inhibitor (AS1517499) impaired the IL4-induced CD20 expression (Figure 1E to G). siRNA against STAT6 or STAT6 inhibitor did not have any effect on cell viability (Online Supplementary Figure S2A, and data not shown). Chromatin immunoprecipitation revealed that STAT6 binds to the CD20 promotor in CLL cells (-2 nucleotides from transcription start-site [TSS]), and the occupancy of this novel site by STAT6 was significantly increased (~2-fold) immediately after IL4 treatment (30 minutes, Figure 1H). We also observed a weaker STAT6 binding at another putative binding site (-197 nucleotides from TSS; Figure 1H; Online Supplementary Figure S2B and C). We also noted higher STAT6 phosphorylation in freshly obtained unstimulated CXCR4dim CD5bright CLL subpopulation in comparison to CXCR4bright CD5dim cells using flow cytometry (Online Supplementary Figure S2D and E). The CXCR4dimCD5bright cells are regarded as an intraclonal CLL cell subpopulation that has recently exited immune niches versus resting CXCR4brightCD5dim cells.3,7 However, we could not reliably detect phosphorylated STAT6 in CLL cells by a less sensitive immunoblotting technique suggesting that its levels were very low in peripheral haematologica | 2021; 106(11)

blood in general. Altogether, these data demonstrate a novel direct role for STAT6 in transcriptional CD20 regulation upon IL4 stimulation. This is potentially a mechanism coupling the regulation of two molecules (IgM and CD20) which are required for BCR signaling3,7,8 and which are both induced by IL4 produced in the immune microenvironment. This mechanism might be a part of the pathway crosstalk by which IL4 alternates the BCR pathway, a phenomenon described by others.10 Ibrutinib leads to a significant reduction of CD20 levels in CLL by interfering with SDF1 signaling.1,3,4 Notably, in vivo therapy with idelalisib as a single agent prominently reduced CD20 mRNA and protein levels within several weeks (Figure 2A to C). Next, we tested if idelalisib interferes with the IL4- or SDF1-dependent induction of CD20. Idelalisib clearly inhibited CD20 induction by IL4, but not by SDF1 (Figure 2D; Online Supplementary Figure S3A). The repression of CD20 induction by idelalisib was observed with doses ≥0.5 mM (Online Supplementary Figure S3B). Therefore, we hypothesized that PI3K is involved in STAT6 activation, while the CXCR4-SDF1 axis does not include STAT6. Indeed, CXCR4 signaling does not induce any STAT6 phosphorylation (Online Supplementary Figure S3C), while idelalisib treatment in vitro (48 hours) significantly impaired STAT6 phosphorylation and CD20 induction by IL4 (Figure 2Ei to Eiv). In order to exclude secondary effects of idelalisib or decreased cell viability, we also performed a short pretreatment of cells by idelalisib (4 hours) and observed an identical decrease in pSTAT6 levels (Figure 2F, and data not shown; idelalisib had no effect on cell viability at 4 hours and a minor effect at 48 hours). In vitro, the treatment of CLL cells with idelalisib (1 or 2 mM) also led to a reduction of CD20 levels prior to IL4 exposure (Online Supplementary Figure S3D; Figure 2Eiii to Eiv, and data not shown), however, this was less pronounced than during IL4 stimulation. This suggests that idelalisib might affect other CD20 regulators besides STAT6 or low-level basal STAT6 activity influences also “basal” CD20 transcription. Indeed, we detected some STAT6 phosphorylation in unstimulated CLL cells (see above). The silencing of PI3Kδ by siRNA decreased the STAT6 phosphorylation after IL4 (Figure 2G; Online Supplementary Figure S3E), indicating a direct role of the PI3Kδ isoform in IL4-induced CD20 expression and a specific on-target effect of idelalisib leading to CD20 downmodulation. This is in line with studies describing the involvement of PI3Kδ in IL4 signaling of normal B cells.11 However, idelalisib does not impair IL4-induced increase in CLL cell viability in vitro,12 suggesting that the pro-survival effect of IL4 is independent of STAT6 or that a weaker STAT6 phosphorylation is sufficient to provide a pro-survival signal. Altogether, our data demonstrate that PI3Kδ inhibition interferes with CD20‘s transcriptional activation by the IL4-STAT6 axis. Besides the effects of idelalisib on STAT6 phosphorylation and CD20 levels, we also noticed a minor decrease in cell surface IL4 receptor (IL4Ra) levels after 48 hours of idelalisib treatment in vitro (Online Supplementary Figure 2997

Letters to the Editor

Figure 3. Schematic overview for CD20 downmodulation by idelalisib via impaired IL4-STAT6 axis. In chronic lymphocytic leukemia (CLL), IL4-STAT6 axis upregulates CD20 gene expression through the STAT6 phosphorylation and its direct binding to CD20 (MS4A1) promoter. The IL4-STAT6-CD20 axis is inhibited by PI3Kδ inhibitor idelalisib. NFκB and FoxO1 represent other two known regulators of CD20 transcription in CLL (NFκB is a positive regulator, and FoxO1 is an indirect negative regulator).1,4,5 This figure was created with tools at BioRender.com.

S3F). However, this decrease is not responsible for the reduced responsiveness to IL4, since incubation of CLL cells with idelalisib for 4 hours also impaired STAT6 phosphorylation after IL4 (Figure 2F), but IL4Ra or CD20 levels remained unchanged during these short-term experiments (Online Supplementary Figure S3G and H). Moreover, we did not observe a significant downmodulation of IL4Ra levels during idelalisib therapy in vivo (Online Supplementary Figure S3I). This supports a direct role of PI3K in STAT6 signaling and the on-target effect of idelalisib. We next asked if CD20 downmodulation by idelalisib might affect the function(s) of CD20 in some signaling pathways other than BCR. To our surprise, CD20 silencing by siRNA significantly impaired the response to IL4 in MEC1 and primary CLL cells (Figure 2H and I; Online Supplementary Figure S3J), and had a minor effect on the phosphorylation of IKK after CpG or CD40L (data not shown). This suggests that CD20 is embedded in several receptor-associated pathways, including the regulation of IL4 signaling propensity. IL4 maintains CD20 levels via STAT6 activation, and CD20 increases responsiveness of CLL cells to IL4 via a still unclear mechanism. We noted that IL4 treatment in CLL cells does not lead to internalization of the IL4 receptor (data not shown), allowing cells to respond to IL4 continuously. These observations of the role of CD20 in T-cell interactions are in line with studies of CD20 in normal B cells, since CD20 knockout in mice 2998

or CD20 loss in humans leads to defects in T-cell dependent immunity.1,13 However, full understanding of this phenome will require further insight into CD20 functions and interaction partners, which is a long-standing question in the field.1 Altogether, here we describe a novel CD20 regulatory axis and reveal for the first time that T- cell interactions via IL4 induce CD20 transcription via STAT6 binding to its promotor (summarized in Figure 3). Then PI3Kδ is involved in CD20 induction by the IL4-STAT6 axis, and consequently, idelalisib therapy represses CD20 in CLL. Idelalisib has been approved in combination with antiCD20 antibodies without a comparison to the adminstration of single agent idelalisib,6 and is currently mainly used in therapy of relapsed/refractory disease and/or in cases of a BTK inhibitor intolerance. Downmodulation of CD20 by idelalisib likely reduces the rituximab-induced apoptosis and CDC, since complement fragment deposition is closely dependent on cell surface CD20 levels,14 and might impair ADCC/ADCP since these are also facilitated by opsonization of target cells with complement.15 Indeed, idelalisib inhibited in vitro the immune cell-mediated mechanisms induced by anti-CD20 antibodies,16 but this requires further investigation and might also include the effect of idelalisib on effector cells. Based on our data the benefit of rituximab addition should be tested in a clinical trial since this might fail to show improvement of progression-free/ overall survival, similarly to the lack of haematologica | 2021; 106(11)

Letters to the Editor

benefit for ibrutinib plus rituximab combination versus ibrutinib alone.2 It would be interesting to also test the combination of idelalisib/PI3Kδ inhibitor with anti-CD20 antibodies whose efficacy is less dependent on CD20 levels such as obinutuzumab. In conclusion, any clinicallyused inhibitor blocking PI3Kδ or interfering with the IL4STAT6 pathway will reduce CD20 expression with potential consequences for combinatorial therapy. Veronika Sandova,1,2 Gabriela Mladonicka Pavlasova,1 Vaclav Seda,1,2 Katerina Amruz Cerna,1 Sonali Sharma,1 Veronika Palusova,1 Yvona Brychtova,2 Sarka Pospisilova,2 Stacey M. Fernandes,3 Anna Panovska,2 Michael Doubek,2 Matthew S. Davids,3 Jennifer R. Brown,3 Jiri Mayer2 and Marek Mraz1,2 1 Central European Institute of Technology, Masaryk University, Brno, Czech Republic; 2Department of Internal Medicine, Hematology and Oncology, University Hospital Brno and Faculty of Medicine, Masaryk University, Brno, Czech Republic and 3Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA Correspondence: MAREK MRAZ - marek.mraz@email.cz doi:10.3324/haematol.2021.278644 Received: February 25, 2021. Accepted: June 22, 2021. Pre-published: July 1, 2021. Disclosures: JRB has served as a consultant for Abbvie, Acerta, Astra-Zeneca, Beigene, Catapult, Dynamo Therapeutics, Eli Lilly, Juno/Celgene, Kite, MEI Pharma, Nextcea, Novartis, Octapharma, Pfizer, Rigel, Sunesis, TG Therapeutics, Verastem; received honoraria from Janssen; received research funding from Gilead, Loxo, Sun and Verastem; and served on data safety monitoring committees for Invectys; MSD has received institutional research funding Ascentage Pharma, Astra-Zeneca, Genentech, MEI Pharma, Novartis, Pharmacyclics, Surface Oncology, TG Therapeutics and Verastem, and consulting fees from AbbVie, Adaptive Biotechnologies, Ascentage Pharma, AstraZeneca, BeiGene, Celgene, Eli Lilly, Genentech, Gilead Sciences, Janssen, MEI Pharma, Merck, Pharmacyclics, TG Therapeutics, and Verastem. Other authors declare no competing financial interests. Contributions: VS performed experiments, analyzed data, and wrote the paper; GMP, VSe, KAC, SS and VP performed experiments; YB, SP, SMF, AP, MD, MSD, JRB and JM provided samples and clinical data; MM designed the study, interpreted the data, and wrote the paper. All the other authors edited and approved the paper for submission. Funding: this work was supported by the Ministry of Health of the Czech Republic, grant no. NU20-03-00292. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 Research and Innovation Program (grant agreement no. 802644). All rights reserved.

haematologica | 2021; 106(11)

Data sharing statement: original data and protocols are available without restrictions. These data can be obtained by contacting of corresponding author.

References 1. Pavlasova G, Mraz M. The regulation and function of CD20: an "enigma" of B-cell biology and targeted therapy. Haematologica. 2020;105(6):1494-1506. 2. Burger JA, Sivina M, Jain N, et al. Randomized trial of ibrutinib vs ibrutinib plus rituximab in patients with chronic lymphocytic leukemia. Blood. 2019;133(10):1011-1019. 3. Pavlasova G, Borsky M, Seda V, et al. Ibrutinib inhibits CD20 upregulation on CLL B cells mediated by the CXCR4/SDF-1 axis. Blood. 2016;128(12):1609-1613. 4. Skarzynski M, Niemann CU, Lee YS, et al. Interactions between ibrutinib and anti-CD20 Antibodies: competing effects on the outcome of combination therapy. Clin Cancer Res. 2016;22(1):86-95. 5. Pyrzynska B, Dwojak M, Zerrouqi A, et al. FOXO1 promotes resistance of non-Hodgkin lymphomas to anti-CD20-based therapy. Oncoimmunology. 2018;7(5):e1423183. 6. 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. 7. Pavlasova G, Borsky M, Svobodova V, et al. Rituximab primarily targets an intra-clonal BCR signaling proficient CLL subpopulation characterized by high CD20 levels. Leukemia. 2018;32(9):20282031. 8. Guo BC, Zhang L, Chiorazzi N, Rothstein TL. IL-4 rescues surface IgM expression in chronic lymphocytic leukemia. Blood. 2016; 128(4):553-562. 9. Venugopal P, Sivaraman S, Huang XK, Nayini J, Gregory SA, Preisler HD. Effects of cytokines on CD20 antigen expression on tumor cells from patients with chronic lymphocytic leukemia. Leuk Res. 2000; 24(5):411-415. 10. Guo BC, Rothstein TL. B cell receptor (BCR) cross-talk: IL-4 creates an alternate pathway for BCR-induced ERK activation that is phosphatidylinositol 3-kinase independent. J Immunol. 2005; 174(9): 5375-5381. 11. Bilancio A, Okkenhaug K, Camps M, et al. Key role of the p110 delta isoforrn of PI3K in B-cell antigen and IL-4 receptor signaling: comparative analysis of genetic and pharmacologic interference with p110 delta function in B cells. Blood. 2006;107(2):642-650. 12. Herman SEM, Gordon AL, Wagner AJ, et al. Phosphatidylinositol 3-kinase-delta inhibitor CAL-101 shows promising preclinical activity in chronic lymphocytic leukemia by antagonizing intrinsic and extrinsic cellular survival signals. Blood. 2010;116(12):2078-2088. 13. Morsy DE, Sanyal R, Zaiss AK, Deo R, Muruve DA, Deans JP. Reduced T-Dependent Humoral Immunity in CD20-Deficient Mice. J Immunol. 2013;191(6):3112-3118. 14. Golay J, Lazzari M, Facchinetti V, et al. CD20 levels determine the in vitro susceptibility to rituximab and complement of B-cell chronic lymphocytic leukemia: further regulation by CD55 and CD59. Blood. 2001;98(12):3383-3389. 15. Campagne MV, Wiesmann C, Brown EJ. Macrophage complement receptors and pathogen clearance. Cell Microbiol. 2007;9(9):20952102. 16. Da Roit F, Engelberts PJ, Taylor RP, et al. Ibrutinib interferes with the cell-mediated anti-tumor activities of therapeutic CD20 antibodies: implications for combination therapy. Haematologica. 2015; 100(1): 77-86.

2999

Letters to the Editor

A novel classification of hematologic conditions in patients with Fanconi anemia Patients with the inherited bone marrow failure (BMF) syndrome Fanconi anemia (FA) have an increased risk of myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML).1,2 The bone marrow (BM) in FA patients is typically hypocellular and it can mimic that of other conditions such as acquired refractory cytopenia of childhood.3 In patients with FA, the diagnosis of MDS cannot be made based on the presence of cytopenia and myelodysplastic features alone (with the exception of distinct subtypes such as cases with a rapid rise in cellularity in which MDS needs to be considered), but requires an elevated blast count and/or other definite signs of transformation, such as duplication of chromosome 3q (3q+), deletion 7q (7q-), or monosomy 7 (-7),4-6 aberrations known to be associated with a poor prognosis in FA patients.5,7,8 In contrast, in the absence of an elevated blast percentage, not every clonal aberration warrants the diagnosis of MDS because not all aberrations, such as duplication of chromosome 1q (1q+), deletion 20q (20q-), or deletion 6p (6p-), are associated with transformation.5,6,9-11 It is increasingly recognized that FA patients display clonal mosaicism generally correlated with a more common occurrence of hematologic neoplasia and poorer prognosis.11 An internationally agreed classification for the diagnosis of hematologic conditions in patients with FA is lacking, but a classification system is important for the clinical care and treatment of patients with this rare disease. Here, we propose an algorithm for the diagnosis of benign and malignant hematologic conditions in FA that is based on blast percentage and cytogenetics (Figure 1). Within this classification, we propose the term “aberration of indeterminate potential in patients with FA” (FAAIP), to classify patients with clonal aberrations, such as duplication of 1q, which remain stable over a prolonged period without transformation. The classification includes: (i) FA-BMF, defined by normal blast percentage and normal cytogenetics in a patient with cytopenia; (ii) FA-AIP, defined by normal blast percentage and AIP; (iii) FA-MDS-without excess blasts or FA-MDS-non-EB, defined by normal blast percentage and clonal aberrations unambiguously associated with transformation; (iv) FA-MDS with excess blasts or FA-MDS-EB, defined by increased blast percentage in BM ≥5% to <20% or in peripheral blood (PB) ≥2% to <20%, (but BM and PB <20% blasts); and (v) FA-AML, defined by increased blast percentage in BM or PB ≥20%. One main intention of BM surveillance is to identify patients with FA-MDS-non EB prior to transformation with elevated blast count (FAMDS-EB or FA-AML) in order to increase overall survival following early hematopoietic stem cell transplantation. We used this diagnostic algorithm to analyze data from 86 patients enrolled in the German FA registry (FAR01)/ Cancer Predisposition Registry (DRKS00017382). FAMDS-non EB and FA-MDS-EB patients were also enrolled in the registry of the European Working Group of Myelodysplastic Syndrome in Childhood (EWOGMDS). The study was approved by the ethical review boards of Hannover Medical School and Freiburg University. Patients were analyzed cytogenetically as part of their clinically indicated cancer and BMF surveillance program. Cytogenetic analyses were performed in our reference laboratory on BM aspirates by karyotyping and fluorescence in situ hybridization. In 62 of the 86 patients, serial cytogenetic analyses were performed (mean, 4.9 3000

analyses per patient). The fluorescence in situ hybridization analyses were employed to identify 3q+, 7q-, -7, or (cryptic) RUNX1 (AML1) aberrations. BM cytology and pathology were reviewed centrally. Clinical characteristics as well as all cytogenetic followup examinations of the 86 FA patients are shown in Table 1 and Online Supplementary Tables S1-S3. Out of the 86 FA patients, 60 (70%) had a normal blast percentage and a normal karyotype at every time point. All of these patients were classified as having FA-BMF. There were no patients with an elevated blast percentage in PB and/or BM who had a normal karyotype. Twenty-six patients showed acquired clonal chromosomal aberrations (30%) (Table 1). The most frequent chromosomal aberration Table 1. Clinical characteristics of patients with Fanconi anemia (n=86).

Characteristics

n (%) a

Age at FA diagnosis median (range) Age at last follow up median (range) Deathb Age at death

8 years (1 month - 25 years) 13 years (2 - 43 years) 7 (8%) median (range) 14 years (3 - 43 years) 39 (45%) 9 years (3 - 42 years)

HSCTc Age at HSCT median (range) Sex Female 38 (44%) Male 48 (56%) Complementation group FANCA 42 (49%) FANCC 8 (9%) FANCD2 4 (5%) FANCP 3 (3%) FANCB 2 (2%) FANCJ 2 (2%) FANCL 2 (2%) FANC:D1, G, I, T (each) 1 (1%) Unknown 20 (23%) Tranformation to AML/MDS 15 (17%) Age at tranformation to FA-MDSd 11 years median (range) (3 - 42 years) FA-MDS-non EB 8 (9%) FA-MDS-EB 2 (2%) FA-AML 5 (6%) FA-AIP 10 (12%) Karyoytype Abnormal 26 (30%) Patients with transformation to AML/MDS 15 (17%) Normal 60 (70%) Patients with transformation to AML/MDS 0

a Date of diagnosis of Fanconi anemia not known for five patients. bDate of death not known for four patients. cDate of transplantation not known for five patients. d Date of transformation not known for two patients. FA: Fanconi anemia; HSCT: hematopoietic stem cell transplantation; AML: acute myeloid leukemia; MDS: myelodysplastic syndrome; EB: excess blasts; AIP: aberration of indeterminate potential.

haematologica | 2021; 106(11)

Letters to the Editor

was a 1q+ (14%) followed by 3q+ (9%), 7q- (9%), -7 (3%), 7p- (3%), and 6p- (3%) (Online Supplementary Table S2). Fifteen of these 26 patients (58%) were diagnosed with a myeloid neoplasia: Eight patients were categorized as having FA-MDS-non EB (FA2, 6, 13, 54, 59, 70, 79, and 89), two patients as having FA-MDS-EB (FA3 and 12) and five patients met the criteria for FA-AML (FA27, 39, 43, 61, and 76). Another ten patients were diagnosed with FA-AIP (FA22, 25, 26, 35, 45, 47, 77, 84, 86, and 87). In patients with FA-related myeloid neoplasia, we observed the expected recurrent pattern of chromosomal aberrations (e.g., 3q+, 7q-, -7). Patients with FA-AIP had a different pattern of chromosomal aberrations, such as 1q+, 6p-, and 7p- (Online Supplementary Tables S1 and S3). Furthermore, patients with FA-AML had different cytogenetic aberrations than patients with de novo AML, which is consistent with previous findings.12 Our dataset was too small to correlate FA-complementation subgroups with hematologic findings.

To further investigate the chronological sequence of aberrations in FA-AIP, FA-MDS-non EB, FA-MDS-EB, and FA-AML over a prolonged period, we analyzed the follow-up examinations of the 26 patients with chromosomal aberrations over the course of several years (Figure 2, Online Supplementary Table S1 ). Twelve of 26 patients showed 1q+, isolated or in combination with additional aberrations (FA2, 6, 12, 22, 26, 27, 47, 70, 79, 84, 86, and 87). Only patients with 1q+ in combination with 3q+ or 7q- developed myeloid neoplasia (FA2, 6, 12, 27, 70, and 79), while patients with 1q+ isolated or in combination with other aberrations (e.g., 6p- or 7p-) did not progress to develop a myeloid neoplasia within the observation period (FA22, 26, 47, 84, 86, and 87) (Figure 2). Several patients displayed a stable clone without transformation for several years. The presence of the recurrent aberrations 3q+, 7q-/-7 or a complex karyotype was always indicative of myeloid neoplasm (FA2, 3, 6, 12, 13, 27, 39, 43, 54, 59, 61, 70, 79, and 89) with the exception of one

Figure 1. Proposed diagnostic algorithm for hematologic conditions in patients with Fanconi anemia. FA: Fanconi anemia; BM: bone marrow; PB; peripheral blood; FA-MDS-EB: myelodysplastic syndrome with excess blasts in a patient with FA; FA-AML: acute myeloid leukemia in a patient with FA; FA-BMF: BM failure in a patient with FA, a normal blast percentage and normal cytogenetics; FA-AIP: aberration of indeterminate potential in a patient with FA; FA-MDS-non-EB: myelodysplastic syndrome without excess blasts in a patient with FA.

haematologica | 2021; 106(11)

3001

Letters to the Editor

Figure 2. Chronology of recurrent cytogenetic events in Fanconi anemia with myelodysplastic syndrome or aberrations of indeterminate potential (n=26). Each bar represents cytogenetic follow-up over the course of several years (recurrent aberration, clone size) as well as additional clinical information. Additional nonrecurrent clonal aberrations of these patients are shown in Online Supplementary Table S2. Patient FA130 showed cytogenetic aberrations associated with myeloid neoplasia in <10% of cells, and no blasts in peripheral blood or bone marrow, which were no longer detectable in the following examination that revealed a normal karyotype. To construct the clone size over time we used the “clonal evolution plot R”.14 FA: Fanconi anemia; HSCT: hematopoietic stem cell transplantation; MDS: myelodysplastic syndrome; AML: acute myeloid leukemia; AIP: aberration of indeterminate potential; EB: excess blasts.

3002

haematologica | 2021; 106(11)

Letters to the Editor

case (FA130): This patient had a small clone (~9% of cells) with 3q+ and 7q- (Figure 2, Online Supplementary Table S1) which was no longer detectable in the following examination that revealed a normal karyotype. This observation may indicate that the presence of a small clone with aberrations affecting <10% of analyzed cells may not be sufficient to establish the diagnosis of FAMDS in all cases. Thus, short-term repeat analysis is recommended to avert unnecessary hematopoietic stem cell transplantation. Clonal evolution (8/15 vs. 1/10) as well as a complex karyotype (5/15 vs. 0/10) occurred more frequently in patients with myeloid neoplasia (n=15) than in patients with FA-AIP (n=10) (P<0.05, Fisher exact test) (Online Supplementary Table S2). In two patients with FAAIP (FA35 and 87) and in three patients with myeloid neoplasms (FA2, 12, and 79), the original clone was either replaced by a new clone or at least one new clone appeared in addition to the original clone. Apart from patient FA130, in one patient with FA-AIP (FA25) and in one FA patient with myeloid neoplasia (FA12) the aberrant clone disappeared and the patients showed a normal karyotype during follow-up examinations. One patient (FA79) developed a 20q- aberration after hematopoietic stem cell transplantation. Notably, this patient had incomplete donor chimerism. Four out of ten patients with FA-AIP (FA2, 6, 12, and 70) developed a myeloid neoplasm over time within the observation period. This finding indicates that the presence of FA-AIP is not a predictor for rapid transformation. It is conceivable that additional remaining patients may transform after a longer time period. Cryptic RUNX1 aberrations were not unambiguously associated with immediate transformation, as noted in patient FA70 who developed FA-MDSnon-EB approximately 2 years after this lesion had been detected. Another patient with a RUNX1 aberration developed FA-AML (FA76). Larger cohorts are required to determine the prognostic value of this aberration and whether immediate hematopoietic stem cell transplantation is indicated because of an isolated RUNX1 aberration in a FA patient. In conclusion, we propose an algorithm for the diagnosis of benign and malignant hematologic conditions in patients with FA, which may serve as a model for other syndromes predisposing to MDS and AML. The concept of AIP may also apply to other inherited BMF syndromes such as Shwachman-Diamond syndrome, since patients with this syndrome commonly develop cytogenetic abnormalities including deletion 20q and isochromosome 7q that are not associated with a high risk of progression to MDS.13 Yvonne Lisa Behrens,1* Gudrun Göhring,1* Randa Bawadi,1 Sümeyye Cöktü,2 Christina Reimer,2 Beatrice Hoffmann,2 Birte Sänger,2 Simon Käfer,1 Felicitas Thol,3 Miriam Erlacher,4,5,6 Charlotte M. Niemeyer,4,5,6 Irith Baumann,7 Reinhard Kalb,8 Detlev Schindler8 and Christian Peter Kratz2 *YLB and GG contributed equally as co-first authors. 1

Department of Human Genetics, Hannover Medical School, Hannover; 2Pediatric Hematology and Oncology, Hannover Medical School, Hannover; 3Department of Hematology, Hemostasis, Oncology and Stem Cell Transplantation, Hannover Medical School, Hannover; 4 Division of Pediatric Hematology and Oncology, Department of Pediatrics and Adolescent Medicine, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg; 5 German Cancer Consortium (DKTK), Freiburg; 6German Cancer Research Center (DKFZ), Heidelberg; 7Institute of Pathology, Kaufbeuren and 8Department of Human Genetics, University of Würzburg, Biocenter, Würzburg, Germany haematologica | 2021; 106(11)

Correspondence: CHRISTIAN P. KRATZ - kratz.christian@mh-hannover.de doi:10.3324/haematol.2021.279332 Received: May 28, 2021. Accepted: June 22, 2021. Pre-published: July 1, 2021. Disclosures: no conflicts of interest to disclose. Contributions: the study was conceived by CPK, YLB, and GG; data and material of the study was generated by YLB, GG, RK, DS, IB, ME, CN, SC, BS, CR, BH, SK, CPK and RB; data collection was performed by YLB, GG, CPK; interpretation and analysis were conducted by YLB, GG, CPK, ME, CN, IB and FT; YLB, GG, CN, and CPK wrote the manuscript; FA diagnostics and complementation analysis was done by DS and RK; the paper was edited by all authors. All authors have read and approved the final manuscript. Acknowledgments: the authors thank the members of the Deutsche Fanconi Hilfe e.V. for their support. Funding: CPK, GG and FT have been supported by the BMBF ADDRess (01GM1909A), DS und RK by BMBF ADDRess (01GM1909B), GG, CN and ME by the BMBF MyPred (01GM1911A). CPK has been supported by the Deutsche Kinderkrebsstiftung (DKS2019.13).

References 1. Alter BP, Giri N, Savage SA, Rosenberg PS. Cancer in the National Cancer Institute inherited bone marrow failure syndrome cohort after fifteen years of follow-up. Haematologica. 2018;103(1):30-39. 2. Tsai FD, Lindsley RC. Clonal hematopoiesis in the inherited bone marrow failure syndromes. Blood. 2020;136(14):1615-1622. 3. Yoshimi A, Niemeyer C, Baumann I, et al. High incidence of Fanconi anaemia in patients with a morphological picture consistent with refractory cytopenia of childhood. Br J Haematol. 2013;160(1):109111. 4. Tonnies H, Huber S, Kuhl JS, Gerlach A, Ebell W, Neitzel H. Clonal chromosomal aberrations in bone marrow cells of Fanconi anemia patients: gains of the chromosomal segment 3q26q29 as an adverse risk factor. Blood. 2003;101(10):3872-3874. 5. Peffault de Latour R, Soulier J. How I treat MDS and AML in Fanconi anemia. Blood. 2016;127(24):2971-2979. 6. Quentin S, Cuccuini W, Ceccaldi R, et al. Myelodysplasia and leukemia of Fanconi anemia are associated with a specific pattern of genomic abnormalities that includes cryptic RUNX1/AML1 lesions. Blood. 2011;117(15):e161-170. 7. Meyer S, Neitzel H, Tonnies H. Chromosomal aberrations associated with clonal evolution and leukemic transformation in fanconi anemia: clinical and biological implications. Anemia. 2012; 2012:349837. 8. Wang Y, Zhou W, Alter BP, et al. Chromosomal aberrations and survival after unrelated donor hematopoietic stem cell transplant in patients with Fanconi anemia. Biol Blood Marrow Transplant. 2018; 24(10):2003-2008. 9. Soulier J. Fanconi anemia. Hematology Am Soc Hematol Educ Program. 2011;2011:492-497. 10. Mehta PA, Harris RE, Davies SM, et al. Numerical chromosomal changes and risk of development of myelodysplastic syndrome-acute myeloid leukemia in patients with Fanconi anemia. Cancer Genet Cytogenet. 2010;203(2):180-186. 11. Reina-Castillon J, Pujol R, Lopez-Sanchez M, et al. Detectable clonal mosaicism in blood as a biomarker of cancer risk in Fanconi anemia. Blood Adv. 2017;1(5):319-329. 12. Rochowski A, Olson SB, Alonzo TA, Gerbing RB, Lange BJ, Alter BP. Patients with Fanconi anemia and AML have different cytogenetic clones than de novo cases of AML. Pediatr Blood Cancer. 2012; 59(5):922-924. 13. Myers KC, Furutani E, Weller E, et al. Clinical features and outcomes of patients with Shwachman-Diamond syndrome and myelodysplastic syndrome or acute myeloid leukaemia: a multicentre, retrospective, cohort study. Lancet Haematol. 2020;7(3):e238-e246. 14. Miller CA, McMichael J, Dang HX, et al. Visualizing tumor evolution with the fishplot package for R. BMC Genomics. 2016; 17(1):880.

3003

Letters to the Editor

The RUNX1 database (RUNX1db): establishment of an expert curated RUNX1 registry and genomics database as a public resource for familial platelet disorder with myeloid malignancy Familial platelet disorder with associated myeloid malignancy (FPD-MM, OMIM:601399)1,2 is a rare cancer predisposition syndrome caused by pathogenic germline variants in RUNX1.3 Despite research dating back over two decades, many challenges remain in improving outcomes for individuals with FPD-MM.4 Firstly, the syndrome may go unrecognized due to poor recognition of family history and/or access to appropriate genetic testing. Secondly, intentional screening or incidental detection (e.g., tumour-sequencing) of RUNX1 variants requires access to expert interpretation. Thirdly, after diagnosis, the relative rarity of the disorder inhibits the collation of sizeable local cohorts, making identification of commonalities in disease course and/or outcome highly challenging. To help overcome these significant challenges, we have developed an interactive public webbased international collaborative database for RUNX1: RUNX1db (https://runx1db.runx1-fpd.org/). RUNX1db is a centralized repository for germline RUNX1 variant information, associated next-generation sequencing (NGS) data, and expert-curated variant information (both germline and somatic). We recently identified, from publications, 140 different families with germline RUNX1 variants.4 While being a rich resource, historically reported variants are largely not classified according to the American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines, only established in 2015.5 Additionally, the Clinical Genome Resource myeloid malignancy variant curation expert panel (ClinGen MM-VCEP) recently created guidelines specific for classification of germline RUNX1 variants.6 Gene-specific guidelines, while important, add additional complexity to the curation of identified variants. Making

available expert knowledge to accurately classify these germline variants prevents both missing pathogenic variants or the misattribution of benign variants as causative in families.7,8 Additionally, variants identified through clinical services and research studies don’t always make it into the public domain due to constraints associated with the reporting of variants through publication or variant repositories. To address some of these challenges, we updated curated variants from publications and undertook an international survey of colleagues, identifying unpublished variants. This study identified an additional 119 families (259 in total), with 164 unique variants. These included ten new variants not previously described Table 1. RUNX1database genomics cohort demographics.

Germline Mutations Age

Parameters

RUNX1 database cohort

Total Individuals/Unique Mutations

120/47

All samples: Median (range), years Malignancy Samples: Median (range), years Pre-leukemic: Median (range), years Gender Males: Total/Malignancy Females: Total/Malignancy Malignancy Subtype Total Samples/Individuals AML: #Individuals MDS: #Individuals MDS/MPN: #Individuals MPN : #Individuals ALL : #Individuals AL : #Individuals Pre-Leukemic (#Samples/#Individuals)

40 (1-76) 43 (3-69) 34 (1-76) 51/23 67/29 62/48 29 14 2 2 4 1 65/56

AML: acute myeloid leukemia; MDS: myelodysplastic syndrome; MDS/MPN: myelodysplastic syndrome/myeloproliferative neoplasm; MPN: myeloproliferative neoplasm; ALL: acute lymphoblastic leukemia; AL: acute leukemia.

Figure 1. Registry of germline RUNX1 mutations. Germline RUNX1 variants currently included in the RUNX1db registry are visualised using the ProteinPaint web application (https://pecan.stjude.cloud/home).12 Variants (displayed as protein changes where possible) are colour-coded according to pathogenicity classification as determined by the MM-VCEP RUNX1-specific recommendations. The number of probands for each variant is indicated within the circle where the number is greater than one. All variants are annotated to RUNX1c; NM_001754.4; LRG_ 482.

3004

haematologica | 2021; 106(11)

Letters to the Editor

Figure 2. RUNX1 database genomics cohort demographic. (A) Breakdown of the number and types of NGS samples currently stored in the RUNX1db. PreLeukemic: thrombocytopenia, asymptomatic Other: includes post-transplant/post-treatment and saliva samples. Both WES and panel data is analysed and stored in the database. (B) Scatter plot displaying the age of the individual when each sample was collected. Major RUNX1db cohorts (malignancy and preleukemic samples) are displayed. The median age for each cohort is represented by the vertical line. Clinical demographics of the malignancy cohort is shown with the number of individuals with different types of FPD-MM malignancy presentation and the (C) gender and (D) age distribution; Adult ≥40years, AYA=15-39 years, children ≤14years. AML: Acute myeloid leukemia; MDS: myelodysplastic syndromes; MDS/MPN: myelodysplastic syndrome/myeloproliferative Neoplasm overlap; MPN: Myeloproliferative Neoplasm; ALL: acute lymphoblastic leukemia; AL: acute undifferentiated leukemia.

(Figure 1 and Table S1). Using this data, we created the first comprehensive RUNX1 germline registry and performed expert curation of all variants according to the RUNX1-specific ACMG classification rules (ALB, CNH, LAG, LM, CDD MM-VCEP members). The registry represents the largest collection of curated and clinically classified RUNX1 germline variants to date, providing a unique clinical resource for researchers, clinical genomics laboratories, and haematologists (Figure 1, Table S1). Utilizing this resource, we have identified 97 pathogenic/likely pathogenic RUNX1 variants, with 54 located within the RUNT domain (RHD)(75% of RHD variants), of which 24 are missense mutations. Only one pathogenic missense variant is observed outside of the RHD, suggesting the RHD is highly intolerant to genetic-variation. Most commonly observed pathogenic germline RUNX1 variations are whole-gene deletions (21 probands), deletion of exons 1-2 (9 probands), and mutation of amino acid p.Arg201 within the RHD (8 probands)(Table S1). Accessibility and update-ability of this information is available through a live-webportal which hosts the registry (https://runx1db.runx1-fpd.org/classification/classifications). Each curated variant has links to patient-phenotypic information and the current clinical classification, including the evidence for each ACMG code assessed and links to external clinical databases, including ClinVar and associated publications. Importantly, expert crowdsourcing allows the real-time updating of the database through user profile accounts. Newly-identified variants can be haematologica | 2021; 106(11)

easily added to the database and are automatically annotated with over 137 parameters required for accurate classification (e.g., population frequency, pathogenicity predictions). These parameters populate a classificationtool that guides users stepwise through the ACMG classification of new variants (or updating current classifications with new information). Once curated and classified, collated information can be exported as an automated classification report summary, flagged for expert-review, shared with other users, and uploaded to ClinVar. In addition to a germline RUNX1 variant registry, RUNX1db has the capacity to house NGS datasets, creating the first international genomics cohort of this rare disease. This initiative intends to enable researchers to answer questions about FPD-MM beyond germline variant detection. For example, family members, heterozygous for RUNX1 mutations, can have varying clinical presentations indicating variable penetrance and expressivity. In almost all cases, germline RUNX1 carriers present with thrombocytopenia and qualitative platelet defects, and progression to hematologic malignancies (HM) is incompletely penetrant with variable age of onset ranging from early childhood to late adulthood.2 Patients develop myeloid malignancies most frequently, and Tcell and, more rarely, B-cell acute lymphoblastic leukaemia (ALL).4 Currently, there is no way to predict which individuals will progress to myelodysplastic syndrome (MDS), acute myeloid leukaemia (AML), or other HM. Accumulation of somatic mutations and additional 3005

Letters to the Editor

germline modifier variants are mechanisms proposed to contribute to this heterogeneity.4 NGS technology is widely used for surveillance and diagnosis of HM,4 accumulating large amounts of data often not utilized beyond RUNX1 variant detection. Individual laboratories often only have small numbers of patients with deleterious RUNX1 germline variants, which makes asking larger questions about commonalities of genotype-phenotype, disease progression, monitoring, treatment and outcome, difficult.9 To accumulate the data required to make evidence-based clinical decisions in FPD-MM, a dedicated resource utilizing the collective wealth of NGS data generated from research and diagnostic laboratories internationally is ideal in standardizing and collating diseasespecific clinical and genomics data. The database has also been designed for the accumulation, sharing and curation of genomics data acquired from individuals with germline RUNX1 mutations both pre- and post-malignancy progression. We have collated 179 NGS datasets, both whole-exome sequencing (WES) and HM gene panel data, from 19 distinct research centres worldwide. This includes NGS from 60 FPD-MM families and 120 individuals, making it the largest FPD-MM NGS dataset (Figure 2). The dataset includes individuals ranging in age from 1-76 years, malignancy phenotypes of AML, MDS, myelodysplastic syndrome/myeloproliferative neoplasm (MDS/MPN), ALL, and pre-leukemic phenotypes including thrombocytopenia and asymptomatic carriers (Table 1). Detailed clinical information for each patient and associated samples are stored on the database and can be updated, enabling specific phenotypic-genotypic cohort studies to be performed on the clinical spectrum of FPDMM. Additionally, the database can be updated easily with new NGS data as available, including longitudinal datasets from serial testing of individual patients. The database allows for a comprehensive, unbiased and customizable review of all RUNX1 germline datasets with all raw sequencing data being analyzed through a standardized bioinformatics pipeline. This is designed to identify both somatic and germline variants and is available on the database as variant level data (VCF, Figure S1). Using the integrated VariantGrid (https://github.com/SACGF/variantgrid) genomics analysis software, we have curated a panel of somatic variants for each dataset (including all malignancy and pre-leukemic samples), prioritizing the identification of potentially pathogenic variants in HM (2,643 variants, 167 samples). Standard filtering criteria were adapted for identifying somatic variants (Online Supplementary Figure S1). Variants that passed all filtering criteria were subsequently manually curated. Variants classified as having no clinical significance (benign/likely benign) according to ACMG/AMP guidelines, were excluded. Remaining variants were either classified as 1) Clinically relevant, 2) Possibly relevant, or 3) Unknown relevance (Online Supplementary Figure S1).10,11 Curated somatic variant data is available through the interactiveoncoplot on the database homepage or variant page. Shared in real-time with the scientific community, this curated dataset has already allowed the selection of secondary mutations to model FPD-MM disease and therapy in vitro and in animals. Importantly, investigators can interrogate the data to answer additional research questions as the software provides a fully automated annotation of variants and allows non-bioinformaticians to filter, sort, analyze, and curate genetic variants stored in the database via a graphical interface (Online Supplementary Figure S2). This project serves as a model for data accumulation for rare cancer predisposition syndromes. The adoption 3006

of a single database that serves as a repository for patient demographic and clinical data, a mutational germline registry, and patient genomics data, which can be interrogated as a large cohort are essential components for the diagnosis and treatment of patients with a rare-disorder such as FPD-MM. This resource is especially useful in FPD-MM, where the genetic cause is well established but variability in clinical presentation and disease development render diagnosis challenging. The aggregation of multiple families, individuals, and disease stages into a centralized database where all data undergo rigorous quality control using a single bioinformatics analysis strategy will aid in the exploration and discovery of the molecular progression of the disorder. The harmonized interpretation of genomic variants is imperative to understanding the mutational profile of a malignancy, which is achieved through a curated list of variants displayed for each sample. Institutional, national, and international ethics and data sharing guidelines may initially limit contributions to initiatives like this that are supported by patient advocates but need to be overcome, given the importance of the work. We envision that information from this database will guide precision-based approaches to patient care plans with reasonable surveillance and adequate counselling and, eventually, the application of new targeted therapies and interventions prior to malignancy development for germline RUNX1 carriers. With the continued accumulation of data and clinical information, this type of gene-specific database can provide the basis to developing evidence-based clinical decisions such as when to watch and wait and when to apply more aggressive therapies such as stem cell transplantation. Finally, we hope that this database will serve as a model from which similar efforts will emerge for other HMs, benefiting all our patients and families. Claire C. Homan,1,2 Sarah L. King-Smith,1,2 David M. Lawrence,1,2,3 Peer Arts,1,2 Jinghua Feng,2,3 James Andrews,1,2 Mark Armstrong,1,2 Thuong Ha,1,2 Julia Dobbins,1,2 Michael W. Drazer,4 Kai Yu,5 Csaba Bödör,6 Alan Cantor,7 Mario Cazzola,8,9 Erin Degelman,10 Courtney D. DiNardo,11° Nicolas Duployez,12,13 Remi Favier,14 Stefan Fröhling,15,16 Jude Fitzgibbon,17 Jeffery M. Klco,18 Alwin Krämer,19 Mineo Kurokawa,20 Joanne Lee,21 Luca Malcovati,8,9° Neil V. Morgan,22 Georges Natsoulis,23 Carolyn Owen,10 Keyur P. Patel,11 Claude Preudhomme,12,13 Hana Raslova,24 Hugh Rienhoff,23 Tim Ripperger,25 Rachael Schulte,26 Kiran Tawana,27 Elvira Velloso,28,29 Benedict Yan,21 Paul Liu,5 Lucy A. Godley,4° Andreas W. Schreiber,2,3,30 Christopher N. Hahn,1,2,31° Hamish S. Scott,1,2,30,31 and Anna L. Brown1,2,31° on behalf of the RUNX1 international data-sharing consortium. 1 Department of Genetics and Molecular Pathology, SA Pathology, Adelaide, SA, Australia; 2Centre for Cancer Biology, SA Pathology and University of South Australia, Adelaide, SA, Australia; 3Australian Cancer Research Foundation (ACRF) Cancer Genomics Facility, Centre for Cancer Biology, SA Pathology, Adelaide, SA, Australia; 4 Section of Hematology/Oncology, Departments of Medicine and Human Genetics, Center for Clinical Cancer Genetics, and The University of Chicago Comprehensive Cancer Center, The University of Chicago, Chicago, IL; 5National Human Genome Research Institute, National Institutes of Health, Bethesda, MD 20892; 6 HCEMM-SE Molecular Oncohematology Research Group, 1st Department of Pathology and Experimental Cancer Research, Semmelweis University, Budapest, Hungary; 7Division of Hematology/Oncology, Boston Children's Hospital and Dana Farber Cancer Institute, Harvard Medical School, Boston, MA; 8Department of Molecular Medicine, University of Pavia, Pavia, Italy; 9Department of Hematology Oncology, Fondazione IRCCS Policlinico San Matteo, haematologica | 2021; 106(11)

Letters to the Editor

Pavia, Italy; 10Division of Hematology and Hematological Malignancies, Foothills Medical Centre, Calgary, AB, Canada; 11 Department of Leukemia, University of Texas MD Anderson Cancer Center, Houston, TX; 12Laboratory of Hematology, Biology and Pathology Center, Centre Hospitalier Regional Universitaire de Lille, Lille, France; 13Jean-Pierre Aubert Research Center, INSERM, Universitaire de Lille, Lille, France; 14Assistance Publique-Hôpitaux de Paris, Armand Trousseau children's Hospital, Paris, France; 15 Department of Translational Medical Oncology, National Center for Tumor Diseases (NCT) and German Cancer Research Center (DKFZ), Heidelberg, Germany; 16German Cancer Consortium (DKTK), Heidelberg, Germany; 17Centre for Haemato-Oncology, Barts Cancer Institute, Queen Mary University of London, London, UK; 18St Jude Children's Research Hospital, Memphis, Tennessee, United States; 19Clinical Cooperation Unit Molecular Hematology/Oncology, German Cancer Research Center (DKFZ) and Dept. of Internal Medicine V, University of Heidelberg, Heidelberg, Germany; 20Department of Hematology & Oncology, Graduate School of Medicine, The University of Tokyo, Japan; 21Department of Haematology-Oncology, National University Cancer Institute, National University Health System, Singapore; 22Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, UK; 23Imago Biosciences, Inc., San Francisco, CA, USA; 24Institut Gustave Roussy, Université Paris Sud, Equipe Labellisée par la Ligue Nationale Contre le Cancer, Villejuif, France; 25Department of Human Genetics, Hannover Medical School, Hannover, Germany; 26Department of Pediatrics, Division of Pediatric Hematology and Oncology, Monroe Carell Jr. Children’s Hospital, Vanderbilt University Medical Center, Nashville, TN, USA; 27 Department of Haematology, Addenbrooke’s Hospital. Cambridge, CB2 0QQ; 28Service of Hematology, Transfusion and Cell Therapy and Laboratory of Medical Investigation in Pathogenesis and Directed Therapy in Onco-Immuno-Hematology (LIM-31) HCFMUSP, University of Sao Paulo Medical School, Sao Paulo, Brazil; 29Genetics Laboratory, Hospital Israelita Albert Einstein, Sao Paulo, Brazil. 30 School of Biological Sciences, University of Adelaide, Adelaide, SA, Australia and 31School of Medicine, University of Adelaide, Adelaide, SA, Australia °Clinical Genome Resources Variation Myeloid Malignancy Expert Panel committee members Correspondence: ANNA L. BROWN - anna.brown@sa.gov.au doi:10.3324/haematol.2021.278762 Received: March 22, 2021. Accepted: July 2, 2021. Pre-published: July 8, 2021. Disclosures: the authors declare no competing financial interests. Contributions: CCH and ALB designed the research, wrote the manuscript, collected NGS and clinical data, curated NGS data, performed ACMG RUNX1 variant classification and analyzed the data. SLK collected and curated NGS data, and performed ACMG RUNX1 variant classification. PA, JD collected and curated NGS data. PA designed Figure 1. MA, TH, JF, AWS designed and performed bioinformatics analysis. DML, JA designed the database and VariantGrid software. MWD, KY, CB, AC, MC, ED, ND, RF, SF, JF, JMK, AK, MK, JL, NVM, GN, CO, KPP, CP, HR, HR, TR, RS, KT, EV, BY, PL, CDD, LAG, LM, ALB contributed NGS data, clinical patient information and scientific insight. CDD, LAG, LM, ALB as members of the MM-VCEP advised on RUNX1 variant classification. ALB, CNH, HSS conceived and designed the study. All authors critically reviewed and approved the manuscript. Acknowledgments: the authors would also like to thank the RUNX1 Research Program for their support in helping to facilitate the development of the database and fostering collaborations. We also thank the patients and their family members for their willingness to participate in

haematologica | 2021; 106(11)

this study and the RUNX1 international data-sharing consortium for their valuable contributions. This project is also proudly supported by funding from the Leukaemia Foundation of Australia, and project grants APP1145278 and APP1164601 from the National Health and Medical Research Council of Australia. This work was produced with the financial and additional support of Cancer Council SA's Beat Cancer Project on behalf of its donors and the State Government of South Australia, through the Department of Health (PRF Fellowship to HSS). PA is supported by a fellowship from The Hospital Research Foundation. Part of this project was undertaken whilst PA was holding a Royal Adelaide Hospital Mary Overton Early Career Fellowship. LM is supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC) (Accelerator Award Project 22796; 5x1000 Project 21267; Investigator Grant 2017 Project 20125). LAG is supported by the Cancer Research Foundation. KY and PL are supported by the Division of Intramural Research, National Human Genome Research Institute, NIH. TR is supported by a grant from the European Hematology Association (EHA) and BMBF MyPred (01GM1911B). CB is supported by the EU’s Horizon 2020 Research and Innovation Program, under grant agreement No. 739593. The RUNX1 international data-sharing consortium includes all co-authors and others, including Michael Doubek (Masaryk University, Czechia), Stephen Langabeer (St. James’s Hospital, Ireland), Koneti Rao (Sol Sherry Thrombosis Research Center, USA), Josée Hébert (Université de Montréal, Canada), Lauren M. Bear (Massachusetts General Hospital, USA), Timothy A. Graubert (Massachusetts General Hospital, USA), Akiko Shimamura (Harvard Medical School, USA), Peter Ganly (Canterbury District Health Board, NZ), Marc H.G.P. Raaijmakers (Erasmus Medical Center Cancer Institute, Netherlands), Peter J.M. Valk (Erasmus Medical Center Cancer Institute, Netherlands), Paula Heller (Instituto de Investigaciones Médicas (IDIM) Alfredo Lanari, Argentina). Funding: this work is supported by a grant from the RUNX1 Research Program.

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. Blood. 2016;128(3):462-463. 2. Brown AL, Hahn CN, Scott HS. Secondary leukemia in patients with germline transcription factor mutations (RUNX1, GATA2, CEBPA). Blood. 2020;136(1):24-35. 3. Song WJ, Sullivan MG, Legare RD, et al. Haploinsufficiency of CBFA2 causes familial thrombocytopenia with propensity to develop acute myelogenous leukaemia. Nat Genet. 1999;23(2):166-175. 4. Brown AL, Arts P, Carmichael CL, et al. RUNX1-mutated families show phenotype heterogeneity and a somatic mutation profile unique to germline predisposed AML. Blood Adv. 2020;4(6):1131-1144. 5. 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. 6. Luo X, Feurstein S, Mohan S, et al. ClinGen Myeloid Malignancy Variant Curation Expert Panel recommendations for germline RUNX1 variants. Blood Adv. 2019;3(20):2962-2979. 7. Brown AL, Hahn C, Hiwase D, Godley LA, Scott HS. Correct application of variant classification guidelines in germline RUNX1 mutated disorders to assist clinical diagnosis. Leuk Lymphoma. 2020;61(1):246-247. 8. Feurstein S, Zhang L, DiNardo CD. Accurate germline RUNX1 variant interpretation and its clinical significance. Blood Adv. 2020;4(24):61996203. 9. Bellissimo DC, Speck NA. RUNX1 mutations in inherited and sporadic leukemia. Front Cell Dev Biol. 2017;5:111. 10. Branford S, Wang P, Yeung DT, et al. Integrative genomic analysis reveals cancer-associated mutations at diagnosis of CML in patients with highrisk disease. Blood. 2018;132(9):948-961. 11. Li MM, Datto M, Duncavage EJ, et al. Standards and guidelines for the interpretation and reporting of sequence variants in cancer: a joint consensus recommendation of the Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists. J Mol Diagn. 2017;19(1):4-23. 12. Zhou X, Edmonson MN, Wilkinson MR, et al. Exploring genomic alteration in pediatric cancer using ProteinPaint. Nat Genet. 2016;48(1):4-6.

3007

Letters to the Editor

mTOR inhibitors sensitize multiple myeloma cells to venetoclax via IKZF3- and Blimp-1-mediated BCL-2 upregulation Physical interference with anti-apoptotic function of BCL-2 family proteins provides a novel therapeutic paradigm for hematological malignancies, the survival of which is often dependent on BCL-2 or MCL-1.1 Among several agents targeting BCL-2 family proteins, venetoclax (ABT-199) was the first agent with selectivity for BCL-2 to enter the clinic.2 Based on randomized phase III studies, venetoclax was approved for the treatment of patients with chronic lymphocytic leukemia and, more recently, for treatment-naive acute myeloid leukemia (AML) patients older than 75 who are unfit for intensive induction therapy.3 Promising signals from phase I trials in multiple myeloma (MM) led to a placebo-controlled phase III trial of venetoclax in combination with bortezomib and dexamethasone.4 In the BELLINI trial, the venetoclax group was superior to the placebo group in terms of progression-free survival, but failed to achieve prolongation of overall survival due to higher incidence of treatment-related deaths.5 The results thus indicate that venetoclax can contribute activity to a standard-ofcare regimen but might be combined more effectively with other, more-targeted therapeutic agents for the treatment of MM. Previous studies have revealed that clinical efficacy of venetoclax is enriched in patients whose tumors carry the t(11;14) chromosomal translocation,4 and could be predicted by the ratio of BCL2/MCL1 mRNA expression in MM cells.6 We confirmed the correlation between the BCL2/MCL1 ratio and the sensitivity to venetoclax in MM cell lines used in this study (Figure 1A). In order to identify optimal combination partners for venetoclax, we performed chemical library screening for compounds that increase the BCL2/MCL1 ratio in MM.1S cells, in which t(11;14) was absent and baseline BCL2 expression was relatively low (Figure 1A). We found that the BCL2/MCL1 ratio was most strikingly increased by mTOR inhibitors (mTORC1-specific inhibitors everolimus and temsirolimus, and a dual mTORC1/2 inhibitor torkinib) among 66 compounds in the library (Figure 1B; Online Supplementary Table S1). Next, we confirmed the increase in the BCL2/MCL1 ratio in other MM cell lines. Everolimus and torkinib significantly upregulated BCL2 expression at mRNA and protein levels in t(11;14)-positive KMS12-BM and KMS-21 cells in a timeand dose-dependent manner (Figure 1C; Online Supplementary Figure S1A). mTOR inhibitors only marginally affected the expression of MCL1 mRNA, and therefore the net effect was to increase the BCL2/MCL1 ratio. The increase was much more robust in t(11;14)-negative MM.1S than t(11;14)-positive MM cells because of the lower baseline expression of BCL2 in the former (Figure 1A; note that pretreatment expression levels were adjusted in Figure 1C to visualize the changes clearly). These results suggest that mTOR inhibitors are strong candidates in combination with venetoclax for the treatment of MM regardless of the presence of t(11;14). Next, we examined the mechanisms by which mTOR inhibitors upregulated BCL2 mRNA expression in MM cells. To this end, we comprehensively analyzed the binding of known transcription factors in the vicinity of the transcription start sites of the BCL2 gene using the ChIP-Atlas platform.7 Among B-cell transcription factors, IKZF3 and Blimp-1, but not IKZF1 or FOXO1, were highly accumulated at acetylated H3K27-enriched 3008

promoter/enhancer regions of BCL2: GRCh37/hg37: 60,985,600-60,987,400, including P1 and P2 promoters,8 in MM cells (Figure 2A; Online Supplementary Figure S1B). This is compatible with the results of biochemical studies, in which IKZF3 is a pivotal transcriptional activator of BCL2 in T lymphocytes9 and co-operates with Blimp1 to maintain the survival of MM cells.10 The expression levels of Blimp-1 and IKZF3, but not IKZF1, were positively correlated with BCL2 expression in primary MM cells, implying their active involvement in BCL2 transcription (Online Supplementary Figure S1C). Consistent with this view, short hairpin RNA (shRNA)-mediated knockdown of IKZF3 or Blimp-1 significantly decreased the abundance of BCL2 mRNA in KMS12-BM cells (Figure 2B, upper panel). Dual inhibition of IKZF3 and Blimp-1 additively suppressed BCL2 mRNA expression despite the fact that sh-IKZF3 negatively affected the expression of Blimp-1 and vice versa due to mutual transcriptional regulation of the two genes in MM cells (Online Supplementary Figure S1D).10 Our attempts to obtain more prominent consequences of IKZF3 and/or Blimp-1 downregulation using other shRNA sequences failed due to induction of massive cell death because the two molecules are indispensable for the survival of MM cells (data not shown). The mTOR inhibitor everolimus markedly increased the abundance of IKZF3 and Blimp-1 in MM cells, both at the mRNA and protein level (Online Supplementary Figure S2A). This increase resulted in >10fold accumulation of IKZF3 and Blimp-1 on the P1 promoter region of BCL2, which was proportional to the level of BCL2 transactivation in MM.1S cells treated with everolimus (Figure 2B, lower panel). As anticipated, BCL2 transactivation resulted in an increased abundance of BCL-2 protein with a reciprocal decrease in BCL-XL expression (Online Supplementary Figure S2B). Everolimusmediated upregulation of BCL-2 and its regulatory proteins was retained even if everolimus was combined with venetoclax (Online Supplementary Figure S2B). Mechanistically, we found that everolimus activated AKT kinase via phosphorylation at serine-473, which in turn phosphorylates and inactivates EZH2 to de-repress BCL2 transcription via erasure of a repressive histone mark, H3K27 trimethylation, on the BCL2 promoter in MM cells (Online Supplementary Figure S2C). This is in line with our previous observation11 and suggests that inhibition of the mTORC1 complex is sufficient for BCL2 upregulation because mTOR inhibitor-mediated AKT activation is stronger when mTORC2 activity is spared.12 Having demonstrated that mTOR inhibitors upregulated BCL-2 expression, we next examined whether everolimus and torkinib could enhance the sensitivity of MM cells to venetoclax-mediated killing. Isobologram analyses of drug interactions revealed that mTOR inhibitors exerted a strong synergy with venetoclax in MM cells under both non-adherent and adherent conditions (Figure 2C). The synergistic effect was confirmed in primary MM cells derived from patients without t(11;14) (Figure 2D, right panel) and even in those from t(11;14)positive patients, which showed higher baseline sensitivity to venetoclax (Figure 2D, left panel). It is well chronicled that MM cells acquire drug resistance through their interaction with stromal cells and/or extracellular matrix proteins.11 The synergistic effect of mTOR inhibitors and venetoclax in the presence of fibronectin (adherent conditions) strongly suggests that this combination could be effective in vivo and potentially overcomes cell adhesionmediated drug resistance. In order to test this notion, we attempted to reproduce the combined effect of everolimus and venetoclax in a murine xenograft model haematologica | 2021; 106(11)

Letters to the Editor

Figure 1. mTOR inhibitors increase the BCL2/MCL1 mRNA expression ratio in multiple myeloma cells. (A) Left panel: we prepared total RNA and whole cell lysates from the indicated cell lines to evaluated the expression of BCL-2, MCL-1 and GAPDH (loading control). The expression levels of BCL2 and MCL1 mRNA were determined by real-time quantitative polymerase chain reaction (qRT-PCR), normalized to that of GAPDH, and quantified by the 2-DDCt method with the values of MM.1S cells set at 1.0. Right panel: we determined the half maximal inhibitory concentration (IC50) of 4 MM cell lines for venetoclax and calculated the Pearson's correlation coefficient with the BCL2/MCL1 ratio. (B) MM.1S cells were cultured for 24 hours in the absence or presence of 66 small molecule inhibitors at IC50 (Online Supplementary Table S1). The categories of target molecules are shown on the x-axis. The relative ratio of BCL2/MCL1 expression is shown on the y-axis. Bars indicate the means of each group of inhibitors. *P<0.05 by one-way ANOVA with the Student-Newman-Keuls multiple comparison test. (C) The indicated MM cell lines were cultured in the presence of 200 nM everolimus for the indicated periods (upper panels) or various concentrations of everolimus for 24 hours (lower panels). The expression level of BCL2 and MCL1 was determined by qRT-PCR and shown as relative mRNA expression with pretreatment values set at 1.0. Whole cell lysates were simultaneously prepared and subjected to immunoblotting for BCL-2, MCL-1 and GAPDH proteins.

haematologica | 2021; 106(11)

3009

Letters to the Editor

Figure 2. mTOR inhibitors enhance the binding of IKZF3 and Blimp-1 to the BCL2 promoter. (A) The chromatin landscape of BCL2 promoter/enhancer regions on a peak browser of ChIP-Atlas (http://chip-atlas.org).7 Data were assembled from the database of human multiple myeloma (MM) cells (Blimp-1 and histone H3 acetylated at lysine-27) and human pre-B cells (IKZF3). Arrowheads indicate the location of P1 and P2 promoters.8 (B) Upper panel: KMS12-BM cells were transduced with short hairpin RNA (shRNA)-expression vectors containing a scrambled sequence (sh-Control), shRNA against IKZF3 (sh-IKZF3), shRNA against Blimp-1 (sh-Blimp-1), or both sh-IKZF3 and sh-Blimp-1 (the nucleotide sequences are available on request). The expression level of IKZF3, Blimp-1 and BCL2 was determined by real-time quantitative polymerase chain reaction (qRT-PCR), normalized to that of GAPDH, quantified by the 2-∆∆Ct method, and shown as foldincreases against the values obtained with sh-Control. *P<0.01 and **P<0.05 by one-way ANOVA with the Student-Newman-Keuls multiple comparison test. Lower panel: chromatin suspensions were prepared from KMS12-BM cells cultured with (+) or without (−) 200 nM everolimus for 24 hours and immunoprecipitated with anti-IKZF3 antibody, anti-Blimp-1 antibody or isotype-matched immunoglobulin (Ig) G. The resulting precipitants were subjected to qRT-PCR with primer 1 (forward: 5'-GTCCGGTATTCGCAGAAGTC-3' [–108 to –88] and reverse: 5'-CTCCTTCATCGTCCCCTCTC-3' [–239 to –219]), which covers the P1 promoter region of the BCL2 gene, and primer 2 (forward: 5'-GTGCCGAGCGCTAGAAGC-3' [–361 to –343] and reverse: 5'-GGGAGAACTTCGTAGCAGTCAT-3' [–467 to –445]), which covers the upstream enhancer region, as indicated in the left panel. The data were normalized to the values of input, quantified by the 2-∆∆Ct method, and Figure 2. Legend continued on the following page.

3010

haematologica | 2021; 106(11)

Letters to the Editor

shown as fold-increases against the values obtained with control IgG immunoprecipitants from untreated KMS12-BM cells. (C) The indicated MM cell lines were treated with venetoclax in combination with everolimus in 96-well plates coated with (adherent conditions) or without (non-adherent conditions) fibronectin for 72 hours. Dose-response curves of each combination were generated to make non-constant normalized isobolograms at half maximal inhibitory concentration (IC50) using CompuSyn software. The isobolograms shown are representative of at least three independent experiments. A combination index (CI) <1.0 indicates the synergism of the two drugs.11 (D) Left panel: CD138-positive cells were isolated from the bone marrow of MM patients carrying t(11;14) and treated with vehicle alone (dimethyl sulfoxide [DMSO]), 50 ng/mL venetoclax, 50 nM everolimus, and the combination of venetoclax and everolimus for 24 hours. Right panel: the same experiments using CD138-positive cells from MM patients who were negative for t(11;14). The y-axis shows the percentage of annexin-V-positive cells assessed by flow cytometry. Bars indicate the means of three samples. P-value was determined by one-way ANOVA with Tukey’s multiple comparison test. No brackets mean P>0.05. Informed consent was obtained in accordance with the Declaration of Helsinki and the protocol was approved by the Institutional Review Board of Jichi Medical University.

Figure 3. Legend on the following page.

haematologica | 2021; 106(11)

3011

Letters to the Editor

Figure 3. The mTOR inhibitor everolimus exhibits synergistic effects with venetoclax in a murine multiple myeloma model. (A) We inoculated 5x105 luciferaseexpressing KMS12-BM cells subcutaneously in the right thigh of male NOD/SCID mice (Charles River Laboratories, Wilmington, MA) and randomized them into four treatment groups when measurable tumors developed (day 0). Each group was treated with the vehicle alone (0.9% NaCl, orally, 5 times a week), everolimus alone (4 mg/kg, orally, twice a week), venetoclax alone (40 mg/kg, orally, 5 times a week), and the combination of everolimus and venetoclax (n=5 each) for 3 weeks. Upper left panel: tumor-derived luciferase activity was measured ex vivo by the IVIS Imaging System after D-luciferin injection. Representative photographs of NOD/SCID mice on day 0 and day 30 are shown (original magnification: x2). Upper middle panel: quantitative data of in vivo bioluminescence imaging shown in the left panel. The signal intensity is shown as photon units (photons/s). P-value was determined by one-way ANOVA with Tukey’s multiple comparison test. Upper right panel: tumor sections were prepared from vehicle-treated (control) and everolimus-treated mice at day 7 and stained with anti-BCL-2 rabbit polyclonal antibody or isotype-matched immunoglobulin G (IgG), followed by FITC-conjugated anti-rabbit IgG antibody (green), and PE-conjugated anti-human CD138 antibody (red). Nuclei were counterstained with DAPI (blue). Only merged images are shown. Data shown are representative of multiple independent experiments. Lower left panel: representative photographs of tumors resected on day 30 (original magnification: x2). Lower right panel: the means ± standard deviation (S.D.) (bars) of the weights of resected tumors shown in the left panel. P-value was determined by one-way ANOVA with Tukey’s multiple comparison test. (B) We inoculated 5x105 luciferase-expressing KMS12-BM cells (transplanted) or culture medium alone (control) subcutaneously in the right thigh of male NOD/SCID mice and randomized them into two treatment groups when measurable tumors developed in transplanted mice (day 0). Each group was treated with the vehicle alone (0.9% NaCl, orally, 5 times a week) or the combination of everolimus (4 mg/kg, orally, twice a week) and venetoclax (40 mg/kg, orally, 5 times a week) for 4 weeks. Left panel: we measured body weights of mice on the indicated days. The means ± S.D. (bars) are shown (n=5). No significant difference was noted between the groups by one-way ANOVA with Tukey’s multiple comparison test. Right panel: we estimated food intake of each group of mice (n=5 in a same cage) by calculating the consumption of supplied diets during a week (the day 0 value means food consumption between day -7 and day 0). (C) We measured the counts of white blood cells (WBC), hemoglobin, and platelets in peripheral blood of recipient mice on day 21 of treatment. The means ± S.D. (bars) are shown (n=5). **P<0.05 by Student's t-test. All animal studies were approved by the Institutional Animal Ethics Committee and performed in accordance with the Guide for the Care and Use of Laboratory Animals formulated by the National Academy of Sciences.

of MM. First, we determined that the maximal tolerated doses of everolimus and venetoclax were 4 mg/kg, twice a week and 40 mg/kg, fives times a week, respectively, for NOD/SCID mice in a pilot experiment (data not shown). We inoculated luciferase-expressing KMS12-BM cells subcutaneously in the right thigh of NOD/SCID mice and, when measurable tumors developed, started the treatment with vehicle alone (0.9% NaCl), everolimus alone, venetoclax alone or the combination of everolimus and venetoclax for randomly assigned groups of mice (n=5 each). The combined treatment with everolimus and venetoclax significantly retarded the growth of inoculated tumors as evidenced by luciferase activity traced ex vivo (Figure 3A, upper panels) and the size of tumors resected on day 30 (Figure 3A, lower panels), whereas either everolimus or venetoclax alone showed only moderate effects at the doses and schedules used. A histopathological examination of resected tumors confirmed the growth-inhibitory effect of the combination of the two drugs and mTOR inhibition-mediated BCL-2 up-regulation in vivo (Figure 3A, upper right panel). Everolimus has already been approved for the treatment of breast cancer, renal cell carcinoma and neuroendocrine tumors by the Food and Drug Adminstration, and is known to cause gastrointestinal toxicity such as mucositis, diarrhea, nausea and vomiting.13 Neither decreased food intake nor weight loss was observed in everolimusor everolimus/venetoclax-treated mice, although food intake slightly declined during experiments in all groups (Figure 3B). Moreover, we measured complete blood counts on day 21 to check for neutropenia, the most common side effect of venetoclax in MM patients,4,5 and other hematological toxicity that might be exacerbated in combination with everolimus. As shown in Figure 3C and the Online Supplementary Figure S2D, tumor implantation caused a significant decrease in leukocytes and platelets in NOD/SCID mice probably due to remote effects of MM cells on hematopoiesis.14 Notably, leukopenia and thrombocytopenia recovered after treatment with the combination of everolimus and venetoclax, likely reflecting their therapeutic effects on the disease. In conclusion, we have shown that mTOR inhibitors can enhance the anti-MM effects of venetoclax via upregulation of BCL-2 expression mainly through mTORC1 inhibition. Farber et al.15 reported that ATP-competitive dual inhibitors of mTORC1/2 augmented the effects of navitoclax via MCL-1 down-regulation in colon and lung cancer cells. The difference between the two observations may stem from the different cellular context or pat3012

terns of mTOR inhibition. The combined effect of mTOR inhibitors and venetoclax in vitro was reproduced in a murine model without obvious hematological and gastrointestinal toxicity. The efficacy and safety of this combination are worthy of investigation in clinical settings. Naoki Osada,1* Jiro Kikuchi,1* Daisuke Koyama,1 Yoshiaki Kuroda,1,2 Hiroshi Yasui,3 Joel D. Leverson4 and Yusuke Furukawa1 1 Division of Stem Cell Regulation, Center for Molecular Medicine, Jichi Medical University, Shimotsuke, Tochigi, Japan; 2Department of Hematology, National Hospital Organization Hiroshimanishi Medical Center, Otake, Hiroshima, Japan; 3The Institute of Medical Science, The University of Tokyo, Minato-ku, Tokyo, Japan and 4Oncology Discovery, AbbVie Inc., North Chicago, IL, USA *NO and JK contributed equally as co-first authors. Correspondence: YUSUKE FURUKAWA - furuyu@jichi.ac.jp doi:10.3324/haematol.2021.278506 Received: February 2, 2021. Accepted: July 2, 2021. Pre-published: July 15, 2021. Disclosures: this study was funded by AbbVie Inc. The authors declare no other potential conflicts of interest. Contributions: NO and JK designed and performed experiments, analyzed data, and drafted the manuscript; DK, YK, HY and JDL provided materials and critically reviewed the manuscript; YF designed and supervised research, and compiled the manuscript. Funding: this work was supported in part by a Grants-in-Aid for Scientific Research from JSPS (to NO, JK, and YF) and research grants from the International Myeloma Foundation Japan (NO, JK, and YF). NO received the Young Scientist Award from Jichi Medical University.

References 1. Valentin R, Grabow S, Davids MS. The rise of apoptosis: targeting apoptosis in hematologic malignancies. Blood. 2018;132(12):12481264. 2. Souers AJ, Leverson JD, Boghaert ER, et.al. ABT-199, a potent and selective BCL-2 inhibitor, achieves antitumor activity while sparing platelets. Nat Med. 2013;19(2):202-208. 3. Bewersdorf JP, Giri S, Wang R, et al. Venetoclax as monotherapy and in combination with hypomethylating agents or low dose cytarabine in relapsed and treatment refractory acute myeloid leukemia: a systematic review and meta-analysis. Haematologica. 2020; 105 (11):2659-2663.

haematologica | 2021; 106(11)

Letters to the Editor

4. Kumar S, Kaufman JL, Gasparetto C, et al. Efficacy of venetoclax as targeted therapy for relapsed/refractory t(11;14) multiple myeloma. Blood. 2017;130(22):2401-2409. 5. Kumar SK, Harrison SJ, Cavo M, et al. Venetoclax or placebo in combination with bortezomib and dexamethasone in patients with relapsed or refractory multiple myeloma (BELLINI): a randomised, double-blind, multicentre, phase 3 trial. Lancet Oncol. 2020;21(12):1630-1642. 6. Siu KT, Huang C, Panaroni C, et al. BCL2 blockade overcomes MCL1 resistance in multiple myeloma. Leukemia. 2019;33(8):20982102. 7. Oki S, Ohta T, Shioi G, et al. ChIP-Atlas: a data-mining suite powered by full integration of public ChIP-seq data. EMBO Rep. 2018;19(12):e46255. 8. Hewitt SM, Hamada S, McDonnell TJ, Rauscher III FJ, Saunders GF. Regulation of the Proto-oncogenes bcl-2 and c-myc by the Wilms' tumor suppressor gene WT1. Cancer Res. 1995;55(22):5386-5389. 9. Romero F, Martinez-A C, Camonis J, Rebollo A. Aiolos transcription factor controls cell death in T cells by regulating Bcl-2 expression and its cellular localization. EMBO J. 1999;18(12):3419-3430. 10. Hung K-H, Su S-T, Chen C-Y, et al. Aiolos collaborates with Blimp-

haematologica | 2021; 106(11)

1 to regulate the survival of multiple myeloma cells. Cell Death Differ. 2016;23(1):1175-1184. 11. Kikuchi J, Koyama D, Wada T, et al. Phosphorylation-mediated EZH2 inactivation promotes drug resistance in multiple myeloma. J Clin Invest. 2015;125(12):4375-4390. 12. Kakiuchi Y, Yurube T, Kakutani K, et al. Pharmacological inhibition of mTORC1 but not mTORC2 protects against human disc cellular apoptosis, senescence, and extracellular matrix catabolism through Akt and autophagy induction. Osteoarthritis Cartilage. 2019; 27(6):965-976. 13. Yao JC, Fazio N, Singh S, et al. Everolimus for the treatment of advanced, non-functional neuroendocrine tumours of the lung or gastrointestinal tract (RADIANT-4): a randomised, placebo-controlled, phase 3 study. Lancet. 2016;387(10022):968-977. 14. Dong M, Blobe GC. Role of transforming growth factor-b in hematologic malignancies. Blood. 2006;107(12):4589-4596. 15. Faber AC, Coffee EM, Costa C, et al. mTOR inhibition specifically sensitizes colorectal cancers with KRAS or BRAF mutations to BCL2/BCL-XL inhibition by suppressing MCL-1. Cancer Discov. 2013;4(1):42-52.

3013

Letters to the Editor

Individuals with sickle cell disease and sickle cell trait demonstrate no increase in mortality or critical illness from COVID-19 - a fifteen hospital observational study in the Bronx, New York Individuals with sickle cell disease (SCD) or sickle cell trait (SCT) could confer different susceptibility to critical illness and mortality associated with COVID-19 compared to the general population. Most COVID-19 SCD and SCT publications to date are case series with small sample sizes, registry studies with limited clinical variables, and without case-matched controls. In this study, we investigated whether patients with SCD or SCT confer different risk profiles of COVID-19 disease compared to the general population and matched controls in one of the largest healthcare systems in New York City. We found that SCD patients with COVID-19 were more likely to visit the emergency department (ED) and to be admitted to the hospital compared to the general population with COVID-19. However, mortality rate, critical illness and other outcomes were not different compared to matched or unmatched controls. Similarly, SCT patients showed no differences in laboratory values and had no increased risk of worse COVID-19 related outcomes compared to the general population or their matched controls. SCD is an inherited red blood cell disorder that cause red blood cells to “sickle” resulting in vaso-occlusive crisis (VOC) and multisystem disease.1 Individuals with SCD have immune-compromised status, chronic anemia, endothelial dysfunction, chronic inflammation, hypercoagulability, and related comorbidities that could increase susceptibility to worse COVID-19 outcomes. SCT is the heterozygous inheritance for sickle hemoglobin that was generally thought to be a benign carrier state, but has been linked to adverse health effects.2 The effects of COVID-19 on SCT individuals are not well documented. The aim of the current study was to examine the clinical outcomes in SCD and SCT associated with COVID19 disease. We used electronic health record data from the Montefiore Health System – a private, non-profit primary and specialty healthcare network of more than 180 locations across Westchester County, the lower Hudson Valley and the Bronx, New York, serving a large lowincome and racially diverse population. Our catchment area was severely impacted by COVID-19 and has a large population of SCD and SCT patients. Data were standardized to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (www.ohdsi.org)3 and searched for the period January 1, 2020 to January 21, 2021, imported into an SQLite data-

base (www.sqlite.org), and queried using the DB Browser (version 3.12.0). The primary outcome was mortality. Secondary outcomes were ED visits, hospitalization, length of stay (LOS), intensive care unit (ICU) admission, invasive mechanical ventilation (IMV), acute kidney injury (AKI), and acute liver injury (ALI). SCD status, SCT status and major comorbidities were based on ICD-10 codes and OMOP concept ID (Online Supplementary Figure S1). SARS-CoV-2 infection was confirmed by a positive real-time polymerase chain reaction test via a nasopharyngeal swab. All ED visits and hospitalizations were primarily due to COVID-19. For comparison, three cohorts with SARS-CoV-2 infection from the same hospital system and same time frame were included as controls: i) general population, ii) SCD-matched controls without SCD, and iii) SCT-matched controls without SCT. Using a nearest neighbor matching algorithm, each SCD or SCT patient was matched with up to four controls from the general population based on age (within 3 years), sex, race, ethnicity, and major comorbidities. This retrospective, observational cohort study was approved by the Einstein-Montefiore Institutional Review Board with an exemption for informed consent and a HIPAA waiver. The study was conducted according to the ethical principles of the Declaration of Helsinki. Among 12,659 COVID-19 patients, 53 had SCD (74% Hb-SS, 21% Hb-SC, and 6% Hb-S/b-thalassemia) and 62 had SCT (Table 1; Online Supplementary Table S1). Chart review was performed to confirm SCD or SCT diagnosis. Compared to the general population (median age 57 years; 30% Black; 42% Hispanic), both SCD (median age 30 years) and SCT cohorts (median age 47 years) were younger, had greater proportion of Black patients (76% and 61%, respectively) and fewer Hispanic patients (25% and 23%, respectively). There were more females with SCT which could be explained by more women knowing their trait status from recommended testing during pregnancy and men being less likely to seek medical care. Essential hypertension (21-40%) was the main comorbidity observed in all groups, but comorbidity burden was greater for SCD and SCT compared to the general population. SCD had lower BMI possibly due to increased metabolic demands and delayed physical and sexual maturation known to occur in SCD patients.4 After adjusting for age, sex, race, ethnicity and comorbidities as covariates in Logistic Regression, patients with SCD were more likely to visit the emergency department (adjusted odds ratio [adj. OR]=3.54, 95% confidence interval [CI]: 1.62-7.73, P=0.001) and to be hospitalized (adj. OR=7.26, 95% CI=3.75 to 14.08, P<0.001) as compared to the general population, but mortality and all other secondary outcomes were not significantly differ-

Table 1. Sample characteristics of COVID-19 patients with sickle cell disease, sickle cell trait, and the general population in this study.

Age in years, median (IQR) Female sex, n (%) Black, n (%) Hispanic, n (%) 2+ comorbidities, n (%)

(a) SCD n = 53

(b) SCT n = 62

P (a vs. c)

P (b vs. c)

30 (20-47) 25 (47%) 40 (76%) 13 (25%) 13 (25%)

47 (32-62) 49 (79%) 38 (61%) 14 (23%) 24 (39%)

57 (39-70) 6525 (52%) 3815 (30%) 5201 (42%) 2534 (20%)

<0.0001 0.49 <0.0001 <0.001 0.02

0.008 <0.001 0.0002 0.001 <0.0001

Note: Between-group comparison of continuous variables were performed using Wilcoxon rank-sum tests, and categorical variables were analyzed with Fisher’s exact test. For additional sample characteristics, including comorbidities, vitals and laboratory values, please see Online Supplementary Table S1. IQR: interquartile range; SCD: sickle cell disease; SCT: sickle cell trait.

3014

haematologica | 2021; 106(11)

Letters to the Editor

Figure 1. Covariate-adjusted Logistic Regression with odds ratios for clinical outcomes among COVID-19 patients with sickle cell disease or sickle cell trait relative to the general population with COVID-19. (A) Compared to the general population, sickle cell disease (SCD) patients with COVID-19 were more likely to visit the emergency department (adjusted odds ratio [adj. OR]=3.54, 95% confidence interval [CI]: 1.62-7.73, P=0.001) and to be hospitalized (adj. OR=7.26, 95% CI: 3.75-14.08, adj. P<0.001), but all other outcomes were not significantly different (P>0.05) in Logistic Regression adjusted for age, gender, race, ethnicity, and comorbidities as covariates. (B) COVID-19 patients with sickle cell trait had no increased risk of COVID-19 related clinical outcomes compared to the general population, except they were less likely to visit the emergency department (adj. OR=0.62, 95% CI: 0.47-0.82, P=0.001). The general population is the reference (vertical dotted line, OR=1). Error bars represent 95% CI. *P<0.01, **P<0.001. ED: emergency department; ICU: intensive care unit; IMV: invasive mechanical ventilation; LOS: length of stay.

ent (Figure 1 and Online Supplementary Table S2). There were no differences in clinical outcomes between SCT patients and the general population, except ED visits were lower in SCT (adj. OR=0.62, 95% CI: 0.47-0.82, P=0.001). With a median age of 30 years, our SCD cohort was relatively young consistent with other reports.5,6 Despite being of young age, SCD patients had more comorbidities than the older general population, consistent with SCD-related complications7 and end organ damage. SCD patients had lower hemoglobin and hematocrit, and higher monocyte count, reticulocytes, leukocytes, aspartate aminotransferase and bilirubin compared to the general population and SCT (Online Supplementary Table S1), consistent with red blood cell dysfunction and hepatobiliary manifestations of SCD.1,8 SCD patients had higher lactate dehydrogenase (LDH) and D-dimer (Online Supplementary Table S1), which could be suggestive of more severe COVID-19 disease9 but other explanations are possible. Elevated LDH, however could result from intravascular hemolysis, ischemia-reperfusion damage and tissue necrosis associated with SCD but could be further elevated in acute VOC.10 More studies are needed to further evaluate the consequences of immunological dysregulation associated with COVID-19 in SCD and SCT patients. Hospital visits were likely associated with SCD-related pain or acute chest syndrome (ACS) triggered by the COVID-19 disease11 as 67% of admitted SCD patients had one or more SCD-related symptoms at admission, including ACS (n=11), pain crisis (n=11), anemia (n=5), and splenic infarct (n=1). Crisis manifestations of SCD might have contributed to favorable outcomes due to haematologica | 2021; 106(11)

proactive seeking of medical care for SCD-related symptoms. In order to limit the potential confounding effects of group differences in demographic variables and preexisting conditions on COVID-19 outcomes, we conducted additional comparisons with age-, sex-, race-, ethnicityand comorbidity-matched controls (Table 2). We found neither significant differences (P>0.05, Wilcoxon ranksum tests) in COVID-19 related outcomes between SCD patients and matched controls, nor between SCT patient and their matched controls. Our findings suggest that individuals with SCD or SCT in this cohort did not carry an added risk of worse COVID-19 outcomes compared to individuals with similar demographics and health status without SCD or SCT. However, SCD patients could have other severe outcomes not evaluated here (e.g., pain or pneumonia).12 The mortality rate in our SCD patients (6-8%) is comparable to those reported in the US6 and UK registries13 (7.3% and 8.4%, respectively). Singh et al.12 found that Black individuals with SCD were more likely to be hospitalized and to develop pneumonia and pain, but no differences in mortality rate compared to matched Black individuals without SCD/SCT were observed, consistent with our findings. Similarly, Alkindi et al.14 reported COVID-19 infection may have triggered the onset of VOC, but it did not significantly influence the morbidity or mortality of SCD patients. The Bronx was disproportionally impacted by the first wave of COVID-19 with more hospitalizations and deaths than any other NYC borough,15 which may explain the high mortality (11%) in our non-SCD population. Although we presented one of the largest single center 3015

Letters to the Editor

Table 2. Comparisons of COVID-19 outcomes between hospitalized individuals with sickle cell disease (SCD) or sickle cell trait (SCT) and their respective age-, sex-, race-, ethnicity- and comorbidity-matched controls without SCD/SCT.

Clinical outcomes, n (%) Length of stay > 6 days Intensive care Invasive mechanical ventilation Acute kidney injury Acute liver injury In-hospital mortality

SCD n = 39

Matched controls n = 91

SCT n = 31

Matched controls n = 109

16 (47%) 4 (10%) 3 (8%) 13 (33%) 12 (34%) 3 (8%)

37 (42%) 17 (19%) 10 (11%) 32 (35%) 23 (32%) 5 (6%)

0.69 0.30 0.75 >0.99 0.83 0.70

10 (33%) 5 (16%) 4 (13%) 16 (52%) 4 (16%) 7 (23%)

38 (37%) 11 (10%) 10 (9%) 49 (45%) 18 (20%) 20 (18%)

0.83 0.35 0.51 0.55 0.78 0.61

Note: Categorical variables were analyzed with two-sided Fisher’s exact tests. Acute kidney injury was defined by KDIGO standards either a 0.3 mg/dL increase within 48 hours or 1.5 times the lowest reading during hospitalization due to lack of data prior to hospitalization. Acute liver injury was defined as serum levels of alanine aminotransferase and aspartate aminotransferase both exceeding 1x upper limit of normal, with the upper reference range as 40 U/L.

cohorts of SCD and SCT with COVID-19, additional multiple institutional data are needed to achieve greater generalizability. As with any retrospective study, there could be unintentional patient selection bias. Sickle cell trait status may be misclassified and, conversely, individuals may be unaware of their trait status. While we carefully reviewed patient charts to confirm trait status, patients were not tested for SCT due to the retrospective nature of the study. Finally, long-term outcomes of SCD and SCT COVID-19 patients should also be explored. In conclusion, although more COVID-19 patients with SCD visited the emergency department and were hospitalized, SCD and SCT did not carry an added risk of COVID-19 related escalated care and death compared to COVID-19 patients in the general population or those with similar demographics and health history. Our study underscores the importance of matched controls in defining risks associated with COVID-19. Wouter S. Hoogenboom,1 Roman Fleysher,1 Selvin Soby,2 Parsa Mirhaji,2 William B. Mitchell,3 Kerry A. Morrone,3 Deepa Manwani3 and Tim Q. Duong1 1 Department of Radiology, Albert Einstein College of Medicine; 2 The Montefiore Einstein Center for Health Data Innovations, Albert Einstein College of Medicine and 3Department of Pediatrics, Division of Hematology and Oncology, Albert Einstein College of Medicine, Bronx, NY, USA Correspondence: WOUTER HOOGENBOOM - wouter.hoogenboom@einsteinmed.org TIM Q. DUONG - tim.duong@einsteinmed.org doi:10.3324/haematol.2021.279222 Received: May 13, 2021. Accepted: July 21, 2021. Pre-published: August 5, 2021. Disclosures: no conflicts of interest to disclose. Contributions: WSH and TD conceived and led the project; RF and SS conducted database building, extraction and coding; WSH queried and analyzed the data, wrote the main manuscript text and created all tables and figures; DM, KAM, and WBM critically reviewed the manuscript and conducted chart review. All authors made a substantial intellectual contribution to the study, interpreted the data, discussed the

3016

results and reviewed, edited and approved the final version of the manuscript. Acknowledgments; we would like to acknowledge the contributions of the Montefiore Einstein Center for Health Data Innovations.

References 1. Rees DC, Williams TN, Gladwin MT. Sickle-cell disease. Lancet.

2010;376(9757):2018-2031. 2. Naik RP, Smith-Whitley K, Hassell KL, et al. Clinical outcomes associated with sickle cell trait: a systematic review. Ann Intern Med. 2018;169(9):619-627. 3. Hripcsak G, Ryan PB, Duke JD, et al. Characterizing treatment pathways at scale using the OHDSI network. Proc Natl Acad Sci U S A. 2016;113(27):7329-7336. 4. Zemel BS, Kawchak DA, Ohene-Frempong K, Schall JI, Stallings VA. Effects of delayed pubertal development, nutritional status, and disease severity on longitudinal patterns of growth failure in children with sickle cell disease. Pediatr Res. 2007;61(5 Pt 1):607-613. 5. Arlet JB, de Luna G, Khimoud D, et al. Prognosis of patients with sickle cell disease and COVID-19: a French experience. Lancet Haematol. 2020;7(9):e632-e634. 6. Panepinto JA, Brandow A, Mucalo L, et al. Coronavirus disease among persons with sickle cell disease, United States, March 20-May 21, 2020. Emerg Infect Dis. 2020;26(10):2473-2476. 7. Gladwin MT, Vichinsky E. Pulmonary complications of sickle cell disease. N Engl J Med. 2008;359(21):2254-2265. 8. Rees DC, Gibson JS. Biomarkers in sickle cell disease. Br J Haematol. 2012;156(4):433-445. 9. Wiersinga WJ, Rhodes A, Cheng AC, Peacock SJ, Prescott HC. Pathophysiology, transmission, diagnosis, and treatment of coronavirus disease 2019 (COVID-19): A Review. JAMA. 2020;324(8):782793. 10. Kato GJ, McGowan V, Machado RF, et al. Lactate dehydrogenase as a biomarker of hemolysis-associated nitric oxide resistance, priapism, leg ulceration, pulmonary hypertension, and death in patients with sickle cell disease. Blood. 2006;107(6):2279-2285. 11. Nur E, Gaartman AE, van Tuijn CFJ, Tang MW, Biemond BJ. Vasoocclusive crisis and acute chest syndrome in sickle cell disease due to 2019 novel coronavirus disease (COVID-19). Am J Hematol. 2020; 95(6):725-726. 12. Singh A, Brandow AM, Panepinto JA. COVID-19 in individuals with sickle cell disease/trait compared with other Black individuals. Blood Adv. 2021;5(7):1915-1921. 13. Telfer P, De la Fuente J, Sohal M, et al. Real-time national survey of COVID-19 in hemoglobinopathy and rare inherited anemia patients. Haematologica. 2020;105(11):2651-2654. 14. Alkindi S, Elsadek RA, Al-Madhani A, et al. Impact of COVID-19 on vasooclusive crisis in patients with sickle cell anaemia. Int J Infect Dis. 2021;106:128-133. 15. Wadhera RK, Wadhera P, Gaba P, et al. Variation in COVID-19 hospitalizations and deaths across New York city boroughs. JAMA. 2020;323(21):2192-2195.

haematologica | 2021; 106(11)

Letters to the Editor

Allogeneic hematopoietic stem cell transplantation from non-sibling 10/10 HLA-matched related donors: a single-center experience Hematopoietic stem cell transplantation (HSCT) from an allogeneic donor (allo-HSCT) is a potentially curative treatment for hematologic malignancies and nonmalignant disorders. Although human leukocyte antigen (HLA) matching between donors and recipients is critical for transplant outcomes,1 with the development of novel methods to overcome the alloreactivity caused by HLA disparity and improvements in the management of transplant-related complications, the use of HLA-haploidentical donors as an alternative source of stem cells has increased continuously over the past decade and grafts from such sources may have a superior graft-versus-leukemia effect compared to those from HLAmatched siblings.2,3 Studies have suggested that various donor-related factors, such as age, sex, ABO compatibility, natural killer cell alloreactivity, donor-recipient cytomegalovirus serostatus and donor-specific anti-HLA antibodies are correlated with recipients’ survival after haploidentical allogeneic HSCT (haplo-HSCT).4 However, the degree of HLA mismatch between donors and recipients may not influence transplant outcomes, especially among patients being treated within the Beijing protocol.5 Given the extreme polymorphism of HLA genes, without consanguineous marriage the probability of finding a non-sibling 10/10 HLA-matched related donor (NSMRD) is quite low. Since recipients and donors have one identical haplotype and one matched haplotype, the risk of graft-versus-host disease (GvHD), relapse and leukemia-free survival after transplantation from these unique donors may have some particularities that cannot simply be extrapolated from ordinary HLA-haploidentical donors, HLA-identical siblings or 10/10 matched unrelated donors. In this study, we describe our experience with allo-HSCT in 23 patients with hematologic malignancies transplanted from NSMRD and compare the outcomes to those of patients who underwent conventional haplo-HSCT. The relatively higher relapse rate and lower acute GvHD incidence show that NSMRD are less immunogenic than conventional haploidentical donors and that individualized treatment with adjustment of the dosage of antithymocyte globulin (ATG) is reasonable for recipients of grafts from such donors. From December 2012 to December 2019, 2,726 HSCT from family members (HLA-matched sibling donors and HLA-haploidentical related donors) were performed in our hospital, and only 23 cases (15 males, 8 females) (0.8%) were transplanted from NSMRD. All patients received the same modified busulfan/cyclophosphamide/ATG-based myeloablative conditioning regimen recommended by Chinese guidelines, except the total dose of ATG was 10.0 mg/kg in patients undergoing conventional haplo-HSCT,6 but it was reduced in NSMRD-HSCT patients at the discretion of each physician. Of these 23 patients transplanted from NSMRD, 13 were treated with ATG at doses ≤6.0 mg/kg, and ten were treated with ATG at doses >6.0 mg/kg. The median ATG dose was 5.0 mg/kg and the interquartile range was 5.0 to 8.4 mg/kg. GvHD prophylaxis consisted of continuous infusion of cyclosporine at 3.0 mg/kg/day starting on day -10 and a short course of methotrexate and mycophenolate mofetil at a dose of 1.0 g/day from days -10 to +30. The median follow-up time of living patients was 19 months (range, 2 to 93 months). The patients’ haematologica | 2021; 106(11)

characteristics are shown in Table 1. The study was approved by the Institutional Review Board of our hospital. All patients achieved engraftment with full donor chimerism at day 30. The median time to neutrophil and platelet engraftment was 12 days (range, 11 to 12) and 12 days (range, 12 to 15), respectively. Eight patients developed acute GvHD, which reached grade I in one patient, grade II in six patients and grade III in one patient. The cumulative incidence of grade II-IV acute GvHD was 38.8% (95% confidence interval [95% CI]: 20.4% to 61.2%) for the whole cohort. This was the same in patients who were treated with ATG at doses ≤6.0 mg/kg as in those who were treated with ATG at doses >6.0 mg/kg. The only patient who experienced grade III acute GvHD was in the ATG high-dose group. In patients who survived and were in remission beyond day 100, chronic Table 1. Characteristics of the study participants at enrollment (total patients N 22).

Variables Age (years), median (IQR) Gender, n (%) Female Male Diagnosis, n (%) Acute lymphocytic leukemia Acute myeloid leukemia Myelodysplastic syndrome CR/CRi at transplant, n (%) Yes No Donor-recipient relationship, n (%) Daughter Son Mother Father Donor-recipient ABO group match, n (%) Matched Major mismatched Minor mismatched Stem cell source, n (%) Bone marrow Bone marrow + peripheral blood Peripheral blood Mononuclear cells, mean±SD CD34, mean±SD Antithymocyte globulin dose, n (%) ≤6.0 mg/kg >6.0 mg/kg Time to granulocyte recovery (days), median (IQR) Time to platelet recovery (days), median (IQR) Acute GvHD grade, n (%) Grade 0 Grade 1 Grade 2 Grade 3-4

37 (27, 51) 8 (35) 15 (65) 7 (30) 9 (39) 7 (31) 18 (78) 5 (22) 3 (13) 8 (35) 4 (17) 8 (35) 14 (61) 5 (21) 4 (18) 1 (4) 14 (61) 8 (35) 10.96 ± 4.5 3.56 ± 1.48 13 (57) 10 (43) 12 (11, 12) 12 (12, 15) 15 (65) 1 (4) 6 (26) 1 (4)

IQR: interquartile range; CR: complete remission; Cri: CR with incomplete count recovery; SD: standard deviation; GvHD: graft-versus-host disease.

3017

Letters to the Editor

Figure 1. Transplant outcomes according to type of donor. (A-E) Differences in probabilities of overall survival (A), cumulative incidences of relapse (B), acute graft-versus-host disease (C), and transplant-related mortality (D) and probabilities of relapse-free survival (E) between patients undergoing haploidentical allogeneic stem cell transplantation and those being transplanted from a non-sibling matched related donor. OS: overall survival; aCvHD: acute graft-versus-host disease; TRM: transplant-related mortality; RFS: relapse-free survival; NSMRD: nonsibling matched related donor, Haplo: haploidentical allogeneic donor; HSCT: hematopoietic stem cell transplantation.

GvHD occurred in five out of the 13 patients in the ATG low-dose group and five out of the ten patients in the ATG high-dose group. The cumulative incidence of chronic GvHD was 43.4% at 2 years (95% CI: 23.9% to 65.1%). Three patients died of transplant-related causes, which included infection (n=1) and extensive chronic GvHD (n=2). Six patients had relapsed at a median time of 5 months (range, 2 to 9) after HSCT and ultimately died. The estimated 2-year overall survival and relapsefree survival for the whole cohort were 61.5% (95% CI: 44.3% to 86.2%) and 62.5% (95% CI: 43.4% to 78.2%), respectively. There was no significant difference in overall survival (P=0.68) and relapse-free survival (P=0.53) between the two groups treated with different doses of ATG. To further compare outcomes by donor type, we performed a propensity-matched analysis in which each patient who underwent NSMRD-HSCT was matched 1:1 with a control patient who received a contemporaneous transplant from a mismatched haplo-donor (Table 2). Age, gender, diagnosis, Disease Risk Index Group,7 donor and recipient blood type, and relationship were selected as factors for propensity-matched analysis. All selected patients were negative for donor-specific anti-HLA anti3018

bodies. The results for patients who underwent haploHSCT from mismatched donors, the overall survival (63.7% vs. 61.5%, P=0.520), relapse-free surival (89.5% vs. 62.5%, P=0.093) and transplant-related mortality (22.2% vs. 23.5%, P=0.740) did not show any statistically significant differences from those for patients with NSMRD-HSCT (Figure 1A, D, E). However, patients who underwent NSMRD-HSCT showed trends to a higher relapse rate (37.5% vs. 10.5%, P=0.093) (Figure 1B) and lower cumulative incidence of grade II-IV acute GvHD (38.8% vs. 52.9%, P=0.310) than patients who accepted haplo-HSCT with mismatched donors (Figure 1C). So, it seems that outcomes after NSMRD-HSCT were closer to those of sibling transplantation (Online Supplementary Table S1, Online Supplementary Figure S1). The probability of finding a fully matched non-sibling related donor through an extended family search is high in regions in which consanguineous marriage is frequently practiced. In a retrospective analysis from Iran by Hamidieh and colleagues, outcomes of 109 patients transplanted from fully matched other-relative donors were reported, and were comparable to those of transplantation from matched sibling donors.8 As consanguineous marriage is prohibited in China, the chance of haematologica | 2021; 106(11)

Letters to the Editor

Table 2. Patients’ baseline and transplant characteristics before and after matching on the propensity score.

Variables Haplo (n=782)

Before matching NSMRD (n=23)

Age (years), median (IQR) 38.98 (27.48, 47.46) 37.45 (27.02, 50.65) Gender, n (%) Female 346 (44) 8 (35) Male 436 (56) 15 (65) Diagnosis, n (%) Acute lymphocytic leukemia 228 (29) 7 (30) Acute myeloid leukemia 393 (50) 9 (39) Mixed phenotype acute leukemia 24 (3) 0 (0) Myelodysplastic syndrome 137 (18) 7 (30) Disease risk index*, n (%) Low risk 161 (21) 4 (17) Intermediate risk 424 (56) 14 (61) High and very high risk 197 (25) 5 (22) Donor-recipient relationship, n (%) Daughter 197 (25) 3 (13) Son 280 (36) 8 (35) Father 240 (31) 8 (35) Mother 65 (8) 4 (17) Donor-recipient ABO group match, n (%) Matched 434 (55) 14 (61) Major mismatched 142 (18) 5 (22) Minor mismatched 161 (21) 4 (17) Major+minor mismatched 45 (6) 0 (0) Donor-recipient HLA disparities, n (%) 5/10 214 (27) 6/10 80 (10) 7/10 218 (28) 8/10 270 (35) 9/10 270 (35) 10/10 0 (0) 23 (100)

P 0.810 0.491

0.013

0.819

0.296

0.799

Haplo (n=23)

After matching NSMRD (n=23)

33.44 (27.26, 53.39) 37.45 (27.02, 50.65) 0.693 0.749 6 (26) 8 (35) 17 (74) 15 (65) 0.212 3 (13) 7 (30) 11 (48) 9 (39) 0 (0) 0 (0) 9 (39) 7 (31) 0.890 3 (13) 4 (17) 14 (61) 14 (61) 6 (26) 5 (22) 0.914 3 (13) 3 (13) 8 (35) 8 (35) 10 (43) 8 (35) 2 (9) 4 (17) 0.085 8 (35) 14 (61) 8 (35) 5 (22) 3 (13) 4 (17) 4 (17) 0 (0) 15 (65) 5 (11) 2 (8) 1 (4) 0 (0) 0 (0) 23 (100)

Haplo: haploidentical allogeneic hematopoietic stem cell donor; NSMRD: non-sibling 10/10 HLA-matched related donor; IQR: interquartile range; HLA: human leukocyte antigen. *Risk was defined according to the refined Disease Risk Index of the Center for International Blood and Marrow Transplant Research.7

finding such donors is very low, and we believe that the scenario is totally different in our study. Although the donor and recipient are 10/10 HLA-matched, theoretically, they only have one identical haplotype, while the other should be defined as a matched haplotype. On the haplotype-matched chromosome, there may be donorrecipient mismatching at additional HLA alleles (such as HLA-DPB1) and non-HLA-linked immune-related genes with polymorphisms (such as TNFA and MICA). These differences may affect the outcome of transplantation.9,10 However, whether further matching at the haplotype level has a significant impact on clinical outcomes is controversial.11 The relapse risk and GvHD incidence in the control group were 10.5% and 52.9%, respectively, which are similar to those in a previous report.5 Although the dose of ATG in the NSMRD cohort was relatively low, there was a trend toward a higher relapse rate and lower GvHD rate in these patients than in patients in the control group, but this trend did not reach statistical significance due to the sample size. Our results suggest that these unique donors are less immunogenic and more similar to HLA-matched sibling donors. The outcome of allohaematologica | 2021; 106(11)

HSCT from these donors might be different. However, a previous study from the Beijing group indicated that with the advent of the ATG protocol, the degree of HLA disparity on the unshared HLA haplotype was not significantly correlated with transplant outcomes, but HLA typing was only performed at the HLA-A, HLA-B, and HLA-DR loci, and only three cases (0.2%) with 6/6 HLAmatched related donors were involved in their study.12 ATG has been widely used for the prevention of GvHD in matched unrelated donor HSCT and haplo-HSCT. However, at high doses, ATG may lead to an increase in fatal infections, relapse, or delayed engraftment due to delayed immune reconstitution of T cells.13 Therefore, the optimal ATG dose remains unclear and should be determined on the basis of a balance of advantages and disadvantages of ATG. The Chinese Society of Hematology recommends that the dose of ATG should be 10.0 mg/kg in haplo-HSCT.6 Given the trend to a lower rate of GvHD observed in NSMRD-HSCT, the lack of significant difference between the groups treated with ≤6.0 or >6.0 mg/kg ATG, and the absence of fatal cases of acute GvHD in the low-dose group, it is reasonable to speculate that the immunosuppression should not be too 3019

Letters to the Editor

deep in this situation. Meanwhile, although a high dose of ATG may not be required, a reduced dose of ATG may still be necessary. In the research by Hamidieh et al. mentioned above, although both haplotypes were identical in the recipient and donor, more than 85% of patients had received ATG in the conditioning regimen, but the authors did not describe the dose.8 Although nonmalignant hematologic diseases, such as aplastic anemia and paroxysmal nocturnal hemoglobinuria, were not included in this study, according to the immunological specificities of NSMRD, such donors should be the first choice for patients with these diseases. However, in the clinical setting of hematologic malignancies, especially in patients with a high risk of relapse, the choice of these donors must be made with caution. Similarly to what was observed in our study, recent studies have shown that, as a consequence of the potential strong graft-versus-leukemia effects, haplo-HSCT, compared to HLA-mathced HSCT, can provide equivalent or even better long-term survival in high-risk leukemia patients because of a lower relapse rate.14,15 The limitations of this study include its retrospective nature and the small sample size. Whether the results could be extrapolated to haplo-HSCT using a post-transplant cyclophosphamide-based regimen is unclear. However, we believe that our data are a useful supplement to the donor selection recommendations for haploHSCT and that individualized treatment with adjustment of the dosage of ATG is reasonable for patients undergoing NSMRD-HSCT. Yaoyao Shen,1,2,3 * Jiaqian Qi,1,2,3* Jia Chen,1,2,3 Yang Xu,1,2,3 Feng Chen,1,2,3 Xiao Ma,1,2,3 Miao Miao,1,2,3 Shengli Xue,1,2,3 Huiying Qiu,1,2,3 Xiaowen Tang,1,2,3 Yue Han,1,2,3* Suning Chen,1,2,3 Aining Sun,1,2,3 Depei Wu1,2,3,4 and Ying Wang1,2,3,4 1 National Clinical Research Center for Hematologic Diseases, Jiangsu Institute of Hematology, The First Affiliated Hospital of Soochow University, Suzhou,; 2Institute of Blood and Marrow Transplantation, Collaborative Innovation Center of Hematology, Soochow University, Suzhou; 3Key Laboratory of Thrombosis and Hemostasis of Ministry of Health, Suzhou, China and 4State Key Laboratory of Radiation Medicine and Protection, Soochow University, Suzhou, China. *YS and JQ contributed equally as co-first authors. Correspondence: YING WANG - yingwang1977@hotmail.com DEPEI WU - drwudepei@163.com doi:10.3324/haematol.2021.278933 Received: April 7, 2021. Accepted: July 21, 2021. Pre-published: July 29, 2021. Disclosures: no conflicts of interest to disclose. Contributions: YS designed and performed the research, analyzed the data, and wrote the manuscript; JQ completed the research and analyzed the data; JC, YX, FC, XM, MM and SX contributed to the data analysis and manuscript writing; HQ, XT, YH, SC and AS contributed to the external validation; YW and DW contributed to the research design, data analysis, writing the

3020

manuscript, and supervision of the study. Acknowledgments: we thank Yi Wang for assistance in collecting the data. All the samples were from Jiangsu Biobank of Clinical Resources. Funding: this work was supported by grants from the National Natural Science Foundation of China (81730003, 81870120, 81670164), the Natural Science Foundation of Jiangsu Province (BK20171205), the Social Development Project of Jiangsu Province (BE2019655), the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and the National Key Research and Development Program (2017ZX09304021, 2017YFA0104500, 2019YFC0840604).

References 1. Shouval R, Fein JA, Labopin M, et al. Outcomes of allogeneic haematopoietic stem cell transplantation from HLA-matched and alternative donors: a European Society for Blood and Marrow Transplantation registry retrospective analysis. Lancet Haematol. 2019;6(11):573-584. 2. Wang Y, Liu QF, Xu LP, et al. Haploidentical vs identical-sibling transplant for AML in remission: a multicenter, prospective study. Blood. 2015;125(25):3956-3962. 3. Passweg JR, Baldomero H, Bader P, et al. Use of haploidentical stem cell transplantation continues to increase: the 2015 European Society for Blood and Marrow Transplant activity survey report. Bone Marrow Transplant. 2017;52(6):811-817. 4. Ciurea SO, Al Malki MM, Kongtim P, et al. The European Society for Blood and Marrow Transplantation (EBMT) consensus recommendations for donor selection in haploidentical hematopoietic cell transplantation. Bone Marrow Transplant. 2020;55(1):12-24. 5. Huang XJ, Liu DH, Liu KY, et al. Haploidentical hematopoietic stem cell transplantation without in vitro T-cell depletion for the treatment of hematological malignancies. Bone Marrow Transplant. 2006;38(4):291-297. 6. Xu L, Chen H, Chen J, et al. The consensus on indications, conditioning regimen, and donor selection of allogeneic hematopoietic cell transplantation for hematological diseases in China-recommendations from the Chinese Society of Hematology. J Hematol Oncol. 2018;11(1):33. 7. Armand P, Kim HT, Logan BR, et al. Validation and refinement of the Disease Risk Index for allogeneic stem cell transplantation. Blood. 2014;123(23):3664-3671. 8. Hamidieh AA, Dehaghi MO, Paragomi P, et al. Efficiency of allogeneic hematopoietic SCT from HLA fully-matched non-sibling relatives: a new prospect of exploiting extended family search. Bone Marrow Transplant. 2015;50(4):545-552. 9. Ishikawa Y, Kashiwase K, Akaza T, et al. Polymorphisms in TNFA and TNFR2 affect outcome of unrelated bone marrow transplantation. Bone Marrow Transplant. 2002;29(7):569-575. 10. Bettens F, Passweg J, Schanz U, et al. Impact of HLA-DPB1 haplotypes on outcome of 10/10 matched unrelated hematopoietic stem cell donor transplants depends on MHC-linked microsatellite polymorphisms. Biol Blood Marrow Transplant. 2012;18(4):608-616. 11. Buhler S, Baldomero H, Ferrari-Lacraz S, et al. High-resolution HLA phased haplotype frequencies to predict the success of unrelated donor searches and clinical outcome following hematopoietic stem cell transplantation. Bone Marrow Transplant. 2019;54(10):1701-1709. 12. Wang Y, Chang YJ, Xu LP, et al. Who is the best donor for a related HLA haplotype-mismatched transplant? Blood. 2014;124(6):843-850. 13. Lindemans CA, Chiesa R, Amrolia PJ, et al. Impact of thymoglobulin prior to pediatric unrelated umbilical cord blood transplantation on immune reconstitution and clinical outcome. Blood. 2014;123(1):126132. 14. Rashidi A, Hamadani M, Zhang MJ, et al. Outcomes of haploidentical vs matched sibling transplantation for acute myeloid leukemia in first complete remission. Blood Adv. 2019;3(12):1826-1836. 15. Yu S, Huang F, Wang Y, et al. Haploidentical transplantation might have superior graft-versus-leukemia effect than HLA-matched sibling transplantation for high-risk acute myeloid leukemia in first complete remission: a prospective multicentre cohort study. Leukemia. 2020;34(5):1433-1443.

haematologica | 2021; 106(11)

CASE REPORTS Massive cerebral venous thrombosis due to vaccine-induced immune thrombotic thrombocytopenia Since the breakthrough of coronavirus disease (COVID-19) more than 3 million people died worldwide1 and different vaccines were developed, tested in phase III clinical trials and used in the general population. Few reports of moderate-to-severe thrombocytopenia and thromboses (especially cerebral-venous and splanchnicvein thromboses) developing approximately 4-14 days after vaccination were reported. These events were related to the adenovirus vector–based DNA vaccines ChAdOx1 nCoV-19 (Oxford–AstraZeneca)2,3,4 or Ad26.COV2.S (Johnson&Johnson/Janssen).5 Recently, this new rare autoimmune syndrome that mimics heparin-induced thrombocytopenia (HIT)3 was defined as vaccine-induced immune thrombotic thrombocytopenia (VITT).4 Even though details on the pathophysiology are still scanty, diagnostic and therapeutic recommendations were proposed by international scientific organizations.6-8 No definite data on risk factors are reported and it is unknown whether or not the therapeutic options currently adopted for HIT are also valid for VITT. With this background, we describe an Italian case of severe VITT-related cerebral venous thrombosis (CVT) and bi-hemispheric hemorrhage, which was successfully treated with argatroban, intravenous immunoglobulin (IVIG) and corticosteroids. The case report is described according to CARE (CAse REport)-statement and checklist.9 A previously healthy 26-year-old female presented to the emergency department 14 days after the first injection of ChAdOx1 nCoV-19 vaccine with a headache nonresponsive to anti-inflammatory drugs. On admission, she had right-sided weakness and visual disturbances. She has been on combined (estrogen-progestogen) contraceptives for more than 10 years but her past medical history was otherwise unremarkable and there was no prior exposure to heparin. While general examination and vital signs were normal, neurological examination found a severe right-sided weakness but no visual field defects. Computerised tomography (CT) scan at admission showed a hyperdense rectus sinus and vein of Galen (Figure 1A). Magnetic resonance imaging (MRI) venography showed multifocal venous thrombosis with bilateral occlusion of parietal cortical veins, straight sinus, vein of Galen, internal cerebral veins and inferior sagittal sinus. Transverse sinuses were also partially involved but still patent (Figure 1B). At the right parietal and left frontoparietal lobes an extensive venous infarction with hemorrhagic transformation was present (Figure 1C). D-dimer was dramatically raised to 12,204 mg/L (reference value <500 mg/L) and the platelet count was 134x109/L. Given her recent exposure to ChAdOx1 nCoV-19 and clinical presentation, she was first treated with fondaparinux (5 mg subcutaneously) and admitted her to the intensive care unit. Her clinical condition rapidly deteriorated with decreased consciousness, right-sided hemiplegia and complete Balint syndrome. In order to perform an extensive hemostasis laboratory work-up before and after therapies, blood was collected at different time points (T0=April 13; T1=April 15, and T2=April 20, 2021) into vacuum-tubes containing 1/10 volumes of trisodium-citrate 0.109 M, K-EDTA or plain tubes. Activated partial thromboplastin time (aPTT), prothrombin time (PT), D-dimer, fibrinogen and factor haematologica | 2021; 106(11)

(F)VIII were obtained. Platelet-factor 4 (PF4)–heparin IgG antibodies (aPF4) were evaluated by a commerciallyavailable enzyme-linked immunosorbant assay (ELISA) (Immucor, Waukesha, WI, USA). Platelet-activating antibodies were evaluated by a platelet-activation test (PAT).2,10 Platelet function was also evaluated by using the Total Thrombus-Formation Analysis device (T-TAS®, Zacros, Fujimori-Kogyo, Tokyo, Japan),11-12 a flow-chamber device that assesses platelet-mediated thrombus formation in capillary channels by means of the following parameters: area under the flow-pressure curve (AUC), occlusion start-time (OST) and occlusion time (OT). Thrombin generation (TG) was measured in plateletpoor plasma (PPP).13 Controls were plasma samples from subjects negative for aPF4 and normal TG. PT, aPTT and fibrinogen were within the normal range; FV-Leiden and G20210A-prothrombin mutations were absent; antithrombin and protein C/S were normal; lupus anticoagulant and antiphospholipid antibodies were negative. Patient serum (T1) was positive to aPF4-heparin ELISA (OD=1.918, reference value <0.4) and was inhibited (OD<0.5) by 100 U/mL heparin. Patient serum (T1) showed strong platelet activation on PPP from two controls in the presence and absence of low-dose heparin, whereas control serum showed no platelet activation. Five days afterwards (T2), the patient serum showed significant reduction of aPF4 reactivity (OD=0.6) and no longer did activate platelets (Figure 2A to C). At T0, platelet thrombus formation was impaired, AUC was smaller and OT longer than the reference range. In contrast, at T1 and T2 thrombus stability improved and TTAS parameters as well as platelet count also improved (Figure 2D to F). Results at the time of hospital admission (T1) showed a marked state of hypercoagulability when compared to control, as indicated by short lag-time (8.5 minutes [min] vs. 21.3 min), increased thrombin-peak (289 nM vs. 115 nM), short time-to-peak (11.8 min vs. 26.2min), increased ETP (2,158 nM/min vs. 1,684 nM/min) and ETP-TM ratio (0.99 vs. 0.79) (Figure 3). FVIII, one of the most potent procoagulants, was higher (200 U/dL) than the upper limit of the reference range (<150 U/dL). In contrast PC, the physiological inhibitor to activated FVIII was normal (88 U/dL). The imbalance between FVIII and PC corresponded to an increased FVIII/PC ratio (2.3), considerably greater than the expected unity and consistently with the hypercoagulability shown by TG. There are potential limitations of the TG assessment. First, measurements (owing to assay complexity and limited blood volumes) were performed only in PPP. Therefore, the potential role of procoagulant platelets in supporting TG could not be assessed. Second, TG could not be assessed on samples obtained during the time course of the disease because soon after the onset of the symptoms and the preliminary diagnosis the patient was treated with anticoagulants, so that TG results would have been unreliable. Considering the clinical conditions and laboratory results, IVIG (1 g/kg o.d. for 2 days) and dexamethasone (40 mg/day, for 4 days) were started.6-8 Owing to the possible need for a sudden decompressive neurosurgical intervention, anticoagulation with fondaparinux was replaced by the short-acting drug argatroban (starting dosage 1 mg/kg/min with an aPTT-ratio [patient/normal] target of 1.5-2.0). Argatroban was subsequently increased to 3 mg/kg/min. The patient’s neurological conditions improved in the next few days. She was awake and fully responsive to stimuli with a progressive recovery of right upper-limb strength, partial optic ataxia and regression of apraxia. 3021

Case Reports

On a follow-up CT scan, the rectus sinus and the vein of Galen showed normal density with oedema in the brain tissue on both hemispheres (Figure 1D). Follow-up MRI venography showed restored venous flow in the rectus sinus and the vein of Galen; right internal cerebral vein and bilateral frontoparietal cortical veins were still occluded (Figure 1E), and the large intraparenchimal venous infarction was unchanged (Figure 1F). In the next few days, platelet count progressively

increased to 339x109/L and D-dimer decreased to normal levels, in parallel with a significant reduction of aPF4 reactivity after 3 days (OD=0.9), and after 1 week (OD=0.6) patient serum was no longer able to activate platelets (Figure 2). Currently (nearly 2 months after the onset of symptoms), the patient has moderate disability: she has no neuropsychological deficits, can walk unassisted for short distances (sustained clonus and spasticity coexist in her right leg) and her right arm almost fully

C F

Figure 1. Neuroradiological findings at baseline and follow-up. (A) baseline computerised tomography (CT) (at admission) shows hyperdense rectus sinus and vein of Galen as signs of thrombosis (arrow). (B) Magnetic resonance imaging (MRI) examination (day 1) confirms complete occlusion of the rectus sinus, vein of Galen, right internal cerebral vein (arrow) and frontoparietal cortical veins on both sides (arrow head) on venous angiography. (C) On coronal T2-FLAIR images extensive venous infarctions with hemorrhagic transformation can be seen in right parietal (arrow) and left frontoparietal (arrow head) regions. (D) On follow-up CT (day 6) rectus sinus and vein of Galen show normal density (arrow) with oedema in brain tissue on both hemispheres (arrow head). Follow-up MRI (day 7) shows (E) restored venous flow in the rectus sinus and the vein of Galen (arrow); right internal cerebral vein and bilateral frontoparietal cortical veins are still occluded. (F) On T2 weighted images the large bilateral venous infarctions are still visible with hemorrhagic transformation in pre- and postcentral gyrus in the left side (arrow).

3022

haematologica | 2021; 106(11)

Case Reports

recovered. Fondaparinux was replaced with oral vitamin K antagonist. In summary, we report a case of 26-year-old female who developed VITT following the first dose of ChAdOx1 nCoV-19. The patient had high-titer aPF4 and signs of platelet activation. Various treatment regimens

have been suggested for this rare syndrome, being all based on HIT-derived approaches. Our treatment protocol, based upon IVIG, dexamethasone and argatroban was successful with almost complete clinical, laboratory and radiological response. VITT should be considered in post-vaccination cases of

Figure 2. Platelet activation test and thrombus formation analysis. (A to C) Results of the platelet activation test (PAT) assessed under various conditions. Citrated blood from two healthy donors was collected and platelet-rich and platelet-poor plasma (PRP, PPP) were prepared by centrifugation (15 minutes [min]) at 200 g or 1,400 g, respectively and kept at 37°C.14 Platelet counts for PRP were 5,6x109/L. Heat-inactivated (56°C, 30 min) serum from patient or controls (negative for aPF4) was incubated with PRP with or without LMWH 0.2 U/mL or unfractionated-heparin (UFH) 100 U/mL. Aggregation was assessed by means of a light-transmission aggregometer (Chrono-log, Mascia-Brunelli, Milano, Italy). Results were expressed as increase of light-transmission (%LT). PAT was considered positive if aggregation occurred in at least one of the two donors in the absence and presence of low LMWH and was inhibited by UFH. Panel (A) shows PAT results at T1 with patient serum that caused platelet activation in presence of LMWH (2), but also in absence of heparin (1) (85%, 77%LT, respectively), in contrast, high levels of UFH inhibited the reaction (3) (0%LT). The negative control serum did not cause aggregation (4) (0%LT). Similar results were obtained with the PRP of the other healthy donor. Panel (B) shows PAT results at T2 with patient serum that behaved like the negative control serum (i.e., it did not activate platelets under any of the conditions) (1-2-3-4) (0%LT). Panel (C) shows PAT results with serum of an asymptomatic subject positive to aPF4 enzyme-linked immunsorbant assay, but negative to PAT (1-2-3-4) (0%LT). Panels (D to F) show results of Total Thrombus Formation Analysis (T-TAS). In order to analyze thrombus formation under flowing conditions, whole blood was applied to type 1 collagen-coated chips at a flow rate of 24 mL/min which corresponds to a wall shear stress of 2,000s-1. Thrombus-formation was analyzed through the following parameters: AUC, defined as the area under the flow pressure curve (D), OST (min), defined as occlusion start time (i.e, the time needed to reach the baseline pressure indicating the onset of the platelet thrombus formation) (E), and OT, defined as occlusion time (i.e., the time at which the occlusion pressure is reached (F). The dotted horizontal lines represent reference ranges (2.5th-97.5th centiles). aPF4: platelet-factor 4 (PF4)–heparin IgG antibodies.

Figure 3. Thrombin generation. Thrombing generation (TG) was assessed as previously described13 following in vitro coagulation activation of platelet poor plasma by calcium chloride with no addition of exogenous triggers. The TG reaction was monitored by means of fluorogenic substrate for 20 minutes (min) in a dedicated fluorimeter (Ascent, Fluoroscan, Thermolab System, Helsinki, Finland). The procedure was carried out in the absence (A) or presence (B) of soluble rabbit thrombomodulin (TM) (2 nM), which acts as the main physiological protein C (PC) activator and is located on endothelial cells. TM was titrated to reduce the ETP of the normal plasma by 50%.13 The assessed parameters were the lag-time, defined as the time (min) needed to start TG, thrombin peak height (nM), the time to reach the peak (min) and the area under the curve, defined as endogenous thrombin potential (ETP) (nM/min). ETP represents the net amounts of thrombin that plasma can generate under the opposing driving forces of the pro- and anticoagulants operating in plasma. ETP results were also expressed as ETP-TM ratio by dividing the ETP in the absence to the ETP in the presence of TM. ETP-TM ratio represents the resistance to the anticoagulant activity of TM and is sensitive to the procoagulant imbalance between factor (F)VIII and PC (the higher the ETP-TM ratio, the greater the FVIII/PC ratio and hence hypercoagulability).13 Solid and broken lines represent patient and control plasma.

haematologica | 2021; 106(11)

3023

Case Reports

thrombosis at unusual sites, even when apparent prothrombotic risk factors are identified (oral contraceptives in our case) and irrespective of the baseline platelet count. Indeed, this patient had mild thrombocytopenia on admission, but historical testing carried out before VITT recorded a platelet count of 275x109/L. Thus, the platelet count had decreased by approximately 50%, in agreement with HIT and VITT diagnostic criteria. Aware of the possible diagnosis of VITT, we initially avoided a potentially detrimental heparin treatment, and this decision is likely to have played a major role in determining the positive outcome.2,3 Another important decision was to promptly start immune modulating therapy which caused the reduction of aPF4 titer and D-dimer. The causative prothrombotic mechanism in the reported patient is likely to be due to the antibodies to PF4 that induced a strong platelet activation even in absence of heparin exposure. The fact that the patient started to improve soon after the antibody titer decreased strongly supports this mechanism. One of the potential mechanisms that can explain the loss of serum activity in the functional assays (PAT) could be the IVIG blockade of FCƴ platelet receptors and/or the antibody suppression. Interestingly, platelets’ ability to promote thrombus formation in vitro was greatly reduced at admission, probably as a consequence of in vivo platelet activation and the formation of exhausted platelets as observed in other pathological conditions such as disseminated intravascular coagulation or sepsis. Laboratory tests correlated well with the clinical and radiological course. All in all, our experience supports the application of an early and multidisciplinary therapeutic approach in cases of VITT, with the possibility to avoid fatalities and obtain a resolution of the syndrome as in this case. Sara Bonato,1 Andrea Artoni,2 Anna Lecchi,2 Ghil Schwarz,1 Silvia La Marca,2 Lidia Padovan,2 Marigrazia Clerici,2 Chiara Gaudino,3 Giacomo Pietro Comi,4,5 Armando Tripodi2 and Flora Peyvandi2,6 1 Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Stroke Unit; 2Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, and Fondazione Luigi Villa; 3Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neuroradiology Unit; 4Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neurology Unit; 5 University of Milan, Dino Ferrari Center, Department of Pathophysiology and Transplantation and 6University of Milan, Department of Pathophysiology and Transplantation, Milan, Italy Correspondence: FLORA PEYVANDI - flora.peyvandi@unimi.it doi:10.3324/haematol.2021.279246 Received: May 26, 2021. Accepted: July 2, 2021. Pre-published: July 15, 2021. Disclosures: no conflicts of interest to disclose. Contributions: SB managed the clinical case, collected clinical data and wrote the clinical part of the manuscript; AA, GS, CG and GPC managed the clinical case, collected clinical data and critically reviewed

3024

the manuscript; AL performed the laboratory assays, interpreted the results and contributed to writing the manuscript; SLM, LP and MC performed the laboratory assays; AT organized the hemostatic assays, interpreted the results and contributed to writing the manuscript; FP designed the study and wrote the manuscript. All authors read and approved the final manuscript. Acknowledgments: the authors thank Luigi Flaminio Ghilardini (Università degli Studi di Milano, Milan) for his help with figures. Funding: this work was partially supported by the Italian Ministry of Health-Bando Ricerca Corrente 2020.

References 1. Wiersinga WJ, Rhodes A, Cheng AC, Peacock SJ, Prescott HC. Pathophysiology, transmission, diagnosis, and treatment of coronavirus disease 2019 (COVID-19): a review. JAMA. 2020;324(08): 782-793. 2. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle PA, Eichinger S. Thrombotic thrombocytopenia after ChAdOx1 nCov19 vaccination. N Engl J Med. 2021;384(22):2092-2101. 3. Schultz NH, Sørvoll IH, Michelsen AE, et al. Thrombosis and thrombocytopenia after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;84(22):2124-2130. 4. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;384(23):2202-2211. 5. Muir KL, Kallam A, Koepsell SA, Gundabolu K. Thrombotic thrombocytopenia after Ad26.COV2.S vaccination. N Engl J Med. 2021; 384(20):1964-1965. 6. Oldenburg J, Klamroth R, Langer F, et al. Diagnosis and management of vaccine-related thrombosis following AstraZeneca COVID19 vaccination: guidance statement from the GTH. Hamostaseologie. 2021;41(3):184-189. 7. Nazy I, Sachs UJ, Arnold DM, et al. Recommendations for the clinical and laboratory diagnosis of vaccine-induced immune thrombotic thrombocytopenia (VITT) for SARS-CoV-2 infections: communication from the ISTH SSC Subcommittee on Platelet Immunology. J Thromb Haemost. 2021;19(6):1585-1588. 8. Gresele P, Marietta M, Ageno W, et al. Management of cerebral and splanchnic vein thrombosis associated with thrombocytopenia in subjects previously vaccinated with Vaxzevria (AstraZeneca): a position statement from the Italian Society for the Study of Haemostasis and Thrombosis (SISET). Blood Transfus. 2021;19 (4):281-3. 9. Riley DS, Barber MS, Kienle GS, et al. CARE guidelines for case reports: explanation and elaboration document. J Clin Epidemiol. 2017;89:218-235. 10.Brodard J, Alberio L, Angelillo-Scherrer A, Nagler M. Accuracy of heparin-induced platelet aggregation test for the diagnosis of heparin-induced thrombocytopenia. Thromb Res. 2020;185:27-30. 11.Hosokawa K, Ohnishi T, Kondo T, et al. A novel automated microchip flow-chamber system to quantitatively evaluate thrombus formation and antithrombotic agents under blood flow conditions. J Thromb Haemost. 2011;9(10):2029-2037. 12.Ghirardello S, Lecchi A, Artoni A, et al. Assessment of platelet thrombus formation under flow conditions in adult patients with COVID-19: an observational study. Thromb Haemost. 2021;121 (8):1087-1096. 13.Tripodi A. Detection of procoagulant imbalance. Modified endogenous thrombin potential with results expressed as ratio of values with-to-without thrombomodulin. Thromb Haemost. 2017; (5);117:830-36. 14.Cattaneo M, Cerletti C, Harrison P, et al. Recommendations for the standardization of light transmission aggregometry: A consensus of the working party from the platelet physiology subcommittee of ssc/isth. J Thromb Haemost. 2013;11(6):1183-1189.

haematologica | 2021; 106(11)

Case Reports

Intracerebral hemorrhage associated with vaccine-induced thrombotic thrombocytopenia following ChAdOx1 nCOVID-19 vaccine in a pregnant woman Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spread globally, causing significant morbidity and mortality. Therefore, highly effective vaccines play an important role in increasing population immunity and preventing severe disease. The ChAdOx1 nCoV-19 (Oxford-AstraZeneca) vaccine consists of a replicationdeficient chimpanzee adenoviral vector containing the SARS-CoV-2 structural surface glycoprotein spike antigen gene.1 No signal for increased thrombotic events was detected in clinical trials of this vaccine involving 23,848 participants,1 which has now been administered to more than 34 million people worldwide.2 As reported to date, 20.8 million doses of the ChAdOx1 nCoV-19 vaccine have been applied in Brazil.3 ChAdOx nCoV-19 has been associated with vaccine-induced thrombotic thrombocytopenia (VITT),4 a rare but serious adverse event. The potential underlying associated mechanism is the formation of anti-heparin/platelet factor 4 (PF4) antibodies, but risk factors other than young age remain unclear.2,5 The immune-driven disease resembles heparin-induced thrombocytopenia (HIT) because platelet-activating antibodies recognize multimolecular complexes between cationic PF4 and anionic heparin, but classically there is no heparin exposure in VITT.5 So far, the incidence of VITT is around one case per 100,000 exposures to the ChAdOx1 nCoV-19 vaccine.2 Clinically, it is characterized by thrombocytopenia and thrombosis in unusual sites, particularly cerebral venous or splanchnic-vein thrombosis. Cerebral venous sinus thrombosis (CVST), a life-threatening event, was described in 72% of the VITT reports.6 Clinical trials did not include pregnant women, and the available data came from accidental pregnancies.7 Pregnant patients with COVID-19 are more likely to die or need intensive care compared with non-pregnant persons of reproductive age.1 Herein we present a fatal case of a 35-year-old pregnant woman who developed intracerebral hemorrhage in the left temporal lobe associated

with VITT 12 days after the off-label ChAdOx1 nCOVID-19 vaccination and we show the characterization of the hemostatic profile. A 35-years-old white pregnant woman at 23 weeks gestation was in prenatal follow-up with normal platelet counts and controlled hypothyroidism due to Hashimoto's disease. She received the first dose of the ChAdOx1 nCov-19 vaccine in late April 2021 (day 0). The next day, she reported having minor symptoms (malaise, chills, tremors, and "cold feeling"). On day 2, she presented a generalized skin rash on her legs, abdomen, and back with spontaneous resolution (see Figure 1A). Prenatal exams on day 3 identified urinary tract infection caused by Morganella morganii treated with trimethoprim-sulfamethoxazole and 148,000/mm3 platelets at the routine complete blood count. On day 7, she presented headache, nausea, nonspecific malaise, polaciuria, pain in the lower limbs and hips. Despite having 121,000/mm3 platelets, she received analgesics and was discharged from the obstetric emergency room. On the following days, she maintained headache and bilateral leg pain, partially responsive to dipyrone. On day 11, she reported a severe headache located in the left maxillary region, and on day 12, she was admitted with 33,000/mm3 platelets and an excruciating headache. During the computed tomography (CT) exam, she became comatose and underwent endotracheal intubation, which precluded the realization of CT cerebral venography, confirming the suspicion of central venous sinus thrombosis. Brain CT showed a large acute intraparenchymal hematoma in the temporal lobe, insula, and temporoparietal transition of the left cerebral hemisphere measuring approximately 10.1x5.4x5.5 cm. The hematoma was surrounded by vasogenic edema with midline shift to the right, herniation of the uncus, marked compressive effect on the midbrain, and subtotal collapse of the supratentorial ventricular system (Figure 1B). The patient received platelet concentrate transfusion and underwent urgent neurosurgery for hematoma drainage and decompressive craniectomy. After receiving critical care procedures, obstetric ultrasound detected fetal death. There were no platelet clumps, signs of erythrocyte fragmentation, or blast cells on blood film. She

Table 1. Laboratory characteristics of the patient.

Laboratory analysis

Reference value

D12

D15*

Hemoglobin (g/dL) Platelet count (per mm3) Leucocytes (per mm3) D-dimer (ng/mL) Fibrinogen (mg/dL) Activated partial thromboplastin time (rel) International normalized ratio Lactate dehydrogenase (UI/L) C-reactive protein (mg/dl) Aspartate aminotransferase (U/L) Alanine aminotransferase (U/L) Thyroid stimulating hormone (mUI/mL) Free thyroxine (ng/dL) Anti-heparin/ PF4 ELISA (OD) IL18 (pg/mL) IL1b (pg/mL)

10.5- 14.8 155,000 -409,000 4,500 -11,000 298-1,653 340-853 <1.25 0.8 – 1.2 81 – 234 <0.3 15 – 37 6 - 45 0.4 – 4.3 0.93 – 1.7 ≤0.4 315±36 4±8

12.7 33,000 12,199 >20,000 221 1.04 1.17 164 6.00 57 37 >10 -

6.2 32,000 10,030 16.54 10.03 0.99 1,075 92

*Functional assay. OD: optical density; ELISA: enzyme-linked immunosorbant assay. IL: interleukin: D: day.

haematologica | 2021; 106(11)

3025

Case Reports

Figure 1. Pictures of skin rash, cerebral hemorrhage, and placenta. (A) Generalized skin rash on legs with spontaneous resolution; (B) brain computed tomography with large acute intraparenchymal hematoma in the temporal lobe, insula, and temporoparietal transition of the left cerebral hemisphere measuring approximately 10.1 x 5.4 x 5.5 cm surrounded by vasogenic edema with midline shift to the right, herniation of the uncus, marked compressive effect on the midbrain and subtotal collapse of the supratentorial ventricular system; (C) cut surface of the placenta shows pale parenchyma with irregular and firm red-brown areas; (D) occlusive thrombus in a decidual vessel surrounded by a recent infarct; (E) chorionic villi encased by perivillous fibrin and red blood cells with some preservation of the trophoblast layer and villous stroma; villous capillaries are sparse; (F) area with massive intervillous fibrin deposition, some leukocytes and red blood cells with degenerative villous configuring an intervillous thrombosis.

received 80 g (1 g/Kg) immunoglobulin (Ig) and 2 g fibrinogen without success. The patient died on day 17 (after vaccination) with refractory intracranial hypertension despite all pressure control measures. SARS-CoV-2 reverse transcription polymerase chain reaction (RT-PCR) of nasopharyngeal swab and placenta was negative. Serology to dengue, Chikungunya, Zika, HIV, hepatitis B and C, cytomeglaovirus, Eppstein-Barr virus, toxoplasmosis, and rubella were negative. Relatives denied past COVID-19 infection, heparin exposure, or hormone therapy. There was no personal or familial history of thrombosis or miscarriage. Laboratory results are depicted in Table 1. The d-dimer level was over 20.000 ng/dL and fibrinogen 221 mg/dL (reference in the second trimester of pregnancy 298-1,653 ng/mL and 340-853 mg/dL, respectively). The results of other blood tests were normal except for alanine aminotransferase and C-reactive protein levels, which were increased. Levels of the proinflammatory cytokines IL-1b 92 pg/mL and IL-18 1,075 pg/ml were increased (compared to healthy controls at 4±b8 pg/mL and 315±36 pg/mL, respectively). Screening for hereditary thrombophilia with factor V Leiden and prothrombin 20210 mutation test was negative. Antinuclear antibodies, anticardiolipin IgG and IgM, lupus anticoagulant, and b2glycoprotein1 IgG antibodies were not detected. There were no signs of hemolysis. Anti-heparin/PF4 enzyme-linked immunosorant assay (ELISA) IgG antibodies were detected with an optical density value above 10 (reference ≤0.4). The pathological study of the placenta revealed intense inflammatory activity with increased intervillous fibrin-thrombotic areas, villous collapse configuring preinfarcted areas, and recent infarctions and focuses of decidual vessel throm3026

bosis (Figure 1C to F). Platelet activation and functional assay were performed as described.9,10 As shown in Figure 2, platelets from the patient exhibited signs of activation. Increased CD63 surface expression, a marker of dense granules secretion, was observed when compared to control (Figure 2A and B). Moreover, although platelet surface expression of CD62p was not different between the patient and controls, elevated levels of circulating sCD62p and platelet-derived microparticles expressing CD62p were observed in the patient compared to controls (Figure 2C to E). Functional activity of patient plasma to activate platelets from healthy volunteers was demonstrated in Figure 2F to H. Patient plasma trigger increased surface expression of CD62p, CD63, and the release of platelet-derived microparticles expressing CD62p. Conversely, plasma-induced platelet activation was inhibited by a high concentration of heparin (100 IU/mL). We present the first report of VITT following ChAdOx1 nCOVID-19 vaccination in a pregnant woman. Despite the lack of image verification of cerebral thrombosis, clinical features were suggestive. Initial brain CT may be negative for CVST in about 30% of cases.11 Besides that, thrombosis was observed in placental vessels. The constellation of signs and symptoms suggests VITT diagnosis complicated by intracranial hemorrhage with a fatal outcome both to fetus and mother. Three independent descriptions of 39 persons presenting VITT after vaccination with ChAdOx1 nCoV-19 revealed a vast majority being women younger than 50 years of age and some of them receiving estrogen-replacement therapy or oral contraceptives. Nevertheless, most of the participants did not have preexisting risk factors for thrombosis.2 Patients presenting VITT have unusually severe haematologica | 2021; 106(11)

Case Reports

Figure 2. Increased platelet activation and plasma-induced platelet activation. The percentage of platelets expressing CD63 (A) and CD62p (C) are shown on platelets isolated from healthy donors (HD) or the patient. Representative contour plot of CD63 expression (mean fluorescence intensity, MFI) by side scatter of HD (grey) or patient (green) platelets (B). The concentration of soluble P-selectin (sCD62P) in the plasma from HD or the patient (D). The percentage of circulating platelet-derived microparticles (MP) expressing CD62p (E). The percentage of platelets expressing CD62p (F) and CD63 (G) after incubation with the patient’s plasma (10%) with (green) or without (grey) the presence of heparin (100 UI/mL). The MFI of CD62p-FITC in platelet-derived microparticles from the supernatant of HD platelets stimulated with the patient’s plasma with (green) or without (grey) the presence of heparin (100 UI/mL) (H). The dotted lines represent the values of the control groups (HD platelets incubated with heterologous HD plasma).

thrombocytopenia, increased frequency of disseminated intravascular coagulation, and atypical thrombotic events. Serum from these patients shows a strong reactivity on the PF4–heparin ELISA and activates platelets in the presence and absence of heparin, but a high concentration of heparin completely inhibits the effect,5 as occurred in our case. In addition, we observed platelet activation confirmed by increased CD63 surface expression, circulating sCD62p levels, and platelet-derived microparticles expressing CD62p. Interestingly, we identified an elevation in the patient´s plasma of IL-1b and IL-18 compared to controls. Of note, IL-1b and IL-18 are central mediators of inflammation and thrombosis, released upon inflammasome activation.12 Indeed, inflammasome activation and IL-1b release play multiple

haematologica | 2021; 106(11)

functions that may favor vein and arterial thrombosis including the induction of procoagulant activity, the promotion of leukocyte adhesion to vascular endothelial cells, induction of neutrophil extracellular traps, and pyroptosis.13 Future studies should address whether or not this is also observed in other VITT cases and participate in the disease pathophysiology. Stroke affects 30 per 100,000 pregnancies, with ischemia, CVST, and hemorrhage causing roughly equal numbers and the highest risk in peripartum and postpartum.14 However, it is unclear if the pregnancy itself could increase the risk for thrombotic events following ChAdOx1 nCov-19 vaccination with the formation of anti-PF4 antibodies. Another element that should be pointed out, in this case, is the history of thyroiditis. Hypothyroidism is an underestimated risk 3027

Case Reports

factor for CVST, although most of patients have additional risk factors, suggesting a multifactorial hypercoagulability.11 This case report draws attention to this possible severe vaccine side effect among pregnant women and the challenges related to an early diagnosis. When symptomatic, pregnant women experience difficulties diagnosing VITT due to avoidance of screening methods involving radiation and presenting features possibly related to pregnancy, such as thrombocytopenia. The presence of persistent headache and thrombocytopenia within 30 days of vaccination should raise the suspicion of VITT. D-dimer, fibrinogen, and imaging exams help to guide prompt treatment, preventing a rapidly progressive disease.5 At least five countries had instituted limitations primarily based on age on which patients should receive the ChAdOx1 nCoV-19 vaccine,2 and Brazil recently temporarily suspended the administration of this vaccine in pregnant women.15 Those either affected by VITT or under investigation for this complication should not receive a second ChAdOx1 nCoV-19 vaccine. Physicians should have a low threshold for recognizing VITT signs and symptoms and requesting ELISA testing for PF4–polyanion antibodies and confirmatory functional tests. Although rare, VITT is a new phenomenon with devastating effects for otherwise healthy young adults, and its association with COVID-19 vaccination requires a thorough risk-benefit analysis especially for pregnant women. Daniela P. Mendes-de-Almeida,1,2,3 Remy MartinsGonçalves,4 Renata Morato-Santos,5 Gustavo Adolpho C. de Carvalho,6,7 Silas A. Martins,6,7 Lohanna Palhinha,4 Vanessa Sandim,8 Elyzabeth Avvad-Portari,9,10 Fernando A. Bozza,11,12 Robson Q. Monteiro,8 Patrícia T. Bozza4 and Pedro Kurtz12,13,14 1 Department of Hematology, Instituto Nacional de Infectologia Evandro Chagas, Fundação Oswaldo Cruz (FIOCRUZ), Rio de Janeiro, Brazil; 2Division of Molecular Carcinogenesis, Research Center, Instituto Nacional de Câncer (INCA), Rio de Janeiro, Brazil; 3Division of Epidemiology, Department of Pediatrics, University of Minnesota, Minneapolis, MNS, USA; 4Laboratory of Immunopharmacology, Instituto Oswaldo Cruz, Fundação Oswaldo Cruz (FIOCRUZ), Rio de Janeiro, Brazil; 5Department of Obstetrics, Universidade Federal do Estado do Rio de Janeiro (UNIRIO), Rio de Janeiro, Brazil; 6 Neurosurgical Deptartment, Hospital Sao Lucas, Rio de Janeiro, Brazil; 7Neurosurgical Deptartment, Hospital Adventista Silvestre, Rio de Janeiro, Brazil; 8Institute of Medical Biochemistry Leopoldo de Meis, Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil; 9Department of Pathologic Anatomy, Instituto Nacional de Saúde da Mulher, da Criança e do Adolescente Fernandes Figueira, Fundação Oswaldo Cruz (FIOCRUZ), Rio de Janeiro, Brazil; 10 Department of Pathologic Anatomy, Histology, and Embryology, Medical Sciences Faculty, Universidade do Estado do Rio de Janeiro (UERJ), Rio de Janeiro, Brazil; 11Laboratory of Intensive Care, Instituto Nacional de Infectologia Evandro Chagas, Fundação Oswaldo Cruz (FIOCRUZ), Rio de Janeiro, Brazil; 12D'Or Institute for Research and Education, Rio de Janeiro, Brazil; 13Department of Neurointensive Care, Instituto Estadual do Cérebro Paulo Niemeyer (IECPN), Rio de Janeiro, Brazil and 14Department of Intensive Care Medicine, Hospital Copa Star, Rio de Janeiro, Brazil Correspondence: PEDRO KURTZ - kurtzpedro@mac.com doi:10.3324/haematol.2021.279407 Received: June 14, 2021.

3028

Accepted: July 8, 2021. Pre-published: July 15, 2021. Disclosures: no conflicts of interest to disclose. Contributions: DPMA, PTB, and PK designed the study; DPMA, RMS, GACC, SAM, and PK attended the patient and analyzed the clinical data; DPMA, RMG, and PTB wrote the manuscript. RMG, LP, and VS performed the functional assay; EAP performed the pathological examination; RQM, FAB, and PTB revised the manuscript and supervised the study. All authors saw and approved the manuscript and its submission. Acknowledgments: we thank the patient´s relatives for the informed consent, Dr. Beatriz Grinsztejn, Dr. Valdiléia Veloso, Dr. Maria de Lourdes Maia and Filipe Santos-Bueno for technical support, and Prof. Logan Spector for the language review. Data sharing statement: the documentation is available on a reasonable request to the corresponding author.

References 1. Voysey M, Clemens SAC, Madhi SA, et al. Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK. Lancet. 2021;397(10269):99-111. 2. Cines DB, Bussel JB. SARS-CoV-2 Vaccine–induced immune thrombotic thrombocytopenia. N Engl J Med. 2021;384(23):2254-2256. 3. Ministério da Saúde. Localiza SUS, Vacinômetro SUS. https://localizasus.saude.gov.br/. Accessed May 29 2021. 4. European Medicines Agency - Sciences Medicine Health. COVID19 Vaccine AstraZeneca: benefits still outweigh the risks despite possible link to rare blood clots with low blood platelets. https://www.ema.europa.eu/en/news/covid-19-vaccineastrazeneca-benefits-still-outweigh-risks-despite-possible-link-rareblood-clots. Accessed May 29 2021. 5. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med 2021; 384(23):2202-2211. 6. See I, Su JR, Lale A, et al. US case reports of cerebral venous sinus thrombosis with thrombocytopenia after Ad26.COV2.S vaccination, March 2 to April 21, 2021. JAMA. 2021;325(24):2448-2456. 7. Male V. Are COVID-19 vaccines safe in pregnancy? Nat Rev Immunol. 2021;21(4):200-201. 8. Zambrano LD, Ellington S, Strid P, et al. Update: characteristics of symptomatic women of reproductive age with laboratory-confirmed SARS-CoV-2 infection by pregnancy status - United States, January 22-October 3, 2020. MMWR Morb Mortal Wkly Rep. 2020; 69(44):1641-1647. 9. Michelson AD, Barnard MR, Krueger LA, Frelinger AL, Furman MI. Evaluation of platelet function by flow cytometry. Methods. 2000; 21(3):259-270. 10. Hottz ED, Azevedo-Quintanilha IG, Palhinha L, et al. Platelet activation and platelet-monocyte aggregate formation trigger tissue factor expression in patients with severe COVID-19. Blood. 2020; 136(11):1330-1341. 11. Idiculla PS, Gurala D, Palanisamy M, Vijayakumar R, Dhandapani S, Nagarajan E. Cerebral venous thrombosis: a comprehensive review. Eur Neurol. 2020;83(4):369-379. 12. Henao-Mejia J, Elinav E, Strowig T, Flavell RA. Inflammasomes: far beyond inflammation. Nat Immunol. 2012;13(4):321-324. 13. Takahashi M. NLRP3 inflammasome as a key driver of vascular disease. Cardiovasc Res. 2021 Jan 23;cvab010. [Epub ahead of print] 14. Swartz RH, Cayley ML, Foley N, et al. The incidence of pregnancyrelated stroke: A systematic review and meta-analysis. Int J Stroke. 2017;12(7):687-697. 15. Fantinato FFST, Cruz LM. Nota Técnica No 627/2021CGPNI/DEIDT/SVS/MS. Nota técnica no. 627/2021 https://saude.rs.gov.br/upload/arquivos/202104/27181903-nota-tecnica-467-2021-cgpni-deidt-svs-ms.pdf. Acessed May 24 2021.

haematologica | 2021; 106(11)