Haematologica, volume 102, issue 8 by Haematologica

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation

Editor-in-Chief Jan Cools (Leuven)

Deputy Editor Luca Malcovati (Pavia)

Managing Director Antonio Majocchi (Pavia)

Associate Editors Hélène Cavé (Paris), Ross Levine (New York), Claire Harrison (London), Pavan Reddy (Ann Arbor), Andreas Rosenwald (Wuerzburg), Juerg Schwaller (Basel), Monika Engelhardt (Freiburg), Wyndham Wilson (Bethesda), Paul Kyrle (Vienna), Paolo Ghia (Milan), Swee Lay Thein (Bethesda), Pieter Sonneveld (Rotterdam)

Assistant Editors Anne Freckleton (English Editor), Cristiana Pascutto (Statistical Consultant), Rachel Stenner (English Editor), Kate O’Donohoe (English Editor), Ziggy Kennell (English Editor)

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

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

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

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation

Information for readers, authors and subscribers Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org). Haematologica publishes editorials, research articles, review articles, guideline articles and letters. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org). Manuscripts should be submitted online at http://www.haematologica.org/. Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail online (www.haematologica.org/content/policies). Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties. Detailed information about subscriptions is available online at www.haematologica.org. Haematologica is an open access journal. Access to the online journal is free. Use of the Haematologica App (available on the App Store and on Google Play) is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org). Rates of the International edition for the year 2017 are as following: Print edition

Institutional Euro 500

Personal Euro 150

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. Edoardo Ascari; Autorizzazione del Tribunale di Pavia n. 63 del 5 marzo 1955. Printing: Tipografia PI-ME, via Vigentina 136, Pavia, Italy. Printed in Luglio 2017.

haematologica calendar of events

Journal of the European Hematology Association Published by the Ferrata Storti Foundation

2nd EBMT International Transplant Course The European Group for Blood and Marrow Transplantation (EBMT) Chairs: M Mohty, J Kuball, R Duarte September 8-10, 2017 Barcelona, Spain

The 4th International Congress on Controversies in Stem Cell Transplantation and Cellular Therapies COSTEM Chairs: N Krรถger, A Nagler October 27-29, 2017 Berlin, Germany

3rd ESH International Conference on New Concepts in Lymphoid Malignancies: Focus on CLL and Indolent Lymphoma European School of Haematology (ESH) Chairs: M Hallek, L Staudt, S Stilgenbauer, A ThomasTikhonenko September 15-17, 2017 Mandelieu, France

Turkish Society of Hematology - EHA Joint Symposium November 1 - 4, 2017 Antalya, Turkey

13th Educational Course of the Lymphoma Working Party on "Treatment of Malignant Lymphoma: State-of-the-Art and Role of Stem Cell Transplantation" The European Group for Blood and Marrow Transplantation (EBMT) Chairs: S Montoto, A Sureda, M Trneny September 21-22, 2017 Prague, Czech Republic

EHA Scientific Meeting on Challenges in the Diagnosis and Management of Myeloproliferative Neoplasms Chairs: J Kiladjian and C Harrison October 12-14, 2017 Budapest, Hungary

28th Congress of the Hellenic Society of Haematology Hellenic Society of Haematology Chairs: P Panayotidis, E Terpos November 2-4, 2017 Athina, Greece

Argentinian Society of Hematology - EHA Joint Education Day November 17 - 18, 2017 Mar del Plata, Argentina

EHA Scientific Meeting on Shaping the Future of Mesenchymal Stromal Cells Therapy Chair: W Fibbe November 23-25, 2017 Amsterdam, The Netherlands

EHA Tutorial on Biology and Management of Myeloid Malignancies October 20-22, 2017 Yerevan, Armenia

Russian Onco-Hematology Society's Conference on Malignant Lymphoma - Joint Symposium October 25-26, 2017 Moscow, Russian Federation

Calendar of Events updated on June 29, 2017

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation

Table of Contents Volume 102, Issue 8: August 2017 Cover Figure Image generated by www.somersault1824.com.

Editorials 1299

p27 in FLT3-driven acute myeloid leukemia: many roads lead to ruin Iris Z Uras, Florian Bellutti and Veronika Sexl

1301

Long non-coding RNAs: another brick in the wall of normal karyotype acute myeloid leukemia? Bruno C. Medeiros

Guideline Article 1304

Recommendations regarding splenectomy in hereditary hemolytic anemias Achille Iolascon et al.

Articles Red Cell Biology & its Disorders

1314

Extracellular glycine is necessary for optimal hemoglobinization of erythroid cells Daniel Garcia-Santos et al.

Hemostasis

1324

Patient-derived anti-β2GP1 antibodies recognize a peptide motif pattern and not a specific sequence of residues Philippe de Moerloose et al.

Platelet Biology & its Disorders

1333

Risk of cardiovascular events and pulmonary hypertension following splenectomy – a Danish population-based cohort study from 1996-2012 Marianne Rørholt et al.

1342

Safety and efficacy of romiplostim in splenectomized and nonsplenectomized patients with primary immune thrombocytopenia Douglas B. Cines et al.

Myeloproliferative Disorders

1352

Bone marrow morphology is a strong discriminator between chronic eosinophilic leukemia, not otherwise specified and reactive idiopathic hypereosinophilic syndrome Sa A. Wang et al.

Chronic Myeloid Leukemia

1361

Single cell immune profiling by mass cytometry of newly diagnosed chronic phase chronic myeloid leukemia treated with nilotinib Stein-Erik Gullaksen et al.

1368

Natural killer-cell counts are associated with molecular relapse-free survival after imatinib discontinuation in chronic myeloid leukemia: the IMMUNOSTIM study Delphine Rea et al.

Haematologica 2017; vol. 102 no. 8 - August 2017 http://www.haematologica.org/

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation Acute Myeloid Leukemia

1378

FLT3 and FLT3-ITD phosphorylate and inactivate the cyclin-dependent kinase inhibitor p27Kip1 in acute myeloid leukemia Ines Peschel et al.

1391

Prognostic and biologic significance of long non-coding RNA profiling in younger adults with cytogenetically normal acute myeloid leukemia Dimitrios Papaioannou et al.

Chronic Lymphocytic Leukemia

1401

Targeted activation of the SHP-1/PP2A signaling axis elicits apoptosis of chronic lymphocytic leukemia cells Elena Tibaldi et al.

Non-Hodgkin Lymphoma

1413

Prognostic relevance of CD163 and CD8 combined with EZH2 and gain of chromosome 18 in follicular lymphoma: a study by the Lunenburg Lymphoma Biomarker Consortium Wendy B.C. Stevens et al.

Plasma Cell Disorders

1424

Lenalidomide/melphalan/dexamethasone in newly diagnosed patients with immunoglobulin light chain amyloidosis: results of a prospective phase 2 study with long-term follow up Ute Hegenbart et al.

1432

Longitudinal fluorescence in situ hybridization reveals cytogenetic evolution in myeloma relapsing after autologous transplantation Maximilian Merz et al.

1439

Serial measurements of circulating plasma cells before and after induction therapy have an independent prognostic impact in patients with multiple myeloma undergoing upfront autologous transplantation Rajshekhar Chakraborty et al.

Cell Therapy & Immunotherapy

1446

Tbet is a critical modulator of FoxP3 expression in autoimmune graft-versus-host disease Shoba Amarnath et al.

1457

Shorter leukocyte telomere length is associated with higher risk of infections: a prospective study of 75,309 individuals from the general population Jens Helby et al.

Letters to the Editor Letters are available online only at www.haematologica.org/content/102/8.toc

e282

Increased vancomycin dosing requirements in sickle cell disease due to hyperfiltration-dependent and independent pathways Jin Han et al. http://www.haematologica.org/content/102/8/e282

e285

NRF2 mediates Îł-globin gene regulation and fetal hemoglobin induction in human erythroid progenitors Xingguo Zhu et al. http://www.haematologica.org/content/102/8/e285

e289

Absence of the spleen and the occurrence of primary red cell alloimmunization in humans Dorothea Evers et al. http://www.haematologica.org/content/102/8/e289

e293

Myelodysplasia and liver disease extend the spectrum of RTEL1 related telomeropathies Shirleny R. Cardoso et al. http://www.haematologica.org/content/102/8/e293

Haematologica 2017; vol. 102 no. 8 - August 2017 http://www.haematologica.org/

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation

e297

E14a2 BCR-ABL1 transcript is associated with a higher rate of treatment-free remission in individuals with chronic myeloid leukemia after stopping tyrosine kinase inhibitor therapy Simone Claudiani et al. http://www.haematologica.org/content/102/8/e297

e300

Amplification of mixed lineage leukemia gene perturbs hematopoiesis and cooperates with partial tandem duplication to induce acute myeloid leukemia Bon Ham Yip et al. http://www.haematologica.org/content/102/8/e300

e305

Mutations in the 3′ untranslated region of NOTCH1 are associated with low CD20 expression levels in chronic lymphocytic leukemia Tamara Bittolo et al. http://www.haematologica.org/content/102/8/e305

e310

Clinical and diagnostic relevance of NOTCH2 and KLF2 mutations in splenic marginal zone lymphoma Yolanda Campos-Martín et al. http://www.haematologica.org/content/102/8/e310

e313

The prognostic value of the depth of response in multiple myeloma depends on the time of assessment, risk status and molecular subtype Carolina Schinke et al. http://www.haematologica.org/content/102/8/e313

e317

Low frequency mutations in ribosomal proteins RPL10 and RPL5 in multiple myeloma Isabel J.F. Hofman et al. http://www.haematologica.org/content/102/8/e317

e321

Dynamics of epigenetic age following hematopoietic stem cell transplantation Friedrich Stölzel et al. http://www.haematologica.org/content/102/8/e321

Case Reports Case Reports are available online only at www.haematologica.org/content/102/8.toc

e324

JAK1 somatic mutation in a myeloproliferative neoplasm Suzanne O. Arulogun et al. http://www.haematologica.org/content/102/8/e324

e328

p.Y317H is a new JAK2 gain-of-function mutation affecting the FERM domain in a myelofibrosis patient with CALR mutation Laura Eder-Azanza et al. http://www.haematologica.org/content/102/8/e328

Haematologica 2017; vol. 102 no. 8 - August 2017 http://www.haematologica.org/

EDITORIALS p27 in FLT3-driven acute myeloid leukemia: many roads lead to ruin Iris Z Uras, Florian Bellutti and Veronika Sexl Institute of Pharmacology and Toxicology, University of Veterinary Medicine, Vienna, Austria E-mail: veronika.sexl@vetmeduni.ac.at

doi:10.3324/haematol.2017.171819

cute myeloid leukemia (AML) is an aggressive, genetically diverse hematopoietic stem cell (HSC) malignancy with a generally poor prognosis: most patients die from the disease or from successive but futile rounds of chemotherapy. The FMS-like tyrosine kinase 3 (FLT3) gene is frequently altered in AML; an internal tandem duplication (FLT3-ITD) is seen in approximately 35% of patients, with an amino acid substitution in the tyrosine kinase domain detected in approximately 10%.1 FLT3-ITD mutations are associated with an extremely poor prognosis in AML. The specific pathways elicited by constitutive FLT3 activation include RAS/ERK/MAPK and PI3K/AKT (Figure 1A). In contrast to the wild-type, FLT3-ITD is a potent activator of STAT5 (Figure 1A); the acquisition of FLT3-ITD ensures leukemic stem cell (LSC) survival by upregulation of MCL-1 via constitutive STAT5 activation.2 The ITD mutations not only constitutively trigger FLT3 kinase activity, but also promote aberrant receptor functions, thereby influencing myeloid differentiation.3 The induction of PIM1, c-MYC and cyclin D3 is likely to contribute to the altered genetic programs downstream of FLT3-ITD necessary for leukemic transformation.4 The clinical importance of FLT3 has stimulated the development of FLT3 tyrosine kinase inhibitors (FLT3-TKI). Unfortunately, the initial high hopes have not been fulfilled due to the rapid development of resistance.5,6 A detailed understanding of the molecular pathways involved in FLT3-ITD signaling may pave the way for improving targeted therapies. In this issue of Haematologica, Peschel and colleagues reveal p27 to be a direct substrate of FLT3. p27 is an inhibitor of cell cycle progression and is abundant in quiescent cells. p27 is rapidly degraded to enable cells to enter the S-phase. Wildtype and FLT3-ITD bind to p27 and phosphorylate it on tyrosine 88, a residue linked to oncogenic transformation of tumor cells. Y88 phosphorylation is a prerequisite for p27 phosphorylation on T187 by the CDK2-cyclin complex, which results in its ubiquitin-dependent degradation via the E3 ligase SCFSKP2. FLT3-TKI treatment significantly reduces pY88-p27 in FLT3-ITD cells, thereby increasing the level of p27 protein and causing cell cycle arrest. Peschel et al. detect p27-Y88 phosphorylation in primary AML blast cells at levels comparable to those observed in FLT3-ITD+ cell lines and hypothesize that reduced levels of p27 upon Y88 phosphorylation and/or an increased localization to the cytoplasm might be useful indicators of disease outcome. Although there is a considerable body of evidence to show that p27 can predict responsiveness in solid tumors,7 the findings in AML are conflicting: while all primary wild-type FLT3 samples have reduced pY88-p27 upon TKI exposure, the levels in ITD+ material may be either increased or decreased. The difference most likely arises from compensatory phosphorylation by Src family kinases. p27 is known to be targeted by oncogenic tyrosine kinases other than FLT3 (see Figure 1B). In chronic myeloid leukemia, BCR-ABL alters p27 functions by two means:8,9 i) a kinase-dependent pathway actihaematologica | 2017; 102(8)

vates SCFSKP2 and promotes degradation of nuclear p27, thereby undermining its CDK inhibitory activity; and ii) a kinaseindependent pathway increases the cytoplasmic level of p27, thereby preventing apoptosis by mechanisms that remain elusive. The prediction of therapeutic response thus represents a challenge. p27 has functions in addition to its role as cell cycle inhibitor and inducer of anti-apoptotic responses. There is recent evidence for an involvement in transcriptional regulation and cell motility.10 p27 has been implicated in differentiation:11 it provokes an erythroid differentiation response and its suppression decreases myeloid differentiation. AML is characterized by a perturbed differentiation and FLT3-ITD activation leads to inhibition of many myeloid transcription factors, such as myeloid Pu.1 and C/EBPα.3 It remains to be determined whether p27 contributes to the pro-survival signals and the maturation arrest downstream of FLT3-ITD. p27 is a predominantly nuclear protein that inhibits certain CDKs, although it is able to shuttle to the cytoplasm. As a cell cycle inhibitor, nuclear p27 is a candidate tumor suppressor but the homozygous loss or silencing of the p27 locus is exceedingly rare.12 The complete deletion of p27 causes spontaneous tumorigenesis, predominantly in the pituitary. A decrease in p27 levels due to p27 degradation occurs in roughly half of carcinomas and correlates with aggressive, high-grade tumors and a poor prognosis. However, when mislocated in the cytoplasm, p27 has been reported to show oncogenic activity. Mice expressing a cytoplasmic p27 mutant lacking the nuclear CDK inhibitory function have a higher rate of spontaneous tumors in many organs, such as lung, retina, pituitary, ovary, adrenals and spleen. A low nuclear:cytoplasmic p27 ratio in solid tumors is an adverse prognostic marker. Although the mechanisms by which p27 exerts its oncogenic effects remain enigmatic, cytoplasmic p27 may represent a therapeutic target. In AML, the cytoplasmic abundance of p27 is regulated by the PIM family members, which are serine/threonine kinases. PIM kinases phosphorylate p27 at T157 and T198 to induce nuclear export and proteasome-dependent degradation. They also repress p27 transcription by phosphorylation and inactivation of forkhead transcription factors FoxO1a and FoxO3a.13 A similar mechanism is employed by mutant FLT3, which induces FoxO3a inactivation and thereby suppresses p27 expression.14 In addition, FLT3 induces PIM1 expression through STAT5 activation and PIM2 through an unknown STAT5-independent mechanism.15 FLT3-ITD thus uses two routes to ‘silence’ p27: direct phosphorylation on Y88 and activation of the STAT5-PIM-FoxO3A pathway. Encouraged by clinical trials of small-molecule CDK inhibitors in AML therapy, the authors propose that preventing p27-Y88 phosphorylation by FLT3-TKI might represent an alternative strategy to inactivate CDKs, thereby inducing G1 arrest. This strategy comes with a caveat: when the LSCs are in a non-cycling stage (dormant), they are more resistant to therapy and may develop TKI-induced resistance. In prac1299

Editorials

Figure 1. Vicious feed-forward loop in FLT3-driven acute myeloid leukemia (AML). (A) Schematic presentation of signaling pathways initiated by FLT3-ITD mutations. (B) p27, an inhibitor of cyclin-dependent kinases (CDK), is a key regulator of cell cycle progression. In a pathological condition, many oncogenic kinases phosphorylate p27 on tyrosine 88. The enhanced pY88-p27 leads to a conformational change that allows for further phosphorylation, thereby leading to its proteasomal degradation. This mechanism liberates CDKs from p27-mediated inhibition and stimulates cellcycle progression. (C) In the proposed model, FLT3-ITD receptors constitutively activate pro-survival pathways through STAT5-dependent and -independent mechanisms. FLT3 and PIM1 inhibit p27 function by direct phosphorylation and/or by transcriptional repression, leading to enhanced CDK6 kinase activity, which in turn promotes the transcription and activity of FLT3 and PIM1. (D) p27 may represent a therapeutic target. The big question in the â&#x20AC;&#x153;cureâ&#x20AC;? of AML remains to be adressed: to degrade or not to degrade p27.

tice, therapeutic failure in AML is due more to the emergence of treatment-resistant clones than to treatmentrelated mortality. One may speculate that FLT3 inhibitors stabilize p27, which in turn restores cell cycle arrest and dormancy, and enhances the risk of developing resistance. A potential strategy to circumventing resistance might be the simultaneous application of a TKI with a specific p27 degrader using novel technologies based on proteolysistargeting chimeras.16 Support for this idea comes from a report showing that deficiency of both p27 and p57 in HSCs induced cycling of dormant cells by activating CDK4/CDK6.17 A recent study found that FLT3-TKI and the CDK4/6 inhibitor palbociclib act synergistically in FLT3-ITD mutant cells.18 The rationale for this combination is that CDK6 inhibition has effects that go beyond cell cycle control, which would be induced by the TKI-mediated stabilization of p27 without palbociclib. The synergistic effects 1300

are mediated by an inhibition of cell-cycle progression in combination with the loss of CDK6-mediated transcription of FLT3, PIM1 and other potential targets.18 As PIM kinases phosphorylate and stabilize FLT3 in vitro,19 the combined treatment disrupts a vicious cycle and feed-forward loop. The same feed-forward loop may explain why leukemic cells become more dependent on FLT3-ITD signaling in the majority of patients, while the ratio of FLT3-ITD mutant alleles is higher after relapse. The finding is consistent with a model in which FLT3-ITD triggers its own expression via PIM1 and CDK6, thereby promoting hyperproliferation of leukemic cells. When p27 function is impaired by FLT3 and PIM1, CDK6 kinase activity is enhanced, which stimulates the transcription and activity of FLT3 and PIM1. Leukemic cells with mutated FLT3-ITD alleles presumably have a selective advantage, which results in the expansion of mutant FLT3 clones. In parallel, FLT3 prevents apoptosis haematologica | 2017; 102(8)

Editorials

by activating STAT5, which stimulates production of cMYC, cyclin D and PIM. It also blocks differentiation, at least partially, via p27 downregulation, an aspect that requires further investigation (Figure 1C). Peschel et al. now add a further layer of complexity to our understanding. In AML cells, p27 partially co-localizes with FLT3 in extended perinuclear structures. The colocalization is accompanied by enhanced p27 Y88-phosphorylation. Cytoplasmic p27 is susceptible to SCFSKP2-triggered proteolysis but is potentially able to exert protooncogenic functions. Although the stabilization of p27 is expected to have a net tumor suppressive effect, the simultaneous increase in the cytoplasmic level of p27 might have unintended consequences that offset the benefits of restoring nuclear p27. Whether AML therapy should aim to stabilize p27 or to degrade it is still unclear (Figure 1D). The importance of learning whether to degrade or not to degrade cannot be overstated. We hope that drug synergy screens will soon provide an answer. Acknowledgments We are grateful to Philipp Jodl and Peter Alexander Martinek for their excellent technical help in generating the figures. We are deeply indebted to Graham Tebb for editing the manuscript. This work was supported by the Austrian Science Fund (FWF) Grant SFB F47 (to VS) and by an ERC-advanced grant (to VS).

References 1. Hawley TS, Fong AZ, Griesser H, Lyman SD, Hawley RG. Leukemic predisposition of mice transplanted with gene-modified hematopoietic precursors expressing flt3 ligand. Blood. 1998;92(6):2003-2011. 2. Yoshimoto G, Miyamoto T, Jabbarzadeh-Tabrizi S, et al. FLT3-ITD upregulates MCL-1 to promote survival of stem cells in acute myeloid leukemia via FLT3-ITD-specific STAT5 activation. Blood. 2009;114(24):5034-5043. 3. Mizuki M, Schwable J, Steur C, et al. Suppression of myeloid transcription factors and induction of STAT response genes by AML-specific Flt3 mutations. Blood. 2003;101(8):3164-3173.

4. Li L, Piloto O, Kim K-T, et al. FLT3/ITD expression increases expansion, survival and entry into cell cycle of human haematopoietic stem/progenitor cells. Br J Haematol. 2007;137(1):64-75. 5. Smith CC, Wang Q, Chin C-S, et al. Validation of ITD mutations in FLT3 as a therapeutic target in human acute myeloid leukaemia. Nature. 2012;485(7397):260-263. 6. Wander SA, Levis MJ, Fathi AT. The evolving role of FLT3 inhibitors in acute myeloid leukemia: quizartinib and beyond. Ther Adv Hematol. 2014;5(3):65-77. 7. Chu IM, Hengst L, Slingerland JM. The Cdk inhibitor p27 in human cancer: prognostic potential and relevance to anticancer therapy. Nat Rev Cancer. 2008;8(4):253-267. 8. Chu S, McDonald T, Bhatia R. Role of BCR-ABL-Y177-mediated p27kip1 phosphorylation and cytoplasmic localization in enhanced proliferation of chronic myeloid leukemia progenitors. Leukemia. 2010;24(4):779-787. 9. Agarwal A, Mackenzie RJ, Besson A, et al. BCR-ABL1 promotes leukemia by converting p27 into a cytoplasmic oncoprotein. Blood. 2014;124(22):3260-3273. 10. Coqueret O. New roles for p21 and p27 cell-cycle inhibitors: a function for each cell compartment? Trends Cell Biol. 2003;13(2):65-70. 11. Munoz-Alonso MJ, Acosta JC, Richard C, et al. p21Cip1 and p27Kip1 Induce Distinct Cell Cycle Effects and Differentiation Programs in Myeloid Leukemia Cells. J Biol Chem. 2005;280(18):18120-18129. 12. Fero ML, Randel E, Gurley KE, Roberts JM, Kemp CJ. The murine gene p27Kip1 is haplo-insufficient for tumour suppression. Nature. 1998;396(6707):177-180. 13. Morishita D, Katayama R, Sekimizu K, Tsuruo T, Fujita N. Pim kinases promote cell cycle progression by phosphorylating and down-regulating p27Kip1 at the transcriptional and posttranscriptional levels. Cancer Res. 2008;68(13):5076-5085. 14. Scheijen B, Ngo HT, Kang H, Griffin JD. FLT3 receptors with internal tandem duplications promote cell viability and proliferation by signaling through Foxo proteins. Oncogene. 2004;23(19):3338-3349. 15. Green AS, Maciel TT, Hospital M-A, et al. Pim kinases modulate resistance to FLT3 tyrosine kinase inhibitors in FLT3-ITD acute myeloid leukemia. Sci Adv. 2015;1(8):e1500221-e1500221. 16. Lai AC, Crews CM. Induced protein degradation: an emerging drug discovery paradigm. Nat Rev Drug Discov. 2017;16(2):101-114. 17. Zou P, Yoshihara H, Hosokawa K, et al. p57(Kip2) and p27(Kip1) cooperate to maintain hematopoietic stem cell quiescence through interactions with Hsc70. Cell Stem Cell. 2011;9(3):247-261. 18. Uras IZ, Walter GJ, Scheicher R, et al. Palbociclib treatment of FLT3ITD+ AML cells uncovers a kinase-dependent transcriptional regulation of FLT3 and PIM1 by CDK6. Blood. 2016;127(23):2890-2902. 19. Natarajan K, Xie Y, Burcu M, et al. Pim-1 kinase phosphorylates and stabilizes 130 kDa FLT3 and promotes aberrant STAT5 signaling in acute myeloid leukemia with FLT3 internal tandem duplication. PLoS One. 2013;8(9):e74653.

Long non-coding RNAs: another brick in the wall of normal karyotype acute myeloid leukemia? Bruno C. Medeiros Stanford University School of Medicine, CA, USA E-mail: brunom@stanford.edu

doi:10.3324/haematol.2017.171744

cute myeloid leukemia (AML) is a complex malignant neoplasm of the hematopoietic system, characterized by multiple somatically (germline mutations occur in ~5% of AML cases) acquired pathologic (driver) mutations and the presence of coexisting competing malignant clones that frequently evolve under â&#x20AC;&#x153;selective pressureâ&#x20AC;? exerted by antileukemic treatment strategies. Genomic events in AML follow specific patterns of mutual exclusivity and co-occurrence and can be grouped into distinct functional categories of mutational and chromosomal abnormalities, including rearrangements involvhaematologica | 2017; 102(8)

ing transcription factors and mutations in tumor suppressor genes, genes encoding myeloid transcription factor, members of the cohesin complex, genes involved in DNA methylation, genes responsible for activated signaling, chromatin-modifying genes, the nucleophosmin 1 (NPM1) gene and members of the spliceosome complex.1 These patterns of genomic associations can be used to segregate AML cases into several non-overlapping cohorts, each with a distinct clinical outcome.2 In addition, these genomic events result in the clustering of AML patients into distinct messenger ribonucleic acid (RNA) expression 1301

Editorials

signatures and microRNA sequencing profiles.1 These expression signatures correlate with unique morphological features, baseline clinical characteristics and underlying genomic abnormalities and have a strong association with prognosis, especially in AML patients with normal karyotype.3 In this issue of Haematologica, Papaioannou D et al. expand on our current understanding of the contribution of a distinct class of RNA molecules to the pathogenesis and prognosis in AML.4 Along with other regulatory RNA molecules, such as microRNAs, short-interfering RNAs, and others, long non-coding RNAs (lncRNAs) are integral for intracellular homeostasis by exerting specific cellular functions, including regulation of gene transcription, progression through the cell cycle, regulation of post-transcriptional mRNA processing, and others.5 Using a wellvalidated cohort of younger de novo AML patients with normal karyotype (NK-AML) enrolled into consecutive studies of cytarabine/anthracycline-based first-line therapy on The Alliance for Clinical Trials in Oncology, the authors defined the global expression of lncRNAs in this cohort of patients. They also defined a lncRNA signature associated with response to therapy and risk of relapse and reported on the baseline and genomic characteristics of NK-AML patients with distinct lncRNA expression signatures. Additionally, the investigators reported several interesting and novel observations. First, it was noted that twothirds of the lncRNAs belong to 1 of 3 categories of these regulatory RNA molecules (processed pseudogenes, intergenic/intervening lncRNAs or antisense lncRNAs). The investigators also identified a 24 lncRNAs signature that was highly correlated with outcomes, whereas low prognostic lncRNA scores (favorable lncRNA scores) were associated with improvements in different survival endpoints (favorable lncRNA score status also associated with longer overall survival (OS; P=0.002, 5-year rates, 52% versus 26%) and longer event-free survival (EFS; P<0.001, 5year rates, 46% versus 16%). Importantly, these differences remained significant in multivariable analyses even after adjustment for other prognostic covariates. In addition, the presence of a low lncRNA prognostic score was associated with other known prognostic variables, such as low white blood cell count at diagnosis, lower frequency of FLT3-internal tandem duplication (ITD) and more frequent classification into the favorable risk category according to the European LeukemiaNet (ELN) classification. Next, several lncRNA expression signatures identified had strong correlation with well-defined prognostic molecular mutations in NK-AML, such as biallelic CEBPA mutations, mutations in the NPM1 gene and presence of FLT3-ITD. Finally, a strong association between high lncRNA prognostic score expression and specific messenger RNA and microRNA expression profiles was observed. Despite these unique observations, several questions remain regarding the value of lncRNA profiling in patients with NK-AML. First, only a highly selected cohort of patients was included in these investigations. For example, adult patients younger than 60 years of age represent one-third of all cases of AML (although, more 1302

limited data have previously demonstrated the prognostic significance of lncRNA expression profiling in older patients with AML, despite a lack of overlap in these expression signatures).6 In addition, NK-AML is observed in approximately half of younger adults with AML. These factors limit their generalization of the findings to less than 20% of AML patients and excluded several high-risk cohorts, such as patients with secondary AML (both antecedent hematologic disorders as well as those with therapy-related AML) and those with adverse-risk karyotypes. Recommendations for post-remission treatment include the consideration of allogeneic hematopoietic cell transplant (alloHCT) in patients with high-risk genomic features. The exclusion of patients receiving alloHCT in first complete remission, including those with an increased risk of relapse (such as patients segregated into the intermediate and adverse ELN cohorts),7 hampers the identification of the optimal post-remission approach in patients with high prognostic lncRNA scores. Also, the study segregates patients with NK-AML according to the updated version of the ELN classification.8 However, the associations of recurrent gene mutations with lncRNA expression studies reported herein do not account for the updated ELN classification, whereby FLT3-ITD patients can be segregated into favorable (mutated NPM1 without FLT3-ITD or with FLT3ITDlow), intermediate (wild-type NPM1 without FLT3ITD or wild-type NPM1 and FLT3-ITDlow or mutated NPM1 and FLT3-ITDhigh) or adverse (wild-type NPM1 and FLT3-ITDhigh) categories. Finally, AML is an oligoclonal malignant disorder where genomic abnormities are acquired serially; the design of the study by Papaioannou D et al. limits the assessment of the combinatorial effect of multiple mutations on the lncRNA expression signature and consequently their prognostic significance. For example, the investigators report that mutation in the NPM1 gene was strongly associated with a lncRNA signature, however, patients with NPM1 often also have co-associated mutations in IDH1/2 (improved clinical outcomes),9 DNMT3A (worse clinical prognosis)10 and FLT3-ITD (variable clinical outcomes depending on the allelic ratio of FLT3-ITD).8 This variability in genomic co-associations may explain why the results failed to demonstrate an association between NPM1 mutational status, FLT3-ITD or biallelic mutations in the CEBPA genes and lncRNA expression prognostic score. In summary, the current report identifies a lncRNA expression signature that allows segregation of younger patients with de novo NK-AML into 2 separate prognostic cohorts and describes expression signatures associated with specific molecular abnormalities. Overall, these results expand on prior studies and highlight the importance of coding and non-coding RNAs, adding another brick in the wall of understanding of the processes involved in leukemic transformation.

References 1. Cancer Genome Atlas Research Network. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med.

haematologica | 2017; 102(8)

Editorials

2013;368(22):2059-2074. 2. Papaemmanuil E, Gerstung M, Bullinger L, et al. Genomic classification and prognosis in acute myeloid leukemia. N Engl J Med. 2016;374(23):2209-2221. 3. Marcucci G, Radmacher MD, Maharry K, et al. MicroRNA expression in cytogenetically normal acute myeloid leukemia. N Engl J Med. 2008;358(18):1919-1928. 4. Papaioannou D, Nicolet D, Volinia S, et al. Prognostic and biologic significance of long non-coding RNA profiling in younger adults with cytogenetically normal acute myeloid leukemia. Haematologica. 2017;102(8):1391-1400. 5. St Laurent G, Wahlestedt C, Kapranov P, et al. The Landscape of long noncoding RNA classification. Trends Genet. 2015;31(5):239-251. 6. Garzon R, Volinia S, Papaioannou D, et al. Expression and prognostic impact of lncRNAs in acute myeloid leukemia. Proc Natl Acad Sci U S

haematologica | 2017; 102(8)

A. 2014;111(52):18679-18684. 7. Röllig C, Bornhäuser M, Thiede C, et al. Long-term prognosis of acute myeloid leukemia according to the new genetic risk classification of the European LeukemiaNet recommendations: evaluation of the proposed reporting system. J Clin Oncol. 2011;29(20):27582765. 8. 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. 9. Patel JP, Gönen M, Figueroa ME, et al Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med. 2012;366(12):1079-1089. 10. Kihara R, Nagata Y, Kiyoi H, et al. Comprehensive analysis of genetic alterations and their prognostic impacts in adult acute myeloid leukemia patients. Leukemia. 2014;28(8):1586-1595.

1303

GUIDELINE ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1304-1313

Recommendations regarding splenectomy in hereditary hemolytic anemias

Achille Iolascon,1,2 Immacolata Andolfo,1,2 Wilma Barcellini,3 Francesco Corcione,4 Loïc Garçon,5 Lucia De Franceschi,6 Claudio Pignata,7 Giovanna Graziadei,8 Dagmar Pospisilova,9 David C. Rees,10 Mariane de Montalembert,11 Stefano Rivella,12 Antonella Gambale,1,2 Roberta Russo,1,2 Leticia Ribeiro,13 Jules-Vives-Corrons,14 Patricia Aguilar Martinez,15 Antonis Kattamis,16 Beatrice Gulbis,17 Maria Domenica Cappellini,8 Irene Roberts18 and Hannah Tamary19 on behalf of the Working Study Group on Red Cells and Iron of the EHA

Department of Molecular Medicine and Medical Biotechnology, University Federico II Naples, Italy; 2CEINGE Biotecnologie Avanzate, Naples, Italy; 3Oncohematology Unit, IRCCS Ca’ Granda Foundation, Ospedale Maggiore Policlinico, Milan, Italy; 4Department of General Surgery, Monaldi Hospital A.O.R.N. dei Colli, Naples, Italy; 5Service d’Hématologie Biologique, CHU Amiens Picardie, Amiens, France; 6Department of Medicine, University of Verona and AOUI-Verona, Italy; 7Department of Translational Medical Sciences, Federico II University of Naples, Italy; 8Department of Clinical Science and Community Health, Fondazione IRCCS Cà Granda, Ospedale Maggiore Policlinico, University of Milan, Italy; 9Department of Pediatrics, Faculty of Medicine and Dentistry, Palacky University Olomouc and University Hospital Olomouc, Czech Republic; 10 Department of Paediatric Haematology, King’s College Hospital, King’s College London School of Medicine, UK; 11Pediatrics Department, Necker Hospital, Paris, France; 12 Department of Pediatrics, Division of Hematology-Oncology, Children’s Blood and Cancer Foundation Laboratories, Weill Cornell Medical College, New York, NY, USA; Department of Pediatrics, Division of Hematology, Children's Hospital of Philadelphia, PA, USA; 13Hematology Service, Hospital and University Center of Coimbra (CHUC), Portugal; 14Red Cell Pathology Unit, Hospital Clínic de Barcelona, Spain; 15Laboratory of Hematology, CHRU de Montpellier, Hôpital Saint Eloi, France; 16First Department of Pediatrics, University of Athens, Greece; 17Department of Clinical Chemistry, Hôpital Erasme, U.L.B., Brussels, Belgium; 18Department of Paediatrics, Children's Hospital, University of Oxford, John Radcliffe Hospital, UK and 19Pediatric Hematology Unit, Schneider Children's Medical Center of Israel, Petah Tiqva, Sackler Faculty of Medicine, Tel Aviv University, Israel 1

Correspondence: achille.iolascon@unina.it ABSTRACT Received: November 28, 2016. Accepted: May 22, 2017. Pre-published: May 26, 2017. doi:10.3324/haematol.2016.161166 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1304 ©2017 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.

1304

ereditary hemolytic anemias are a group of disorders with a variety of causes, including red cell membrane defects, red blood cell enzyme disorders, congenital dyserythropoietic anemias, thalassemia syndromes and hemoglobinopathies. As damaged red blood cells passing through the red pulp of the spleen are removed by splenic macrophages, splenectomy is one possible therapeutic approach to the management of severely affected patients. However, except for hereditary spherocytosis for which the effectiveness of splenectomy has been well documented, the efficacy of splenectomy in other anemias within this group has yet to be determined and there are concerns regarding short- and long-term infectious and thrombotic complications. In light of the priorities identified by the European Hematology Association Roadmap we generated specific recommendations for each disorder, except thalassemia syndromes for which there are other, recent guidelines. Our recommendations are intended to enable clinicians to achieve better informed decisions on disease management by splenectomy, on the type of splenectomy and the possible consequences. As no randomized clinical trials, case control or cohort studies regarding splenectomy in these disorders were found in the literature, recommendations for each disease were based on expert opinion and were subsequently critically revised and modified by the Splenectomy in Rare Anemias Study Group, which includes hematologists caring for both adults and children.

haematologica | 2017; 102(8)

Splenectomy in hereditary hemolytic anemias

Introduction Hereditary hemolytic anemias are a group of disorders with a variety of causes, including red cell membrane defects, enzyme disorders, congenital dyserythropoietic anemias, and hemoglobinopathies. Given the rarity of these disorders, their optimal management has yet to be determined. Splenectomy has been suggested as a possible therapeutic approach to manage severely affected patients, based on the evidence that abnormal or damaged red blood cells passing through the spleen red pulp are removed by the splenic macrophage system. However, although splenectomy has been commonly used in recent decades in the clinical management of patients with severe hematologic phenotypes, its efficacy in many of these disorders has yet to be determined. Additionally, concerns remain regarding short- and long-term infectious complications, and increased risk of cardiovascular complications in later life, including thrombosis and pulmonary hypertension.1 We first reviewed the literature in PubMed in order to generate recommendations regarding splenectomy in hereditary hemolytic anemias. Since no randomized clinical trials, case-control or cohort studies were identified, the level of evidence considered was lowered to that from non-analytic studies and case series. Expert recommendations were developed, and subsequently critically revised and modified by the Splenectomy in Rare Anemias Study Group in order to achieve the greatest possible agreement, which was classified as "full consensus” (100% agreement) or “consensus” (>80% agreement). None of the core statements achieved a degree of consensus below 80%. The Grading of Recommendation Assessment, Developing and Evaluation (GRADE) system was used to rate the quality of evidence and strength of recommendations (see Online Supplementary Information).2 We first present general considerations on the possible complications of splenectomy, including post-splenectomy infections and acute and long-term thromboembolic complications, and types of splenectomy (laparatomic, laparoscopic and partial). We then discuss the advantages and complications of splenectomy for each of the specific hereditary hemolytic anemia disorders, concluding with specific recommendations when possible.

Splenectomy complications Post-splenectomy infections Given the role of the spleen in immune competence and blood filtration, there is a risk of overwhelming postsplenectomy infection (OPSI), which is highest with encapsulated micro-organisms such as Streptococcus pneumoniae, Neisseria meningitidis and Haemophilus influenza.3 Asplenia is also an important risk factor for serious infections with Plasmodium, Capnocytophaga canimorsus and C. cynodegmi (after an animal bite), Babesia spp. (after a tick bite), and Bordetella holmesii.4-6 The risk of post-splenectomy sepsis may vary according to the indication for splenectomy (intermediate risk in spherocytosis and higher in other inherited anemias),7 patient’s age at the time of surgery (highest before the age of 5 years),3 and time since the splenectomy was performed (risk highest during the first year after the intervention). However, the risk probahaematologica | 2017; 102(8)

bly remains elevated for life.8,9 Given the high risk of OPSI at a young age, splenectomy should not normally be performed before 5 years of age. It is difficult to estimate the current risk of OPSI in subjects aged >5 years as most studies are retrospective and include patients with heterogeneous diseases who were not fully immunized. It is possible that the widespread use of conjugated vaccines will significantly reduce the risk of OPSI. In fact, a recent retrospective study in which 141 consecutive children undergoing splenectomy during 1991-2010 were analyzed indicated that ten of the 11 patients who developed post-splenectomy sepsis had an additional underlying immune deficiency.10 However, in a recent prospective, multicenter cohort study of German patients with severe sepsis or septic shock, S. pneumoniae sepsis was more frequent among splenectomized patients than among those with a normal functional spleen (42% versus 12%, respectively; P<0.001). It is of note that less than half of the OPSI patients in this study had received pneumococcal vaccination before splenectomy, despite national and international guidelines.11 Strategies to reduce the development of OPSI include: (i) patient’s education, including advice to take urgent action in response to febrile episodes; (ii) vaccination and (iii) prophylactic anti-microbial therapy. Detailed guidelines regarding the prevention and treatment of infections in splenectomized or asplenic patients are available through the British Committee for Standards in Haematology12 and American Academy of Pediatrics (Red Book 30th edition, 2015), to which the reader is referred.

Post-splenectomy thromboembolic complications It has been reported that, following splenectomy, there is an increased risk of early and late venous and arterial thrombosis1,13 including acute splenic and portal vein thrombosis (SPVT)14 and delayed severe life-long complications.15

Acute splenic and portal vein thrombosis Acute SPVT after splenectomy is an early and lifethreatening complication, which can lead to bowel ischemia and/or portal hypertension. This complication has been related to stasis in the splenic vein remnant.14 The risk varies depending on the underlying disorder. In 2008, Krauth et al. reviewed prospective and retrospective studies and found that 11/89 patients with hemolytic disease developed SPVT (12.3%) while only 2/118 patients (1.7%) of those with immune thrombocytopenia had SPVT. None of 122 patients who underwent splenectomy because of trauma developed this complication.14,15 Large spleen size has also been identified as a risk factor for the development of SPVT. Although there are no welldesigned randomized trials comparing the risk of SPVT after open splenectomy versus laparoscopic splenectomy, the surgical approach does not seem to affect the incidence of SPVT.16 Screening for thrombophilia has not been shown to allow early identification of patients at risk of SPVT after splenectomy.17 A study in which contrast-enhanced computed tomography was used for the diagnosis of SPVT showed that the median time between splenectomy and the appearance of asymptomatic SPVT was 6 days (range, 3-11 days).18 A Canadian study of 40 patients suggested that an appropriate time for Doppler/ultrasound surveillance to diagnose SPVT is 1 week following splenectomy.19 1305

A. Iolascon et al.

Most patients with documented SPVT are treated with anticoagulant therapy (i.e. intravenous heparin followed by oral anticoagulants) for a variable period ranging between 3 to 6 months. In one study, the majority of treated patients had documented complete (57/90, 63.3%) or at least partial (13.3%) resolution of the thrombus. However, 7.7% of this latter population showed persistence of thrombus, while 15.5% developed a cavernoma or portal hypertension.14 The role of prophylactic antithrombotic therapy is unknown as strategies in the various cohorts or case reports were extremely heterogeneous in terms of duration of prophylaxis. A randomized study comparing different durations of postoperative antithrombotic prophylaxis after laparoscopic abdominal surgery was terminated early due to the low incidence of thrombosis.16

Late cardiovascular events Arterial and venous thromboembolism An increased risk of vascular complications after splenectomy was first described in a group of 740 World War II veterans whose spleen was removed because of trauma. It was seen that these subjects had a significantly increased risk of death from ischemic heart disease compared to controls (relative risk, 1.85).15 Data from a Danish Registry identifying all splenectomized patients from 1996-2005 showed that the long-term (>1 year) risk of venous thromboembolism remained approximately 3-fold higher in patients who had undergone splenectomy because of trauma than in the general population.20 The contribution of other trauma complications to thrombotic risk was not, however, evaluated. The long-term risk of venous thromboembolism was highest in patients splenectomized for malignant hematologic disorders and hemolytic anemias. Pulmonary arterial hypertension Splenectomy has been reported to be a risk factor for the development of pulmonary arterial hypertension,21 particularly in patients with hemolytic disorders.22-24 Loss of splenic function is associated with increased numbers of platelets and also enhances their activation, promoting pulmonary microthrombosis and adhesion of red cells to the endothelium.25 The spleen also plays a critical function in the removal of senescent and damaged erythrocytes.

Splenectomy recommendations -For prompt diagnosis of SPVT at least one Doppler ultrasound study should be carried out on day 7 after splenectomy (agreed by 86% of experts) (grade 2 recommendation, grade C evidence).

Surgical approach Laparotomic splenectomy The traditional approach to splenectomy has been by laparotomy. This approach allows for a careful search for an accessory spleen which, if left behind, may cause recurrence of anemia.10 The disadvantages of open splenectomy are mainly surgical morbidity and abdominal wall scarring.26 Laparoscopic splenectomy has become feasible with progress in minimally invasive techniques.27

Laparoscopic splenectomy Laparoscopic splenectomy is currently considered the gold standard technique for removal of a normal sized or slightly enlarged spleen and is preferred to open splenectomy. Compared to open splenectomy, laparoscopic splenectomy: (i) is less traumatic; (ii) is associated with fewer complications; (iii) requires shorter hospital stays; (iv) has a better cosmetic outcome; and (v) overall, has a lower cost. However, it should only be performed by experienced surgeons.28 Nowadays, laparoscopic splenectomy is possible and safe also for massively enlarged spleens, but in such cases is associated with longer operating times and longer stays in hospital.29 Perioperative splenic artery embolization has been found to be useful and reduces the complications of massive spleen laparoscopic splenectomy.30 Moreover, selective splenic artery embolization, carried out in steps, will reduce spleen size and alleviate cytopenias.31 A preoperative assessment of splenic size by ultrasound is recommended. Although three-dimensional computed tomography is considered to be more accurate, it does not provide significant advantage in estimating spleen size and its use should be limited to those cases in which additional information about the anatomy is required prior to surgery.32

Table 1. Summary of splenectomy recommendations for hemolytic disorders.

Disease

When splenectomy recommended? *

Hereditary spherocytosis

Patient is transfusion-dependent or suffers severe anemia. Patient has moderate disease: decision based on spleen size and quality of life parameters. No need to perform cholecystectomy. Consider if patient is transfusion-dependent or severely anemic. Cholecystectomy should be performed at time of splenectomy. Consider if patient is transfusion-dependent and/or has massive splenomegaly and/or has symptomatic splenomegaly. Contraindicated. Consider if patient is transfusion-dependent and/or has symptomatic splenomegaly. Patient has had two acute splenic sequestration crises and/or has massive splenomegaly and/or suffers symptomatic hypersplenism. Consider only if patient has transfusion-dependent anemia and/or symptomatic splenomegaly.

Pyruvate kinase deficiency Splenectomy in congenital non-spherocytic hemolytic anemia due to G6PD deficiency Hereditary stomatocytosis Congenital dyserythropoietic anemia type II Sickle cell disease Unstable hemoglobin

*For all indications splenectomy should be performed after 6 years of age. G6PD: glucose-6-phosphate dehydrogenase.

1306

haematologica | 2017; 102(8)

Splenectomy in hereditary hemolytic anemias

Partial splenectomy In an attempt to reduce the infectious risk of total splenectomy, especially for children less than 6 years of age who suffer severe anemia or are transfusion-dependent, partial splenectomy has been increasingly used in recent years. In partial splenectomy, usually 80-90% of the enlarged spleen is removed. Partial splenectomy was initially performed using an open splenectomy approach but laparoscopic and robotic approaches have recently been introduced.33 Nevertheless, partial splenectomy should always be performed by an experienced surgeon.

Disease-specific recommendations A detailed discussion of the clinical pictures of and diagnostic approaches to hereditary hemolytic disorders is beyond the scope of this manuscript. Data available on the advantages and complications of splenectomy in hereditary hemolytic anemias and expert recommendations are summarized in Table 1. It should be appreciated that data for some disorders are so sparse that no recommendations could be generated.

Hereditary spherocytosis Following thalassemia syndrome and sickle cell disease (SCD), hereditary spherocytosis (HS) is the most common form of congenital hemolytic anemia with an incidence of approximately 1:2000 and a dominant transmission in about 70-80% of cases. HS is caused by mutations in genes encoding α- and β-spectrin and other proteins involved in the attachment of the cytoskeleton to the overlying lipid bilayer (ankyrin, band 3 and protein 4.2). Defects in these structural proteins render the red blood cells spherical, rigid and susceptible to premature destruction in the spleen.34-37 Clinically, patients with HS are grouped into three categories according to disease severity: mild, moderate and severe (Table 2).36

Long-term thrombotic complications of splenectomy In 1997, Schilling found that the rate of arteriosclerotic events (stroke, myocardial infarction, coronary or carotid artery surgery) in patients older than 40 years of age with HS was 5.6-fold higher in asplenic patients than in HS patients with an intact spleen, with the first event occurring one or more decades following splenectomy.38 This was further confirmed in his follow-up study in which the hazard ratio for arterial events was 7.2 in HS patients who underwent splenectomy compared to affected patients who did not undergo splenectomy. In addition, affected patients who underwent splenectomy had a hazard ratio of 3 for developing venous events as compared to HS patients who did not undergo splenectomy.39

However, only a few patients with HS who developed stroke, pulmonary emboli or pulmonary arterial hypertension following splenectomy have been reported.24,40-42 Moreover, Buchanan et al. studied 39 adults with HS and found no evidence of thrombotic manifestations despite a long follow up (median 25 years).43

Splenectomy Splenectomy in HS usually results in disappearance of anemia and a clear decrease of hemolytic markers. In the large HS series reported by Mariani et al., the median hemoglobin increase after splenectomy was 3 g/dL (10.8 to 13.9 g/dL), associated with a decrease of reticulocyte count (from 337 to 51×109/L) and unconjugated bilirubin (from 32.5 to 12 mmol/L).44 Due to increasing awareness of post-splenectomy complications, the rate of splenectomy has declined in the last decade. During the period from 1980 to 2005, splenectomy was performed in only 20% of HS patients.44 In general, splenectomy is not indicated in patients with mild HS, whereas it is usually necessary in severe cases, albeit delayed if possible until the age of 6 years (Table 2). In the intermediate categories the indications for splenectomy are less clear. One indication is symptomatic/painful splenomegaly with associated thrombocytopenia or leukopenia that affects the patient’s quality of life. For young adult patients, unacceptable cutaneous jaundice (usually in patients with concomitant Gilbert genotype) may become a social problem that balances a decision towards splenectomy.

Available guidelines for splenectomy Guidelines for the diagnosis and management of HS, including splenectomy were published on behalf of the General Haematology Task Force of the British Committee for Standards in Haematology in 2004 and updated in 2011.45,46 In agreement with previous recommendations, the laparoscopic approach is preferred if trained surgeons are available; in children undergoing splenectomy, the gall bladder should be removed concomitantly if symptomatic gallstones are present (Table 3). We referred to previously published guidelines when considering two particular clinical situations: (i) whether splenectomy should accompany cholecystectomy when biliary stones are present; and (ii) the role of partial splenectomy, particularly in children younger than 6 years of age with severe HS.

Splenectomy and cholecystectomy It was previously suggested that there is an increased risk of intrahepatic choledocholithiasis following splenectomy44 and the 2004 guidelines suggested that for children with HS who require cholecystectomy the spleen should

Table 2. Indications for splenectomy in hereditary spherocytosis based on severity of disease*.

Disease severity

Hemoglobin (g/dL)

Reticulocyte count (%)

Bilirubin (mmol/L)

Severe Moderate Mild

<8 8 - 12 11 - 15

> 10 >6 3-6

> 51 > 34 17 - 34

Indication for splenectomy Indicated, delay until age >6 years Individually tailored based on spleen size and quality of life parameters Not indicated

*severity of disease based on reference 35.

haematologica | 2017; 102(8)

1307

A. Iolascon et al. Table 3. Splenectomy guidelines in hereditary spherocytosis –2011 update45 and the authors’ recommendations.

Guidelines 2011

The laparoscopic approach to splenectomy is associated with less pain, shorter hospital stay and better cosmetic appearance; but is dependent on the availability of appropriately trained surgeons, and suitable equipment (grade 1 recommendation, grade B evidence). In children undergoing splenectomy, the gall bladder should be removed concomitantly if there are symptomatic gallstones (grade C evidence, grade 2 recommendation). In children who require cholecystectomy for symptoms of gallstones the use of concurrent splenectomy is controversial. It may be associated with a decreased risk of common bile duct stones in the future, but is also associated with a risk of post-splenectomy sepsis (grade 2 recommendation, grade C evidence).

2 3

4 5

Authors' recommendations

When splenectomy is indicated, ideally it should be done after the age of 6 years (grade 2 recommendation, grade C evidence). Partial splenectomy is theoretically associated with a decreased risk of post-splenectomy sepsis, but it is possible that further surgery may be needed for either recurrence of hematologic problems or symptomatic cholelithiasis (grade 2 recommendation, grade C evidence).

No change.

No change. In children > 6 years of age concomitant splenectomy is indicated according to severity of anemia (Table 1). No change. No consensus among our experts.

* Reference 45 The Grading of Recommendation Assessment, Developing and Evaluation (GRADE) system was used to rate quality of evidence and strength of recommendations.

always be removed.45 This recommendation was based on expert opinion despite little supportive data in the literature; the recommendation was changed in subsequent guidelines to indicate that this issue remains controversial.46 In a recent study, of 32 pediatric patients with HS who underwent cholecystectomy, 27 underwent synchronous splenectomy. However, none of the five patients who underwent cholecystectomy without splenectomy experienced signs or symptoms consistent with gallstones over a median follow-up of 15.6 years.47 Similarly, in a recent study of children aged 4-17 years studied during 2009-2012, simultaneous cholecystectomy (for cholelithiasis) and splenectomy was performed in fewer than half of the patients.48 We recommend that indications for splenectomy when cholecystectomy is required should not differ from those for splenectomy when cholecystectomy is not planned (Tables 1 and 2).

Partial splenectomy In an attempt to reduce the infectious risk following total splenectomy, especially for children less than 6 years of age who suffer severe anemia or are transfusion-dependent, partial splenectomy has been increasingly performed in recent years. Several studies indicate that partial splenectomy reduces the rate of hemolysis and increases red blood cell lifespan while maintaining efficient splenic phagocytic function.49-50 In a recently published follow-up study, Pincez et al. reported on 79 HS children who underwent partial splenectomy using an open splenectomy approach between 1985 and 2013. In this population, 39 children were less than 5 years of age at the time of splenectomy (mean age at surgery, 4.3±0.6 years) and most were transfusion–dependent (31/39). Following partial splenectomy (mean follow up of 12±0.9 years) there were drastic reductions in transfusion rate and increases in hemoglobin levels that were compatible with normal growth while maintaining efficient spleen function in 96% of cases.51 On the other hand, this approach reduced but did not totally suppress hemolysis and was associated with later development of gallstones, and splenic remnant regrowth. Finally, 50% of the 39 severely affected young 1308

HS children required total splenectomy in a median of 5 years following partial splenectomy at an age when total splenectomy was much safer.51 Falling hemoglobin levels and discomfort due to spleen remnant regrowth were the most common indications for this procedure. A recently published systematic review and meta-analysis comparing HS patients undergoing total splenectomy (1941 children) versus partial splenectomy (283 children) confirmed that although total splenectomy was more effective than partial splenectomy in increasing hemoglobin levels (increases of 3.6 g/dL and 2.2 g/dL, respectively) and in reducing reticulocyte counts (by 12.5% and 6.5%, respectively), the outcome following partial splenectomy was stable for at least 6 years. There were no cases of OPSI.52 In this metaanalysis, with an overall short follow-up, recurrence of symptoms following partial splenectomy was uncommon (5-10%) and secondary splenectomy was indicated in only 5% of children. Thus, partial splenectomy still needs to be evaluated in larger series with longer term follow-up. In view of these conflicting data no recommendations regarding partial splenectomy in HS could be generated by the group. Splenectomy recommendations - Splenectomy is recommended in HS patients who are transfusion-dependent or suffer severe anemia (agreed by 100% of experts) (Grade 2 recommendation, grade C evidence). -Indications for splenectomy in the intermediate forms of HS should be individually tailored based on spleen size and quality of life parameters (agreed by 95% of experts) (grade 2 recommendation, grade C evidence).

Pyruvate kinase deficiency Pyruvate kinase (PK) deficiency is the most common glycolytic defect causing congenital non-spherocytic hemolytic anemia, having an incidence of 1:20,000 in white individuals.53 PK converts phosphoenolpyruvate to pyruvate, generating 50% of the total red cell ATP. PKdeficient red blood cells are damaged due to lack of energy haematologica | 2017; 102(8)

Splenectomy in hereditary hemolytic anemias

Table 4. Indications for splenectomy in pyruvate kinase deficiency based on severity of disease*.

Disease severity

Age at diagnosis

Clinical manifestations

Median hemoglobin (g/dL)

Transfusion requirement

Indication for splenectomy

Severe

Birth/infancy

Most patients suffer severe neonatal jaundice requiring exchange transfusion, median age at diagnosis 4, almost all transfusion-dependent Moderate anemia occasional exacerbations Lifelong history of mild anemia

6.8

Transfusion-dependent

Indicated after age of 6 years

Confined to exacerbations

Not indicated

Rare

Not indicated

Moderate Mild

Variable, childhood to adult Variable, childhood to adult

Severity of disease based on references 54 and 55.

to support membrane ion transport and to maintain membrane structure and are, therefore, cleared by the spleen and liver. Clinically, PK deficiency has been categorized into mild, moderate and severe forms (Table 4).54-55

Results of splenectomy Small retrospective studies suggest that splenectomy may result in a moderate increase in hemoglobin levels of approximately 1.8 g/dL (range, 0.4â&#x20AC;&#x201C;3.4 g/dL), together with a conspicuous rise of reticulocytes (up to 50-70%, a typical feature of PK deficiency) and increased amounts of indirect bilirubin, even if the anemia becomes less severe.54 Analysis of the results of a recent international, multicenter registry study involving 144 patients56 suggested that splenectomy was performed mainly to reduce transfusion burden and resulted in improved anemia, and thus enhanced quality of life. The median pre-splenectomy hemoglobin concentration was 7 g/dL and, surgery reduced the transfusion burden in 91% of cases. Fifty-three patients (66%) underwent cholecystectomy at a median age of 14 years (range, 2.660.4 years), of whom 35 (37%) were splenectomized. Importantly this study showed that transfusion-dependency and moderate anemia persisted despite splenectomy in more than half of the patients, suggesting that surgery is less effective in PK deficiency than in HS.

Indications for splenectomy Splenectomy may be beneficial in patients with high transfusion requirements. It may also be considered in patients with low transfusion requirements who may subsequently become transfusion independent following splenectomy, although this is difficult to predict. Although splenectomy does not arrest hemolysis, it reduces and sometimes eliminates the transfusion requirement in most transfusion-dependent cases. The response to surgery of other affected family members may help in predicting the therapeutic efficacy of splenectomy.54,55 Optimal timing of surgery is unclear and needs to be considered individually weighing up the life-long risks (infection, thromboembolism) against the likely benefits.

Cholecystectomy and splenectomy As for HS, splenectomy should also be considered in patients requiring cholecystectomy to avoid second surgery. However, in contrast to HS, in PK deficiency gallstones are also common in splenectomized patients, and therefore cholecystectomy should accompany splenectomy. haematologica | 2017; 102(8)

Partial splenectomy Partial splenectomy was reported to be unsuccessful in two patients with PK deficiency and effective in one patient, who achieved an increase in baseline hemoglobin and reduction in transfusion rate.57 Other therapeutic options for PK deficiency that should be considered include HLA-matched sibling allogeneic transplantation,58 and new therapies which are still under investigation, including the enzyme activator (AG-348)59 and gene therapy.60 Splenectomy recommendations -Splenectomy should be considered in patients with PK deficiency who are transfusion-dependent or do not tolerate the anemia (agreed by 91% of experts) (grade 2 recommendation, grade C evidence). -Cholecystectomy should be performed together with splenectomy (agreed by 86% of experts) (grade 2 recommendation, grade C evidence).

Congenital non-spherocytic hemolytic anemia due to glucose-6-phosphate dehydrogenase deficiency Patients with glucose-6-phosphate dehydrogenase deficiency rarely suffer hemolytic anemia in the steady state and hemolysis is triggered by an exogenous factor. Some mutations of glucose-6-phosphate dehydrogenase do, however, result in chronic hemolysis without precipitating causes. These mutations are more severe than the more commonly occurring polymorphic forms of the enzyme. The severity of anemia ranges from borderline to transfusion-dependent. In 2004, Hamilton et al. identified nine transfusion-dependent patients in the literature: seven responded to splenectomy and became transfusion-independent.61 Luzzatto and Poggi suggested that splenectomy should be performed if splenomegaly becomes a physical encumbrance, or if there is evidence of hypersplenism and if the anemia is severe.62 Splenectomy recommendations -Splenectomy should be considered in patients with congenital non-spherocytic hemolytic anemia due to glucose-6-phosphate dehydrogenase deficiency who are transfusion-dependent and/or have symptomatic splenomegaly. (agreed by 100% of experts) (grade 2 recommendation, grade C evidence)

1309

A. Iolascon et al.

Pyrimidine-5’-nucleotidase deficiency Deficiency of erythrocyte pyrimidine-5’-nucleotidase is the most common inherited abnormality of nucleotide metabolism causing hemolytic anemia of moderate severity.55,63 Transfusions are rarely required. Splenectomy has been associated with variable increases in hemoglobin levels.64-67 Portosplenomesentric venous thrombosis was described in one patient with pyrimidine-5’-nucleotidase following splenectomy due to trauma.68 As there are insufficient data regarding the efficacy and complications of splenectomy in this disorder no recommendations could be made.

Hereditary stomatocytosis Hereditary stomatocytosis (HSt), comprising both dehydrated and overhydrated types, is a dominantly inherited disorder in which there is altered red blood cell membrane permeability to monovalent cations (Na+ and K+), with consequent changes in intracellular cation content and red cell volume. Dehydrated HSt is the most common form of HSt, with an incidence of approximately 1:50,000 births. Overhydrated HSt is a very rare subtype, with only 20 cases having been reported worldwide. The recent identification of genes mutated in HSt has improved the diagnosis and understanding of the pathophysiology of this group of disorders. To date, a total of five different genes encoding membrane proteins have been reported to be responsible for red blood cell volume alterations: three lead to overhydrated HSt [AE1 (also termed SLC4A1), RHAG and GLUT1 (also termed SCL2A1)69,70 and two to dehydrated HSt (PIEZO1 and KCNN4, encoding the Gardos channel)]71-75. Reviews on the clinical picture and molecular pathogenesis have recently been published.37,76

Splenectomy A high risk of thromboembolic complications following splenectomy in HSt was first described by Stewart and colleagues (1996) in nine splenectomized adults.77 In their seven families four had overhydrated HSt and the remaining dehydrated HSt. Both groups suffered from serious late thrombotic complications, sometimes recurrent over years, including deep vein thrombosis, pulmonary emboli, superficial thrombophlebitis, portal vein thrombosis, intracardiac mural thrombosis, arterial thrombosis, and pulmonary arterial hypertension. Four of the nine patients died. No such complications were observed in six affected before splenectomy. Since this original observation there have been at least four additional reports of individuals with overhydrated or dehydrated HSt who developed severe thrombotic complications.78-81 Given the retrospective, incomplete, and anecdotal nature of the description of thromboembolic complications following splenectomy in HSt, it is impossible to estimate the precise risk of this procedure, but it is apparent that there is a high risk and that splenectomy, which is only partially effective in overhydrated HSt and ineffective in dehydrated HSt, should be avoided. In patients with HSt who were mistakenly misdiagnosed as having HS and underwent splenectomy, life-long anticoagulation should be considered.

1310

Splenectomy recommendations - In patients with hereditary stomatocytosis, splenectomy is contraindicated (agreed by 100% of experts) (grade 2 recommendation, grade C evidence).

Congenital dyserythropoietic anemia Congenital dyserythropoietic anemia (CDA) is a group of rare red blood cell disorders characterized by ineffective erythropoiesis, pathognomonic cytopathology of nucleated red blood cells in bone marrow and increased iron absorption with secondary hemochromatosis.82 Based on morphological criteria, subsequently supported by genetic analysis, three types have been described (I-III). More recently CDA IV was defined and several additional unclassified patients have been characterized.83 CDA II is the most common subtype, with more than 200 patients described in the literature, followed by CDA I with approximately 100 patients.82

Splenectomy in congenital dyserythropoietic anemia I Two case series of 13 severely anemic, mostly transfusion-dependent patients who had undergone splenectomy were identified.84,85 Six of those patients became transfusion-independent following splenectomy, while seven had no improvement in hemoglobin levels. Long-term followup of six patients revealed that three had died, one due to pulmonary arterial hypertension and the other two due to overwhelming sepsis. Because of inconsistent responses and possible complications, splenectomy should probably be reserved for patients manifesting worsening anemia, and/or significant thrombocytopenia or leukopenia or for patients with massive, painful splenomegaly. Due to paucity of data no recommendation regarding splenectomy in CDA I could be made.

Splenectomy in congenital dyserythropoietic anemia II Heimpel et al. reported on 22 patients who underwent splenectomy at a median age of 19.9 years.86 Hemoglobin concentration increased in all patients from an average of 9.2 g/dL to 10.3 g/dL. However, in all but one patient, hemoglobin levels remained below sex-matched reference values.86 Russo et al. reported that 36% (41/112) of their patients with CDA II underwent splenectomy and most (14/17) showed a similar, moderate increase in hemoglobin concentration (9.3±1.2 g/dL to 10.6±1.6 g/dL).87 To date thrombosis has not been documented in patients with CDA II following splenectomy. Splenectomy recommendations -In patients with CDA II splenectomy should be considered in severely anemic patients and/or in those with symptomatic splenomegaly (agreed by 95% of experts) (grade 2 recommendation, grade C evidence).

Thalassemic syndromes The authors decided not to discuss thalassemic syndromes since revised guidelines have recently been presented by the Thalassaemia International Federation (please visit: www.tif.org).88

haematologica | 2017; 102(8)

Splenectomy in hereditary hemolytic anemias

Sickle cell disease SCD is a hereditary hemoglobinopathy with a worldwide distribution: projected numbers of births in 2010 were 237,381 in Africa, 11,143 in the USA and 1,939 in Europe.89 SCD is caused by a point mutation in the Î˛-globin gene resulting in the synthesis of a pathological hemoglobin, HbS.90 Cyclic polymerization/depolymerization of deoxy-HbS generates dense, dehydrated red cells that play a central role in the acute and chronic clinical manifestations of SCD, in which intravascular sickling leads to vasoocclusion and impaired blood flow with ischemic/reperfusion injury.90 Some organs, such as the spleen, have been shown to be more vulnerable to damage from HbS polymerization than others organs, due to their peculiar anatomic organization mainly characterized by sluggish circulation, low pH and local high pro-oxidant environment.90,91

approach in SCD children are the size of the spleen, with its local adhesions, and the longer duration of the operation. Although laparoscopic splenectomy has some positive aspects, such as the shorter duration of hospitalization, its impact on the incidence of post-operative severe, acute, SCD-related complications, including acute chest syndrome, is still unclear.98,99 To date there are no real advantages, in terms of hematologic phenotype and infective risk, from partial splenectomy rather than total splenectomy in children with SCD.100 Thus, guidelines on the clinical management of acute splenic sequestration crisis in SCD are needed. The guidelines should present parameters for and timing of splenectomy, with accompanying surgical approaches. Indications are likely to vary depending on environmental factors, including the availability of safe blood transfusions and the spectrum of infections. Future multicenter studies should be designed to address these issues.

Splenectomy An acute splenic sequestration crisis is defined as acute abdominal pain and distension associated with spleen enlargement, a decrease in hemoglobin levels of at least 2 g/dL and stable or high reticulocyte count compared to that of the patient in steady state.92 Even though the mortality rate of patients with SCD has declined since the introduction of neonatal screening for the disease, vaccination programs and parental education, acute splenic sequestration crisis is still a life-threatening complication.93 The clinical management of such crises, with acute splenic sickling and spleen blood entrapment, is based on rapid correction of hypovolemic shock by infusion of crystalloids and packed red cells. Although international guidelines and consensus statements on the management of acute splenic sequestration crisis are not available, splenectomy is usually recommended after two such crises requiring urgent transfusion.92-95 Hypersplenism, defined as chronic splenic enlargement with lowered hemoglobin concentration and decreased platelet and leukocyte counts, is the second major indication for splenectomy in SCD patients.92-95 Splenectomy is usually indicated if there is hypersplenism, pressure effects of the spleen or failure to thrive. Transfusion is often ineffective in such children because of red blood cell sequestration in the enlarged spleen. Although splenectomy is the treatment for recurrent acute splenic sequestration crises and hypersplenism in SCD, there is no evidence that it increases hemoglobin level, decreases hemolysis or improves patientsâ&#x20AC;&#x2122; survival.96 Splenectomy may, however, increase thromboembolism. Limited evidence is available about any possible increased incidence of pulmonary arterial hypertension or pain frequency in SCD patients.97 Laparoscopic splenectomy is generally used in children with SCD. It is performed after preoperative transfusion or exchange transfusion to decrease the percentage of HbS. The major limitations to using a laparoscopic

References 1. Crary SE, Buchanan GR. Vascular complications after splenectomy for hematologic disorders. Blood. 2009;114(14):2861-2868.

haematologica | 2017; 102(8)

Splenectomy recommendations -In SCD children, splenectomy is recommended after two acute splenic sequestration crises (agreed by 100% of experts) (grade 2 recommendation, grade C evidence). -In SCD children, splenectomy is also recommended in patients with massive splenomegaly and/or hypersplenism (agreed by 90% of experts) (grade 2 recommendation, grade C evidence).

Unstable hemoglobin Globin mutations that destabilize hemoglobin tetramers constitute a very rare cause of hemolytic anemia. The clinical pattern of hemolytic anemia related to unstable hemoglobin is extremely variable. Severely affected patients, particularly those with hyperunstable hemoglobin, present early in childhood and may require chronic transfusion therapy. Thrombotic complications following splenectomy have been described in nine patients (with Hb Bridlington/HbTaybe, Hb Taybe, Hb Mainz, Hb Olmsted, Hb Madrid and Hb Perth).101-106 The thrombotic events, including pulmonary emboli, pulmonary arterial hypertension, arterial stroke and priapism, occurred even 4-32 years after splenectomy. The majority of patients (7/9) had no or only partial improvement in hemoglobin levels. Given the anecdotal data, splenectomy should be considered only when there is severe anemia and/or massive or symptomatic splenomegaly. Splenectomy recommendations -In patients with unstable hemoglobin splenectomy should be considered when the spleen is very large and/or there is evidence of hypersplenism (agreed by 95% of experts) (grade 2 recommendation, grade C evidence).

2. Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336(7650):924-926. 3. Eraklis AJ, Kevy SV, Diamond LK, Gross RE. Hazard of overwhelming infection after

splenectomy in childhood. N Engl J Med. 1967;276(22):1225-1229. 4. Schnitzer B, Sodeman TM, Mead ML, Contacos PG. An ultrastructural study of the red pulp of the spleen in malaria. Blood. 1973;41(2):207-218.

1311

A. Iolascon et al. 5. Rosner F, Zarrabi MH, Benach JL, Habicht GS. Babesiosis in splenectomized adults. Review of 22 reported cases. Am J Med. 1984;76(4):696-701. 6. Rubin LG, Schaffner W. Clinical practice. Care of the asplenic patient. N Engl J Med. 2014;371(4):349-356. 7. Bisharat N, Omari H, Lavi I, Raz R. Risk of infection and death among post-splenectomy patients. J Infect. 2001;43(3):182-186. 8. Styrt B. Infection associated with asplenia: risks, mechanisms, and prevention. Am J Med. 1990;88(5N):33N-42N. 9. Kristinsson SY, Gridley G, Hoover RN, Check D, Landgren O. Long-term risks after splenectomy among 8,149 cancer-free American veterans: a cohort study with up to 27 years follow-up. Haematologica. 2014;99(2):392-398. 10. Luoto TT, Pakarinen MP, Koivusalo A. Longterm outcomes after pediatric splenectomy. Surgery. 2016;159(6):1583-1590. 11. Theilacker C, Ludewig K, Serr A, et al. Overwhelming postsplenectomy infection: a prospective multicenter cohort study. Clin Infect Dis. 2016;62(7):871-878. 12. Davies JM, Lewis MP, Wimperis J, et al. Review of guidelines for the prevention and treatment of infection in patients with an absent or dysfunctional spleen: prepared on behalf of the British Committee for Standards in Haematology by a working party of the Haemato-Oncology task force. Br J Haematol. 2011;155(3):308-317. 13. Rodeghiero F, Ruggeri M. Short- and longterm risks of splenectomy for benign haematological disorders: should we revisit the indications? Br J Haematol. 2012;158(1):16-29. 14. Krauth MT, Lechner K, Neugebauer EA, Pabinger I. The postoperative splenic/portal vein thrombosis after splenectomy and its prevention â&#x20AC;&#x201C; an unresolved issue. Haematologica. 2008;93(8):1227-1232. 15. Robinette CD, Fraumeni JF Jr. Splenectomy and subsequent mortality in veterans of the 1939-45 war. Lancet. 1977;2(8029):127-129. 16. Tincani E, Piccoli M, Turrini F, Crowther MA, Melotti G, Bondi M. Video laparoscopic surgery: is out-of-hospital thromboprophylaxis necessary? J Thromb Haemost. 2005;3 (2):216-220. 17. Manouchehri N, Kaneva P, Seguin C, Artho GP, Feldman LS. Screening for thrombophilia does not identify patients at risk of portal or splenic vein thrombosis following laparoscopic splenectomy. Surg Endosc. 2016;30(5):2119-2126. 18. Ikeda M, Sekimoto M, Takiguchi S, et al. Total splenic vein thrombosis after laparoscopic splenectomy: a possible candidate for treatment. Am J Surg. 2007;193(1):21-25. 19. Tran T, Demyttenaere SV, Polyhronopoulos G, et al. Recommended timing for surveillance ultrasonography to diagnose portal splenic vein thrombosis after laparoscopic splenectomy. Surg Endosc. 2010;24(7):16701678. 20. Thomsen RW, Schoonen WM, Farkas DK, Riis A, Fryzek JP, Sorensen HT. Risk of venous thromboembolism in splenectomized patients compared with the general population and appendectomized patients: a 10-year nationwide cohort study. J Thromb Haemost. 2010;8(6):1413-1416. 21. Hoeper MM, Niedermeyer J, Hoffmeyer F, Flemming P, Fabel H. Pulmonary hypertension after splenectomy? Ann Intern Med. 1999;130(6):506-509. 22. Atichartakarn V, Likittanasombat K, Chuncharunee S, et al. Pulmonary arterial hypertension in previously splenectomized patients with beta-thalassemic disorders. Int J

1312

Hematol. 2003;78(2):139-145. 23. Chou R, DeLoughery TG. Recurrent thromboembolic disease following splenectomy for pyruvate kinase deficiency. Am J Hematol. 2001;67(3):197-199. 24. Hayag-Barin JE, Smith RE, Tucker FC Jr. Hereditary spherocytosis, thrombocytosis, and chronic pulmonary emboli: a case report and review of the literature. Am J Hematol. 1998;57(1):82-84. 25. Atichartakarn V, Angchaisuksiri P, Aryurachai K, Chuncharunee S, Thakkinstian A. In vivo platelet activation and hyperaggregation in hemoglobin E/beta-thalassemia: a consequence of splenectomy. Int J Hematol. 2003;77(3):299-303. 26. Wilhelm MC, Jones RE, McGehee R, Mitchener JS, Sandusky WR, Hess CE. Splenectomy in hematologic disorders. The ever-changing indications. Ann Surg. 1988;207(5):581-589. 27. Tulman S, Holcomb GW 3rd, Karamanoukian HL, Reynhout J. Pediatric laparoscopic splenectomy. J Pediatric Surg. 1993;28(5):689-692. 28. Wood JH, Partrick DA, Hays T, Sauaia A, Karrer FM, Ziegler MM. Contemporary pediatric splenectomy: continuing controversies. Pediatr Surg Int. 2011;27(11):1165-1171. 29. Al-Mulhim AS. Laparoscopic splenectomy for massive splenomegaly in benign hematological diseases. Surg Endosc. 2012;26(11):3186-3189. 30. Van Der Veken E, Laureys M, Rodesch G, Steyaert H. Perioperative spleen embolization as a useful tool in laparoscopic splenectomy for simple and massive splenomegaly in children: a prospective study. Surg Endosc. 2016;30(11):4962-4967. 31. Luzzatto L. Recent advances in the pathogenesis and treatment of paroxysmal nocturnal hemoglobinuria. F1000Res. 2016;5. 32. Pietrabissa A, Marconi S, Peri A, et al. From CT scanning to 3-D printing technology for the preoperative planning in laparoscopic splenectomy. Surg Endosc. 2016;30(1):366371. 33. Slater BJ, Chan FP, Davis K, Dutta S. Institutional experience with laparoscopic partial splenectomy for hereditary spherocytosis. J Pediatr Surg. 2010;45(8):1682-1686. 34. Perrotta S, Gallagher PG, Mohandas N. Hereditary spherocytosis. Lancet. 2008;372(9647):1411-1426. 35. Gallagher PG. Abnormalities of the erythrocyte membrane. Pediatr Clin North Am. 2013;60(6):1349-1362. 36. Barcellini W, Bianchi P, Fermo E, et al. Hereditary red cell membrane defects: diagnostic and clinical aspects. Blood Transfus. 2011;9(3):274-277. 37. Andolfo I, Russo R, Gambale A, Iolascon A. New insights on hereditary erythrocyte membrane defects. Haematologica. 2016;101 (11):1284-1294. 38. Schilling RF. Spherocytosis, splenectomy, strokes, and heat attacks. Lancet. 1997;350(9092):1677-1678. 39. Schilling RF, Gangnon RE, Traver MI. Delayed adverse vascular events after splenectomy in hereditary spherocytosis. J Thromb Haemost. 2008;6(8):1289-1295. 40. Bruguier A, Clement MC, Texier P, Ponsot G. [Cerebral ischemic accident and hereditary spherocytosis]. Arch Fr Pediatr. 1983;40(8): 653-654. 41. Jardine DL, Laing AD. Delayed pulmonary hypertension following splenectomy for congenital spherocytosis. Intern Med J. 2004;34(4):214-216. 42. Smedema JP, Louw VJ. Pulmonary arterial

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

hypertension after splenectomy for hereditary spherocytosis. Cardiovasc J Afr. 2007; 18(2):84-89. Crary SE, Ramaciotti C, Buchanan GR. Prevalence of pulmonary hypertension in hereditary spherocytosis. Am J Hematol. 2011;86(12):E73-E76. Mariani M, Barcellini W, Vercellati C, et al. Clinical and hematologic features of 300 patients affected by hereditary spherocytosis grouped according to the type of the membrane protein defect. Haematologica. 2008;93(9):1310-1317. Bolton-Maggs PH, Stevens RF, Dodd NJ, et al. Guidelines for the diagnosis and management of hereditary spherocytosis. Br J Haematol. 2004;126(4):455-474. Bolton-Maggs PH, Langer JC, Iolascon A, Tittensor P, King MJ, General Haematology Task Force of the British Committee for Standards in Haematology. Guidelines for the diagnosis and management of hereditary spherocytosis â&#x20AC;&#x201C; 2011 update. Br J Haematol. 2012; 156(1):37-49. Ruparel RK, Bogert JN, Moir CR, et al. Synchronous splenectomy during cholecystectomy for hereditary spherocytosis: is it really necessary? J Pediatr Surg. 2014;49(3): 433-435. Rogulski R, Adamowicz-Salach A, Matysiak M, et al. Laparoscopic splenectomy for hereditary spherocytosis - preliminary report. Eur J Haematol. 2016;96(6):637-642. Bader-Meunier B, Gauthier F, Archambaud F, et al. Long-term evaluation of the beneficial effect of subtotal splenectomy for management of hereditary spherocytosis. Blood. 2001;97(2):399-403. Buesing KL, Tracy ET, Kiernan C, et al. Partial splenectomy for hereditary spherocytosis: a multi-institutional review. J Pediatr Surg. 2011;46(1):178-183. Pincez T, Guitton C, Gauthier F, et al. Longterm follow-up of subtotal splenectomy for hereditary spherocytosis: a single-center study. Blood. 2016;127(12):1616-1618. Guizzetti L. Total versus partial splenectomy in pediatric hereditary spherocytosis: a systematic review and meta-analysis. Pediatr Blood Cancer. 2016;63(10):1713-1722. Beutler E, Gelbart T. Estimating the prevalence of pyruvate kinase deficiency from the gene frequency in the general white population. Blood. 2000;95(11):3585-3588. Zanella A, Fermo E, Bianchi P, Valentini G. Red cell pyruvate kinase deficiency: molecular and clinical aspects. Br J Haematol. 2005;130(1):11-25. Zanella A, Fermo E, Bianchi P, Chiarelli LR, Valentini G. Pyruvate kinase deficiency: the genotype-phenotype association. Blood Rev. 2007;21(4):217-231. Grace RF, Morton DH, Barcellini W, et al. The phenotypic spectrum of pyruvate kinase deficiency (PKD) from the PKD Natural History Study (NHS): description of four severity groups by anemia status. Blood. 2015;126 (23):2136. Grace RF, Zanella A, Neufeld EJ, et al. Erythrocyte pyruvate kinase deficiency: 2015 status report. Am J Hematol. 2015;90(9):825830. Tanphaichitr VS, Suvatte V, Issaragrisil S, et al. Successful bone marrow transplantation in a child with red blood cell pyruvate kinase deficiency. Bone Marrow Transplant. 2000;26 (6):689-690. Grace RF, Rose C, Layton DM, et al. Effects of AG-348, a pyruvate kinase activator, on anemia and hemolysis in patients with pyruvate kinase deficiency: data from the DRIVE PK

haematologica | 2017; 102(8)

Splenectomy in hereditary hemolytic anemias study. Blood. 2016;128(22):402. 60. Garcia-Gomez M, Calabria A, Garcia-Bravo M, et al. Safe and efficient gene therapy for pyruvate kinase deficiency. Mol Ther. 2016;24(7):1187-1198. 61. Hamilton JW, Jones FG, McMullin MF. Glucose-6-phosphate dehydrogenase Guadalajara â&#x20AC;&#x201C; a case of chronic non-spherocytic haemolytic anaemia responding to splenectomy and the role of splenectomy in this disorder. Hematology. 2004;9(4):307-309. 62. Luzzatto L, Poggi V. Glucose 6 phosphate dehydrognase deficiency. In: Orkin SH, Fisher DE, Ginsburg D, Look AT, Nathan DG, Lux SE, et al., eds. Nathan and Oski's Hematology and Oncology of Infancy and Childhood. Canada: Elsevier/Saunders, 2015:609-629. 63. Rees DC, Duley JA, Marinaki AM. Pyrimidine 5' nucleotidase deficiency. Br J Haematol. 2003;120(3):375-383. 64. McMahon JN, Lieberman JE, Gordon-Smith EC, Egan EL. Hereditary haemolytic anaemia due to red cell pyrimidine 5'-nucleotidase deficiency in two Irish families with a note on the benefit of splenectomy. Clin Lab Haematol. 1981;3(1):27-34. 65. Ozsoylu S, Gurgey A. A case of hemolytic anemia due to erythrocyte pyrimidine 5'nucleotidase deficiency. Acta Haematol. 1981;66(1):56-58. 66. Dvilansky A, Hezkelson L, Wolfson M, Nathan I, Bashan N, Meyerstein N. Haemolytic anaemia due to pyrimidine-5'nucleotidase deficiency. Int J Tissue React. 1984;6(4):351-354. 67. Rees DC, Duley J, Simmonds HA, et al. Interaction of hemoglobin E and pyrimidine 5' nucleotidase deficiency. Blood. 1996;88(7):2761-2767. 68. Al-Jafar HA, Taqi A, Madda JP, Abdullah TA. Splenectomy complicated by sustained extreme thrombocytosis and extensive portosplenomesenteric vein thrombosis in pyrimidine 5'-nucleotidase deficiency. BMJ Case Rep. 2013;2013. 69. Bruce LJ, Robinson HC, Guizouarn H, et al. Monovalent cation leaks in human red cells caused by single amino-acid substitutions in the transport domain of the band 3 chloridebicarbonate exchanger, AE1. Nat Genet. 2005;37(11):1258-1263. 70. Bruce LJ, Guizouarn H, Burton NM, et al. The monovalent cation leak in overhydrated stomatocytic red blood cells results from amino acid substitutions in the Rh-associated glycoprotein. Blood. 2009;113(6):13501357. 71. Zarychanski R, Schulz VP, Houston BL, et al. Mutations in the mechanotransduction protein PIEZO1 are associated with hereditary xerocytosis. Blood. 2012;120(9):1908-1915. 72. Albuisson J, Murthy SE, Bandell M, et al. Dehydrated hereditary stomatocytosis linked to gain-of-function mutations in mechanically activated PIEZO1 ion channels. Nat Commun. 2013;4:1884. 73. Andolfo I, Alper SL, De Franceschi L, et al. Multiple clinical forms of dehydrated hereditary stomatocytosis arise from mutations in PIEZO1. Blood. 2013;121(19):3925-3935. 74. Rapetti-Mauss R, Lacoste C, Picard V, et al. A mutation in the Gardos channel is associated with hereditary xerocytosis. Blood. 2015;126(11):1273-1280.

haematologica | 2017; 102(8)

75. Andolfo I, Russo R, Manna F, et al. Novel Gardos channel mutations linked to dehydrated hereditary stomatocytosis (xerocytosis). Am J Hematol. 2015;90(10):921-926. 76. Badens C, Guizouarn H. Advances in understanding the pathogenesis of the red cell volume disorders. Br J Haematol. 2016;174(5):674-685. 77. Stewart GW, Amess JA, Eber SW, et al. Thrombo-embolic disease after splenectomy for hereditary stomatocytosis. Br J Haematol. 1996;93(2):303-310. 78. Bergheim J, Ernst P, Brinch L, Gore DM, Chetty MC, Stewart GW. Allogeneic bone marrow transplantation for severe postsplenectomy thrombophilic state in leaky red cell membrane haemolytic anaemia of the stomatocytosis class. Br J Haematol. 2003;121(1):119-122. 79. Jais X, Till SJ, Cynober T, et al. An extreme consequence of splenectomy in dehydrated hereditary stomatocytosis: gradual thromboembolic pulmonary hypertension and lungheart transplantation. Hemoglobin. 2003;27(3):139-147. 80. Perel Y, Dhermy D, Carrere A, et al. Portal vein thrombosis after splenectomy for hereditary stomatocytosis in childhood. Eur J Pediatr. 1999;158(8):628-630. 81. Yoshimoto A, Fujimura M, Nakao S. Pulmonary hypertension after splenectomy in hereditary stomatocytosis. Am J Med Sci. 2005;330(4):195-197. 82. Gambale A, Iolascon A, Andolfo I, Russo R. Diagnosis and management of congenital dyserythropoietic anemias. Expert Rev Hematol. 2016;9(3):283-296. 83. Iolascon A, Heimpel H, Wahlin A, Tamary H. Congenital dyserythropoietic anemias: molecular insights and diagnostic approach. Blood. 2013;122(13):2162-2166. 84. Heimpel H, Schwarz K, Ebnother M, et al. Congenital dyserythropoietic anemia type I (CDA I): molecular genetics, clinical appearance, and prognosis based on long-term observation. Blood. 2006;107(1):334-340. 85. Shalev H, Al-Athamen K, Levi I, Levitas A, Tamary H. Morbidity and mortality of adult patients with congenital dyserythropoietic anemia type I. Eur J Haematol. 2017;98(1):1318. 86. Heimpel H, Anselstetter V, Chrobak L, et al. Congenital dyserythropoietic anemia type II: epidemiology, clinical appearance, and prognosis based on long-term observation. Blood. 2003;102(13):4576-4581. 87. Russo R, Gambale A, Langella C, Andolfo I, Unal S, Iolascon A. Retrospective cohort study of 205 cases with congenital dyserythropoietic anemia type II: definition of clinical and molecular spectrum and identification of new diagnostic scores. Am J Hematol. 2014;89(10):E169-E175. 88. Cappellini M-D, Cohen A, Porter J, Taher A, Viprakasit V. Guidelines for the management of transfusion dependent thalassaemia (TDT). TIF publication. 2014. 89. Piel FB, Hay SI, Gupta S, Weatherall DJ, Williams TN. Global burden of sickle cell anaemia in children under five, 2010-2050: modelling based on demographics, excess mortality, and interventions. PLoS Med. 2013;10(7):e1001484. 90. De Franceschi L, Cappellini MD, Olivieri O.

Thrombosis and sickle cell disease. Semin Thromb Hemost. 2011;37(3):226-236. 91. Platt OS. The acute chest syndrome of sickle cell disease. N Engl J Med. 2000;342(25):19041907. 92. Lesher AP, Kalpatthi R, Glenn JB, Jackson SM, Hebra A. Outcome of splenectomy in children younger than 4 years with sickle cell disease. J Pediatr Surg. 2009;44(6):1134-1138. 93. Brousse V, Elie C, Benkerrou M, et al. Acute splenic sequestration crisis in sickle cell disease: cohort study of 190 paediatric patients. Br J Haematol. 2012;156(5):643-648. 94. Al-Salem AH. Indications and complications of splenectomy for children with sickle cell disease. J Pediatr Surg. 2006;41(11):19091915. 95. Machado NO, Grant CS, Alkindi S, et al. Splenectomy for haematological disorders: a single center study in 150 patients from Oman. Int J Surg. 2009;7(5):476-481. 96. Owusu-Ofori S, Remmington T. Splenectomy versus conservative management for acute sequestration crises in people with sickle cell disease. Cochrane Database Syst Rev. 2015(9):CD003425. 97. Mouttalib S, Rice HE, Snyder D, et al. Evaluation of partial and total splenectomy in children with sickle cell disease using an Internet-based registry. Pediatr Blood Cancer. 2012;59(1):100-104. 98. Goers T, Panepinto J, Debaun M, et al. Laparoscopic versus open abdominal surgery in children with sickle cell disease is associated with a shorter hospital stay. Pediatr Blood Cancer. 2008;50(3):603-606. 99. Alwabari A, Parida L, Al-Salem AH. Laparoscopic splenectomy and/or cholecystectomy for children with sickle cell disease. Pediatr Surg Int. 2009;25(5):417-421. 100. Englum BR, Rothman J, Leonard S, et al. Hematologic outcomes after total splenectomy and partial splenectomy for congenital hemolytic anemia. J Pediatr Surg. 2016;51(1): 122-127. 101. Hill QA, Farrar L, Lordan J, Gallienne A, Henderson S. A combination of two novel alpha globin variants Hb Bridlington (HBA1) and Hb Taybe (HBA2) resulting in severe hemolysis, pulmonary hypertension, and death. Hematology. 2015;20(1):50-52. 102. Juul MB, Vestergaard H, Petersen J, Frederiksen H. Thrombosis in Hb Taybe [codons 38/39 (-ACC) (alpha1)]. Hemoglobin. 2012;36(6):600-604. 103. Lode HN, Krings G, Schulze-Neick I, et al. Pulmonary hypertension in a case of HbMainz hemolytic anemia. J Pediatr Hematol Oncol. 2007;29(3):173-177. 104. Thuret I, Bardakdjian J, Badens C, et al. Priapism following splenectomy in an unstable hemoglobin: hemoglobin Olmsted beta 141 (H19) Leu-->Arg. Am J Hematol. 1996;51(2):133-136. 105. Kim BJ, Park KW, Koh SB, et al. Stroke induced by splenectomy in hemoglobin Madrid: autopsy clues to the underlying mechanism. Blood Coagul Fibrinolysis. 2005;16(2):141-144. 106. Gyan E, Darre S, Jude B, et al. Acute priapism in a patient with unstable hemoglobin Perth and factor V Leiden under effective oral anticoagulant therapy. Hematol J. 2001;2(3):210211.

1313

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Red Cell BIology & its Disorders

Ferrata Storti Foundation

Extracellular glycine is necessary for optimal hemoglobinization of erythroid cells Daniel Garcia-Santos,1* Matthias Schranzhofer,1* Richard Bergeron,2 Alex D. Sheftel3,4 and Prem Ponka1 *Co-first authors

Lady Davis Institute for Medical Research, Jewish General Hospital, and the Department of Physiology, McGill University, Montréal, Quebec; 2Ottawa Hospital Research Institute, University of Ottawa, Ontario; 3Spartan Bioscience Inc., Ottawa and 4High Impact Editing, Ottawa, Ontario, Canada 1

Haematologica 2017 Volume 102(8):1314-1323

ABSTRACT

Correspondence: prem.ponka@mcgill.ca

Received: August 31, 2016. Accepted: May 9, 2017. Pre-published: May 11, 2017. doi:10.3324/haematol.2016.155671 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1314 ©2017 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.

1314

ertebrate heme synthesis requires three substrates: succinyl-CoA, which regenerates in the tricarboxylic acid cycle, iron and glycine. For each heme molecule synthesized, one atom of iron and eight molecules of glycine are needed. Inadequate delivery of iron to immature erythroid cells leads to a decreased production of heme, but virtually nothing is known about the consequence of an insufficient supply of extracellular glycine on the process of hemoglobinization. To address this issue, we exploited mice in which the gene encoding glycine transporter 1 (GlyT1) was disrupted. Primary erythroid cells isolated from fetal livers of GlyT1 knockout (GlyT1-/-) and GlyT1-haplodeficient (GlyT1+/-) embryos had decreased cellular uptake of [2-l4C]glycine and heme synthesis as revealed by a considerable decrease in [2-l4C]glycine and 59Fe incorporation into heme. Since GlyT1-/- mice die during the first postnatal day, we analyzed blood parameters of newborn pups and found that GlyT1-/- animals develop hypochromic microcytic anemia. Our finding that Glyt1-deficiency causes decreased heme synthesis in erythroblasts is unexpected, since glycine is a non-essential amino acid. It also suggests that GlyT1 represents a limiting step in heme and, consequently, hemoglobin production.

Introduction In vertebrates, the enzyme 5-aminolevulinate synthase (ALAS; localized in mitochondria) catalyzes the first step of the heme synthesis pathway, namely a condensation reaction between glycine and succinyl-CoA resulting in 5-aminolevulinic acid (ALA).1 In the subsequent enzymatic steps a total of eight molecules of ALA are used to assemble one tetrapyrrole macrocycle. In the final step, which takes place in mitochondria, the enzyme ferrochelatase inserts a ferrous ion (Fe2+) into the ring structure of protoporphyrin IX to produce heme. Shemin and co-workers found that eight of the porphyrin carbon atoms came from the α-carbon atom of each glycine and the remaining 26 came from acetate.2 Thus, the biosynthesis of one molecule of heme requires one atom of iron and eight molecules of glycine. The insufficient delivery of iron to differentiating erythroid cells leads to impaired production of heme. Although defects in glycine transport can cause sideroblastic anemia, thus far there has been no published study directly examining the impact of glycine restriction on the rate of heme synthesis. Glycine is the simplest amino acid in nature3 and, in animals, it is the main component of extracellular structural proteins such as collagen and elastin.4 In mammals, glycine is classified as a non-essential amino acid,5 which is synthesized from three distinct substrates: (i) serine, via the enzyme serine hydroxymethyltransferase,6 (ii) choline, through sarcosine formation7 and (iii) threonine, in a pathway involving the enzyme threonine dehydrogenase.8 However, it has been shown, in humans9 and pigs,5 that the amounts of glycine synthesized in vivo are insufficient haematologica | 2017; 102(8)

GlyT1 is necessary for hemoglobinization

to meet cellular metabolic demands. Therefore, many specialized cells possess different systems to transport glycine actively through their cell membranes: glial cells,10 enterocytes,11 hepatocytes,12 placental cells13 and erythroid cells.14 In human erythrocytes, the major glycine uptake pathway, described in the 1980s, is a Na+- and Cl--dependent active uptake mechanism dubbed the â&#x20AC;&#x153;System Glyâ&#x20AC;?.14 This system was first described in avian erythrocytes as a highaffinity transporter selective for glycine and sarcosine.15 Subsequently, it has been shown that System Gly is present in the reticulocytes of all vertebrates analyzed. In some mammalian species its activity was shown to be lost during erythrocyte development to mature red blood cells.16 With a considerable delay, the two genes (GLYT1 and GLYT2) encoding the glycine carrier system were cloned17 and shown to possess 12 transmembrane domains with amino and carboxy-terminal ends intracellularly oriented.18 Although the System Gly was first described in erythroid cells, it has been extensively studied primarily in the central nervous system. Glycine and D-serine are required as co-agonists of NMDA-type glutamate receptors located at excitatory glutamatergic synapses.19 The levels of glycine within the synapses are controlled by haematologica | 2017; 102(8)

Figure 1. GlyT1 expression increases during erythroid differentiation in vitro and in vivo. Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of (A) GlyT1 and (B) globin mRNA expression in primary erythroid cells kept either in a non-differentiating state (0 h) or induced to differentiate for 24 to 72 h in vitro. (C) Western blot analysis of GlyT1 protein expression in primary erythroid cells differentiated for 0 h, 24 h and 48 h. (D) qRT-PCR and (E) western blot analysis of GlyT1 expression in Ter119- and Ter119+ cells isolated from mouse bone marrow. qRT-PCR-based results are presented as fold-change relative to the 0 h sample. Statistical analysis was done using one way analysis of variance followed by a Bonferroni multiple comparison test (n=3). The P-value refers always to the value of the previous time point; *P<0.05; **P<0.01; ***P<0.001.

glycine transporters20 that directly affect glycine homeostasis in the central nervous system and, consequently, control the balance between neuronal inhibition and excitation in several neural circuits.21 Inhibition of glycine reuptake by glycine transporters is the rationale for the design of a third generation of anti-schizophrenia drugs.22 It was initially suggested that glycine transport systems provide glycine mainly for the synthesis of glutathione in immature red blood cells,14 whereas glycineâ&#x20AC;&#x2122;s crucial importance as one of the substrates for heme synthesis was inconceivably disregarded. Indeed, much earlier Shemin and co-workers had demonstrated that in vitro incubation of human reticulocytes with heavy nitrogen [15N]glycine resulted in the formation of significant amounts of [15N]heme.23 This seminal work was the first to demonstrate that reticulocytes internalize glycine and utilize its backbone for heme biosynthesis. Somewhat surprisingly, the role of glycine transporters in supplying glycine for hemoglobin synthesis in developing red blood cells has never been assessed. We hypothesized that the System Gly (more specifically glycine transporter 1, GlyT1), which is responsible for most glycine uptake in red blood cells,14 actively transports and supplies glycine for heme biosynthesis. 1315

D. Garcia-Santos et al.

In the present study, we demonstrate that GlyT1 is expressed in fetal liver cells, and its expression increases during erythropoietin-mediated induction of erythroid differentiation. We also show that, compared to wild-type cells, GlyT1 knockout (GlyT1-/-) fetal liver cells internalize less [2-l4C]glycine and incorporate less 59Fe from 59Fe-transferrin into heme. Moreover, GlyT1-/- fetal liver cells have significantly lower hemoglobin levels as compared to wild-type cells. Finally, newborn mice with a homozygous GlyT1 defect have hypochromic microcytic anemia, whereas adult mice heterozygous for a GlyT1 defect exhibit only mild anemia. Our data show that glycine uptake by erythroblasts is limiting for heme synthesis and thus required for normal erythroid development. These observations are of particular importance in light of the increasing number of clinical trials employing GlyT1 inhibitors for the treatment of schizophrenia.22

Methods Study approval All the animal studies performed in this work were approved by

the Lady Davis Institute animal care committee following the guidelines of the Canadian Council on Animal Care.

Animals 129SvEv mice haplodeficient for GlyT1 (GlyT1+/-), generated in Professor Joseph Coyleâ&#x20AC;&#x2122;s laboratory,24 were maintained and bred in the animal facility at the Lady Davis Institute for Medical Research. All GlyT1 knockout mice die during the first postnatal day;24,25 viable neonates were sacrificed within 1-2 h after birth. As indicated, pregnant animals were sacrificed at day E12.5, 14.5 and 16.5 post coitum to determine the embryonic phenotype. For studies on erythroid cells in vitro, fetal livers were isolated between E12.5 and E13.5. Genotypes were determined by polymerase chain reaction analysis of tail DNA using a previously described protocol.24

Culture of primary mouse erythroblasts Erythroid cells were isolated and cultured as previously described.26 Briefly, cells were grown from fetal livers obtained from E12.5-13.5. Primary fetal liver cells were kept either in an undifferentiated state (0 h) or induced to terminal differentiation (24, 48 or 72 h). The composition of the medium used to maintain the cells in such states is described in further detail in the Online Supplementary Information.

Figure 2. [2-14C]glycine uptake and incorporation into heme is disrupted in GlyT1-depleted erythroid cells. Erythroid cells isolated from GlyT1+/+, GlyT1+/- and GlyT-/embryos were either kept in a non-differentiating state (0 h) or induced to differentiate for 24 and 48 h. (A) Giemsa staining of GlyT1+/+, GlyT1+/- and GlyT-/- undifferentiated (0 h) and differentiated (24 h and 48 h) fetal liver cells (B) Total uptake of [2-14C]glycine and (C) [2-14C]glycine incorporation into heme. [2-14C]glycine values are presented relative to the wild-type cells at indicated time intervals. Statistical analysis was done using one way analysis of variance followed by a Bonferroni multiple comparison test (n=3). The P value always refers to the value of the previous time point; *P<0.05; **P<0.01; ***P<0.001.

1316

haematologica | 2017; 102(8)

GlyT1 is necessary for hemoglobinization

Hemolgobin assay Hemoglobin measurements were performed as described previously.27 Briefly, 2-5x105 cells were transferred in triplicate into a 96well microtiter plate with conical bottomed wells and washed with 100 mL phosphate-buffered saline. Cells were lysed in 50 mL H2O, following which 125 mL of dye solution [0.5 mg/mL ophenylene-diamine-dihydrochloride (Sigma-Aldrich, St. Louis, MO, USA), 50 mM citric acid, 0.1 M Na2HPO4 and 1 mL/mL of 30% H2O2) were added. The reaction was stopped within 3-5 min with 25 mL 8 N H2SO4 and the optical density of samples at 492 nm was determined. The results were normalized to total cell volume.

Glycine uptake and its incorporation into heme

Cells were incubated with 4-5 mCi of [2-l4C]glycine (100 mCi/mL; NEN, Boston, MA, USA) for 1 h, following which cells were washed three times with 1 mL phosphate-buffered saline. Heme was extracted as previously described.28 A detailed protocol is described in the Online Supplementary Information.

Iron uptake from iron-transferrin Radioactive iron-transferrin (59Fe2-Tf) was made from 59FeCl3 (PerkinElmer, Santa Clara, CA, USA; 2 mCi) as described previously.28 For measurements of total cellular 59Fe uptake, cells were incubated with 2 mM (final concentration; saturating) 59Fe2-Tf for 3 h, following which samples were washed twice with ice-cold phosphate-buffered saline and collected by centrifugation (200 x g for 5 min at 4°C). The 59Fe in heme and non-heme fractions was measured as described in the Online Supplementary Methods.

Statistical analysis Statistical analysis was done using SPSS v15.0 (IBM Software, Markham, ON, Canada) software. Data were evaluated using one-way analysis of variance (ANOVA) and the Student t test. Multiple comparisons were performed using Bonferroni and Dunnett tests. Error bars of graphs represent standard deviations (n=3).

Results GlyT1 expression increases during erythroid differentiation To analyze the expression of GlyT1 mRNA during erythroid differentiation, we used primary mouse erythroid cells derived from fetal livers.26 We expanded erythroid progenitors isolated from mouse fetal livers and kept them either in a non-differentiated state (0 h) or, after expanding them, stimulated their terminal differentiation using high concentrations of erythropoietin (24-72 h). At the indicated time intervals, mRNA levels of GlyT1 and β-globin were measured using quantitative real-time polymerase

chain reaction analysis. The results shown in Figure 1A demonstrate a strong increase in GlyT1 mRNA within the first 48 h, which matches the maximal rate of hemoglobin synthesis in fetal liver cells at this stage of differentiation.26 In the following 24 h, when hemoglobinization slows down,26 GlyT1 mRNA expression declines to its initial levels. This expression pattern of GlyT1 mRNA closely resembles the expression profile obtained for β-globin mRNA (Figure 1B), suggesting a possible link between GlyT1 and globin synthesis in differentiating fetal liver cells. GlyT1 protein levels were also increased in fetal liver cells after 48 h of differentiation (Figure 1C). We then analyzed GlyT1 expression levels in vivo using the erythroid surface marker Ter11929 to separate erythroid cells (Ter119+) from non-erythroid cells (Ter119-) obtained from mouse bone marrow. Quantitative real-time polymerase chain reaction analysis revealed a 3-fold increase in GlyT1 mRNA expression in the Ter119+ fraction compared to the Ter119- one (Figure 1D). In agreement with that, GlyT1 protein levels also increased in Ter119+ cells (Figure 1E). Of interest, previous studies determined that the components of System Gly (currently known as GlyT1 and GlyT2) are localized in red blood cell membranes.30,31 Our finding of augmented GlyT1 expression, both in vitro and in vivo, suggests that this transporter is likely required for erythroid differentiation.

Glycine uptake and its incorporation into heme are decreased in GlyT1-depleted erythroid cells The increase in GlyT1 expression depicted in Figure 1 suggests that an increase in the transporter’s levels in erythroid cells is needed to satisfy cellular demands for glycine during erythroid differentiation. To address this issue, we first determined that primary erythroid cells isolated from wild-type (GlyT1+/+), haplodeficient (GlyT1+/-) and knockout (GlyT1-/-) fetal livers showed similar morphology in their undifferentiated (0 h) and differentiated (48 h) state, as demonstrated by Giemsa staining (Figure 2A). The total uptake of [2-l4C]glycine was significantly reduced in undifferentiated (0 h) and differentiated (24 h and 48 h) erythroid cells derived from both haplodeficient and knockout animals when compared to that in wildtype cells (Figure 2B). Non-differentiated GlyT1+/- and GlyT1-/- cells reached only 80±4.9% and 55±8.3%, respectively, of the amount of glycine taken up by GlyT1+/+ erythroblasts. At 48 h of differentiation, these values further decreased to 60±1.9% and 15±1.3%, respectively. Similarly, [2-l4C]glycine incorporation into heme was significantly decreased in both GlyT1-/- and GlyT1+/- fetal liver cells (Figure 2B).

Table 1. GlyT1-/- and GlyT+/- newborn mice exhibit microcytic hypochromic anemia.

Genotypes +/+

GlyT1 GlyT1+/GlyT1-/-

RBC (1012/L)

Hemoglobin (g/dL)

Hematocrit (%)

MCV (fL)

MCH (pg)

3.86±0.09 3.75±0.06 3.80±0.12

14.24±0.38 13.16±0.19* 10.76±0.28***

42.34±0.87 39.18±0.55* 32.72±0.83***

109.73±1.21 104.86±1.09* 86.72±0.95***

36.85±0.48 35.20±0.39* 28.46±0.27***

Values are mean±standard error. Statistical analysis was done using one-way analysis of variance followed by a Dunnett multiple comparison test; *P<0.05; ***P<0.001.We determined red blood cell indices on GlyT1+/+ (n=20), GlyT1+/- (n=36) and GlyT-/- (n=25) newborn animals by automated analysis. RBC, red blood cell number; MCV: mean corpuscular volume; MCH, mean corpuscular hemoglobin.

haematologica | 2017; 102(8)

1317

D. Garcia-Santos et al.

Figure 3. 59Fe uptake from 59Fe-Tf and 59Fe incorporation into heme, as well as hemoglobin levels, are significantly reduced in GlyT1-/- erythroid cells. Erythroid cells isolated from GlyT1+/+, GlyT1+/- and GlyT-/- embryos were either kept in a non-differentiating state (0 h) or induced to differentiate for 24 and 48 h. (A) Total uptake of 59 Fe from 59Fe-Tf, (B) 59Fe incorporation into heme and (C) hemoglobin levels. 59Fe and hemoglobin values are presented relative to the wild-type cells at indicated time intervals. Statistical analysis was done using one way analysis of variance followed by a Bonferroni multiple comparison test (n=3); *P<0.05; **P<0.01; ***P<0.001.

Iron uptake and its incorporation into heme is significantly reduced in differentiating GlyT1-/erythroblasts Erythroid cells import iron exclusively via the transferrin (Tf)-transferrin receptor (TfR) pathway.1 Hence, we used 59 Fe2-Tf to investigate whether the restriction in glycine supply, shown in Figure 2B, would also affect the uptake of 59Fe and its incorporation into heme. As depicted in Figure 3, there were significant reductions in both 59Fe uptake (Figure 3A) and its incorporation into heme (Figure 3B) in GlyT1-/- and GlyT1+/- cells when compared to wildtype controls. After 48 h of differentiation, GlyT1+/+ fetal liver cells accumulated massive amounts of hemoglobin (a more than 4-fold increase in relation to the 0 h time point), which was expected based on earlier reports.26,27,32 Importantly, at 48 h, GlyT1-/- fetal liver cells had significantly lower hemoglobin levels compared to their wildtype counterparts (Figure 3C). In agreement with these results, we observed that cell surface TfR1 levels decreased at 24 h and 48 h in erythropoietin-treated GlyT1-/- and GlyT+/- cells compared to GlyT1+/+cells (Figure 4). Of interest, in non-differentiated cells (0 h), the lack or reduction of GlyT1 expression did not significantly affect total 59Fe uptake (Figure 3A), 59Fe incorporation into heme 1318

(Figure 3B) or hemoglobin levels (Figure 3C), despite the reduction in the total glycine uptake by GlyT1-/- and GlyT1-/+ cells at 24 and 48 h (Figure 2). Additionally, in differentiated erythroid cells (48 h), both alleles of GlyT1 were necessary to achieve a rate of heme biosynthesis comparable to that seen in wild-type cells.

GlyT1-/- newborn mice have microcytic hypochromic anemia We then investigated whether GlyT1 deficiency in vivo results in a phenotype similar to that seen in our in vitro model. Congruent with earlier reports,24,25 mice lacking the gene encoding GlyT1 appear normal at birth, but, 6 to 14 h later, start displaying motor sensory deficits and severe respiratory malfunction, which ultimately leads to death. We sacrificed the newborn pups shortly after birth and observed that GlyT1-deficient mice (GlyT1-/-) were anemic with hemoglobin levels 25% lower than their wild-type (GlyT1+/+) siblings (Table 1). Moreover, the hematocrit was decreased from 42.3 Âą 0.9% in GlyT1+/+ to 32.7 Âą 0.8% in GlyT1-/- animals. This reduction was most likely due to a smaller red cell volume since there was no significant reduction in the total red blood cell counts (Table 1). These differences are further corroborated by the observation that GlyT1 knockout embryos isolated between E12 and haematologica | 2017; 102(8)

GlyT1 is necessary for hemoglobinization

Figure 4. Cell surface transferrin receptor levels are reduced in differentiating erythroid cells depleted of GlyT1. Erythroid cells were isolated from GlyT1+/+, GlyT1+/- and GlyT1-/- embryos and either kept in a non-differentiated state (0 h) or induced to differentiate for 24 to 48 h. TfR surface expression was analyzed by flow cytometry. (A) Representative TfR (CD71) fluorescence histograms obtained from cells of each genotype at indicated time points. (B) TfR quantification based on mean fluoresce intensity displayed as values relative to those in 0 h wild-type cells. Statistical analysis was done using one way analysis of variance followed by a Dunnett multiple comparison test in relation to the respective GlyT+/+ sample (n=6); *P<0.05; **P<0.01;***P<0.001.

Table 2. Adult GlyT+/- mice exhibit mild microcytic hypochromic anemia.

Genotypes GlyT1+/+ GlyT1+/-

RBC (x1012/L)

Hemoglobin (g/dL)

Hematocrit (%)

MCV (fL)

MCH (pg)

11.04±0.13 11.48±0.10*

16.59±0.19 16.49±0.14

54.09±0.74 54.31±0.43*

48.89±0.30 47.42±0.31**

15.04±0.09 14.38±0.11***

Values are mean±standard error. Statistical analysis was done using one way analysis of variance followed by a Dunnett multiple comparison test; *P<0.05; ***P<0.001.We determined red blood cell indices on GlyT1+/+ (n=19) and GlyT1+/- (n=26) adult animals (3- to 4 month old mice) by automated analysis. RBC: red blood cell number; MCV: mean corpuscular volume; MCH: mean corpuscular hemoglobin.

E16.5 were paler than their wild-type siblings (Figure 5). Moreover, GlyT-/- animals had a significantly increased percentage of reticulocytes in the blood as compared to GlyT1+/+ and GlyT1+/- animals (Online Supplementary Figure S1A,B). It is noteworthy that compared to GlyT1+/+ animals, haplodeficient (GlyT1+/-) pups also showed a mild anemia, with significantly lower hemoglobin levels and decreased hematocrit, mean corpuscular volume and mean corpuscular hemoglobin (Table 1). Collectively, these data clearly demonstrate that the lack or deficiency of GlyT1 leads to hypochromic microcytic anemia in newborn mice.

Adult mice heterozygous for GlyT1 exhibit mild anemia Disruption of one GlyT1 allele is sufficient to cause a significant reduction in glycine uptake and heme synthesis during erythroid differentiation in vitro (Figures 2 and 3). This is in line with the in vivo data showing that haplodeficient GlyT1 newborn pups were anemic (Table 1). In order to evaluate whether adult GlyT1 haplodeficient mice also show any signs of anemia, we analyzed blood from 3- to 4-month old mice. As shown in Table 2, both haematologica | 2017; 102(8)

mean corpuscular volume and mean corpuscular hemoglobin were slightly reduced in heterozygous animals, whereas red blood cell counts were slightly increased compared to those in wild-type mice. No significant differences in hemoglobin levels or hematocrit were observed between these two genotypes (Table 2).

Discussion The requirement of the α-carbon of glycine for heme synthesis in vertebrates was established by the groups of Shemin and Neuberger who discovered that one carbon atom and the nitrogen atom of each pyrrole ring of heme is derived from the α-atom and the associated nitrogen atom of glycine33 and that each of the four methylene bridge carbon atoms of heme is derived from the α-carbon of glycine.33 However, the first suggestion that extracellular glycine is used in heme synthesis came from Shemin and co-workers who demonstrated that reticulocytes incubated with [15N]glycine synthesized significant amounts of [15N]heme.23 1319

D. Garcia-Santos et al.

Our first clue that GlyT1 may be required for efficient hemoglobin synthesis was the matching pattern of GlyT1 mRNA expression with that of β-globin during erythroid differentiation (Figure 1A,B). A similar expression profile was obtained in an erythroid global gene expression analysis that showed increased GlyT1 expression at the polychromatic erythroblast stage in mouse primitive, fetal and adult erythroid cells.34 The same analysis did not detect expression of GlyT2, another member of the System Gly, in erythroblasts at any stage of their development. We did not, therefore, examine GlyT2 expression in our study. Our GlyT1 in vitro expression data were supported by an in vivo experiment in which we also found the increase of GlyT1 mRNA and protein levels in mouse Ter119+ bone marrow cells (Figure 1D,E). Our data show for the first time that the absence of GlyT1 leads to a significant reduction in glycine uptake and its incorporation into heme in erythroid cells (Figure 2B,C), associated with diminished 59Fe incorporation into heme (Figure 3B). Decreases in these values were more prominent in differentiated fetal liver cells from Glyt1-/animals, but a lesser reduction in these parameters was also seen in differentiating fetal liver cells from mice haplodeficient for the GlyT1 gene, in particular at 48 h of differentiation. Reduced hemoglobin levels, particularly in differentiating fetal liver cells from Glyt1-/- mice (Figure 3C), can be explained by a decrease in heme synthesis in these cells. However, haplodeficient GlyT1 fetal liver cells have hemoglobin levels similar to those of wild-type controls, which suggest that one functional GlyT1 allele is sufficient to fully support hemoglobinization. Of interest, sarcosine, a competitive inhibitor of GlyT1,35 also decreases 59 Fe incorporation into heme from 59Fe-Tf (Online Supplementary Figure S2). These observations suggest that, similarly to what happens in other cell types,5,9 endogenous glycine sources are insufficient to meet erythroid cell demands for optimal hemoglobinization. However, in undifferentiated erythroid cells, de novo glycine synthesis3 and/or glycine transported by systems other than System Gly14 (band 3, system L, system ASC and simple passive diffusion through the membrane) seem to be supplying sufficient levels of glycine to support housekeeping heme synthesis. It was demonstrated that System Gly (GlyT1 and GlyT2) accounts for 42% of the total glycine uptake in erythroid cells.14 Band 3 and system L account for 16% and 15% of glycine uptake, respectively, while system ASC accounts for 11% of uptake.14 The remaining glycine uptake occurs by passive diffusion.14 King and Gunn also observed System Gly-independent glycine transport via the band 3 system in erythroid cells.36 The total glycine uptake in erythroid cells is not, therefore, entirely via GlyT1. The existence of alternative glycine transport systems (apart from GlyT1) explains glycine transport to erythroid cells for heme synthesis-unrelated pathways. It seems likely that GlyT1-independent glycine uptake could be responsible for at least some hemoglobin synthesized in GlyT1-/- erythroid cells (Figure 3C) and allows in utero survival of GlyT1-/- fetuses (Table 1). Nevertheless, GlyT1 is involved in the transport of glycine, which is specifically committed for heme synthesis and hemoglobin production in erythroid cells. GlyT1 knockdown caused a decrease in total 59Fe uptake by fetal liver cells after 24 and 48 h of differentiation (Figure 3A). At the same time points, we observed a reduc1320

tion in cell surface TfR levels (Figure 4) in these cells. As erythroid cells can take up iron only via the transferrin/TfR pathway,1 it can be expected that the reduction in TfR levels will lead to a decrease in iron uptake and its incorporation into heme. Additionally, heme enhances TfR expression at the transcriptional level in erythroid cells.27,37 The reduction in TfR levels observed under glycine restriction is, therefore, most likely due to a decrease in cellular heme levels in fetal liver cells from GlyT1-/- mice. Our findings that the ablation of GlyT1 causes a decrease in heme synthesis in erythroblasts and hypochromic microcytic anemia in newborn mice suggest that the transport of glycine to erythroid cells represents a limiting step that controls the rate of heme synthesis. A decrease in heme levels is also well known to cause a reduction in globin synthesis that occurs by welldescribed mechanisms involving both transcription38 and translation.39 This proposal seems to be in conflict with the idea that in erythroid cells iron acquisition from transferrin is the rate-limiting step in heme synthesis.28 However, all things considered, these two ideas may not be mutually exclusive. The uptake of iron by immature erythroid cells is controlled by a negative feedback mechanism in which non-hemoglobin-bound heme (“uncommitted heme”) inhibits iron acquisition from transferrin.1,40 Importantly in this context, heme was also shown to inhibit glycine uptake by erythroid cells.28 The exact molecular mechanisms by which heme affects iron or glycine trafficking in erythroid cells are unknown, but it is tempting to speculate that heme interacts with a yet-to-be-identified entity involved in both iron and glycine conveyance or the regulation of these processes. Nothing is known about a potential relationship between GlyT1-mediated transport of glycine through the plasma membrane and the import of this amino acid into mitochondria. Nevertheless, some important clues are emerging from the studies of a subtype of hereditary sideroblastic anemias, all of which are characterized by the presence of mitochondrial iron deposits in erythroblasts. Most cases of hereditary X-linked sideroblastic anemia are caused by mutations in the gene encoding erythroid-specific ALAS.41 However, a certain proportion of patients with hereditary sideroblastic anemia exhibit autosomal recessive inheritance. Guernsey et al.42 described that at least some such patients have a defect in the gene encoding the erythroid-specific mitochondrial carrier protein, SLC25A38. They demonstrated that this transporter is important for the biosynthesis of heme in eukaryotes and conjectured that this protein may translocate glycine into mitochondria.42 Fernández-Murray et al. have recently provided convincing evidence that Hem25, the yeast homolog of human SLC25A38, provides glycine for ALAS.43 They also demonstrated that expression of human SLC25A38 in yeast cells in which the HEM25 gene was deleted restored heme to a normal level, leading the authors to conclude that Hem25 and SLC25A38 are mitochondrial glycine importers. Taking together these findings and our data, it is tempting to speculate that erythroid cells have a unique pathway to convey glycine from the extracellular milieu to mitochondria, which would involve both GlyT1 and SLC25A38. Such a mechanism would optimize glycine delivery for heme synthesis in developing erythroid cells. The reduced heme and hemoglobin synthesis occurring in GlyT1-/- fetal liver cells in vitro is recapitulated in vivo. haematologica | 2017; 102(8)

GlyT1 is necessary for hemoglobinization

Figure 5. GlyT1-/- embryos are paler than their wild-type and heterozygous counterparts. Embryos were isolated at embryonic days 12.5 (E12.5), 14.5 (E14.5) and 16.5 (E16.5). Genotype was determined by polymerase chain reaction analysis as described in the Methods section.

GlyT1-/- newborn mice display hypochromic microcytic anemia, as documented by significantly lower hemoglobin levels and decreased hematocrit, mean corpuscular volume and mean corpuscular hemoglobin (Table 1). These changes clearly indicate impairment in hemoglobin synthesis. It has already been described that wild-type mice show mean cell volumes between 100 to 110 fL at birth, and these values drop during the first month of life, stabilizing between 48 and 63 fL.44 On the other hand, GlyT1+/- adult mice have only a 5% reduction in mean corpuscular volume and mean corpuscular hemoglobin, when compared to their wild-type counterparts (Table 2). Therefore, similarly to what was observed in vitro (Figure 3C), one functional GlyT1 allele is almost sufficient to support normal hemoglobinization in vivo. Glycine is an inhibitory neurotransmitter that acts by directly activating ionotropic glycine receptors, which are predominantly found at inhibitory synapses in the spinal cord, brainstem and retina.45 Glycine transporters control haematologica | 2017; 102(8)

the levels of glycine at the inhibitory and excitatory synapses,20 and GlyT1-mediated glycine reuptake facilitates the activation of NMDA receptors.21 Thus, clinical conditions associated with deficient NMDA receptor signaling, such as schizophrenia, may potentially be treated by inhibition of glycine reuptake in the synapses.21 The aim of therapeutic GlyT1 inhibitors is to restore glycine homeostasis in specific regions of the central nervous system and alleviate the clinical symptoms of schizophrenia. This new approach represents an alternative to current therapies, which are based primarily on dopamine receptor blockage.46 A wide variety of GlyT1 inhibitors have been developed and tested in different pre-clinical models.47,48 The GlyT1 inhibitor bitopertin (RG-1678; RO-4917838) is one of these inhibitors. During the preparation of our study for publication, bitopertin was reported to cause anemia in rats.49 Similarly to what we observed in GlyT1-/- newborn mice (Table 1), adult rats treated with bitopertin showed 1321

D. Garcia-Santos et al.

decreases in mean corpuscular volume, mean corpuscular hemoglobin and hemoglobin levels, when compared to those of controls.49 Additionally, as expected, bitopertin administration led to an increase in erythropoietin levels in rats.49 Our study, for the first time, presents compelling evidence that glycine uptake by erythroblasts represents a limiting step in heme and, consequently, hemoglobin production (Figures 2 and 3). Concerns regarding the side effects that could be associated with long-term therapy with GlyT1 inhibitors are focused on damage to intestinal epithelial cells50 and adverse skin reactions.51 Importantly, our data show that GlyT1 suppression causes impairment of hemoglobin synthesis in vitro and hypochromic microcytic anemia in vivo (Tables 1 and 2). Long-term inhibition of glycine uptake by erythroblasts in humans could lead to the development of sideroblastic anemias due to defective heme biosynthesis, which could consequently cause hepatic and systemic iron overload.41 Umbritch et al. performed a randomized, double-blinded, placebo-controlled bitopertin trial involving 323 patients with schizophrenia.52 Their study suggested a

References 1. Ponka P. Tissue-specific regulation of iron metabolism and heme synthesis: distinct control mechanisms in erythroid cells. Blood. 1997;89(1):1-25. 2. Wittenberg J, Shemin D. The location in protoporphyrin of the carbon atoms derived from the alpha-carbon atom of glycine. J Biol Chem. 1950;185(1):103-116. 3. Wang W, Wu Z, Dai Z, et al. Glycine metabolism in animals and humans: implications for nutrition and health. Amino Acids. 2013;45(3):463-477. 4. Wu G. Amino acids: metabolism, functions, and nutrition. Amino Acids. 2009;37(1):1-17. 5. Wu G. Functional amino acids in growth, reproduction, and health. Adv Nutr. 2010;1(1):31-37. 6. Shemin D. Some aspects of the biosynthesis of amino acids. Cold Spring Harb Symp Quant Biol. 1950;14:161-167. 7. Soloway S, Stetten D Jr. The metabolism of choline and its conversion to glycine in the rat. J Biol Chem. 1953;204(1):207-214. 8. Dale RA. Catabolism of threonine in mammals by coupling of L-threonine 3-dehydrogenase with 2-amino-3-oxobutyrateCoA ligase. Biochim Biophys Acta. 1978;544(3):496-503. 9. Melendez-Hevia E, De Paz-Lugo P, Cornish-Bowden A, Cardenas ML. A weak link in metabolism: the metabolic capacity for glycine biosynthesis does not satisfy the need for collagen synthesis. J Biosci. 2009;34(6):853-872. 10. Zafra F, Gimenez C. Characteristics and adaptive regulation of glycine transport in cultured glial cells. Biochem J. 1989;258(2):403-408. 11. Christie GR, Ford D, Howard A, Clark MA, Hirst BH. Glycine supply to human enterocytes mediated by high-affinity basolateral GLYT1. Gastroenterology. 2001;120(2): 439-448. 12. Mailliard ME, Kilberg MS. Sodium-depen-

1322

13.

14. 15.

16.

17.

18.

19. 20.

21.

22.

23.

modest but clinically relevant effect of bitopertin in improving the negative symptoms associated with schizophrenia. However, at all bitopertin doses tested, a significant number of patients showed a more than 2 g/dL reduction in hemoglobin levels.52 In summary, our data show for the first time that the transport of glycine to erythroid cells is a limiting step in their synthesis of heme. We have also demonstrated that extracellular restriction of glycine due to defective GlyT1 leads to a decrease in hemoglobin levels both in vivo and in vitro. Collectively, our data reveal a uniqueness of erythroid cells in terms of their high demand for extracellular substrates for heme biosynthesis and also raise the question of how glycine and iron transport are coordinated in these cells. Acknowledgments The authors thank Dr. Rhoda Blostein for helpful discussions Funding The research of PP was funded by grants from the Canadian Institutes of Health Research (MOP-14100 and MOP-126064).

dent neutral amino acid transport by human liver plasma membrane vesicles. J Biol Chem. 1990;265(24):14321-14326. Van Winkle LJ, Haghighat N, Campione AL, Gorman JM. Glycine transport in mouse eggs and preimplantation conceptuses. Biochim Biophys Acta. 1988;941(2): 241-256. Ellory JC, Jones SE, Young JD. Glycine transport in human erythrocytes. J Physiol. 1981;320:403-422. Vidaver GA, Shepherd SL. Transport of glycine by hemolyzed and restored pigeon red blood cells. Symmetry properties, trans effects of sodium ion and glycine, and their description by a single rate equation. J Biol Chem. 1968;243(23):6140-6150. Blostein R, Drapeau P, Benderoff S, Weigensberg AM. Changes in Na+-ATPase and Na,K-pump during maturation of sheep reticulocytes. Can J Biochem Cell Biol. 1983;61(1):23-28. Guastella J, Brecha N, Weigmann C, Lester HA, Davidson N. Cloning, expression, and localization of a rat brain high-affinity glycine transporter. Proc Natl Acad Sci USA. 1992;89(15):7189-7193. Olivares L, Aragon C, Gimenez C, Zafra F. Analysis of the transmembrane topology of the glycine transporter GLYT1. J Biol Chem. 1997;272(2):1211-1217. Wolosker H. NMDA receptor regulation by D-serine: new findings and perspectives. Mol Neurobiol. 2007;36(2):152-164. Eulenburg V, Armsen W, Betz H, Gomeza J. Glycine transporters: essential regulators of neurotransmission. Trends Biochem Sci. 2005;30(6):325-333. Harvey RJ, Yee BK. Glycine transporters as novel therapeutic targets in schizophrenia, alcohol dependence and pain. Nat Rev Drug Discov. 2013;12(11):866-885. Kristensen AS, Andersen J, Jorgensen TN, et al. SLC6 neurotransmitter transporters: structure, function, and regulation. Pharmacol Rev. 2011;63(3):585-640. Shemin D, London IM, Rittenberg D. The

24.

25.

26.

27.

28.

29.

30. 31.

32.

33.

in vitro synthesis of heme from glycine by the nucleated red blood cell. J Biol Chem. 1948;173(2):799. Tsai G, Ralph-Williams RJ, Martina M, et al. Gene knockout of glycine transporter 1: characterization of the behavioral phenotype. Proc Natl Acad Sci USA. 2004;101 (22):8485-8490. Gomeza J, Hulsmann S, Ohno K, et al. Inactivation of the glycine transporter 1 gene discloses vital role of glial glycine uptake in glycinergic inhibition. Neuron. 2003;40(4):785-796. Schranzhofer M, Schifrer M, Cabrera JA, et al. Remodeling the regulation of iron metabolism during erythroid differentiation to ensure efficient heme biosynthesis. Blood. 2006;107(10):4159-4167. Garcia-Santos D, Schranzhofer M, Horvathova M, et al. Heme oxygenase 1 is expressed in murine erythroid cells where it controls the level of regulatory heme. Blood. 2014;123(14):2269-2277. Ponka P, Schulman HM. Regulation of heme synthesis in erythroid cells: hemin inhibits transferrin iron utilization but not protoporphyrin synthesis. Blood. 1985;65(4):850-857. Kina T, Ikuta K, Takayama E, et al. The monoclonal antibody TER-119 recognizes a molecule associated with glycophorin A and specifically marks the late stages of murine erythroid lineage. Br J Haematol. 2000;109(2):280-287. Tunnicliff G. Membrane glycine transport proteins. J Biomed Sci. 2003;10(1):30-36. Tunnicliff G. Amino acid transport by human erythrocyte membranes. Comp Biochem Physiol Comp Physiol. 1994;108 (4):471-478. Dolznig H, Boulme F, Stangl K, et al. Establishment of normal, terminally differentiating mouse erythroid progenitors: molecular characterization by cDNA arrays. FASEB J. 2001;15(8):1442-1444. Wittenberg J, Shemin D. The utilization of glycine for the biosynthesis of both types

haematologica | 2017; 102(8)

GlyT1 is necessary for hemoglobinization

34. 35.

36.

37.

38.

39.

40.

of pyrroles in protoporphyrin. J Biol Chem. 1949;178(1):47-51. Kingsley PD, Greenfest-Allen E, Frame JM, et al. Ontogeny of erythroid gene expression. Blood. 2013;121(6):e5-e13. Zhang HX, Lyons-Warren A, Thio LL. The glycine transport inhibitor sarcosine is an inhibitory glycine receptor agonist. Neuropharmacology. 2009;57(5-6):551-555. King PA, Gunn RB. Glycine transport by human red blood cells and ghosts: evidence for glycine anion and proton cotransport by band 3. Am J Physiol. 1991;261(5 Pt 1):C814-821. Garcia dos Santos D, Schranzhofer M, Chun NL, Hamdi A, Ponka P. Transcriptional induction of transferrin receptors by heme in erythroid cells. Blood. 2015;126(23):3352-3352. Katsumura KR, DeVilbiss AW, Pope NJ, Johnson KD, Bresnick EH. Transcriptional mechanisms underlying hemoglobin synthesis. Cold Spring Harb Perspect Med. 2013;3(9):a015412. Chen JJ. Regulation of protein synthesis by the heme-regulated eIF2alpha kinase: relevance to anemias. Blood. 2007;109(7):26932699. Ponka P, Schulman HM, Martinez-Medellin

haematologica | 2017; 102(8)

41.

42.

43.

44. 45.

46.

47.

J. Haem inhibits iron uptake subsequent to endocytosis of transferrin in reticulocytes. Biochem J. 1988;251(1):105-109. Bottomley SS, Fleming MD. Sideroblastic anemia: diagnosis and management. Hematol Oncol Clin North Am. 2014;28(4):653-670. Guernsey DL, Jiang H, Campagna DR, et al. Mutations in mitochondrial carrier family gene SLC25A38 cause nonsyndromic autosomal recessive congenital sideroblastic anemia. Nat Genet. 2009;41(6):651-653. Fernandez-Murray JP, Prykhozhij SV, Dufay JN, et al. Glycine and folate ameliorate models of congenital sideroblastic anemia. PLoS Genet. 2016;12(1):e1005783. Roscoe BJML, Green EL. Biology of the laboratory mouse. New York: Blakiston Division, McGraw-Hill, 1966. Avila A, Vidal PM, Dear TN, et al. Glycine receptor alpha2 subunit activation promotes cortical interneuron migration. Cell Rep. 2013;4(4):738-750. Coyle JT, Balu D, Benneyworth M, Basu A, Roseman A. Beyond the dopamine receptor: novel therapeutic targets for treating schizophrenia. Dialogues Clin Neurosci. 2010;12(3):359-382. Black MD, Varty GB, Arad M, et al.

48.

49.

50.

51.

52.

Procognitive and antipsychotic efficacy of glycine transport 1 inhibitors (GlyT1) in acute and neurodevelopmental models of schizophrenia: latent inhibition studies in the rat. Psychopharmacology (Berl). 2009;202(1-3):385-396. Roberts BM, Shaffer CL, Seymour PA, et al. Glycine transporter inhibition reverses ketamine-induced working memory deficits. Neuroreport. 2010;21(5):390-394. Winter M, Funk J, Korner A, et al. Effects of GlyT1 inhibition on erythropoiesis and iron homeostasis in rats. Exp Hematol. 2016. Howard A, Tahir I, Javed S, et al. Glycine transporter GLYT1 is essential for glycinemediated protection of human intestinal epithelial cells against oxidative damage. J Physiol. 2010;588(Pt 6):995-1009. Fuziwara S, Inoue K, Denda M. NMDAtype glutamate receptor is associated with cutaneous barrier homeostasis. J Invest Dermatol. 2003;120(6):1023-1029. Umbricht D, Alberati D, Martin-Facklam M, et al. Effect of bitopertin, a glycine reuptake inhibitor, on negative symptoms of schizophrenia: a randomized, double-blind, proof-of-concept study. JAMA Psychiatry. 2014;71(6):637-646.

1323

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Hemostasis

Ferrata Storti Foundation

Patient-derived anti-β2GP1 antibodies recognize a peptide motif pattern and not a specific sequence of residues Philippe de Moerloose,1 Céline Fickentscher,1 Françoise Boehlen,1 Jean-Marie Tiercy,2 Egbert K.O. Kruithof1 and Karim J. Brandt1*

Division of Angiology and Hemostasis and 2National Reference Laboratory for Histocompatibility, Transplantation Immunology Unit, Department of Genetic and Laboratory Medicine, University Hospital of Geneva and Faculty of Medicine, Switzerland; *K.J.B is currently affiliated with Division of Cardiology, Department of Internal Medicine, Faculty of Medicine, University of Geneva, Switzerland

Haematologica 2017 Volume 102(8):1324-1332

ABSTRACT

Correspondence: karim.brandt@hcuge.ch

Received: April 10, 2017. Accepted: May 25, 2017. Pre-published: May 26, 2017. doi:10.3324/haematol.2017.170381 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1324 ©2017 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.

1324

ntiphospholipid antibody syndrome is an autoimmune disease characterized by the presence of so-called antiphospholipid antibodies and clinical manifestations such as recurrent thromboembolic or pregnancy complications. Although the main antigenic determinant for antiphospholipid antibodies has been identified as the β-2-glycoprotein 1 (β2GP1), the precise epitope recognized by antiphospholipid antibodies still remains largely unknown. In the study herein, we wanted to identify a sequence in domain I of β2GP1 able to induce the proliferation of CD4+ T cells isolated from antiphospholipid antibody syndrome patients, but not from healthy donors, and to interact with antiphospholipid antibodies. We have characterized a sequence in domain I of β2GP1 that triggers CD4+ T-cell proliferation. A comparison of this sequence with the previously reported binding of antiphospholipid antibodies to discontinuous epitope R39-R43 reveals the presence of an indeterminate motif in β2GP1, in which the polarity determines the characteristics and specificity of antiphospholipid antibodies-interacting motifs. Using point mutations, we characterized the main antiphospholipid antibodies-interacting motif as fffζζFxC, but also established fffζζFxf-related motifs as potential antiphospholipid antibodies epitopes, in which f represents nonpolar residues and ζ polar residues, with charges of the residues not being involved. Of specific importance, these different motifs are present at least once in all antiphospholipid antibodies-related receptors described so far. We have further demonstrated, in vitro, that peptides and domains of β2GP1 containing these motifs were able to interact with antiphospholipid antibodies and inhibit their monocyte activating activity. These results established that the antiphospholipid antibodies-interacting motifs are determined by the polarity, but not by the sequence or charge, of amino acids. These data could also contribute to the future development of more sensitive and specific diagnostic tools for antiphospholipid antibody syndrome determination and potential peptide- or β2GP1 domainbased clinical therapies. Introduction Antiphospholipid syndrome (APS) is described as a common risk factor for recurrent thromboembolic events and/or pregnancy complications resulting from circulating antiphospholipid antibodies (aPLA).1 It is now widely accepted that the plasma phospholipid binding protein β-2-Glycoprotein 1 (β2GP1) is the main antigenic target for aPLA.2 β2GP1 is a protein of 43 kDa composed of 5 short consensus repeat domains called “sushi” domains. Two different conformations exist for β2GP1: a circular plasma conformation and a fishhook conformation.3 Epitopes within domains I and V are involved in maintaining a circular conformation, haematologica | 2017; 102(8)

aPLA recognizes fffζζFxf motif

Table 1. Clinical and laboratory profiles of the 9 patients providing the aPLA. aPLA 1 aPLA 2 aPLA 3 aPLA 4 aPLA 5 aPLA 6 aPLA 7 aPLA 8 aPLA 9

aCL IgG

aβ2GP1 IgG

Clinical manifestations

+ + ND + ND + + + +

> 100 > 60 > 100 30.2 > 60 46 > 60 > 60 > 57

POSITIVE POSITIVE POSITIVE POSITIVE POSITIVE POSITIVE POSITIVE POSITIVE POSITIVE

Arterial thrombosis Fetal loss Venous thrombosis Recurrent fetal loss Arterial thrombosis Arterial thrombosis Fetal loss Fetal loss Thromboembolism

LA: lupus anticoagulant; aCL IgG: anticardiolipin IgG (normal value < 5GPL); aβ2GP1 IgG: anti-β-2-glycoprotein 1 lgG; ND: not done (patient under vitamin K treatment); aPLA: antiphospholipid antibodies; Recurrent fetal loss: >2 fetal losses.

whereas binding of domain V to anionic surfaces induces a fishhook conformation and exposure of a cryptic epitope in domain I.3,4 This cryptic epitope is described as being located around residues 39 and 43; however, Iverson et al. have identified additional residues involved in the recognition of residues by pathogenic anti-β2GP1 antibodies in domain I.5,6 Ioannou et al. have also studied mutations including residues R39 to R43 describing complex, and probably discontinuous, epitopes.7 Their data suggest that the epitope(s) are not “classical” and that several epitopes are present in domain I and could potentially be present elsewhere in β2GP1. Humoral immunophysiology studies of APS and the treatment of APS patients with an anti-CD20 monoclonal antibody (rituximab) have aroused interest in B cells as therapeutic targets. Anti-CD20-treated APS patients have a normal distribution of anti-β2GP1, anti-cardiolipin (aCL) and Lupus anticoagulant (LAC) antibody titers and improved clinical manifestations.8 The isotype of antiβ2GP1 antibody is mainly immunoglobulin G (IgG), suggesting that the production of these antibodies requires antigen-specific CD4+ T helper cells.9 Hattori et al. have shown that β2GP1 induce an in vitro proliferative response of T cells from APS patients. These β2GP1-specific CD4+ T cells are able to induce the production of anti-β2GP1 antibodies by autologous peripheral blood B cells through human leukocyte antigen-D related (HLA-DR) interactions.10,11 The identity of the principal T-cell epitopes on β2GP1 has not been established as of yet. It seems that all 5 β2GP1 domains are able to induce a T-cell proliferative response depending on the APS patient.11 Moreover, analysis of T-cell responses to a β2GP1-derived peptide library have shown that CD4+ T cells are reactive to different peptides independently of HLA.12 In the study herein, we have investigated and identified an immunodominant β2GP1specific CD4+ T-cell epitope using a peptide-associated major histocompatibility complex (pMHC) II tetramerbased assay. We have shown that the immunodominant β2GP1-specific CD4+ T-cell epitope shares a common peptide motif, which is present in the β2GP1 peptide sequence R39-R43. We have further determined that the characteristic fffζζFxC motif, in which f represents nonpolar residues (AVILMFWCPG) and ζ polar residues (YTSHKREDQN), as well as motifs closely related to fffζζFxC are not only present several times in β2GP1 but also in every receptor described for aPLA.13 haematologica | 2017; 102(8)

Methods Ethics statement Buffy coats of blood from healthy donors were provided by the Geneva Hospital Blood Transfusion Center. In accordance with the ethical committee of the Geneva Hospital and with the Declaration of Helsinki, the blood bank obtained informed consent from the donors, who were informed that part of their blood would be used for research purposes.

Patient characteristics All patients had an APS, as defined by the revised Sapporo criteria.2 Control antibodies and peripheral blood mononuclear cells (PBMC) were isolated from the blood plasma of healthy volunteers. The characteristics of the patients used in this study are listed in Table 1.

Recombinant β2GP1 fusion proteins and peptide libraries Recombinant fusion proteins, corresponding to sushi domains I and II, and domains III, IV and V, respectively, of β2GP1 were generated. Briefly, a series of β2GP1 complementary DNA (cDNA) constructs, encoding these domains, were inserted into the vector pcDNA3.1_mycHis_A_A130 between cloning sites BamHI / XhoI. The plasmids were prepared by Life Technology (Carlsbad, CA, USA) and transfected into HEK293 cells. The fusion proteins were purified using nickel resin affinity chromatography (GE Healthcare) and dialysed with Amicon® Ultra (Millipore, Billerica, MA, USA). The concentration was then adjusted to 5mg/ml in peripheral blood smear (PBS). The peptides libraries were generated by Mimotopes (Clayton, VIC, Australia). Lyophilized non water-soluble peptides were reconstituted in 50% dimethyl sulfoxide (DMSO) and 7.5% acetic acid before dilution in PBS. All peptides had 95% purity as assessed by analytical reverse phase high performance liquid chromatography (RP-HPLC). Native β2GP1 was purified from human plasma with 96% purity (Prospecbio, NJ, USA).

T-cell proliferation assays Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation over Lymphoprep™ (Axis-Shield PoC) according to the manufacturer’s instructions. T-cell proliferation was evaluated by 5,6-carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution assays (eBioscience). PBMC were stained with 0.1 mM CFSE (eBioscience), according to the manufacturer’s instructions. Cells were cultured in the presence of anti1325

P. de Moerloose et al. A

Figure 1. Domain I-II and reduced β2GP1 induce proliferation of T cells from APS patients. (A) β2GP1 recombinant domains are purified and analyzed by Western blot. Data are representative of 3 independent experiments. (B-D) Flow cytometry analysis of the proliferation assay performed on PBMC of APS patients in the presence of domains I-II, II-IV, V β2GP1, reduced β2GP1 and tetanus toxin (TTx). CD3-CD28 is positive proliferation control. CD4+CFSElow population is gated on CD3+ cells. Data are mean ± SEM of 9 different donors. Statistical significance was determined by Mann-Whitney U analysis. HC: healthy controls; APS: Antiphospholipid syndrome; WB: western blot; ac-MYC: anti-c-Myc; CFSE: carboxyfluorescein diacetate succinimidyl ester.

gens (10mg/ml) for 10 days. T-cell proliferation was assessed by flow cytometry evaluation of CFSE dilution. Proliferation was expressed as the cell division index (defined as the number of CFSElow T cells cultured with antigen/number of CFSElow T cells without antigen). In all cases, the culture medium consisted of X-VIVO 15 (Lonza, Walkersville, MD, USA).

Statistical analysis When required, the significance of differences between groups was assessed using the nonparametric Mann-Whitney U test. *P≤0.05; **P≤0.005; ***P≤0.0005. All data are represented as mean ± SEM of at least 3 independent experiments.

Results Flow cytometry Single-cell suspensions stained with CFSE (eBioscience) were incubated with anti-CD32 (Biolegend) to prevent nonspecific antibody binding, then stained with antibodies against AF488-CD3, AF647-CD4, (BD Biosciences) and major histocompatibility complex (MHC) class II tetramer (Benaroya Research Institute). Staining was assessed with ACCURI C6 flow cytometer (BD Biosciences).

ELISA epitope mapping assay

MaxiSorp™ 96 well plates (Nunc) were coated with 10 mg/ml recombinant domains or peptides of β2GP1 prior to incubation with aPLA or control IgG. Secondary anti-human antibodies conjugated to IR800CW (Rockland) were used. Protein- or peptidebound antibodies were detected and quantified by the Odyssey system (Li-Cor Biosciences). 1326

Identification of β2GP1 domains recognized by T cells

It was previously demonstrated that β2GP1-specific T cells are present in APS patients, although the specific domain(s) and peptide(s) inducing T-cell proliferation were not well defined.10,12 To identify β2GP1 domain-specific T cells in APS patients, PBMCs were isolated from 9 APS patients (Table 1), and 3 healthy controls (HC) were stimulated with native β2GP1, reduced β2GP1 and recombinant β2GP1 domain I-II, domain III-IV, and domain V (Figure 1A). We tested the proliferation of PBMC by the CFSE dilution assay. In contrast with HC, we detected CD4+ T-cell proliferative responses in primary culture to β2GP1 domain I-II and reduced β2GP1 from APS patients (Figure 1B,C). While the cryptic epitope exposed by reduced β2GP1 was recognized by β2GP1 reactive T cells haematologica | 2017; 102(8)

aPLA recognizes fffζζFxf motif

Figure 2. Peptide 1 of β2GP1 carries antigenic determinant for CD4+ T-cell proliferation. (A) Representation of the peptide library of domain I-II. Peptides in red are peptides contained in Pool III. Residues in green represent R39-R43 epitopes. Underlining represents a signal peptide. (B) Table of peptide distribution for proliferation assay. (C) Flow cytometry analysis of proliferation assay performed on PBMC of APS patients in the presence of pool I, II or III. CD3-CD28 shows positive proliferation control. (D) Flow cytometry analysis of proliferation assay performed on the PBMC of APS patients in the presence of p1, p4, p6, p10 and p11. CD3-CD28 is positive proliferation control. CD4+CFSElow population is gated on CD3+ cells. Data are mean ± SEM of 9 different donors. Statistical significance was determined by Mann-Whitney U analysis. (E) Flow cytometry analysis of pMHC class II tetramer staining loaded with p1 on PBMS of APS patients pulsed with recombinant domain I-II. pMHC-OVA (ovalbumin) and pMHC-TTx (Tetanus toxin) are negative controls. Populations are gated on CD3+ cells. HC: healthy controls; APS: antiphospholipid syndrome.

in the normal T-cell repertoire of APS patients, native β2GP1 was unable to significantly stimulate T-cell proliferation (Figure 1C). The proliferation controls performed with tetanus toxoid (TTx) and CD3-CD28-beads, i.e., specific and unspecific T-cell proliferation stimuli, showed that the proliferative ability of T cells from HC and APS were unaltered (Figure 1D). These results suggest that the core sequence(s) of the β2GP1 peptide(s) inducing T-cell proliferation were present in β2GP1 domain I-II.

Identification of a core peptide fragment recognized by T cells

Having identified candidate regions of β2GP1 containing T-cell determinants, we then tested proliferative responses of APS patients against a pool of β2GP1 peptides. For this purpose, we tested the proliferation of PBMC from APS patients and HC to the pools of peptides containing 4 to 5 β2GP1 domain I-II peptides from a library of 13 synthetic overlapping 20-mer peptides preshaematologica | 2017; 102(8)

ent in the 138 amino-acid sequence of β2GP1 domain I-II (Figure 2A,B). By CFSE dilution assay, we identified proliferative responses in primary cultures to β2GP1 domain III peptide pool III (Figure 2C). T-cell proliferations of HC and APC triggered by CD3-CD28 have shown similar responses. β2GP1 domain I-II peptide pool III was composed of 5 peptides consistently insoluble in water with the cryptic character of epitope recognized by β2GP1 reactive T cells. Interestingly, this pool of peptides contained the region targeted by aPLA, i.e., peptide n°6 (p6).14 We then examined the responses to individual β2GP1 peptides contained in pool III. T-cell proliferation response against leader peptide n°1 (p1) was significantly increased (but not against peptides n°4, 6, 10 or 11), while T-cell proliferations of HC showed a similar response to CD3-CD28 (Figure 2D). To confirm the presence of circulating T cells specific for p1, we defined the HLA class II haplotype profiles of APS patients used previously (Table 2). We then used pMHCII-p1 tetramer DRB3* 02:02 staining to detect 1327

P. de Moerloose et al. Table 2. Haplotype profiles of 6 patients from Table 1.

DRB1* Patient 1 Patient 2 Patient 3 Patient 4 Patient 5 Patient 6

04:03/09:01/13:01 03:01/11:04 04:03/01:01/13:01 04:01/13:01

DRB3* 02:02 02:02 01:01 02:02

Haplotype DRB4* 01:03

01:03

DQB1*

DPB1*

03:02/03:05 03:02/06:03 02:01/03:01 03:02/03:05 05:01/06:03 03:02/06:03

04:01/05:01/14:01 03:01/04:02 04:01/09:01 NT 04:01/04:02

Clinical manifestations Pregnancy loss Thrombosis + + + +

+ + + + -

Table 3. Different motifs present in antiphospholipid antibodies (aPLA)-related receptors and proteins.

Motif aPLA-related receptors

Names and References β2GP127 TLR228 ApoER229

Names and References β2GP1 TLR830 ANXA231

Motif aPLA-related receptors

Names and References β2GP1 TLR132 TLR433 TLR632 TLR734 TLR935

Motif aPLA-related receptors

Names and References βGPIbα36

fffζζFxC Localization

Number

Orientation

Domain I Extracellular Extracellular

NH2 => COOH

fffζζFxCf Localization

2 1 2

Number

Orientation

Domain I Extracellular Annexin repeats

COOH => NH2

fffζζFxCf Localization

2 1 1

Number

Orientation

Domain III Extracellular Extracellular Extracellular Extracellular Extracellular

1 1 3 1 2 2

Localization Extracellular

Number 1

fffζζFxCf

specific circulating T cells in β2GP1 domain I-II pulsed T-cell population from APS patients. Although, as expected, no population was observed with negative control pMHC-OVA (tetramer with a peptide of ovalbumin), a pMHC-p1 positive population was strongly present in β2GP1 domain I-II specific pulsed T-cell population (Figure 2E). pMHC-TTx staining was used as a positive control (Figure 2E). These results demonstrated that the immunodominant β2GP1-specific CD4+ T-cells’ epitope in patients with APS is present in a sequence of amino acids corresponding to MISPVLILFSSFLCHVAIAG, which correlates with the signal peptide of β2GP1.

Characterization of common motif shared by p1 and p6 of domain I-II To identify a potential consensus motif between p1 and p6, which was previously reported as containing the aPLA binding discontinuous epitope R39-R43, we aligned their sequences and observed a common motif present around a phenylalanine (Figure 3A). This motif seemed to be determined by intrinsic physical properties (polar and 1328

COOH => NH2

Orientation COOH => NH2

nonpolar) and not by amino acid sequences. To investigate the ability of p1 to interact with patient-derived antiβ2GP1 antibodies and to identify critical residues in the motif, we generated an alanine scanning library of p1 and p6 including the common motif (Figure 3B) to perform an ELISA epitope mapping assay. As shown in Figure 3B, aPLA are able to bind p1 as potently as p6. The mutations I7A, L8A F9A, S11A, F12A and C14A of p1 significantly decreased the binding of aPLA. Similarly, mutations R58A (R39), G59A (G40), G60A (G41), M61A (M42), R62A (R43), F64A and C66A were involved in the binding of aPLA to p6 (Figure 3B) (numbering corresponds to the numbering in the mature β2GP1 protein, i.e., without the signal peptide). aPLA interacted similarly with both peptides, although the sequences of amino acids present in the motif were not the same. It seems that the aPLA interacting motif, fffζζFxC, is dependent on polar or nonpolar properties more than hydropathy classes or charge of residues. Alanine is nonpolar, but it decreased the ability of aPLA to bind I7A, L8A, F9A, and C14A p1-mutated peptides as well as R58A, G59A, G60A, M61A, and C66A haematologica | 2017; 102(8)

aPLA recognizes fffζζFxf motif

Figure 3. aPLA epitope contains fffζζFxf motifs. (A) Alignment of p1, p6, p4, anti-sense p3 and CD40 consensus sequence. Red color shows nonpolar residues while green color shows polar residues in the motif. Underlining represents R39-43 epitopes and a frame represents conserved residue. (B) Binding of aPLA to microplate wells coated with p1 and p6 with consecutive point mutations. Substituted residues are in red. Frame represents conserved residue. Data are mean ± SEM of 9 different donors. (C) Representative picture of IgG ctrl (control) or aPLA binding to peptide-coated well. Quantification of aPLA binding to p1-, p6-, p4- and p3-coated wells. Data are mean ± SEM of 9 different donors. Statistical significance was determined by Mann-Whitney U analysis. aPLA: antiphospholipid antibodies; IgG: immunoglobulin G; A.U: arbitrary unit.

p6-mutated peptides. These data suggest that the volume of amino acids and the ability to form electrostatic or hydrogen bonds may also contribute to interactions between aPLA and the motif. We then performed bioinformatical analysis using Prosite resource (SIB Swiss Institute of Bioinformatics) to find not only the fffζζFxC motif (N- to C-term) but also the fxFζζfff motif which is the reverse motif (C- to Nterm) in the human proteome. Four sequences were present in domain I of β2GP1; 2 fffζζFxC motifs and 2 fxFζζfff motifs (Table 3). fxFζζfff motifs were present in p1 and p3 (Figure 3A). We thus assessed the ability of aPLA to interact with p3. We also included a slightly different motif that was present in p4 to investigate whether a degenerated motif was able to interact with aPLA. As expected, aPLA bound to p3 with a similar efficiency compared to p1 and p6 (Figure 3C). We further observed that p4, although less potent in interactions with aPLA, bound significantly to the aPLA (Figure 3C). Altogether, these data demonstrated that aPLA interact with a fffζζFxC motif, and more generally with a fffζζFxf motif, in which the polarity of the residues was more important than the hydropathy classes. This motif could also interact with aPLA even if the motif was slightly modified by insertion or deletion of an amino acid before or after the key phenylalanine residue. Finally, aPLA were able to bind motifs in sense or anti-senses. haematologica | 2017; 102(8)

aPLA-related receptors described possess closely related motifs Vlachoyiannopoulos et al.15 identified a consensus motif between CD40 and p3 capable of interacting with aPLA. Although this sequence had one additional nonpolar amino acid between the polar and phenylalanine (fxFfζζfff) amino acids, it was strongly related to the initial motif (Figure 3A). In addition, it seemed that both polar amino acids present in the fffζζFxf motif were not fully required (Figure 3B). Thus we performed additional bioinformatical analysis relative to fxFζfff and fxFfζζfff motifs. Among the proteins containing these motifs, we identified β2GP1 domain I, toll-like receptor (TLR)1, TLR4, TLR6, TLR7, TLR8, TLR9 and GPIbα (Table 3). In addition to identified proteins containing the motifs fffζζFxf and fxFζζfff, all potential aPLA receptors described in the literature possess at least 1 of these 2 motifs. These results suggest that fffζζFxC and other closely related motifs constitute the potential links between controversial data obtained by different research groups working on aPLA.

Functional characterization of recombinant domains of β2GP1, peptides 1 and 6

To examine the functional ability of recombinant domains of β2GP1, p1, p4 and p6 to inhibit the activity of aPLA isolated from APS patients studied in our previous 1329

P. de Moerloose et al. A

Figure 4. Motif containing domains and peptides inhibits aPLA activity. (A) Inhibition of tumor necrosis factor (TNF) production induced by aPLA by domains I-II, IIIIV, V, β2GP1 of reduced β2GP1. (B) Representative picture of IgG ctrl (control) or aPLA binding to domains I-II, III-IV, V and β2GP1-coated well. Quantification of aPLA binding to domains I-II, III-IV, V or β2GP1-coated well. aPLA to well coated with domains I-II, III-IV, V, β2GP1 of reduced β2GP1. (C) Inhibition by domain I-II, p1, p4, p6, p10 and p11 of TNF production induced by aPLA. (D) Representative picture of IgG ctrl or aPLA binding to domains I-II-, p1-, p4-, p6-, p10- and p11-coated well. Quantification of aPLA binding to domains I-II, p1, p4, p6, p10 and p11-coated well. aPLA to well coated with domains I-II, III-IV, V, β2GP1 of reduced β2GP1. Data are mean ± SEM of 9 different donors. Statistical significance was determined by Mann-Whitney U analysis. aPLA: antiphospholipid antibodies; IgG: immunoglobulin G; US: unstimulated; NS: non significant.

experiments (Table 1), we evaluated tumor necrosis factor (TNF) production in human monocytes activated with aPLA which had previously been treated with peptides or recombinant domains. While incubation of human monocytes with aPLA pretreated with β2GP1 and reduced β2GP1 have no effect on aPLA activity, incubation of the aPLA with domain I-II, domain III-IV and domain V significantly decreased the production of TNF induced by aPLA (Figure 4A). This inhibition of TNF production involved the binding of aPLA to β2GP1, and reduced β2GP1, domain I-II and domain III, but not domain V (Figure 4B). Similarly, pre-incubation with peptides n°1, 4 and 6 (p1, p4 and p6) diminished the production of TNF induced by aPLA, whereas p10 and p11 had no effect (Figure 4C). This inhibition showed that p1, p4 and p6, but not p10 or p11, interact with aPLA (Figure 4D). We observed that p1 and p6 had similar aPLA binding activity, while p4 was significantly less potent in its interaction with aPLA (Figure 4D). These results were consistent with the presence of 4 aPLA interaction motifs in domain I-II, while p1 and p6 had only 1 motif (considering that p1 and revp1 were not simultaneously available) (Figure 5). As shown in Figure 3C, the interaction of p4 with aPLA was less effective due to a slightly degenerated motif (Figure 4D and 3A). Taken together, our results clearly revealed that the motifs recog1330

nized by aPLA, which were present in p1, p4 and p6 (Figure 3A), but not in p10 or p11 (Figure 2A), were able to functionally decrease the activity of aPLA. They further demonstrated that the aPLA interacting motif (fxFζfff) present in domain III-IV (Table 3) efficiently interacted with aPLA and inhibited its activity, although to a lesser extent than domain I-II (Figure 4B), confirming the potential targets exposed in Table 3.

Discussion The present study was undertaken to investigate the Tcell response associated with the production of autoantibodies against β2GP1 and their related epitopes in APS patients. We established that the β2GP1 signal peptide (p1) induced the proliferation of CD4+ T cells from APS patients but not from healthy donors. Furthermore, autoreactive CD4+ T cells were detected in APS patient blood samples and bound HLA-DRB3* 02:02 tetramer loaded with p1, thus endorsing the fact that specific HLADR alleles are associated with susceptibility to APS.10,16,17 We further observed that the signal peptide contained a motif that was similar to that of the R39-R43 epitope. We thus identified the aPLA interacting motif (fffζζFxf) to be haematologica | 2017; 102(8)

aPLA recognizes fffζζFxf motif

Figure 5. Representation of domain I-II. Frames represent motifs containing sequence. Red color shows nonpolar residues while green color shows polar residues in the motif. Violet color shows phenylalanine present in these motifs. Pink color and asterisks show point mutations already described. Dom: domain.

dependent on polar or nonpolar properties more than on hydropathy classes of amino acids. The recombinant domains of β2GP1, as well as individual peptides containing this motif, or closely related motifs, were able to inhibit the activity of aPLA on human monocytes. This study provides an appealing explanation for the controversial results obtained on potential aPLA-associated receptors. fffζζFxC and fffζζFxf-related motifs are indeed present in all aPLA-associated receptors described in the literature13 (Table 3). Bioinformatics analyses have revealed, however, that the fffζζFxC motif is present and accessible in 16 different proteins within the whole human proteome (Online Supplementary Table S1). Although it is well accepted that the β2GP1 domain I is the target of aPLA, the exact sequence of the epitope remains largely unknown. Herein, we have identified fffζζFxf as being a common aPLA interacting motif. While the signal peptide (p1), containing an aPLA interacting motif, was not present in the mature β2GP1, CD4+ T cells were able to recognize the p1 sequence. This suggests that a signal peptide could trigger autoimmune responses. Consistently, it was shown that a polymorphism in the signal peptide of the cytotoxic T-lymphocyte antigen-4 (CTLA-4) leads to autoimmune predispositions.18 Interestingly, the aPLA interacting motif present in p1 was also present in revp1, corresponding to a palindromic form of p1 (Figure 5). Furthermore, APS patients autoreactive Tcell proliferation was triggered by p1 (Figure 2D and 2E). Smith et al.19 suggested that complementary sequences of proteins may have a function in the induction of autoimmunity. This theory was confirmed by the discovery that autoimmunity can be initiated through an immune response against a peptide that is antisense or complementary to the autoantigen, which then induced anti-idiotypic antibodies (autoantibodies) that cross-reacted with the autoantigen.20 Furthermore, as a potential origin of APS pathogenicity, the palindromic sequence of p1 could explain the aPLA interactions with p3 and domain III, in which the binding sequences were antisense. Six sense and antisense epitopes are present in β2GP1. Five of these are found in domain I (Table 3 and Figure 5) and could be the reason for its identification as the main interacting protein of aPLA. To address the question of aPLA binding epitopes, several research groups performed epitope mapping using point mutations of domain I.6,7 They identified D8, D9, K19, S38, R39, G40, M42, R43 and T50 as residues participating in, or contributing to, aPLA domain I interactions. While 44% of these mutations were present in β2GP1 domain I motifs, the proportion increased to 100% if the mutations directly adjacent to motifs were included (Figure 5). Furthermore the adjacent mutations were at the same position as the motifs (Figure 5).6,7 While additional haematologica | 2017; 102(8)

investigation is required in order to define in further detail the exact sequence required to attain the highest affinity for aPLA, these parallels suggest that the motifs proposed could be slightly larger than those exposed in the study herein. Zager et al. have further shown, through phage display techniques based on an affinity for anti-β2GP1, that epitopes recognized by aPLA contained the fffζζ part of the motifs.21 They suggested consistently that polarity of the residues mainly mediate the interaction between epitope and aPLA.21 Another research group highlighted a sequence homology between residues 239-245 of CD40 and residues 26-32 of β2GP1.15 Residues 26-32 of β2GP1 corresponded to p3 and belonged to the fffζζ part of the fffζζfFxf motif (Figure 3 and Figure 5). Moreover, the hexapeptide TLRVYK, described by Blank et al. as an aPLA binding epitope also corresponded to the fffζζ part of the motifs.22 Mice immunized with a panel of microbial preparations have been shown to produce high titers of antiβ2GP1 antibodies. This panel of bacteria is composed of 6 different preparations, which include Streptococcus pneumonia, Shigella dysenteriae and haemophilus influenzae.22 Employing bioinformatical analysis (Prosite ressource, SIB), we found that all bacteria strains used in different preparations carry the fffζζFxC motif. In fact, not only the bacteria strains but also the virus strains carry the fffζζFxC motif. Several cases or clinical studies including patients with established APS or APS with thromboembolic phenomena have revealed a correlation between increased aPLA levels and HIV, Epstein-Barr virus, hepatitis C virus or herpesvirus-6 infection.23-26 Our bioinformatical analysis has revealed that all these virus strains carry the fffζζFxC motif. Taken together these data demonstrate that the aPLA interaction motif fffζζFxC, and by extension fffζζFxf-related motifs characterized in the present study were strongly related to other previously depicted epitopes, thereby placing emphasis on polarity properties of residues and not on particular sequences. The characterization of the aPLA epitope performed herein leads to a better understanding of the surprising number of candidate receptors described for aPLA.13 We have indeed shown that all TLRs suggested as aPLA receptors carry at least 1 motif in their extracellular region. Thus, TLR2 has 1 fffζζ FxC motif and TLR4 3 fffζFxf motifs. As presented in Figure 4D, the fffζζ FxC motif may be more potent than the fffζFxf motif in binding aPLA, but the presence of up to 3 fffζFxf motifs on TLR4 may explain why TLR2 and TLR4 were proposed as aPLA receptors. Annexin A2, apolipoprotein E receptor 2 (ApoER2) and GPIbα also have at least 1 motif accessible by aPLA (Table 3). The most described aPLA interacting protein, β2GP1, has no less than 5 motifs which are mainly present in domain I, but also in domain III. Several other targets carrying fffζζFxf motifs are important for APS, 1331

P. de Moerloose et al.

such as proteins involved in coagulation processes. Thus, thrombin, proteinase-activated receptor 2 (PAR-2 and 3) and the complements C5, C4a and C4b are some of the aPLA-targeted proteins carrying fffζζFxf motifs (data not shown). Although additional developments are necessary in order to find the exact association of residues needed to obtain the highest aPLA affinity, the study herein offers the opportunity to provide an accurate tool for the detection of β2GP1 antibodies for diagnostic purposes. Furthermore, aPLA interacting motifs present in peptides

References 1. Giannakopoulos B, Krilis SA. The pathogenesis of the antiphospholipid syndrome. N Engl J Med. 2013;368(11):1033-1044. 2. Miyakis S, Lockshin MD, Atsumi T, et al. International consensus statement on an update of the classification criteria for definite antiphospholipid syndrome (APS). J Thromb Haemost. 2006;4(2):295-306. 3. Agar C, van Os GM, Morgelin M, et al. Beta2-glycoprotein I can exist in 2 conformations: implications for our understanding of the antiphospholipid syndrome. Blood. 2010;116(8):1336-1343. 4. de Laat B, van Berkel M, Urbanus RT, et al. Immune responses against domain I of beta(2)-glycoprotein I are driven by conformational changes: domain I of beta(2)-glycoprotein I harbors a cryptic immunogenic epitope. Arthritis Rheum. 2011; 63(12):39603968. 5. de Laat B, Derksen RH, van Lummel M, Pennings MT, de Groot PG. Pathogenic antibeta2-glycoprotein I antibodies recognize domain I of beta2-glycoprotein I only after a conformational change. Blood. 2006; 107(5):1916-1924. 6. Iverson GM, Reddel S, Victoria EJ, et al. Use of single point mutations in domain I of beta(2)-glycoprotein I to determine fine antigenic specificity of antiphospholipid autoantibodies. J Immunol. 2002; 169(12):70977103. 7. Ioannou Y, Pericleous C, Giles I, et al. Binding of antiphospholipid antibodies to discontinuous epitopes on domain I of human beta(2)-glycoprotein I: mutation studies including residues R39 to R43. Arthritis Rheum. 2007;56(1):280-290. 8. Khattri S, Zandman-Goddard G, Peeva E. Bcell directed therapies in antiphospholipid antibody syndrome--new directions based on murine and human data. Autoimmun Rev. 2012;11(10):717-722. 9. Cousins L, Pericleous C, Khamashta M, et al. Antibodies to domain I of beta-2-glycoprotein I and IgA antiphospholipid antibodies in patients with 'seronegative' antiphospholipid syndrome. Ann Rheum Dis. 2015;74(1):317-319. 10. Hattori N, Kuwana M, Kaburaki J, et al. T cells that are autoreactive to beta(2)-glycoprotein I in patients with antiphospholipid syndrome and healthy individuals. Arthritis Rheum. 2000;43(1):65-75. 11. Arai T, Yoshida K, Kaburaki J, et al. Autoreactive CD4(+) T-cell clones to beta2glycoprotein I in patients with antiphospholipid syndrome: preferential recognition of the major phospholipid-binding site. Blood. 2001;98(6):1889-1896. 12. Ito H, Matsushita S, Tokano Y, et al.

1332

13.

14. 15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

have the ability to inhibit aPLA activity and represent a prevention strategy for APS as an alternative to the use of anticoagulants. Finally, p1-tetramers associated with inducers of cell death could be used to specifically disrupt autoreactive T cells in APS patients, thus representing a potential therapeutic approach. Funding This work was supported by two grants: the Swiss National Funds (n°310030–127639) and an unrestricted grant from the ISTH2007 Presidential Fund.

Analysis of T cell responses to the β2-glycoprotein I-derived peptide library in patients with anti-β2-glycoprotein I antibody-associated autoimmunity. Hum Immunol. 2000;61(4):366-377. Brandt KJ, Kruithof EKO, de Moerloose P. Receptors involved in cell activation by antiphospholipid antibodies. Thromb Res. 2013;132(4):408-413. de Laat B, de Groot PG. Autoantibodies directed against domain I of beta2-glycoprotein I. Curr Rheumatol Rep. 2011;13(1):70-76. Vlachoyiannopoulos PG, Mavragani CP, Bourazopoulou E, Balitsari AV, Routsias JG. Anti-CD40 antibodies in antiphospholipid syndrome and systemic lupus erythematosus. Thromb Haemost. 2004;92(6):13031311. Tanimura K, Jin H, Suenaga T, et al. β2-Glycoprotein I/HLA class II complexes are novel autoantigens in antiphospholipid syndrome. Blood. 2015;125(18):2835-2844. Domenico Sebastiani G, Minisola G, Galeazzi M. HLA class II alleles and genetic predisposition to the antiphospholipid syndrome. Autoimmun Rev. 2003;2(6):387-394. Anjos S, Nguyen A, Ounissi-Benkalha H, Tessier MC, Polychronakos C. A common autoimmunity predisposing signal peptide variant of the cytotoxic T-lymphocyte antigen 4 results in inefficient glycosylation of the susceptibility allele. J Biol Chem. 2002;277(48):46478-46486. Smith LR, Bost KL, Blalock JE. Generation of idiotypic and anti-idiotypic antibodies by immunization with peptides encoded by complementary RNA: a possible molecular basis for the network theory. J Immunol. 1987;138(1):7-9. Pendergraft WF, Preston GA, Shah RR, et al. Autoimmunity is triggered by cPR-3(105201), a protein complementary to human autoantigen proteinase-3. Nat Med. 2004; 10(1):72-79. Žager U, Lunder M, Hodnik V, et al. Significance of K(L/V)WX(I/L/V)P Epitope of the B2Gpi in Its (Patho)Physiologic Function. EJIFCC. 2011;22(4):118-124. Blank M, Krouse I, Fridkin M, et al. Bacterial induction of autoantibodies to b2-glycoprotein-1 accounts for the infectious etiology of antiphospholipid syndrome. J Clin Invest. 2002;109:797-804. Toyoshima M, Maegaki Y, Yotsumata K, Takei S, Kawano Y. Antiphospholipid syndrome associated with human herpesvirus-6 infection. Pediatr Neurol. 2007;37(6):449451. Grunewald T, Burmester GR, Schuler-Maue W, Hiepe F, Buttgereit F. Anti-phospholipid antibodies and CD5+ B cells in HIV infection. Clin Exp Immunol. 1999;115(3):464471.

25. Abdel-Wahab N, Lopez-Olivo MA, PintoPatarroyo GP, Suarez-Almazor ME. Systematic review of case reports of antiphospholipid syndrome following infection. Lupus. 2016;25(14):1520-1531. 26. Durkin ML, Marchese D, Robinson MD, Ramgopal M. Catastrophic antiphospholipid syndrome (CAPS) induced by influenza A virus subtype H1N1. BMJ Case Rep. 2013;2013:bcr2013200474. 27. McNeil HP, Simpson RJ, Chesterman CN, Krilis SA. Antiphospholipid antibodies are directed against a complex antigen includes a lipid-binding inhibitor of coagulation : B2glycoprotein I (apolipoprotein H). Proc Natl Acad Sci USA. 1990;87:4120-4124. 28. Satta N, Dunoyer-Geindre S, Reber G, et al. The role of TLR2 in the inflammatory activation of mouse fibroblasts by human antiphospholipid antibodies. Blood. 2007; 109(4):1507-1514. 29. Lutters BC, Derksen RH, Tekelenburg WL, et al. Dimers of beta 2-glycoprotein I increase platelet deposition to collagen via interaction with phospholipids and the apolipoprotein E receptor 2'. J Biol Chem. 2003;278(36):33831-33838. 30. Döring Y, Hurst J, Lorenz M, et al. Human antiphospholipid antibodies induce TNFα in monocytes via Toll-like receptor 8. Immunobiology. 2010;215(3):230-241. 31. Romay-Penabad Z, Montiel-Manzano MG, Shilagard T, et al. Annexin A2 is involved in antiphospholipid antibody-mediated pathogenic effects in vitro and in vivo. Blood. 2009;114:3074-3083. 32. Brandt KJ, Fickentscher C, Boehlen F, Kruithof EK, de Moerloose P. NF-kappaB is activated from endosomal compartments in antiphospholipid antibodies-treated human monocytes. J Thromb Haemost. 2014;12(5): 779-791. 33. Pierangeli SS, Vega-Ostertag ME, Raschi E, et al. Toll-like receptor and antiphospholipid mediated thrombosis: in vivo studies. Ann Rheum Dis. 2007;66(10):1327-1333. 34. Hurst J, Prinz N, Lorenz M, et al. TLR7 and TLR8 ligands and antiphospholipid antibodies show synergistic effects on the induction of IL-1β and caspase-1 in monocytes and dendritic cells. Immunobiology. 2009;214(8):683-691. 35. Aguilar-Valenzuela R, Nickerson K, RomayPenabad Z, et al. Involvement of TLR7 and TLR9 in the Production of Antiphospholipid Antibodies. Arthritis Rheum. 2011;63(10): S281-S282. 36. Shi T, Giannakopoulos B, Yan X, et al. Antiβ2-glycoprotein I antibodies in complex with β2-glycoprotein I can activate platelets in a dysregulated manner via glycoprotein Ib-IX-V. Arthritis Rheum. 2006;54(8):25582567.

haematologica | 2017; 102(8)

ARTICLE

Platelet Biology & its Disorders

Risk of cardiovascular events and pulmonary hypertension following splenectomy – a Danish population-based cohort study from 1996-2012

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Marianne Rørholt,1 Waleed Ghanima,2,3 Dora Körmendiné Farkas4 and Mette Nørgaard 4

Department of Research and Otorhinolaryngology, Østfold Hospital Trust, Norway; Department of Research and Medicine, Østfold Hospital Trust, Norway; 3Institute of Clinical Medicine, University of Oslo, and Department of Haematology Oslo University Hospital, Norway and 4Department of Clinical Epidemiology, Aarhus University Hospital, Denmark 1 2

Haematologica 2017 Volume 102(8):1333-1341

ABSTRACT

plenectomized patients are at increased risk of cardiovascular events, but it remains unclear whether this is due to lack of the spleen or due to the underlying disease leading to splenectomy. We aimed to assess the risk of myocardial infarction, pulmonary hypertension, and stroke following splenectomy. We identified patients splenectomized in Denmark between 1996 and 2012. We constructed two comparison cohorts: an age- and sex-matched general population cohort and a disease-matched cohort based on the splenectomy-related underlying disease. We computed 5-year cumulative incidences and adjusted hazard ratios of myocardial infarction, pulmonary hypertension, and stroke for the three cohorts. The study included 5,306 splenectomized patients, 53,060 members of the general population, and 11,651 disease-matched patients. During the 5-year follow-up, 1.3% of splenectomized patients had a myocardial infarction versus 1.8% of the population cohort. The adjusted hazard ratio for myocardial infarction in splenectomized patients versus the population cohort was 1.24 (95% confidence interval: 1.011.52). The 5-year cumulative incidence of pulmonary hypertension was 0.4% among splenectomized subjects and 0.2% in the population cohort [adjusted hazard ratio 3.25 (95% confidence interval: 1.93-5.45)], while that of stroke was 3.3% among splenectomized patients versus 2.6% in the population cohort [adjusted hazard ratio 2.04 (95% confidence interval: 1.78-2.35)]. When comparing splenectomized subjects with the disease-matched cohort, only stroke risk was elevated, with 5-year risks of 3.0% and 2.3%, respectively [adjusted hazard ratio 1.56 (95% confidence interval: 1.26-1.92)]. In conclusion, splenectomized patients were at increased risk of stroke. Additionally, we found that underlying splenectomy-related diseases explained the increased risk of myocardial infarction and pulmonary hypertension following splenectomy. Introduction Splenectomy is a relatively common surgical procedure performed for various medical and surgical conditions.1 According to the National Hospital Discharge Survey, approximately 22,000 splenectomies are performed annually in the USA, with trauma and incidental splenectomy as the primary surgical indications and hematologic disorders as the primary medical indications.2 Splenectomy is known to be associated with both postoperative and long-term complications.1,3-5 Common short-term complications have been well described and include postoperative infections, bleeding, and venous thromboembolism.3,6 The most serious long-term consequence is a lifelong increased susceptibility to infections by encapsulated bacteria. Among such infections, pneumococcal sepsis has a particularly high case fatality rate.1,5 haematologica | 2017; 102(8)

Correspondence: mn@clin.au.dk

Received: September 22, 2016. Accepted: May 31, 2017. Pre-published: June 1, 2017. doi:10.3324/haematol.2016.157008 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1333 ©2017 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.

1333

M. RĂ¸rholt et al.

Over the past 40 years, research has suggested that splenectomized patients are also at increased long-term risk of atherosclerotic events and pulmonary hypertension (PH).4,5,7 Suggested underlying mechanisms include hypercoagulability, increased platelet counts, platelet activation, disturbance and activation of the endothelium, and altered lipid profiles.4 Thus, loss of the filtering function of the spleen may permit particulate matter and damaged cells to persist in the bloodstream, thereby perturbing and activating the vascular endothelium and shifting vascular homostasis towards enhanced coagulation.4,7 Since several underlying diseases for which splenectomy is performed may be associated with increased risks of venous and arterial thrombosis,8-10 it remains unclear whether increased cardiovascular risk arises from removal of the spleen or from the underlying indication for the splenectomy.11 If lack of a spleen is the cause, then effects would be expected across the underlying reasons for splenectomy. Yet, Kristinsson et al. found no increased risks of myocardial infarction (MI) or ischemic stroke in 8,149 cancer-free veterans who underwent splenectomy for various reasons compared with the risk in four million

hospitalized veterans.5 They did not, however, take into account the underlying reason for splenectomy. In an earlier Danish study, the mortality risk among splenectomized patients more than 1 year after the operation, regardless of the indication, was 2-fold higher than that in the general population. Compared with un-splenectomised patients with similar indications, the risk of death associated with splenectomy was not increased.12 Elevated risks of cardiovascular complications may have important clinical implications. Data on these risks are needed to understand and potentially prevent postsplenectomy death. We therefore conducted a nationwide population-based cohort study on the long-term risks of cardiovascular events following splenectomy. We investigated the risks of MI, PH, and stroke among patients splenectomized for a variety of indications and compared these risks with those in the general population. We then examined whether the risks were related to the splenectomy and its consequences or to the underlying diseases by comparing outcomes among patients who underwent splenectomy with outcomes among non-splenectomized patients with similar diseases.

Figure 1. Cumulative risk of myocardial infarction following splenectomy (in years). (A) Cumulative risk of myocardial infarction (MI) in splenectomized patients compared with the risk in the general population cohort. (B) Cumulative risk of MI in splenectomized patients compared with the risk in the diseasematched cohort.

1334

haematologica | 2017; 102(8)

Risk of post-splenectomy events

Methods We used the Danish National Patient Registry (DNPR) to identify patients who underwent splenectomy between January 1, 1996 and December 31, 2012. The DNPR contains information on all admissions to Danish hospitals since 1977 and hospital outpatient clinic visits since 1995. Data include dates of admission and discharge, surgical procedures coded according to the Danish version of the Nordic Medico-Statistical Committee (NOMESCO) Classification of Surgical Procedures, and up to 20 discharge diagnoses coded by physicians according to the International Classification of Diseases, Eighth Revision (ICD-8) until the end of 1993, and thereafter the ICD-10. We classified splenectomized patients into eight subgroups, according to all previous diagnoses recorded in the patient registry, using the following hierarchy: (i) traumatic rupture of the spleen; (ii) idiopathic thrombocytopenic purpura; (iii) other/unspecified thrombocytopenia; (iv) hematopoietic cancer; (v) hereditary hemolytic anemia; (vi) abdominal cancer; (vii) splenomegaly/other splenic diseases only; and (viii) other indications.6,12 Accordingly, if a patient had traumatic splenic rupture before the date of splenectomy, he/she was categorized into this first indication group, regardless of the presence/absence of the other indications. Patients in the “other indications” group had none of the selected indications before splenectomy. We excluded splenectomized patients with a prior diagnosis of coronary artery disease, MI, PH, or stroke before the date of surgery. Surgical and diagnostic codes are summarized in Online Supplementary Table S1. We used the Danish Civil Registration System (CRS)13 to select ten members of the general population for each splenectomized

patient, matched by age, sex, and calendar year of splenectomy. We also constructed a disease-matched comparison cohort, using the CRS and DNPR to identify up to five patients diagnosed with the same underlying disease in the same calendar year as the corresponding splenectomized patient. Members of this comparison cohort could not have a procedure code for splenectomy, or a diagnosis code for coronary artery disease, MI, PH, or stroke before study inclusion. As a comparison group for patients splenectomized due to trauma, we identified trauma patients who underwent surgery for acute injury of the spleen, liver, or gallbladder, with no recorded splenectomy. In our comparison of the splenectomy and disease-matched cohorts, we excluded the “other indications” subgroup from the splenectomy cohort. The date of splenectomy represented the “index date” for the matched sets of patients. To address potential confounding, we retrieved information on the presence of the following diagnoses recorded prior to the splenectomy/index date: chronic obstructive pulmonary disease, pulmonary embolism, heart failure, diabetes, atrial fibrillation, hypertension and obesity (see Online Supplementary Table S1 for diagnostic codes). We also retrieved information on all diagnoses of venous thromboembolism recorded prior to cardiovascular outcomes. Splenectomized patients and members of their comparison cohorts were followed from their surgery/index dates to occurrence of any long-term outcomes of interest, death, or end of follow-up (31 December, 2012), whichever occurred first. We used cumulative incidence functions with death as a competing event, and plotted the cumulative risks of MI, PH and

Table 1. Demographic characteristics of the splenectomized cohort (overall and by indication for splenectomy), the disease-matched comparison cohort, and the general population comparison cohort.

INDICATION FOR SPLENECTOMY, N (%) Overall Population Disease- Traumatic Abdominal ITP** Hematopoietic Splenomegaly/ Hereditary Nonspecific Other splenectomized comparison matched rupture cancers N (%) cancers other splenic hemolytic thrombocytopenia N (%) cohort cohort comparison N (%) N (%) N (%) disease anemias N (%) N (%) N (%) cohort N (%) N (%) N (%) Total, n (%) 5306 (100) 53,060 (100) 11,651 (100) Age – years at 58.4 58.4 58.4 Median (IQR) (40.1-70.3) (40.2-70.3) (40.2-70.3) Sex Female 2485 (46.8) 24,850 (46.8) 5700 (48.9) Male 2821 (53.2) 28,210 (53.2) 5951 (51.1) Year of splenectomy/index date 1996-2001 2113 (39.8) 21,130 (39.8) 4645 (39.9) 2002-2007 1796 (33.9) 17,960 (33.9) 3935 (33.8) 2008-2012 1397 (26.3) 13,970 (26.3) 3071 (26.4) Comorbid conditions COPD* 329 (6.2) 2197 (4.1) 746 (6.4) Diabetes 210 (4.0) 1282 (2.4) 503 (4.3) Hypertension 437 (8.2) 2610 (4.9) 972 (8.3) Atrial fibrillation 130 (2.5) 1034 (1.9) 399 (3.4) PE 54 (1.0) 187 (0.4) 117 (1.0) Heart failure 64 (1.2) 436 (0.8) 205 (1.8) Obesity 123 (2.3) 742 (1.4) 270 (2.3) 5-year mortality 39.4 % 9.5% 38.3%

1033 (19.5) 880 (16.6) 379 (7.1) 36.6 70.8 46.4 (22.1-53.9) (78.5-95.2) (28.9-60.8)

417 (7.9) 63.9 (55.5-71.7)

321 (6.0) 48.0 (32.6-63.0)

216 (4.1) 13.1 (9.7-26.7)

300 (29.0) 733 (71.0)

439 (49.9) 239 (63.0) 441 (50.1) 140 (36.9)

183 (43.9) 234 (56.1)

149 (46.4) 172 (53.6)

121 (56.0) 95 (44.0)

38 (50.6) 37 (49.3)

1016 (51.2) 969 (48.8)

427 (41.3) 399 (38.6) 207 (20.0)

392 (44.6) 176 (46.4) 298 (32.8) 117 (30.9) 199 (22.6) 86 (22.7)

137 (32.9) 144 (34.5) 136 (32.6)

100 (31.2) 124 (38.6) 97 (30.2)

79 (36.6) 90 (41.7) 47 (21.8)

29 (38.7) 30 (40.0) 16 (21.3)

773 (38.9) 603 (30.4) 609 (30.7)

28 (6.7) 18 (4.3) 30 (7.2) 11 (2.6) 11 (2.6) 6 (1.4) 8 (1.9) 50.4%

23 (7.2) 12 (3.7) 23 (7.2) 8 (2.5) 2 (0.6) 2 (0.6) 8 (2.5) 22.3%

14 (6.5) 3 (1.4) 0 (0) 0 (0) 0 (0) 1 (0.5) 2 (0.9) 4.1%

8 (10.7) 8 (10.7) 6 (8.0) 3 (4.0) 2 (2.7) 0 (0) 4 (5.3) 44.0%

149 (7.5) 107 (5.4) 233 (11.7) 59 (3.0) 22 (1.1) 30 (1.5) 61 (3.1) 50.1%

37 (3.6) 16 (1.5) 31 (3.0) 14 (1.4) 3 (0.3) 5 (0.5) 13 (1.3) 17.3%

58 (6.6) 31 (3.5) 93 (10.6) 27 (3.1) 14 (1.6) 16 (1.8) 15 (1.7) 66.4%

12 (3.2) 15 (4.0) 21 (5.5) 8 (2.1) 0 (0) 4 (1.1) 12 (3.2) 7.2%

75 (1.4) 1985 (37.4) 56.0 62.9 (43.7-66.5) (51.6-71.8

IQR: Interquartile range; COPD: chronic obstructive pulmonary disease; PE: pulmonary embolus; ITP: idiopathic thrombocytopenic purpura.

haematologica | 2017; 102(8)

1335

M. RĂ¸rholt et al. Table 2. Five-year risks (cumulative incidence rates with death as a competing event) and adjusted hazard ratios, with 95% confidence intervals, of myocardial infarction, pulmonary arterial hypertension, and stroke in 5,306 splenectomized patients compared with 53,060 age- and gender-matched members of the general population, overall and stratified by indication for splenectomy.

Myocardial infarction Pulmonary arterial hypertension Splenectomized General Adjusted Splenectomized General Adjusted patients, population hazard patients, population hazard 5-year risk, % comparison ratio* 5-year risk, % comparison ratio* cohort, cohort, 5-year risk, % 5-year risk, % Overall Traumatic rupture Abdominal cancer Idiopathic thrombocytopenic purpura Hematopoietic cancers Splenomegaly/splenic disease Non-specific thrombocytopenia Hereditary hemolytic anemia Other indications

Stroke Splenectomized General Adjusted patients, population hazard 5-year risk, % comparison ratio** cohort, 5-year risk, %

1.28 (0.99-1.63) 0.54 (0.21-1.21) 1.50 (0.82-2.55) 1.13 (0.38-2.71)

1.75 (1.64-1.88) 0.57 (0.43-0.74) 2.92 (2.56-3.31) 1.00 (0.70-1.39)

1.24 (1.01-1.52) 0.74 (0.34-1.60) 0.79 (0.47-1.33) 1.51 (0.70-3.25)

0.35 (0.21-0.56) 0.21 (0.05-0.74) 0.12 (0.01-0.68) -

0.16 (0.12-0.20) 0.06 (0.02-0.13) 0.29 (0.19-0.43) 0.03 (0.00-0.15)

3.25 (1.93-5.46) 1.75 (0.31-10.01) 2.21 (0.46-10.58) 2.47 (0.25-24.08)

3.34 (2.86-3.89) 2.78 (1.88-3.96) 3.08 (2.03-4.48) 2.09 (0.93-4.09)

2.62 (2.47-2.77) 0.95 (0.76-1.16) 4.58 (4.13-5.06) 1.24 (0.90-1.66)

2.04 (1.78-2.35) 3.12 (2.19-4.47) 1.34 (0.96-1.87) 1.70 (0.99-2.91)

1.86 (0.83-3.66) 1.73 (0.66-3.79) 2.95 (0.56-9.17) -

2.14 (1.69-2.67) 0.81 (0.52-1.21) 0.94 (0.39-1.97) 0.17 (0.05-0.47) 2.31 (2.08-2.54)

1.12 (0.53-2.38) 2.59 (1.15-5.83) 1.73 (0.35-8.43) 3.35 (0.65-17.36) 1.44 (1.08-1.93)

0.58 (0.12-1.94) 0.77 (0.15-2.57) -

0.14 (0.05-0.31) 0.18 (0.07-0.42) -

5.13 (0.53 - 49.28) 9.44 (0.95-94.29) -

0.56 (0.05-2.86) 0.20 (0.14-0.27)

4.40 (2.62-6.88) 4.51 (2.52-7.36) 3.08 (0.58-9.51) 0.49 (0.05-2.52) 3.98 (3.12-5.00)

2.83 (2.32-3.42) 1.59 (1.16-2.13) 2.14 (1.23-3.48) 0.57 (0.31-1.00) 3.27 (3.01-3.55)

2.15 (1.30-3.54) 5.00 (2.85-8.75) 1.02 (0.23-4.52) 4.69 (1.65-13.29 1.96 (1.57-2.44)

1.50 (1.00-2.16)

0.49 (0.23-0.94)

2.92 (1.33 - 6.43)

*adjusted for age, sex, chronic obstructive pulmonary disease, pulmonary embolism, heart failure, diabetes, hypertension, and obesity. **additionally adjusted for venous thromboembolism as a time-varying covariate.

stroke for the three study cohorts. Only the first of these outcomes were included in the analyses. We computed the 5-year mortality and 5-year cumulative incidence of MI, PH, and stroke in the three cohorts as a measure of 5-year risk, treating death as a competing event. We used stratified Cox regression analysis to compute the adjusted hazard ratio (aHR) overall and separately for all indications for splenectomy [with the corresponding 95% confidence intervals (CI)] for each outcome, with adjustment for age, sex, and the following comorbid conditions: chronic obstructive pulmonary disease, PE, heart failure, diabetes, hypertension, and obesity while we censored patients who died. In the PH analysis, we additionally included venous thromboembolism as a time-varying covariate. We also analyzed the data categorizing stroke as ischemic or hemorrhagic because of their different underlying mechanisms. Because more than two-thirds of all unspecified strokes are known to be ischemic strokes, we re-categorized unspecified strokes as ischemic strokes.14 In a separate analysis, we additionally adjusted for atrial fibrillation to explore whether this could be an intermediate step. All analyses were conducted using SAS 9.2 software. The study was approved by the Danish Data Protection Agency (Jr n. 1-1602-1-08).

splenectomy, 53,060 matched members of the general population, and 11,651 members of a disease-matched cohort. Their characteristics and 5-year mortality rates are summarized in Table 1. The four most frequently recorded indications for splenectomy were traumatic rupture of the spleen [1,033 patients (19.5%)], abdominal cancers [880 patients (16.6%)], hematopoietic cancers [417 patients (7.9%)], and idiopathic thrombocytopenic purpura [379 patients (7.1%)]. In total, 1,985 patients (37.4%) had none of the specified indications. Of the comorbid conditions examined, hypertension and chronic obstructive pulmonary disease were most frequently reported, with a prevalence of 8.2% and 6.2%, respectively, among splenectomized patients overall. The prevalence of comorbidity was generally higher than in the general population comparison cohort (Table 1). Among the general population, 4.9% had hypertension and 4.1% had chronic obstructive pulmonary disease. Patients with other indications were older and had a high prevalence of comorbid conditions. The splenectomy cohort and the disease-matched cohort had almost equal rates of the specified comorbid conditions (Table 1).

Study outcomes Results Patientsâ&#x20AC;&#x2122; characteristics We identified 5,306 patients who had undergone 1336

We followed the splenectomized cohort for a median of 3.8 years (maximum 17.0 years). Figure 1 illustrates the cumulative risk of MI for the splenectomy and comparison cohorts over the entire follow-up period. After 5 years haematologica | 2017; 102(8)

Risk of post-splenectomy events

Figure 2. Cumulative risk of pulmonary hypertension following splenectomy (in years). (A) Cumulative risk of pulmonary hypertension (PH) in splenectomized patients compared with the risk in the general population cohort. (B) Cumulative risk of PH in splenectomized patients compared with the risk in the disease-matched cohort.

of follow-up, the risk of a first-time MI was 1.3% (95% CI: 1.0%-1.6%) among splenectomized patients compared with 1.8% (95% CI: 1.6%-1.9%) among members of the general population cohort, taking death into account as a competing event. However, the unadjusted HR comparing splenectomized patients with the general population cohort was 1.28 (95% CI: 1.04 - 1.57) and the aHR was 1.24 (95% CI: 1.01-1.52) (Table 2). In contrast, splenectomized patients and the disease-matched cohort had similar 5-year risks of MI (Table 3). Considering death as a competing event, the 5-year risk of MI was 1.2% (95% CI: 0.8%-1.6%) in the splenectomy cohort and 1.4% (95% CI: 1.2%-1.6%) in the disease-matched cohort, with an unadjusted HR for MI of 0.92 (95% CI: 0.68 - 1.24) and an aHR of 0.95 (95% CI: 0.70-1.28). Compared with the disease-matched cohort, the relative risk of MI did not vary substantially between subgroups of splenectomized patients. The 5-year risk of PH was 0.4% (95% CI: 0.2%-0.6%) in splenectomized patients compared with 0.2% (95% CI 0.1%-0.2%) in the general population cohort (Figure 2), with an aHR of 3.25 (95% CI: 1.93-5.46) (Table 2). However, the cumulative incidences were similar in the splenectomy and disease-matched cohorts (Figure 2) with an aHR of 1.03 (95% CI: 0.55-1.93) (Table 3). Comparing haematologica | 2017; 102(8)

splenectomized patients with those in the diseasematched cohort, the aHR of PH was 7.89 (95% CI: 0.7187.99) among patients with splenomegaly/splenic disease and 4.16 (95% CI: 0.58-29.88) for patients with idiopathic thrombocytopenic purpura. However, the statistical precision of these estimates was low (Table 3). Among splenectomized patients, 3.3% (95% CI: 2.9%3.9%) had a stroke within the first 5 years of follow-up, compared with 2.6% (95% CI: 2.5%-2.8%) of people in the general population cohort (Table 2 and Figure 3), taking death into account as a competing event. The unadjusted HR was 2.05 (95% CI: 1.79-2.36) and the aHR was 2.04 (95% CI: 1.78-2.35) (Table 2). Compared with the general population cohort, the 5-year risk of stroke was consistently higher in all splenectomy indication subgroups, except for the subgroup with non-specific thrombocytopenia (Table 2). The cumulative incidence of stroke was higher in splenectomized patients than in the disease-matched cohort (Figure 3). The 5-year stroke risk was 3.0% (95% CI: 2.4%-3.7%) and 2.3% (95% CI: 2.0%-2.6%) in the two groups, respectively (Table 3). After 10 years of follow-up the stroke risk remained higher in splenectomized patients than in the disease-matched cohort [4.9% (95% CI: 4.1-5.8) versus 4.0 % (95% CI: 3.54.4), respectively]. The unadjusted HR was 1.48 (95% CI: 1337

M. RĂ¸rholt et al.

1.22-1.80) and the aHR was 1.53 (95% CI; 1.26-1.86) (Table 3). In the disease-matched comparisons, the 5-year stroke risk was higher in splenectomized patients than in non-splenectomized ones in all sub-groups except for those with idiopathic thrombocytopenic purpura or nonspecific thrombocytopenia (Table 3). For patients splenectomized because of traumatic rupture of the spleen, the aHR for stroke was 3.12 (95% CI: 2.19-4.47) compared with the general population cohort (Table 2) and 1.95 (95 % CI: 1.06-3.58) compared with disease-matched patients who underwent surgery for acute injury of the spleen, liver, or gallbladder, with no recorded splenectomy (Table 3). When we analyzed the risk of ischemic and hemorrhagic stroke separately, we found that splenectomized patients had a 2-fold increased risk of ischemic stroke compared with the general population cohort [aHR 2.05 (95% CI: 1.76 - 2.37)] (Table 4) and a 50% increased risk of ischemic stroke, compared with the disease-matched cohort [aHR 1.56 (95% CI: 1.26 - 1.92)]. For haemorrhagic stroke, the aHR were similarly increased: 1.77 (95% CI: 1.17 - 2.70) for the splenectomy cohort compared with the general population cohort and 1.37 (95% CI: 0.81 - 2.31)

for the splenectomy cohort compared with the diseasematched cohort. Including atrial fibrillation as a covariate did not substantially change the aHR for stroke overall, ischemic stroke, or hemorrhagic stroke (data not shown).

Discussion Our nationwide population-based study showed that splenectomized patients had, as expected, higher risks of MI, PH, and stroke than people in the general population. When we compared splenectomized patients with a disease-matched cohort, we found similar risks of MI and PH. This indicates that the underlying medical conditions for which the splenectomy was performed caused the increased risk of MI and PH. However, splenectomized patients had a 50% higher risk of ischemic stroke and a 30% increased risk of hemorrhagic stroke compared with patients in the disease-matched cohort. This suggests that the increased risk of stroke was a consequence of the splenectomy, rather than of the underlying disease leading to splenectomy.

Figure 3. Cumulative risk of stroke following splenectomy (in years). (A) Cumulative risk of stroke in splenectomized patients compared with the risk in the general population cohort. (B) Cumulative risk of stroke in splenectomized patients compared with the risk in the disease-matched cohort.

1338

haematologica | 2017; 102(8)

Risk of post-splenectomy events

The increased risk of cardiovascular events following splenectomy was first suggested in a study of 745 World War II servicemen who had been splenectomized because of trauma.15 By the end of 1974, the risk of death due to ischemic heart disease was nearly doubled following splenectomy.15 However, the majority of cardiovascular deaths in splenectomized servicemen (36 out of a total of 41) occurred more than 15 years after the splenectomy. This is in accordance with our finding that splenectomized patients did not have higher risk of MI than the general population cohort during our 17-year follow-up period. When we took death into account as a competing event the 5-year risk of MI was in fact slightly higher in the background population than in splenectomized patients. Still, we found an overall increased hazard ratio of MI and PH when comparing splenectomized subjects with the background population illustrating that although splenectomized people have a higher risk of MI, they also have a higher mortality than the background population and, therefore, fewer of them will live long enough to actually develop MI or PH. Two previous studies, both restricted to patients with hereditary spherocytosis, indicated that splenectomy increased the risk of arteriosclerotic events (stroke, MI, and coronary or carotid artery surgery) 5- to 7-fold.16,17 In our study, we categorized hereditary spherocytosis with other hereditary hemolytic anemias and found that the risks of MI and stroke were less than 70% increased, when comparing splenectomized and non-splenec-

tomized patients with hemolytic anemia. This was substantially lower than previous findings.16,17 Still, due to our low statistical precision, we could not rule out a 7-fold increased risk. Latency since splenectomy and risk of cardiovascular outcomes were addressed in the large study based on US Veterans Affairs data with up to 27 years of follow-up.5 Comparing splenectomized veterans with other veterans, the risk of being hospitalized with MI was not increased at any time during the follow-up period. Our study thus corroborates earlier findings that the absolute risk of MI in general is not increased following splenectomy. Because we stratified splenectomy by underlying indication, our study extended these findings. We observed that, compared with risk in the general population, the risk of MI following splenectomy was only increased for patients who were splenectomized due to hematologic disorders. The effect nearly vanished in comparisons with patients with similar hematologic disorders. Increased incidences of PH following splenectomy were previously observed in patients with sickle cell anemia and thalassemia18 and also in patients referred for lung transplantation.7 When we compared our splenectomized and disease-matched cohorts, we did not find an increased risk of PH. This also suggests that it is not the absence of a spleen, but factors related to the underlying indication for splenectomy that may be the primary causes of PH. Pulmonary embolism is a risk factor for PH, and it is well documented that many of the underlying conditions for

Table 3. Five-year risks (cumulative incidence rates with death as a competing event) and adjusted hazard ratios with 95% confidence intervals of myocardial infarction, pulmonary arterial hypertension, and stroke in 3,321 splenectomised patients with a known underlying indication compared with 11,651 members of a disease-matched cohort. Patients with other indications for splenectomy (N=1985) were not included in the overall analyses.

Myocardial infarction Splenectomized Disease-matched Adjusted patients, comparison hazard 5-year risk, % cohort, ratio* 5-year risk,% Overall

1.16 (0.82-1.59) Traumatic 0.54 rupture (0.21-1.21) Abdominal 1.50 cancer (0.82-2.55) Idiopathic 1.13 thrombocytopenic (0.38-2.71) purpura Hematopoietic 1.86 cancers (0.83-3.66) Splenomegaly/ 1.73 splenic disease (0.66-3.79) Non-specific 2.95 thrombocytopenia (0.56-9.17) Hereditary hemolytic anemia

Pulmonary arterial hypertension Splenectomized Disease-matched Adjusted patients, comparison hazard 5-year risk, % cohort, ratio* 5-year risk, %

1.40 (1.18-1.64) 0.69 (0.27-1.55) 1.59 (1.24-2.02) 1.30 (0.83-1.95)

0.95 (0.70-1.28) 0.55 (0.18-1.70) 1.07 (0.58-1.99) 0.75 (0.33-1.74)

0.28 (0.13-0.53) 0.21 (0.05-0.74) 0.12 (0.01-0.68) -

0.28 (0.19-0.40) 0.14 (0.01-0.74) 0.28 (0.15-0.49) 0.37 (0.16-0.79)

1.80 (1.26-2.51) 1.22 (0.68-2.04) 1.49 (0.57-3.28) 0.29 (0.06-1.03)

1.04 (0.46-2.37) 2.20 (0.79-6.09 0.65 (0.08-5.67) 1.57 (0.13-15.15)

0.58 (0.12-1.94) 0.77 (0.15-2.57) -

0.12 (0.03-0.43) 0.39 (0.13-0.97) 0.27 (0.03-1.42) 0.45 (0.13-1.25)

0.56 (0.05-2.86)

1.03 (0.55-1.93) 1.27 (0.11 - 15.17) 0.86 (0.17-4.32) 4.16 (0.58-29.88) 7.89 (0.71-87.99) 1.30 (0.13-12.60)

Stroke Splenectomized Disease-matched Adjusted patients, comparison hazard 5-year risk, % cohort, ratio*â&#x20AC;&#x2122; 5-year risk,% 2.99 (2.42-3.65) 2.78 (1.88-3.96) 3.08 (2.03-4.48) 2.09 (0.93-4.09)

2.32 (2.04-2.64) 1.53 (0.78-2.73) 2.53 (2.06-3.06) 2.56 (1.87-3.42)

1.53 (1.26-1.86) 1.95 (1.06-3.58) 1.37 (0.90-2.09) 1.02 (0.57-1.80)

4.40 (2.62-6.88) 4.51 (2.52-7.36) 3.08 (0.58-9.51) 0.49 (0.05-2.52)

2.46 (1.80-3.29) 2.33 (1.53-3.40) 2.46 (1.16-4.61) 0.82 (0.34-1.72)

1.63 (0.91-2.91) 1.99 (1.02-3.85) 0.92 (0.19-4.40 2.00 (0.65- 6.10)

haematologica | 2017; 102(8)

1339

M. RĂ¸rholt et al. Table 4. Five-year risks (cumulative incidence rates with death as a competing event) and adjusted hazard ratios with 95% confidence intervals of ischemic stroke and hemorrhagic stroke in 5,306 splenectomized patients compared with 53,060 members of an age- and gender-matched general population comparison cohort, overall and stratified by indication for splenectomy.

Ischemic stroke Splenectomized General Adjusted patients, population hazard 5-year risk, % comparison cohort, ratio* 5-year risk, % (95% CI) Overall

2.94 (2.48-3.46) Traumatic rupture 1.98 (1.23-3.02) Abdominal cancer 2.96 (1.93-4.34) Immune thrombocytopenia 1.56 (0.59-3.43) Hematopoietic cancers 3.67 (2.05-6.01) Splenomegaly/splenic disease 3.87 (2.04-6.59) Non-specific thrombocytopenia 3.08 (0.58-9.51) Hereditary hemolytic anemia 0.49 (0.05-2.52) Other indications 3.74 (2.91-4.72)

2.35 (2.21-2.49) 0.80 (0.64-1.00) 4.17 (3.74-4.63) 1.03 (0.73-1.43) 2.62 (2.12-3.18) 1.41 (1.01-1.92) 1.98 (1.11-3.28) 0.47 (0.23-0.87) 2.95 (2.69-3.21)

2.05 (1.76-2.35) 2.82 (1.90-4.19) 1.47 (1.05-2.06) 1.41 (0.76-2.64) 2.09 (1.21-3.60) 5.18 (2.82-9.50) 1.21 (0.27-5.46) 6.29 (2.18-18.20) 2.02 (1.60-2.40)

which splenectomy is performed are associated with increased risk of venous thromboembolism and/or PH, including malignancies, trauma, myeloproliferative neoplasms, idiopathic thrombocytopenic purpura, and hemolytic anemia.19,20 Some studies have also shown that splenectomy is a risk factor for chronic thromboembolic pulmonary hypertension (defined by the absence of thrombus resolution after acute pulmonary embolism), particularly in patients splenectomized for a hemolytic disorder.11,21 Unfortunately, even our large cohort did not allow us to study specific types of PH such as chronic thromboembolic pulmonary hypertension.10 Moreover, a previous case series demonstrated that chronic thromboembolic pulmonary hypertension may occur more than 20 years after splenectomy for trauma.21 Accordingly, our follow-up may not have been sufficiently long to capture such cases. Still, when we included venous thromboembolism as a time-varying covariate in comparisons of the splenectomized and general population cohorts, the HR was not substantially lowered. Our study additionally highlighted that the absolute risk of PH was very low. Our finding of a nearly 2-fold increased risk of stroke among splenectomized patients compared to a diseasematched cohort with traumatic rupture of the spleen, and a nearly 3-fold increased risk of stroke compared with the general population are in line with previous research.18,22 A nationwide cohort study from Taiwan considered 11,273 patients with splenic injury during 1998-2010, including 5,294 patients who were splenectomized. Compared with a control cohort from the background population splenectomized subjects had a 2-fold higher incidence of stroke while patients with splenic injury but no splenectomy 1340

Splenectomized patients, 5-year risk, %

Hemorrhagic stroke General population comparison cohort 5-year risk,%

Adjusted hazard ratio* (95% CI)

0.40 (0.25-0.61) 0.80 (0.38-1.52) 0.12 (0.01-0.68) 0.53 (0.11-1.79) 0.74 (0.21-2.02) 0.64 (0.13-2.15) -

0.27 (0.22-0.32) 0.14 (0.08-0.24) 0.41 (0.29-0.58) 0.20 (0.09-0.41) 0.22 (0.10-0.42) 0.18 (0.07-0.42) -

1.77 (1.17-2.70) 5.05 (2.06-12.35) 0.28 (0.04-2.08) 3.27 (0.81-13.13) 2.92 (0.78-11.01) 7.98 (1.44-44.12) -

0.25 (0.08-0.61)

0.10 (0.02-0.35) 0.33 (0.25-0.43)

1.10 (0.51-2.40)

only had a 20% increased incidence.22 As comparisons for those splenectomized due to trauma we similarly used patients with traumatic injury of the spleen, the liver or the gallbladder who were not splenectomized and found a higher risk of stroke in the splenectomized patients. Although we cannot completely rule out confounding by indication, our study extends the findings from the Taiwanese study by showing an increased risk of stroke across varying underlying reasons for splenectomy. In the Taiwanese study, patients with splenectomy had higher prevalences of liver cirrhosis, hypertension, hyperlipidemia, diabetes, and chronic obstructive pulmonary disease compared with the control cohort. This suggests that lifestyle may differ between splenectomised subjects and the general population and thus may confound comparisons between splenectomized people and the background population. Nevertheless, the prevalence of these factors did not differ between splenectomized subjects and those with splenic injury who were not splenectomized.22 The US Veterans study showed no increase in risk of hospitalization due to ischemic stroke;5 however, the risk of death due to stroke was nearly doubled in splenectomized veterans (standardized mortality ratio 1.89; 95% CI: 0.913.90). The mechanisms underlying the increased risk of stroke following splenectomy remain unclear. MI and stroke have broadly comparable risk factors.23 We did not, however, observe an increased risk of MI, which speaks against a generally increased risk of arteriosclerosis caused by platelet activation, disturbance and activation of the endothelium, and altered lipid profiles.4 Several potential study weaknesses should be considhaematologica | 2017; 102(8)

Risk of post-splenectomy events

ered in interpreting our data. As discussed above, one major weakness is that we cannot rule out confounding by disease severity. Furthermore, our study relied on diagnoses recorded in the DNPR and it is well known that coding errors occur.24 We did not validate the underlying diagnoses in splenectomized or disease-matched subjects. However, diagnoses in the DNPR are validated on an ad hoc basis,24 and it has been shown that the surgical procedures used to identify splenectomized patients have high validity.25 Moreover, the positive predictive value of a diagnosis of MI in the DNPR was previously found to be above 90%,26 and that of acute ischemic stroke was found to be 97%.14 We, therefore, do not think that misclassification of the underlying disease constitutes a major source of bias in our study. We were able to adjust for selected comorbid conditions such as chronic obstructive pulmonary disease, diabetes, hypertension, and obesity, which are known to be associated with increased risk of cardiovascular events. However, we based our comorbidity information on hospital-related diagnoses and did not capture diagnoses made by general practitioners. It has been recognized that the diagnosis of obesity may be severely underreported in the DNPR.27 The prevalence of hypertension in our general population cohort was 5%, which is lower than expected based on an age-adjusted

References 1. Di Sabatino A, Carsetti R, Corazza GR. Post-splenectomy and hyposplenic states. Lancet. 2011;378(9785):86-97. 2. DeFrances CJ, Cullen KA, Kozak LJ. National Hospital Discharge Survey: 2005 annual summary with detailed diagnosis and procedure data. Vital Health Stat. 2007;(165):1-209. 3. Cadili A, de Gara C. Complications of splenectomy. Am J Med. 2008;121(5):371375. 4. Crary SE, Buchanan GR. Vascular complications after splenectomy for hematologic disorders. Blood. 2009;114(14):2861-2868. 5. Kristinsson SY, Gridley G, Hoover RN, Check D, Landgren O. Long-term risks after splenectomy among 8,149 cancer-free American veterans: a cohort study with up to 27 years follow-up. Haematologica. 2014;99(2):392-398. 6. Thomsen RW, Schoonen WM, Farkas DK, Riis A, Fryzek JP, Sorensen HT. Risk of venous thromboembolism in splenectomized patients compared with the general population and appendectomized patients: a 10-year nationwide cohort study. J Thromb Haemost. 2010;8(6):1413-1416. 7. Hoeper MM, Niedermeyer J, Hoffmeyer F, Flemming P, Fabel H. Pulmonary hypertension after splenectomy? Ann Intern Med. 1999;130(6):506-509. 8. Yusuf HR, Hooper WC, Grosse SD, Parker CS, Boulet SL, Ortel TL. Risk of venous thromboembolism occurrence among adults with selected autoimmune diseases: a study among a U.S. cohort of commercial insurance enrollees. Thromb Res. 2015;135(1):50-57. 9. Ramagopalan SV, Wotton CJ, Handel AE, Yeates D, Goldacre MJ. Risk of venous thromboembolism in people admitted to hospital with selected immune-mediated

haematologica | 2017; 102(8)

10.

11.

12.

13.

14.

15. 16.

17. 18.

prevalence.28 Consequently, residual confounding is likely to be present in comparisons of splenectomized patients with the general population. Although we cannot rule out residual confounding in the comparisons with a diseasematched cohort either, the reported prevalences of comorbid diseases were similar between the splenectomized and disease-matched cohorts so that we assume residual confounding was smaller in these comparisons. More than 30% of the patients in our splenectomy cohort had another underlying diagnosis than the indications that we a priori had specified as the major underlying causes of splenectomy. This â&#x20AC;&#x153;other groupâ&#x20AC;? was not included in the indication-matched analyses; when compared with their matched cohort from the general population the relative estimates did not suggest that this group had a higher relative risk of the outcomes than those with selected underlying indications. Finally, even in our nationwide study, the statistical precision in some of our strata did not allow us to make firm conclusions. In conclusion, our study showed that splenectomy is associated with an increased risk of stroke, across the underlying indications for splenectomy. In contrast, any increased risk of MI and PH in splenectomized patients seemed to be related to the underlying indication rather than to the splenectomy itself.

diseases: record-linkage study. BMC Med. 2011;9:1. Galie N, Humbert M, Vachiery JL, et al. 2015 ESC/ERS Guidelines for the diagnosis and treatment of pulmonary hypertension: The Joint Task Force for the Diagnosis and Treatment of Pulmonary Hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS): Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC), International Society for Heart and Lung Transplantation (ISHLT). Eur Heart J. 2016;37(1):67-119. Peacock AJ. Pulmonary hypertension after splenectomy: a consequence of loss of the splenic filter or is there something more? Thorax. 2005;60(12):983-984. Yong M, Thomsen RW, Schoonen WM, Farkas DK, Riis A, Fryzek JP, et al. Mortality risk in splenectomised patients: a Danish population-based cohort study. Eur J Intern Med. 2010;21(1):12-16. Schmidt M, Pedersen L, Sorensen HT. The Danish Civil Registration System as a tool in epidemiology. Eur J Epidemiol. 2014;29(8): 541-549. Krarup LH, Boysen G, Janjua H, Prescott E, Truelsen T. Validity of stroke diagnoses in a national register of patients. Neuroepidemiology. 2007;28(3):150-154. Robinette CD, Fraumeni JF Jr. Splenectomy and subsequent mortality in veterans of the 1939-45 war. Lancet. 1977;2(8029):127-129. Schilling RF, Gangnon RE, Traver MI. Delayed adverse vascular events after splenectomy in hereditary spherocytosis. J Thromb Haemost. 2008;6(8):1289-1295. Schilling RF. Spherocytosis, splenectomy, strokes, and heat attacks. Lancet. 1997;350(9092):1677-1678. Taher A, Isma'eel H, Mehio G, et al. Prevalence of thromboembolic events among 8,860 patients with thalassaemia major and intermedia in the Mediterranean area and Iran. Thromb Haemost. 2006;96

(4):488-491. 19. Severinsen MT, Engebjerg MC, Farkas DK, et al. Risk of venous thromboembolism in patients with primary chronic immune thrombocytopenia: a Danish populationbased cohort study. Br J Haematol. 2011;152(3):360-362. 20. Blann AD, Lip GY. Venous thromboembolism. BMJ. 2006;332(7535):215-219. 21. Jais X, Ioos V, Jardim C, et al. Splenectomy and chronic thromboembolic pulmonary hypertension. Thorax. 2005;60(12):10311034. 22. Lin JN, Lin CL, Lin MC, et al. Increased risk of hemorrhagic and ischemic strokes in patients with splenic injury and splenectomy: a nationwide cohort study. Medicine (Baltimore). 2015;94(35):e1458. 23. Glynn RJ, Rosner B. Comparison of risk factors for the competing risks of coronary heart disease, stroke, and venous thromboembolism. Am J Epidemiol. 2005;162 (10):975-982. 24. Schmidt M, Schmidt SA, Sandegaard JL, Ehrenstein V, Pedersen L, Sorensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol. 2015;7:449-490. 25. Sorensen HT, Sabroe S, Olsen J. A framework for evaluation of secondary data sources for epidemiological research. Int J Epidemiol. 1996;25(2):435-442. 26. Buch P, Rasmussen S, Gislason GH, et al. Temporal decline in the prognostic impact of a recurrent acute myocardial infarction 1985 to 2002. Heart. 2007;93(2):210-215. 27. Sogaard M, Heide-Jorgensen U, Norgaard M, Johnsen SP, Thomsen RW. Evidence for the low recording of weight status and lifestyle risk factors in the Danish National Registry of Patients, 1999-2012. BMC Public Health. 2015;15:1320. 28. Kronborg CN, Hallas J, Jacobsen IA. Prevalence, awareness, and control of arterial hypertension in Denmark. J Am Soc Hypertens. 2009;3(1):19-24.e2.

1341

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Platelet Biology & its Disorders

Ferrata Storti Foundation

Safety and efficacy of romiplostim in splenectomized and nonsplenectomized patients with primary immune thrombocytopenia

Douglas B. Cines,1 Jeffrey Wasser,2 Francesco Rodeghiero,3,4 Beng H. Chong,5 Michael Steurer,6 Drew Provan,7 Roger Lyons,8 Jaime Garcia-Chavez,9 Nancy Carpenter,10 Xuena Wang11 and Melissa Eisen11

Haematologica 2017 Volume 102(8):x1342-1351

Perelman University of Pennsylvania School of Medicine, Philadelphia, PA, USA; University of Connecticut Health Center, Farmington, CT, USA; 3Hematology Project Foundation, Vicenza, Italy; 4San Bortolo Hospital, Vicenza, Italy; 5St George Hospital/University of New South Wales, Sydney, Australia; 6Medical University of Innsbruck, Austria; 7Barts and the London School of Medicine and Dentistry, London, UK; 8Texas Oncology and US ONCOLOGY Research, San Antonio, TX; 9Centro Médico Nacional La Raza, Mexico City, Mexico; 10Amgen Limited, Uxbridge, UK and 11Amgen Inc., Thousand Oaks, CA, USA 1 2

ABSTRACT

Correspondence: dcines@mail.med.upenn.edu

Received: December 15, 2016. Accepted: April 12, 2017. Pre-published: April 14, 2017. doi:10.3324/haematol.2016.161968 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1342 ©2017 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.

1342

rimary immune thrombocytopenia is an autoimmune disorder characterized by increased platelet destruction and insufficient platelet production without another identified underlying disorder. Splenectomy may alter responsiveness to treatment and/or increase the risk of thrombosis, infection, and pulmonary hypertension. The analysis herein evaluated the safety and efficacy of the thrombopoietin receptor agonist romiplostim in splenectomized and nonsplenectomized adults with primary immune thrombocytopenia. Data were pooled across 13 completed clinical studies in adults with immune thrombocytopenia from 2002-2014. Adverse event rates were adjusted for time of exposure. Results were considered different when 95% confidence intervals were non-overlapping. Safety was analyzed for 1111 patients (395 splenectomized; 716 nonsplenectomized) who received romiplostim or control (placebo or standard of care). At baseline, splenectomized patients had a longer median duration of immune thrombocytopenia and a lower median platelet count, as well as a higher proportion with >3 prior immune thrombocytopenia treatments versus nonsplenectomized patients. In each treatment group, splenectomized patients used rescue medications more often than nonsplenectomized patients. Platelet response rates (≥50x109/L) for romiplostim were 82% (310/376) for splenectomized and 91% (592/648) for nonsplenectomized patients (P<0.001 by Cochran-Mantel-Haenszel test). Platelet responses were stable over time in both subgroups. Exposure-adjusted adverse event rates were higher for control versus romiplostim for both splenectomized (1857 versus 1226 per 100 patient-years) and nonsplenectomized patients (1052 versus 852 per 100 patient-years). In conclusion, responses to romiplostim were seen in both splenectomized and nonsplenectomized patients, and romiplostim was not associated with an increase in the risk of adverse events in splenectomized patients. clinicaltrials.gov Identifier: 00111475(A)(B), 00117143, 00305435, 01143038, 00102323, 00102336, 00415532, 00603642, 00508820, 00907478, 00116688, and 00440037. Introduction In primary immune thrombocytopenia (ITP), increased platelet destruction and suboptimal platelet production result in low platelet counts, with bleeding symptoms that range from minimal to severe.1,2 The thrombopoietin (TPO) receptor agonist romiplostim stimulates platelet production.3 Romiplostim is approved in The USA for the treatment of chronic ITP in adults who have had an insufficient haematologica | 2017; 102(8)

Romiplostim in primary ITP by prior splenectomy

Table 1. Studies included in this analysis.

Study identifier* Parent studies 00111475(A) 00111475(B) 00117143 00305435 01143038 00102323 00102336 00415532 00603642 00508820 00907478 Extension studies 00116688 00440037

Study design

Control

No. Splenectomized / Nonsplenectomized**

Reference

Phase 1 dose-finding*** Phase 2 dose-finding*** Phase 2 dose-finding*** Phase 2 dose-finding*** Phase 2 Phase 3 Phase 3 Phase 3 Phase 3 Phase 3 Phase 4

None Placebo None None None Placebo Placebo SOC Placebo None None

19/5 14/7 13/3 3/9 0/75 63/0 0/62 0/229 15/19 208/198 60/109

6 6 7 8 9 10 10 11 12 13 14

Open-label extension Open-label extension

None None

94/197 17/27

15 16

*Registry number from clinicaltrials.gov Identifier. **Number of adults who received study treatment. The total of 335 patients in the 2 open-label extension studies previously participated in a parent study and are counted twice in the table. If a patient received control in the parent study and romiplostim in the open-label extension study, data from the parent study were included in the control group and data from the extension study were included in the romiplostim group. ***Dose-finding studies were not included in efficacy analyses because they used off-label doses of romiplostim. SOC: standard of care.

response to corticosteroids, intravenous immunoglobulins (IVIGs), or splenectomy,4 and in Europe for adults with chronic ITP who are refractory to other treatments (e.g., corticosteroids, IVIGs).5 Across 13 clinical studies of romiplostim in patients with primary ITP,6-16 romiplostim improved platelet counts, reduced bleeding, and reduced the use of concomitant medications compared with placebo or standard of care (SOC). The frequencies of on-study splenectomy and serious adverse events (AEs) were lower in patients treated with romiplostim compared with SOC.11 The spleen is an important site of antigen presentation, antibody production, and perpetuation of the autoimmune response in patients with ITP.17-19 Splenectomy, which reduces clearance of antibody-coated platelets and may attenuate antibody production, induces a complete clinical remission in approximately two-thirds of patients.20-22 Treatment guidelines include splenectomy as a second-line therapy for adults with ITP (after first-line therapy with corticosteroids, intravenous anti-D, or IVIG),1,2 based on extensive clinical experience and evidence to support the benefit/risk profile for splenectomy. However, approximately one-third of patients relapse or fail to respond to splenectomy and may require additional therapy.20,21 The impact of splenectomy on the safety and efficacy of treatment with romiplostim is not well described.23 Splenectomized patients were enrolled in some of the romiplostim ITP studies, but the benefits and risks in this subpopulation were not specifically addressed in previous pooled analyses.24,25 One concern about inducing sustained responses in patients with a prior splenectomy involves the potential to increase the risk of thrombosis, including portal and mesenteric venous thrombosis in the perioperative period after splenectomy in general26,27 and specifically in patients with ITP.28,29 The latter is complicated somehaematologica | 2017; 102(8)

what by the finding that ITP itself carries a slightly increased risk of thrombosis.28-32 Increased thrombosis after splenectomy may be associated with the eventual development of pulmonary hypertension,26,33-35 or atherosclerosis.26 The mechanism by which splenectomy predisposes to thromboembolism is unclear but, in theory, a sustained increase in platelet count could contribute to this risk.30 Splenectomy also increases the risk of infection, especially overwhelming sepsis caused primarily by encapsulated bacteria.30,36 Postsplenectomy infection is more common in the first 2 years after splenectomy. However, infection remains a lifelong risk because the spleen enhances clearance of antibody- and complement-coated particulate antigens, and it produces opsonins and immunologic memory cells.30,37,38 The long-term risk of postsplenectomy infection may be reduced by vaccination or prophylactic antibiotics.30,38 Therefore, on the one hand, splenectomy removes a major site of platelet sequestration,39,40 which may improve responsiveness to stimulation of platelet production by romiplostim. On the other hand, patients who have undergone splenectomy may comprise a subset of patients with more protracted, severe, and unresponsive disease who may also prove to be refractory to romiplostim and may be at greater risk of thrombosis and infection. This analysis was conducted to evaluate the safety and efficacy of romiplostim in splenectomized and nonsplenectomized adults with primary ITP.

Methods Patients and studies Methods for the pooled analyses were reported previously.24,25 Data were analyzed from 13 studies of romiplostim in adults with 1343

D.B. Cines et al.

Figure 1. Patient disposition. *One nonsplenectomized patient entered the extension study but did not receive romiplostim; the data from this patient were excluded.

primary ITP conducted between 2002 and 2014, including 5 controlled studies, 6 single-arm studies, and 2 open- label extension studies (Table 1).6-16 These studies were conducted in compliance with regulatory obligations, including institutional review board and informed consent regulations. Patients were diagnosed with primary ITP per American Society of Hematology guidelines.41 Patients received subcutaneous romiplostim, placebo, or SOC, along with concomitant or rescue medications (such as IVIG, but excluding other TPO mimetics and investigational products) as required and allowed per study. Platelet counts were targeted to be in the range of 50-200x109/L. In some earlier studies, the dose of romiplostim was adjusted between 1 and 15 mg/kg/week;10,13 in all other studies, the dose was adjusted between 1 and 10 mg/kg/week.

Assessments and statistical methods Measurements of platelet counts and use of other ITP treatments were documented at each visit. AE assessments were based on symptoms at any time during the study. Investigators evaluated AEs as to causality and severity (1=mild to 5=fatal). A serious AE was fatal, life-threatening, required (or prolonged) hospitalization, resulted in significant disability/incapacity, or was another significant complication. Amgen monitored serious AEs on an ongoing basis, and all AEs were reviewed quarterly. Other safety data, including laboratory values, were reviewed on an ad hoc basis for individual patients and on an ongoing basis for emerging trends. Bone marrow findings from a long-term study of bone marrow morphology were reported separately;14 1344

other data from that study not related to bone marrow findings were included in this pooled analysis. Bone marrow findings from AE reports in all other studies and available bone marrow biopsy results from 1 of the open-label extension studies15 were reported using the modified Bauermeister grading scale. The total number of bone marrow biopsies performed was unknown because results were only reported if the outcome was considered to be an AE, with the exception of the extension study noted above. Data from the placebo and SOC treatment arms were pooled. Unless otherwise indicated, results were adjusted for study duration and reported as events per 100 patient-years (calculated as 100x the number of events/patient-year), in order to reflect the unequal study duration between patients who received romiplostim and those who received placebo/SOC. When patients were enrolled in 2 consecutive studies, data from the parent and extension studies were combined. For patients who initially received placebo or SOC and then romiplostim, data prior to the first dose of romiplostim were included in the placebo/SOC group, and data beginning on the day the first dose of romiplostim was given were included in the romiplostim group, regardless of any subsequent change in treatment. Analyses were performed separately for patients who were splenectomized or nonsplenectomized before the parent study. Comparisons between splenectomized and nonsplenectomized patients included prespecified tests for platelet responses (P-values from Cochran-MantelHaenszel tests) and ad hoc analysis for other endpoints (95% confidence intervals). haematologica | 2017; 102(8)

Romiplostim in primary ITP by prior splenectomy

Table 2. Patient characteristics.

Characteristic Age, years, median (95% CI) Female, n (%) Years since ITP diagnosis, median (95% CI) >3 prior ITP therapies, n (%) (95% CI) Baseline platelet count x109/L, median (95% CI)

Splenectomized (N=395)

Nonsplenectomized (N=716)

52.0 (50.0, 55.0) 254 (64.3) 8.7 (7.7, 9.7) 150 (38.0) (33.2, 42.8) 14.0 (12.0, 15.3)

53.0 (52.0, 56.0) 431 (60.2) 1.6 (1.4, 2.0) 84 (11.7) (9.4, 14.1) 19.3 (18.0, 21.0)

CI: confidence interval; ITP: immune thrombocytopenia.

Table 3. Exposure-adjusted rates of AEs per 100 pt-yr (95% CI).

AE category

Any AE Any serious AE Any fatal AE Any treatment-related AE Any treatment-related serious AE

Romiplostim* Splenectomized Nonsplenectomized N=391** N=655** (702.0 pt-yr) (1129.7 pt-yr) 1226 (1201, 1253) 68.1 (62.1, 74.5) 1.6 (0.8, 2.8) 123.1 (115.0, 131.6) 9.3 (7.1, 11.8)

852 (835, 869) 44.1 (40.3, 48.1) 2.7 (1.9, 3.9) 82.1 (76.9, 87.6) 5.2 (4.0, 6.7)

Placebo/SOC Splenectomized Nonsplenectomized N=27** N=106** (11.2 pt-yr) (97.7 pt-yr) 1857 (1613, 2127) 133.9 (75.0, 220.9) 26.8 (5.5, 78.3) 133.9 (75.0, 220.9) 0.0 (0.0, 32.9)

1052 (989, 1119) 94.2 (75.9, 115.5) 5.1 (1.7, 11.9) 155.6 (131.8, 182.4) 18.4 (10.9, 29.1)

Results Studies and patients Results from a total of 1111 patients enrolled in a parent study were analyzed, including 395 who were splenectomized and 716 who were nonsplenectomized (Figure 1). Of the 1111 patients, 978 (368 splenectomized, 610 nonsplenectomized) received only romiplostim, 65 (4 splenectomized, 61 nonsplenectomized) received only placebo/SOC, and 68 (23 splenectomized, 45 nonsplenectomized) received placebo/SOC in a parent study and romiplostim in an open-label extension study. Safety outcomes for the latter 68 patients were analyzed in both groups; events before the switch were attributed to placebo/SOC and events after the switch were attributed to romiplostim. Thus, the safety analyses included data for 1046 patients in the romiplostim group (391 splenectomized and 655 nonsplenectomized) and 133 in the placebo/SOC group (27 splenectomized and 106 nonsplenectomized). A similar proportion of splenectomized and nonsplenectomized patients discontinued the parent study. Median age and sex were similar between the two cohorts, but splenectomized patients had a longer duration of ITP and lower baseline platelet counts versus nonsplenectomized patients, and a higher proportion of splenectomized patients had received more than 3 prior ITP treatments (Table 2).

of 50-200x109/L, median platelet counts were maintained within the target range in most splenectomized and nonsplenectomized patients (Figure 2A). A platelet response (at least 1 platelet count â&#x2030;Ľ50x109/L without rescue medication use in the previous 4 weeks) was attained in 82% of splenectomized patients and 91% of nonsplenectomized patients (Figure 2B). A sustained platelet response was attained in 68% of splenectomized patients and 80% of nonsplenectomized patients (Figure 2B). For this analysis, a sustained platelet response was defined as platelet counts â&#x2030;Ľ50x109/L for 9 out of 12 weeks (75% of weekly assessments), with no use of rescue medication during the 4 weeks prior to each qualifying platelet count. The qualifying criterion for sustained response was any 12-week interval that started with a platelet count â&#x2030;Ľ50x109/L. Patients with more than 1 sustained platelet response were counted only once in the analysis. Response rates and sustained response rates were lower in splenectomized patients than in nonsplenectomized patients (P<0.001 by Cochran-Mantel-Haenszel test both for any platelet response and for sustained platelet response). The use of rescue medication was higher in splenectomized than in nonsplenectomized patients within each treatment group (Figure 2C). In both splenectomized and nonsplenectomized patients, the use of corticosteroids, IVIG, anti-D, and rituximab decreased from baseline with long-term treatment and follow up (Figure 3).

Exposure Efficacy Response based on platelet count was analyzed in 1024 patients treated with romiplostim (376 splenectomized; 648 nonsplenectomized); 22 patients in dose-finding studies received off-label doses of romiplostim and were not included in efficacy analyses. Using a target platelet count haematologica | 2017; 102(8)

Splenectomized patients received romiplostim for up to 281 weeks and nonsplenectomized patients for up to 283 weeks, with mean (standard deviation [SD]) treatment durations of 87.3 (75.5) and 82.2 (60.0) weeks, respectively. The mean (SD) dose of romiplostim used most frequently was 4.9 (4.0) mg/kg for splenectomized patients 1345

D.B. Cines et al. A

B (P<0.001)

and 4.4 (3.4) Îźg/kg for nonsplenectomized patients (Figure 4A). Mean doses of romiplostim during >50 months of treatment were also similar comparing splenectomized to nonsplenectomized patients (Figure 4B). Total exposure to romiplostim was 702.0 patient years in splenectomized patients and 1129.7 patient-years in nonsplenectomized patients. Total exposure to placebo/SOC was 11.2 patientyears in splenectomized patients and 97.7 patient-years in nonsplenectomized patients.

Safety In each treatment group, exposure-adjusted rates of AEs, serious AEs, treatment-related AEs, and treatment-related serious AEs were higher among splenectomized versus 1346

(P<0.001)

Figure 2. Platelet count and rescue medication use. (A) Median platelet counts. (B) Platelet response rates to romiplostim without rescue medication use in the previous 4 weeks. Patients with more than 1 sustained platelet response were counted only once. (C) Rescue medication use per 100 pt-yr. Excludes a controlled study of nonsplenectomized patients,11 in which rescue medication use was reported inconsistently; thus, the placebo/SOC group for rescue medication use included only placebo. BL: baseline; CI: confidence interval; pt-yr: patient-year(s); Q1: quartile 1; Q3: quartile 3. SOC: standard of care.

nonsplenectomized patients (Table 3). The confidence intervals around the rates for splenectomized and nonsplenectomized patients did not overlap for AE rates and serious AE rates in either the romiplostim group or in the placebo/SOC group. Exposure-adjusted rates of AEs and serious AEs (Table 3) were lower with romiplostim treatment versus treatment with placebo/SOC in both splenectomized and nonsplenectomized patients. For each comparison, the confidence intervals around the rates for romiplostim and placebo/SOC did not overlap, showing a significant reduction in the event rates. The rate ratio for exposure-adjusted AE rates for romiplostim compared with placebo/SOC was 0.81 among nonsplenectomized patients (852 versus haematologica | 2017; 102(8)

Romiplostim in primary ITP by prior splenectomy

Table 4. Rates of AEs of interest per 100 pt-yr (95% CI).

AE of interest

Hemorrhagic Thrombotic Reticulin*** Any infection Possible opportunistic infection**** Systemic infection****

Romiplostim* Splenectomized Nonsplenectomized N=391** N=655** (702.0 pt-yr) (1129.7 pt-yr) 266.1 (254.2, 278.4) 6.3 (4.6, 8.4) 0.4 (0.0, 1.3) (N=331; 560.6 pt-yr) 156.7 (147.6, 166.2) 8.7 (6.6, 11.2) 2.1 (1.2, 3.5)

140.8 (134.0, 147.9) 4.6 (3.4, 6.0) 0.6 (0.2, 1.3) (N=546; 866.7 pt-yr) 124.8 (118.4, 131.5) 4.5 (3.4, 5.9) 0.4 (0.1, 0.9)

Splenectomized N=27** (11.2 pt-yr)

Placebo/SOC Nonsplenectomized N=106** (97.7 pt-yr)

482.1 (362.2, 629.1) 8.9 (0.2, 49.7) 0.0 (0.0, 32.9)

238.5 (208.8, 271.2) 5.1 (1.7, 11.9) 0.0 (0.0, 3.8)

196.4 (123.1, 297.4) 0.0 (0.0, 32.9) 0.0 (0.0, 32.9)

112.6 (92.5, 135.7) 5.1 (1.7, 11.9) 3.1 (0.6, 9.0)

*Any event reported after the first dose of romiplostim. **Of the 1111 patients, 978 (368 splenectomized, 610 nonsplenectomized) received only romiplostim, 65 (4 splenectomized, 61 nonsplenectomized) received only placebo/SOC, and 68 (23 splenectomized, 45 nonsplenectomized) received placebo/SOC in a parent study and romiplostim in an open-label extension study. Safety outcomes for the latter 68 patients were analyzed in both groups: events before the switch were attributed to placebo/SOC and events after the switch were attributed to romiplostim. ***AEs reported as bone marrow reticulin fibrosis, myelofibrosis, or reticulin increase across 12 studies; excluded 1 single-arm romiplostim study of immune thrombocytopenia specifically designed for bone marrow assessment (reported separately14). ****Reported terms for infection AEs were reviewed to identify possible opportunistic infections and systemic infections (Online Supplementary Table S3). AE: adverse event; CI: confidence interval; SOC: standard of care; pt-yr: patientyear(s).

1052 per 100 patient-years) and 0.66 among splenectomized patients (1226 versus 1857 per 100 patient-years). Exposure-adjusted rates of serious AEs in the romiplostim group were 53% lower than in the placebo/SOC group among nonsplenectomized patients (44.1 versus 94.2 per 100 patient-years) and 49% lower among splenectomized patients (68.1 versus 133.9 per 100 patient-years). The exposure-adjusted rate of any serious AE tended to be highest among splenectomized patients in the placebo/SOC group, followed in turn by nonsplenectomized patients in the placebo/SOC group, splenectomized patients in the romiplostim group, and nonsplenectomized patients in the romiplostim group. The relatively low number of fatal AEs made comparisons difficult. Several prespecified AEs of interest were analyzed separately (Table 4). Splenectomized patients had a higher incidence of bleeding than nonsplenectomized patients within each treatment group. However, the incidence of bleeding was lower in the romiplostim group than in the placebo/SOC group in both cohorts. AE rates for infections were similar between the romiplostim and placebo/SOC groups, with a small increase in infections in splenectomized patients in each treatment group. Rates of AEs reported as bone marrow reticulin fibrosis, myelofibrosis, or reticulin increase were too small for meaningful comparisons between treatment groups or across subgroups. The overall rate of any thrombotic AE was similar between the romiplostim and placebo/SOC groups and similar between splenectomized and nonsplenectomized patients within each treatment group. Types of thrombotic AEs (Online Supplementary Table S1) were also similar between treatment groups and between splenectomized and nonsplenectomized patients within each treatment group. The most commonly reported thrombotic AEs in the romiplostim and placebo/SOC groups were deep vein thrombosis (1.0 versus 0.9 events per 100 patient-years) and pulmonary embolism (0.8 versus 0.9 events per 100 patient-years). The median platelet count at the last measurement before a thrombotic AE was 150.0x109/L in patients who received romiplostim and 52.5x109/L in patients who received only haematologica | 2017; 102(8)

placebo/SOC, and the platelet counts before a thrombotic event were similar between splenectomized and nonsplenectomized patients in the romiplostim group. The percentage of time that patients in the romiplostim group had a platelet count of 50x109/L or greater, or 200x109/L or greater, was similar between patients with or without thrombotic AEs (Online Supplementary Table S2). Neutralizing antibodies to romiplostim were reported for 6 patients: 2 before romiplostim treatment and 4 after treatment was initiated. All 6 patients had a platelet response and a sustained platelet response, and none lost response to romiplostim following the detection of neutralizing antibodies.

Discussion In this analysis of 1111 patients across 13 completed clinical studies of the TPO receptor agonist romiplostim, more than 1 in 3 patients underwent splenectomy before they entered the parent study. Platelet response rates following treatment with romiplostim were high in both splenectomized and nonsplenectomized patients, and median platelet counts were generally maintained in the target range with long-term romiplostim treatment in both populations. The observed response rate was higher in nonsplenectomized patients than in splenectomized patients. A sustained platelet response for 9 out of 12 weeks was achieved by 68% of splenectomized and 80% of nonsplenectomized patients given romiplostim in this pooled analysis. In 2 pivotal phase 3 studies, 38% of splenectomized and 61% of nonsplenectomized patients given romiplostim achieved durable platelet responses.10 In those studies, a durable platelet response was defined as platelet counts â&#x2030;Ľ50x109/L for â&#x2030;Ľ6 of the last 8 weeks. Although the same ratio (75% of the weekly assessments) was used to define sustained platelet responses in this pooled analysis, the start of the 12-week interval was based on a platelet count of â&#x2030;Ľ50x109/L. Thus, the pivotal studies examined the proportion of patients who achieved durable platelet responses during a specific 8 week interval after dose stabilization, while this analysis examined the 1347

D.B. Cines et al. A

Figure 3. Use of other ITP medications in the romiplostim group: prevalence by postbaseline quarter. (A) Corticosteroids. (B) IVIG; (C) Anti-D; (D) Rituximab. Ig: immunoglobulin.

proportion of patients who sustained the response over any 12-week interval that began with a platelet response. Interpretation of these results is complicated by the fact that the efficacy analysis pooled data across 9 studies (4 dose-finding studies used off-label doses of romiplostim and were not included in the efficacy analyses), and the splenectomized and nonsplenectomized patients were not balanced for other baseline characteristics. Splenectomized patients had more severe disease at baseline, as shown by the longer duration of ITP, a higher proportion that had used more than 3 prior ITP treatments, and a lower baseline platelet count at study entry compared with nonsplenectomized patients. The timing of splenectomy relative to other prior treatments was uncertain. Therefore, patients might have undergone splenectomy as second-line therapy or after multiple lines of therapy had failed.1,2,41 Although the platelet responses to prior treatments were not recorded in these studies, the fact that 38% of splenectomized patients and 12% of nonsplenectomized patients received more than 3 prior lines of therapy suggests that the splenectomized patients in this analysis included more patients with treatment-resistant ITP. Exposure-adjusted rates of AEs and serious AEs 1348

were lower in the romiplostim group than in the placebo/SOC group. The difference in AE rates between the treatment groups was greater among splenectomized patients than among nonsplenectomized patients. Specifically, although splenectomized patients appeared to have more severe disease than nonsplenectomized patients at baseline, the responses to romiplostim were similar between the cohorts. Moreover, treatment with romiplostim was associated with a reduction of approximately 70% in the use of rescue medication compared with the placebo/SOC group in both splenectomized and nonsplenectomized patients. This reduction appeared to occur after initial dose titration for romiplostim, based on analyses of ITP medication use over time. The substantially greater use of concomitant medications in the placebo/SOC group may have contributed to the higher rates of AEs and serious AEs compared with patients treated with romiplostim. Notwithstanding the higher rate of AEs in splenectomized patients, the rate of serious AEs among this cohort who were treated with romiplostim was lower than that in nonsplenectomized patients in the placebo/SOC group. Thus, treatment had a greater impact on reducing the rate of serious AEs than did a prior history haematologica | 2017; 102(8)

Romiplostim in primary ITP by prior splenectomy

Figure 4. Romiplostim dosing. (A) Most frequent romiplostim dose. (B) Romiplostim dose over time by splenectomy status. SD: standard deviation.

of splenectomy. No new safety issues or major increases in AEs were seen among romiplostim-treated splenectomized patients compared with romiplostim-treated nonsplenectomized patients in this analysis, which included the examination of specific AEs such as bleeding, infection, thrombosis, and bone marrow fibrosis. Previous research has shown that splenectomized patients have approximately a 2-fold to 4-fold greater risk of thrombotic events than nonsplenectomized patients,26,27 and the risk may remain elevated more than 1 year after splenectomy.27 In this analysis, patients with a prior history of thrombotic events were excluded from some of the studhaematologica | 2017; 102(8)

ies but were permitted to enroll in other studies.25 Therefore, we cannot speculate on how many patients with prior thrombotic events were not enrolled in these studies, which could have influenced the risk of subsequent thrombotic events after initiation of romiplostim or placebo/SOC treatment. Rates of bone marrow reticulin fibrosis, myelofibrosis, or reticulin AEs were too low for meaningful comparisons between the splenectomized and nonsplenectomized patients, or between the romiplostim and placebo/SOC groups, but no worrisome signals were observed. These analyses were limited by the fact that in most of the stud1349

D.B. Cines et al.

ies, bone marrow data were only collected for patients with a documented bone marrow AE. Additionally, the actual time at which bone marrow AEs occurred could not be evaluated; only the time of detection by bone marrow biopsy was known. One of the studies included in this pooled analysis was a single-arm study that was designed to evaluate bone marrow findings from biopsies 1, 2, or 3 years after initiation of treatment with romiplostim. A separate publication of those findings14 showed that bone marrow changes were observed in a small proportion of patients who received long-term romiplostim treatment, including 3 out of 60 splenectomized patients and 6 out of 109 nonsplenectomized patients. Several limitations of this pooled analysis should be noted. First, comparisons of efficacy and safety between splenectomized and nonsplenectomized patients were not among the planned analyses within the individual romiplostim studies. Some of the phase 3 studies enrolled only splenectomized patients or only nonsplenectomized patients; other studies enrolled both splenectomized and nonsplenectomized patients. The studies that enrolled both splenectomized and nonsplenectomized patients did not collect data for splenectomy during the study, which made it impossible to examine splenectomy rates in the romiplostim and placebo/SOC groups. In a 52-week study that was designed to address this question in nonsplenectomized patients, the rates of splenectomy on-study were significantly lower for romiplostim than SOC (9% versus 36%).11 Second, investigators were not asked to report if an AE was related to prior splenectomy, so it was not possible to confirm the relatedness of splenectomy to safety outcomes. Third, comparisons between the romiplostim and placebo/SOC groups were limited by the differences in exposure to study treatment in each group. Eight out of 13 studies in this pooled analysis were single-arm evaluations of romiplostim, including 2 long-term open-label extension studies. Consequently, the sample sizes in this pooled analysis were much larger for romiplostim than for placebo/SOC and the total exposure to placebo/SOC was limited to 6 months or less for individual patients, whereas exposure to romiplostim could continue for several years. To address these discrepancies, analyses of AEs and rescue medication use were based on exposure-adjusted

References 1. Provan D, Stasi R, Newland AC, et al. International consensus report on the investigation and management of primary immune thrombocytopenia. Blood. 2010;115(2):168-186. 2. Neunert C, Lim W, Crowther M, Cohen A, Solberg L, Jr., Crowther MA. The American Society of Hematology 2011 evidencebased practice guideline for immune thrombocytopenia. Blood. 2011;117(16): 4190-4207. 3. Molineux G. The development of romiplostim for patients with immune thrombocytopenia. Ann N Y Acad Sci. 2011;1222:55-63. 4. NPLATEÂŽ (romiplostim) prescribing information. Thousand Oaks, CA: Amgen, Inc., 2016.

1350

rates. After subdividing the population by treatment and splenectomy status, total exposure to placebo/SOC was only 11.2 patient-years in splenectomized patients and 97.7 patient-years in nonsplenectomized patients, versus exposure to romiplostim of 702.0 and 1129.7 patientyears, respectively. For the patients who received placebo/SOC in a controlled study and later received romiplostim in an extension study, efficacy and safety outcomes after the first dose of romiplostim were considered to be related to that treatment, but it is possible that SOC treatment could have impacted subsequent efficacy or safety outcomes. Fourth, the use of medications prior to study entry could have impacted efficacy and safety during the studies, but one of the pivotal studies (approximately 40% of patients in the analysis) did not collect information on prior medications. Moreover, in the other studies the only information available was whether a patient had received a medication previously, without details on the nature, intensity, or duration of prior treatment. Lastly, patients were permitted to receive rescue treatment such as IVIG, but not other investigational products or other TPO mimetics, which complicated the interpretation of platelet responses. Notwithstanding these limitations, the analyses reported herein address an important data gap regarding the safety and efficacy of romiplostim in splenectomized patients for whom therapeutic options are more limited.23 In conclusion, this pooled analysis of data from 13 completed studies showed that romiplostim has a favorable safety profile, regardless of splenectomy status before treatment with romiplostim. Long-term treatment with romiplostim maintained platelet counts in the target range in both splenectomized and nonsplenectomized patients, and no new safety signal emerged. Importantly, romiplostim did not increase the risk of thromboembolic events in splenectomized patients compared to placebo/SOC. Based on this post hoc analysis, romiplostim appears to be a treatment option for patients with ITP with or without a history of failed splenectomy. Funding This research, including both the original studies and the pooled analysis, was funded by Amgen Inc.

5. NPlate (romiplostim) summary of product characteristics. Breda, The Netherlands: Amgen B.V., 2013. 6. Bussel JB, Kuter DJ, George JN, et al. AMG 531, a thrombopoiesis-stimulating protein, for chronic ITP. N Engl J Med. 2006;355(16):1672-1681. 7. Newland A, Caulier MT, Kappers-Klunne M, et al. An open-label, unit dose-finding study of AMG 531, a novel thrombopoiesis-stimulating peptibody, in patients with immune thrombocytopenic purpura. Br J Haematol. 2006;135(4):547-553. 8. Shirasugi Y, Ando K, Hashino S, et al. A phase II, open-label, sequential-cohort, dose-escalation study of romiplostim in Japanese patients with chronic immune thrombocytopenic purpura. Int J Hematol. 2009;90(2):157-165. 9. Newland A, Godeau B, Priego V, et al.

10.

11.

12.

13.

Remission and platelet responses with romiplostim in primary immune thrombocytopenia: final results from a phase 2 study. Br J Haematol. 2016;172(2):262-273. Kuter DJ, Bussel JB, Lyons RM, et al. Efficacy of romiplostim in patients with chronic immune thrombocytopenic purpura: a double-blind randomised controlled trial. Lancet. 2008;371(9610):395-403. Kuter DJ, Rummel M, Boccia R, et al. Romiplostim or standard of care in patients with immune thrombocytopenia. N Engl J Med. 2010;363(20):1889-1899. Shirasugi Y, Ando K, Miyazaki K, et al. Romiplostim for the treatment of chronic immune thrombocytopenia in adult Japanese patients: a double-blind, randomized Phase III clinical trial. Int J Hematol. 2011;94(1):71-80. Janssens A, Tarantino M, Bird RJ, et al.

haematologica | 2017; 102(8)

Romiplostim in primary ITP by prior splenectomy

14.

15.

16.

17.

18.

19.

20.

21.

Romiplostim treatment in adults with immune thrombocytopenia of varying duration and severity. Acta Haematol. 2015;134(4):215-228. Janssens A, Rodeghiero F, Anderson D, et al. Changes in bone marrow morphology in adults receiving romiplostim for the treatment of thrombocytopenia associated with primary immune thrombocytopenia. Ann Hematol. 2016;95(7):1077-1087. Kuter DJ, Bussel JB, Newland A, et al. Longterm treatment with romiplostim in patients with chronic immune thrombocytopenia: safety and efficacy. Br J Haematol. 2013;161(3):411-423. Shirasugi Y, Ando K, Miyazaki K, et al. An open-label extension study evaluating the safety and efficacy of romiplostim for up to 3.5 years in thrombocytopenic Japanese patients with immune thrombocytopenic purpura (ITP). Int J Hematol. 2012;95(6):652-659. Audia S, Rossato M, Santegoets K, et al. Splenic TFH expansion participates in B-cell differentiation and antiplatelet-antibody production during immune thrombocytopenia. Blood. 2014;124(18):2858-2866. Audia S, Samson M, Guy J, et al. Immunologic effects of rituximab on the human spleen in immune thrombocytopenia. Blood. 2011;118(16):4394-4400. Daridon C, Loddenkemper C, Spieckermann S, et al. Splenic proliferative lymphoid nodules distinct from germinal centers are sites of autoantigen stimulation in immune thrombocytopenia. Blood. 2012;120(25):5021-5031. Kojouri K, Vesely SK, Terrell DR, George JN. Splenectomy for adult patients with idiopathic thrombocytopenic purpura: a systematic review to assess long-term platelet count responses, prediction of response, and surgical complications. Blood. 2004;104(9):2623-2634. Vianelli N, Palandri F, Polverelli N, et al. Splenectomy as a curative treatment for immune thrombocytopenia: a retrospective analysis of 233 patients with a minimum

haematologica | 2017; 102(8)

22.

23.

24.

25.

26. 27.

28.

29.

30.

follow up of 10 years. Haematologica. 2013;98(6):875-880. Vianelli N, Galli M, de Vivo A, et al. Efficacy and safety of splenectomy in immune thrombocytopenic purpura: longterm results of 402 cases. Haematologica. 2005;90(1):72-77. Perdomo J. Role of romiplostim in splenectomized and nonsplenectomized patients with immune thrombocytopenia. Immunotargets Ther. 2016;5:1-7. Cines DB, Gernsheimer T, Wasser JS, et al. Integrated analysis of long-term safety in patients with chronic immune thrombocytopaenia (ITP) treated with the thrombopoietin (TPO) receptor agonist romiplostim. Int J Hematol. 2015;102(3):259270. Rodeghiero F, Stasi R, Giagounidis A, et al. Long-term safety and tolerability of romiplostim in patients with primary immune thrombocytopenia: a pooled analysis of 13 clinical trials. Eur J Haematol. 2013; 91(5):423-436. Crary SE, Buchanan GR. Vascular complications after splenectomy for hematologic disorders. Blood. 2009;114(14):2861-2868. Thomsen RW, Schoonen WM, Farkas DK, Riis A, Fryzek JP, Sorensen HT. Risk of venous thromboembolism in splenectomized patients compared with the general population and appendectomized patients: a 10-year nationwide cohort study. J Thromb Haemost. 2010;8(6):14131416. Boyle S, White RH, Brunson A, Wun T. Splenectomy and the incidence of venous thromboembolism and sepsis in patients with immune thrombocytopenia. Blood. 2013;121(23):4782-4790. Doobaree IU, Nandigam R, Bennett D, Newland A, Provan D. Thromboembolism in adults with primary immune thrombocytopenia: a systematic literature review and meta-analysis. Eur J Haematol. 2016; 97(4):321-330. Rodeghiero F, Ruggeri M. Short- and longterm risks of splenectomy for benign

31. 32.

33.

34.

35.

36.

37. 38. 39.

40.

41.

haematological disorders: should we revisit the indications? Br J Haematol. 2012; 158(1):16-29. Rodeghiero F. Is ITP a thrombophilic disorder? Am J Hematol. 2016;91(1):39-45. Langeberg WJ, Schoonen WM, Eisen M, Gamelin L, Stryker S. Thromboembolism in patients with immune thrombocytopenia (ITP): a meta-analysis of observational studies. Int J Hematol. 2016;103(6):655664. Hoeper MM, Niedermeyer J, Hoffmeyer F, Flemming P, Fabel H. Pulmonary hypertension after splenectomy? Ann Intern Med. 1999;130(6):506-509. Jais X, Ioos V, Jardim C, et al. Splenectomy and chronic thromboembolic pulmonary hypertension. Thorax. 2005;60(12):10311034. Peacock AJ. Pulmonary hypertension after splenectomy: a consequence of loss of the splenic filter or is there something more? Thorax. 2005;60(12):983-984. Kristinsson SY, Gridley G, Hoover RN, Check D, Landgren O. Long-term risks after splenectomy among 8,149 cancer-free American veterans: a cohort study with up to 27 years follow-up. Haematologica. 2014;99(2):392-398. Cadili A, de Gara C. Complications of splenectomy. Am J Med. 2008;121(5):371375. Di Sabatino A, Carsetti R, Corazza GR. Post-splenectomy and hyposplenic states. Lancet. 2011;378(9785):86-97. Navez J, Hubert C, Gigot JF, et al. Does the site of platelet sequestration predict the response to splenectomy in adult patients with immune thrombocytopenic purpura? Platelets. 2015;26(6):573-576. Psaila B, Bussel JB. Refractory immune thrombocytopenic purpura: current strategies for investigation and management. Br J Haematol. 2008;143(1):16-26. George JN, Woolf SH, Raskob GE, et al. Idiopathic thrombocytopenic purpura: a practice guideline developed by explicit methods for the American Society of Hematology. Blood. 1996;88(1):3-40.

1351

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Myeloproliferative DIsorders

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1352-1360

Bone marrow morphology is a strong discriminator between chronic eosinophilic leukemia, not otherwise specified and reactive idiopathic hypereosinophilic syndrome Sa A. Wang,1 Robert P. Hasserjian,2 Wayne Tam,3 Albert G. Tsai,4 Julia T. Geyer,3 Tracy I. George,5 Kathryn Foucar,5 Heesun J. Rogers,6 Eric D. Hsi,6 Bryan A. Rea,7 Adam Bagg,7 Carlos E. Bueso-Ramos,1 Daniel A. Arber,8 Srdan Verstovsek9 and Attilio Orazi3

Department of Hematopathology, M.D. Anderson Cancer Center, Houston, TX; Department of Pathology, Massachusetts General Hospital, Boston, MA; 3Department of Pathology and Laboratory Medicine, Weill Cornell Medical College, New York, NY; 4 Department of Pathology, Stanford University, CA; 5Department of Pathology, University of New Mexico, Albuquerque, NM; 6Department of Laboratory Medicine, Cleveland Clinic, OH; 7Department of Pathology and Laboratory Medicine, the University of Pennsylvania, Philadelphia, PA; 8Department of Pathology and Laboratory Medicine, University of Chicago, IL and 9Department of Leukemia, M.D. Anderson Cancer Center, Houston, TX, USA 1 2

ABSTRACT

Correspondence: swang5@mdanderson.org

Received: January 25, 2017. Accepted: May 4, 2017. Pre-published: May 11, 2017. doi:10.3324/haematol.2017.165340 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1352 ÂŠ2017 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.

1352

hronic eosinophilic leukemia, not otherwise specified can be difficult to distinguish from idiopathic hypereosinophilic syndrome according to the current World Health Organization guideline. To examine whether the morphological features of bone marrow might aid in the differential diagnosis of these two entities, we studied a total of 139 patients with a diagnosis of chronic eosinophilic leukemia, not otherwise specified (n=17) or idiopathic hypereosinophilic syndrome (n=122). As a group, abnormal bone marrow morphological features, resembling myelodysplastic syndromes, myeloproliferative neoplasm or myelodysplastic/myeloproliferative neoplasm, were identified in 40/139 (27%) patients: 16 (94%) of those with chronic eosinophilic leukemia and 24 (20%) of those with hypereosinophilic syndrome. Abnormal bone marrow correlated with older age (P<0.001), constitutional symptoms (P<0.001), anemia (P=0.041), abnormal platelet count (P=0.002), organomegaly (P=0.008), elevated lactate dehydrogenase concentration (P=0.005), abnormal karyotype (P<0.001), as well as the presence of myeloid neoplasm-related mutations (P<0.001). Patients with abnormal bone marrow had shorter survival (48.1 months versus not reached, P<0.001), a finding which was independent of other confounding factors (P<0.001). The association between abnormal bone marrow and shorter survival was also observed in hypereosinophilic syndrome patients alone. In summary, most patients with chronic eosinophilic leukemia, not otherwise specified and a proportion of those with idiopathic hypereosinophilic syndrome show abnormal bone marrow features similar to the ones encountered in patients with myelodysplastic syndromes, myelodysplastic/myeloproliferative neoplasm or BCR-ABL1-negative myeloproliferative neoplasm. Among patients who are currently considered to have idiopathic hypereosinophilic syndrome, abnormal bone marrow is a strong indicator of clonal hematopoiesis. Similar to other myeloid neoplasms, bone marrow morphology should be one of the major criteria to distinguish patients with chronic eosinophilic leukemia, not otherwise specified or clonal hypereosinophilic syndrome from those with truly reactive idiopathic hypereosinophilic syndrome. haematologica | 2017; 102(8)

BM morphology of CEL,NOS/idiopathic HES

Introduction Hypereosinophilia is defined by the presence of ≥1.5 x109/L eosinophils in the peripheral blood (PB) and may be reactive, neoplastic or idiopathic.1-3 Chronic eosinophilic leukemia, not otherwise specified (CEL, NOS)4 is a myeloproliferative neoplasm (MPN), characterized by an expansion of eosinophils but lacking well-defined molecular genetic alterations such as BCR-ABL1 and rearrangements of PDGFRA, PDGFRB, FGFR1 and PCM1-JAK2. In idiopathic hypereosinophilic syndrome (HES), there is tissue/organ damage related to an eosinophilic infiltrate/activation, but the cause of the hypereosinophilia is unknown. Due to substantial overlapping of their features, idiopathic HES was described alongside CEL, NOS in the World Health Organization (WHO) classification monograph;4 this categorization remains largely unmodified in the 2016 WHO revision.5 According to the current guidelines, CEL, NOS, can only be reliably separated from idiopathic HES by the presence of increased blasts in bone marrow (BM) and/or PB, or proof of clonality. Clonality was mainly determined by chromosomal analysis or testing for mutations well known to occur in MPN, such as JAK2, MPL, CALR and KIT. However, these mutations are uncommon in the eosinophilic diseases.6,7 More recently, next-generation sequencing (NGS) approaches have been applied to these eosinophilic disorders. Anderson and colleagues conducted whole-exome sequencing of nine patients with idiopathic HES8 and identified somatic missense mutations in three of them. The mutations they found involved the spliceosome gene PUF60 and the cadherin gene CDH17. More recently, using NGS with a gene panel targeted to somatic mutations commonly associated with myeloid neoplasms, we detected the presence of mutations at a relatively high (≥10%) variant allele frequency (VAF) in 25-30% of cases of idiopathic HES.7 These mutations mostly occurred in genes involved in DNA methylation and chromatin modification, such as AXSL1, TET2, EZH2, and DNMT3A. While such mutations would imply that some idiopathic HES are clonal stem cell neoplasms, they have also been reported in some aging individuals without evidence of a myeloid neoplasm,9,10 mandating caution in the use of mutations as definitive proof of a neoplastic myeloid process. On the other hand, the detection of mutations by NGS relies on the testing panel used, which may vary for the number of genes sequenced as well as the depth of sequencing. Although some studies have suggested that abnormal eosinophil morphology is associated with clonal eosinophilia, it is generally felt that cytological abnormalities lack sufficient specificity to differentiate a neoplastic process from a reactive eosinophilia.4,11-14 As a result, BM morphology is not an integral part of the diagnosis and classification of hypereosinophilia. This is in apparent contrast to the situation in other myeloid neoplasms, in which abnormal BM features play a major role in establishing the diagnosis. In particular, BM morphology represents a “gold standard” in the diagnosis of myelodysplastic syndromes (MDS) and MDS/MPN. With regards to MPN, morphology has become one of the major criteria in the WHO classification (2016)5 of essential thrombocythemia, polycythemia vera, and primary myelofibrosis. In contrast, in the case of CEL, NOS or of HES with clonal eosinophilia, there is limited published information in haematologica | 2017; 102(8)

relation to BM morphology. In our previous study,7 with molecular genetic information for patients, we observed some BM features that appeared to be preferentially present in cases with molecular genetic alterations. In this study, we conducted a thorough review of BM morphology of a large series of CEL, NOS and idiopathic HES cases collected from seven large medical centers in the USA using a defined set of morphological criteria, blinded to the original diagnosis and molecular genetic data. We used the morphological features to define an “abnormal” BM morphology, and correlated the morphological results with clinical and laboratory features, cytogenetics, mutation data, and patients’ outcomes. We sought to determine whether morphology can be utilized in the distinction of CEL, NOS and clonal HES from truly reactive idiopathic HES.

Methods Patients Cases were collected from MD Anderson Cancer Center, Stanford University Medical Center, Cleveland Clinic, Massachusetts General Hospital, Weill-Cornell Medical College, the Hospital of the University of Pennsylvania and the University of New Mexico between 2005 and 2014. All included patients had persistent hypereosinophilia (≥1.5x109/L) and did not have acute leukemias, chronic myeloid leukemia, MDS, chronic myelomonocytic leukemia, systemic mastocytosis, or rearrangements of PDGFRA, PDGFRB, FGFR1 or PCM1-JAK2. For idiopathic HES, every patient had “end-organ damage” according to the definition by the working group on eosinophil disorders.15 Lymphocytic/Tcell variant HES1 was excluded based on the identification of aberrant T cells by flow cytometry with or without TCR gene rearrangement polymerase chain reaction studies. Clinical information was retrieved from the electronic medical records. This study was approved by the Institutional Review Boards of all participating institutions.

Bone marrow morphology and histology BM was assessed for the following morphological parameters (Table 1): cellularity; megakaryocyte numbers, morphology and distribution; fibrosis; dysgranulopoiesis; dyserythropoiesis; myeloid:erythroid (M:E) ratio; and eosinophil morphology. PB smears were also reviewed for eosinophil morphology and evidence of dysgranulopoiesis. Hypercellularity was defined by a cellularity at least 20% higher than the age-appropriate cellularity, and overall ≥70% in patients age 50-60 years; ≥60% in patients >60 years; and ≥90% in patients <30 years of age. Megakaryocyte morphology was recorded as predominantly MDS-like (small with hypolobated/non-lobated nuclei or separated nuclear lobes), MPN-like (medium-sized to large megakaryocytes with hyperlobulated, hyperchromatic, or bulbous nuclei, often with clustering and increase in numbers), mixed MDS and MPN-like, or within normal limits (WNL). In order to define dysgranulopoiesis and dyserythropoiesis strictly, the features had to be seen in ≥20% of cells of the assessed lineage. Myelofibrosis grade was assessed according to the European Bone Marrow Fibrosis Consensus criteria.16 For eosinophil morphology, abnormal features were markedly abnormal granulation (hypogranulation or uneven granulation), cytoplasmic vacuoles, abnormal nuclear lobation (non-lobated or multilobated), unusually large size or markedly increased immature forms. These features had to be observed in at least 20% of the eosinophils on the BM smears, since mild nuclear hypersegmentation and mild abnormal granulation in the PB can be seen 1353

S.A. Wang et al.

with eosinophil activation17 or treatment with hydroxyurea.18 All cases were assessed by the members from the respective institution using the same set of criteria, which were developed by members of the Bone Marrow Pathology study group after having reviewed representative cases as a group. The features were reassessed by one observer (SAW); cases with borderline morphological abnormalities or discrepancy were again centrally reviewed by the group and scored by consensus. There was some disagreements on some of the parameters for approximately 10% of cases (n=13), but all members agreed on “abnormal” or “not abnormal” morphology for all cases. The disagreements mainly related to eosinophilic morphology, since the criteria were not previously defined; disagreements on scoring megakaryocyte morphology were present in a smaller subset of cases and centered on whether the features were MDS-like or mixed MDS/MPN-like. All morphology reviews were blinded to clinical features, molecular genetic data, original diagnoses and patients’ outcomes.

Cytogenetics, fluorescence in situ hybridization and molecular testing Conventional cytogenetic analysis was performed on G-banded metaphase cells prepared from unstimulated BM aspirate cultures using standard techniques. Twenty metaphases were analyzed and the results were reported using the International System for Human Cytogenetic Nomenclature. Fluorescence in situ hybridization (FISH) and/or molecular genetic methods for detecting BCRABL1, PDGFRA, PDGFRB, and FGFR1 were performed at the respective institutions as part of the routine clinical work-up, if indicated.

Targeted next-generation sequencing Targeted NGS had been performed on 57 patients previously7 and was performed in an additional 19 cases on DNA samples extracted from frozen unfractionated BM cells collected at the time of diagnosis, using the same method we described previous-

Table 1. Bone marrow morphological findings of patients with a diagnosis of chronic eosinopil leukemia, not otherwise specified or idiopathic hypereosinophilic syndrome

Morphological features Eosinophil percentage, median (range) • Patients with ≥10% BM eosinophils Cellularity, median (range) • Hypercellularity* MDS-like megakaryocytes** MPN-like megakaryocytes** Mixed MDS and MPN-like megakaryocytes** Dyserythropoiesis ** Dysgranulopoiesis ** Abnormal eosinophils *** M:E ratio, median (range) • M:E ratio >10 MF2 or MF3 fibrosis Morphologically abnormal****

Patients (n=139) 21% (4-91%) 136/139 (98%) 60% (1-100%) 43/129 (31%) 17/137 (12%) 2/137 (2%) 6/137 (4%) 9/135 (7%) 9/135 (7%) 25/134 (19%) 3.3(0.7-31.7) 16/129 (12%) 13/114 (11%) 40/139 (29%)

*At least 20-30% higher than age-appropriate cellularity or ≥90%; ** dysplastic/abnormal cells ≥20% of respective lineage, *** markedly abnormal granulation (hypogranulation or uneven granulation), cytoplasmic vacuoles, and/or abnormal nuclear lobation (monolobated or multinucleated); unusually large size or markedly increased immature forms, present in ≥20% cells, **** ≥20% abnormal megakaryocytes, erythroids or myeloid cells, or two of the following: hypercellularity, abnormal eosinophils, M:E ratio >10, or MF1-3 fibrosis. BM: bone marrow; MDS: myelodysplastic syndrome; MPN: myeloproliferative neoplasm; M:E: myeloid: erythroid; MF: myelofibrosis.

1354

ly.7 The coding sequences of 44 genes (sequencing >90% gene coding regions), including ABL1, ASXL1, BCOR, BRAF, CALR, CBL, CEBPA, DNMT3A, ETV6, EZH2, FAM5C, FLT3 (ITD and TKD), GATA1, GATA2, HNRNPK, IDH1, IDH2, IKZF1, JAK1, JAK2, KDM6A, KIT, KRAS, MPL, NFE2, NOTCH1, NPM1, NRAS, PHF6, PTPN11, RAD21, RUNX1, SEPBP1, SF3B1, SH2B3, SMC1A, SMC3, STAG2, SUZ12, TET2, TP53, U2AF1, WT1, and ZRSR2, were investigated specifically for this study. Variant calling was performed with Illumina MiSeq Reporter software 1.3.17. using human genome build 19 (hg 19) as a reference.

Statistical analyses Data for continuous variables were reported as medians and ranges. Data for nominal variables were reported as the number of patients unless otherwise specified. Survival was calculated from the date of diagnosis to the date of last follow-up or death not attributable to causes that were clearly not associated with disease (e.g., car accident, suicide). Patients who underwent hematopoietic stem cell transplantation were censored at the time of the procedure. Distribution of survival was estimated by Kaplan-Meier curves. Multivariable analysis was performed using a Cox regression model. Fisher exact and χ2 tests were used for categorical comparisons. All P values are two-tailed and considered statistically significant when <0.05. No adjustments for multiplicity were made.

Results Patients, clinical data and molecular genetic data A total of 139 patients were included in the study: these patients met the criteria for CEL, NOS (17 patients) or idiopathic HES (122 patients) after applying the exclusion and inclusion criteria and had sufficient material for morphological assessment. An abnormal karyotype was seen in 16 of the 17 CEL, NOS patients; detailed karyotype information on these cases was published previously.7 In brief, five patients had a complex karyotype, one had two cytogenetic abnormalities, nine had a single abnormality, and one was identified by FISH as having a del(9p) abnormality. Three patients had ≥5% BM blasts, including one patient with a normal BM karyotype who was also diagnosed as having CEL, NOS according to the WHO Classification criteria. The clinical and laboratory features of these patients as a group are shown in Table 2. NGS was performed in 76 patients. In total, mutations were found in 21/76 patients (27.6%). The mutation data and frequency are shown in Figure 1. In brief, the mutations, in decreasing frequency, were: ASXL1 (7/76, 9.2%); TET2 (5/76, 6.6%); EZH2 (5/76, 6.6%), DNMT3A (5/76, 6.6%), NOTCH1 (4/76, 5.3%), SETBP1 (3/76, 4.0%); CBL (2/76, 2.6%); U2AF1 (2/76, 2.6%), and one each (1.3%) of TP53, JAK2 exon 13, NRAS, BCOR, GATA2, CSF3R and ETV6. Two or more mutations were found in 8/76 (11%) patients. Overall, mutations were identified in 18/70 (25.7%) tested cases of idiopathic HES and 3/6 tested cases of CEL, NOS (50%).

Bone marrow morphology BM morphology was evaluated in conjunction with PB smears, blinded to all clinical, laboratory, molecular genetic data and patients’ outcome. Increased BM eosinophils were seen in the majority of the cases, accounting for a median of 21% (range, 4-91%) of BM cells; only three patients had <10% eosinophils in the BM. In two-thirds of the cases (71%), BM was unremarkable except for haematologica | 2017; 102(8)

BM morphology of CEL,NOS/idiopathic HES

Table 2. Clinical, molecular genetic features and survival comparison of patients with morphologically abnormal bone marrow or bone marrow within normal limits.

Total

Age Gender (male:female) White blood cell count (x109/L) Eosinophils % Absolute eosinophil count Hemoglobin (g/L) Platelet count (x109/L) • <140 • >450 • 140-450 Clinical presentations • Constitutional symptoms • Allergy/hypersensitivity • Muscular/joints/fasciitis • Thrombotic events • Skin rashes/dermatitis • Endocrine (thyroid, pancreas) • Gastrointestinal symptoms • Pulmonary/upper respiratory • Heart/pericardium • CNS/peripheral neuropathy • Organomegaly • Elevated LDH Abnormal karyotype Mutations • Two or more mutations • TP53, EZH1, SEPBP1, NRAS, CSF3R, JAK2 Patients’ outcomes** • Death • Survival (months)

Morphologically within normal limits (n=99)

(n=139)

Morphologically abnormal (n=40)

53.2(13.5-90.0) 79:60 13.8(5.3-193.2), 37(10-92) 4.8 (1.5-177.7) 12.9 (6-18.0) 273 (29-1744) 21/132 (16%) 16/132 (12%) 97/132 (73%)

64.0(28-89.5) 28:12 29.7(5.4-193.2) 39(10-92) 11.2(1.6-177.7) 12.2(6-15.3) 236 (29-1744) 13/39 (33%) 6/39 (15%) 20/39 (51%)

47.7(13.5-90.0) 51:48 11.5(5.3-143.1) 36(12-88) 3.9(1.5-113) 13.1(6.6-18.0) 279(72-734) 9/93 (10%) 10/93 (11%) 74/93 (80%)

<0.001 0.059 <0.001 0.235 <0.001 0.041 0.476 0.002

36/137 (26%) 32/137 (23%) 32/137 (23%) 6/137 (4%) 46/137 (34%) 6/137 (4%) 33/137 (24%) 25/137 (18%) 22/137 (16%) 11/137 (8%) 17/123 (14%) 30/83 (36%) 16/133 (12%) 21/76 (28%) 8/76 (10%) 6/76 (8%)

19/40 (48%) 2/40 (5%) 5/40 (13%) 4/40 (10%) 13/40 (33%) 0/40 (0%) 5/40 (13%) 1/40 (3%) 2/40 (5%) 6/40 (15%) 10/36 (28%) 17/30 (57%) 15/39 (38%) 12/20 (60%) 7/20 (35%) 6/20 (30%)

17/97 (18%) 30/97 (31%) 27/97 (28%) 2/97 (2%) 33/97 (34%) 6/97 (6%) 28/97 (29%) 24/97 (25%) 20/97 (21%) 5/97 (5%) 7/87 (8%) 13/53 (25%) 1/94 (1%) 9/56 (16%) 1/56 (2%) 0/56 (0%)

<0.001 <0.001 0.075 0.060 1.0 0.180 0.049 0.001 0.012 0.080 0.008 0.005 <0.001 <0.001 <0.001 <0.001

26/139 (19%)

18/40 48.1 (1-120.1)

8/99 Not reached (0-277.2)

<0.001 <0.001

Note: *P values are for comparison between morphologically normal vs. abnormal bone marrows; **patients’ outcomes: censored for unrelated death if known and at the time of hematopoietic stem cell transplant. CNS: central nervous system; LDH: lactate dehydrogenase.

Figure 1. Mutations detected in 21 patients with a diagnosis of chronic eosinophilic leukemia, not otherwise specified/idiopathic hypereosinophilic syndrome.

haematologica | 2017; 102(8)

1355

S.A. Wang et al.

increased eosinophils (Figure 2). In contrast, in one-third of cases (29%), besides the increased eosinophils, a number of other changes in BM were observed; these are shown in Table 1. The most common abnormalities (Figure 3) were BM hypercellularity, abnormal eosinophils, abnormal megakaryocytes, a markedly elevated M:E ratio ≥10; moderate to marked fibrosis, dysgranulopoiesis, and dyserythropoiesis.

Cases were considered to be morphologically abnormal if they showed overtly abnormal megakaryocytes (resembling those in MDS or MPN), significant dysgranulopoiesis or dyserythropoiesis, or increased (≥5%) BM blasts. These included 25 cases with abnormal megakaryocytes, most of which showed MDS-like morphology (Online Supplementary Table S1). Of these 25 cases, 15 also had abnormal eosinophils; three had ≥5% BM blasts, six

Figure 2. Many cases with idiopathic hypereosinophilic syndrome. show unremarkable bone marrow morphology. Bone marrow (BM) cellularity is either age-appropriate [(A) patient aged 48 years] or only slightly increased [(B) patient aged 45 years], with increased BM eosinophils and normalappearing megakaryocytes. (C) Eosinophils in peripheral blood (PB) may show mild uneven granulation (D) but are unremarkable on BM smear. No dysgranulopoiesis or dyserythropoiesis (BM biopsy, hematoxylin & eosin, original magnification x400; PB and BM smears Wright-Giemsa, original magnification x1000).

Figure 3. Morphologically abnormal bone marrow. (A, B) Bone marrow (BM) hypercellularity with increased eosinophils and neutrophilic granulocytic elements; frequent small hypolobated MDS-like megakaryocytes (A, arrows) or mixed MDS- and MPN-like megakaryocytes (B). (C) Peripheral blood (PB) shows abnormal eosinophils with multiple lobes and marked hypogranulation or agranulation. (D) The same changes are also observed in the BM from the same case. In addition, dysplastic erythroids and granulocytes (arrows) are evident. (E, F) A case with decreased megakaryocytes, hypercellularity with disrupted BM topography (E) and a BM smear showing markedly increased immature eosinophils and dyserythropoiesis (F, arrows). (BM biopsy: hematoxylin & eosin, original magnification x400; PB and BM smears: Wright-Giemsa, original maginification x1000)

1356

haematologica | 2017; 102(8)

BM morphology of CEL,NOS/idiopathic HES

had MF2 or MF3 myelofibrosis; 19 had hypercellularity, six had dysgranulopoiesis, and eight showed dyserythropoiesis. Of patients whose BM did not show abnormal megakaryocytes or had insufficient megakaryocytes for assessment, in four cases the BM was concluded to be abnormal, including three with marked dysgranulopoiesis and one with marked dyserythropoiesis (2 of these also showed abnormal eosinophils; 2 with hypercellularity and 1 with MF2 fibrosis). An additional 11 cases were scored as “abnormal” because of the presence of at least two of the following changes: BM hypercellularity (n=10); MF2 or MF3 fibrosis (n=6); abnormal eosinophils (n=4); M:E ratio >10 (n=1); and markedly decreased/near absence of megakaryocytes (n=2), of which one with a subset of MDS-like megakaryocytes (see Online Supplementary Table S1). There were also increased macrophages/histiocytes, stromal cells, vessels, and a disarrayed distribution of the BM cellular elements in some of these cases. These 11 cases were centrally reviewed, and one example is shown in Figure 3E, F. Thus, 40/139 cases (29%) were considered to have abnormal BM morphology and 99 had either normal morphology or only one morphological abnormality that did not include significant dysplasia, abnormal megakaryocytes, or excess blasts. In total, 16 of the 17 (94%) cases of CEL, NOS and 24 of 122 (22%) of the HES cases were morphologically abnormal. If the current WHO definitions of CEL, NOS and HES were used to anchor the reviewed cases as “true positives” for each diagnosis, abnormal morphology would have a sensitivity of 94.1% (95% confidence interval: 71.3-99.8%) and a specificity of 84.7% (95% confidence interval: 77.8-90.2%) for CEL, NOS.

Correlation of bone marrow morphology with clinical features The clinical presentations of the 40 patients with abnormal BM differed from those of the 99 patients who lacked significantly abnormal BM findings (Table 2). The patients with abnormal BM were older and presented with a higher white blood cell count and a higher absolute eosinophil count. These patients also had lower hemoglobin levels and more commonly had abnormal platelet counts (thrombocytopenia or thrombocytosis) (P=0.002). Clinically, more patients with abnormal BM morphology presented with constitutional symptoms (19/40 versus 17/97, P<0.001), but fewer with allergy/hypersensitivity (2/40 versus 30/97, P<0.001); cough, bronchitis, or pneumonitis (1/40 versus 24/97, P<0.001); gastrointestinal symptoms (5/40 versus 28/97, P=0.049); and heart failure, myocardial infarction or pericardial effusion (2/40 versus 20/97, P=0.012). Skin rashes and various forms of dermatitis were common in both groups of patients. An abnormal BM also correlated with more frequent organomegaly and elevated lactate dehydrogenase levels (Table 2).

Correlation of bone marrow morphology with molecular genetic data Of 17 patients with abnormal cytogenetics and/or increased BM blasts (CEL, NOS by the current WHO criteria), 16 had an abnormal BM. The one patient with CEL-NOS without abnormal BM morphology was a 64year old female who presented with chest pain, and was found to have elevated troponin, and pericardial and pleural effusions. This patient had del(16)(q23q24) and her BM showed 40% eosinophils, but otherwise had normal haematologica | 2017; 102(8)

morphology. FISH was negative for both CBFB and CHIC2. The patient was alive at 24 months of follow-up. Mutations detected by NGS were more frequent in patients with abnormal BM (12/20 versus 9/56, P<0.001). Mutations involving two or more genes were significantly more common in patients with abnormal BM (7/20 versus 1/56, P<0.001). Moreover, TP53, EZH2, SETBP1, NRAS, JAK2 exon 13 and CSF3R were only seen in patients with abnormal BM. Of patients who lacked abnormal BM findings, mutations included single gene mutations in TET2 (n=2), DNMT3A (n=3, 2 with 5-10% VAF); ASXL1 (n=1), CBL (n=1), or NOTCH1 (n=1). The only patient who had two mutations (TET2, VAF 25% and DNMT3A, VAF 15%) but no abnormal BM findings, was a 24-year old man who presented with fever and chest pain, dizziness and sensory abnormalities in both hands. The patient had abnormal magnetic resonance imaging findings in the brain and lungs, likely due to eosinophilic infiltrates. He showed some response to corticosteroids, but did not tolerate imatinib. He was alive at 28 months of follow-up.

Correlation of bone marrow morphology with outcome data These patients were treated with various agents recommended for people with idiopathic HES/CEL, NOS, including corticosteroids with or without interferon, hydroxyurea for cytoreduction, cyclosporine, methotrexate, and alemtuzumab. Tyrosine kinase inhibitors, mostly imatinib and dasatinib in some cases, were used in 54/126 (43%) patients over the course of the disease. Hypomethylating agents, single-agent chemotherapy, and high-dose chemotherapy were also used in some patients when their disease progressed or was refractory to other treatment modalities. A total of seven patients underwent hematopoietic stem cell transplantation. The median follow-up for all 139 patients was 38.9 months (range, 0 - 405.3 months). Three unrelated deaths (1 due to suicide, 1 due to a car accident and 1 due to diffuse large B-cell lymphoma), were censored at the time of death. Of 40 patients with abnormal BM, there were 18 deaths, including three due to progression of acute myeloid leukemia. The other causes of death included infection, bleeding and organ failure. Among the five patients in this group who received a hematopoietic stem cell transplant, four were alive and one died of disease recurrence. In contrast, of 99 patients without abnormal BM morphology, none experienced acute myeloid leukemia progression. There were eight deaths in this group, including four due to myocardial infarction or heart failure, two due to chronic obstructive lung disease, one due to the complication of a bone fracture as a result of long-term steroid use, and another of unknown cause. Both patients who received a hematopoietic stem cell transplant were alive at the last follow-up. The median overall survival for patients with an abnormal BM was 48.1 months (range, 1-120.1 months), which was significantly inferior to that of patients with a normal BM (not reached; range, 0-277.2) (Kaplan-Meier log rank, P<0.001) (Figure 4A). Survival was also compared in patients with a normal karyotype and <5% BM blasts, who would be considered as having idiopathic HES by the current WHO criteria. Within this group of patients, abnormal BM morphology remained a predictor of an inferior survival (median 125.5 months versus not reached, Kaplan-Meier log rank, P<0.001) (Figure 4B). 1357

S.A. Wang et al.

The prognostic significance of abnormal morphology was tested in multivariable analysis. The variables included age, gender, white blood cell count, absolute eosinophil count, hemoglobin level, lactate dehydrogenase concentration, organomegaly, constitutional symptoms, heart or brain involvement, cytogenetics, mutations, and the presence of two or more mutations. In the final multivariable Cox regression model, only age, thrombocytopenia, heart and/or brain involvement and abnormal BM morphology emerged as significant prognostic factors (Table 3); abnormal karyotype, mutations, and other clinical and laboratory parameters were not independently significant. Multivariable analysis was also performed in the subset of 122 patients with a normal karyotype and <5% BM blasts, who would be classified as having idiopathic HES by the current WHO criteria; abnormal morphology remained an independent predictor for inferior survival (Table 3).

Discussion In this study, we reviewed the BM of 139 patients who presented with hypereosinophilia without recurrent molecular genetic alterations or a known reactive cause. According to the current WHO classification criteria, 17 (12%) of these patients would be classified as having CEL, NOS, either due to the presence of an abnormal karyotype and/or increased BM blasts. However, abnormal BM morphology, with features resembling MDS, MDS/MPN or MPN, was observed in 40 of these patients, including 16 of 17 (94%) patients who were classified as having CEL, NOS and 24 of the 122 (20%) patients who were classified as having idiopathic HES. The criteria used to assess BM morphology were in part derived from what we had observed previously7 by comparing cases with molecular genetic abnormalities versus no identifiable abnormalities. These included increased blasts, hypercellularity, abnormal megakaryocytes, dyserythropoiesis and dysgranulopoiesis, markedly elevated M:E ratio and fibrosis, and abnormal eosinophils. The definitions of â&#x20AC;&#x153;abnormalâ&#x20AC;? BM findings were similar to those characteristically found in other myeloid neoplasms, including MDS, MPN and MDS/MPN, except for the inclusion of eosinophil morphology. In these patients, a BM eosinophilic infiltrate was invariably present, with only three patients having <10% eosinophils in the BM. However, in two-thirds of the patients, an increase in BM eosinophils either did not significantly alter or only led to a slight increase in BM cellularity. In patients with significant BM hypercellularity, this was frequently due to increased neutrophils and their precursors, megakaryocytes, and in some, erythroid precursors, or less commonly to an increased number of macrophages. Additional changes included increased stromal cells, histiocytes, vessels, and disarrayed cellular distribution, which are often referred by others as altered BM topography.19 Recognizing that cytological eosinophil atypia may be seen in reactive eosinophilia,4,11-14 we arbitrarily considered eosinophil morphology abnormal only if at least 20% of the eosinophils were involved. Interestingly, we found that mild atypical changes in reactive eosinophils were more frequently observed in PB than in BM (Figure 2), suggesting that BM smears may be more reliable for the assessment of eosinophil morphology. Nevertheless, of the 25 patients who showed significant numbers of abnor1358

mal eosinophils, 22 also had other BM abnormalities and only three patients had it as the sole alteration. Of these latter three patients, two had a long-standing history of allergy and gastrointestinal symptoms and one patient had deep venous thrombosis, endocardial fibrosis and BuddChiari syndrome. All three of these patients had a normal karyotype, and two tested by NGS were negative for mutations. Our findings suggest that alterations in eosinophil morphology can be used in conjunction with other BM findings in morphological assessment, but, if it is the sole alteration, may be unreliable to differentiate a neoplastic process from reactive eosinophilia.4,11-14 Prior to this study, there have been very few published studies with descriptions of megakaryocytes in patients with hypereosinophilia,20 even in well-defined entities such as PDGFRA- and PDGFRB-rearranged myeloid/lymphoid neoplasms.21-23 In our series, abnormal megakaryocyte morphology was frequently observed, with cytological features mostly resembling those of MDS-type megakaryocytes or mixed small and large megakaryocytes, with only a few cases showing megakaryocyte morphology similar to that present in BCR/ABL1-negative MPN. Some patients showed dysgranulopoiesis and/or

Figure 4. Comparison of survival of patients with chronic eosinophilic leukemia, not otherwise specified /idiopathic hypereosinophilic syndrome. (A) All patients (n=139): patients with morphologically abnormal bone marrow (ABN) had a median survival of 48.1 months, which is significantly inferior to that of patients with a normal BM (WNL) (unreached, P<0.001). (B) For patients who would otherwise be classified as having idiopathic HES (a normal karyotype and/or <5% blasts, n=122), an abnormal BM was also significantly associated with a shorter survival (P<0.001).

haematologica | 2017; 102(8)

BM morphology of CEL,NOS/idiopathic HES

Table 3. Factors independently predicting an inferior survival of patients with chronic eosinophilic leukemia, not otherwise specified/idiopathic hypereosinophilic syndrome (n=139) as well as of patients with idiopathic hypereosinophilic syndrome (n=122) only in multivariable analysis*

All patients (CEL,NOS/idiopathic HES)(n=139) Variables

Hazard ratio (95% CI)

Age (per year increase)

1.050 (1.019-1.082) 7.875 (2.888-21.473) 4.411 (1.803-10.785) 7.818 (2.795-21.869)

Heart and/or brain involvement 9

Platelet counts <140 x10 /L Abnormal bone marrow morphology

0.001 < 0.001 0.001 < 0.001

Idiopathic HES Patients with a normal karyotype and <5% blasts (n=122) Hazard ratio P (95% CI) 1.061 (1.021-1.103) 5.260 (1.636-16.912) 7.575 (2.153-26.651) 7.043 (2.191-22.639)

0.003 0.005 0.002 0.001

*bone marrow morphology was co-analyzed with age, gender, organomegaly, increased lactate dehydrogenase concentration, karyotype, mutation, hemoglobulin levels, platelet count, white blood cell count, absolute eosinophil count and brain/heart involvement

dyserythropoiesis. These findings in a patient with eosinophilia suggest a clonal neoplastic process. On the other hand, the presence of dysplastic changes in association with thrombocytopenia and anemia, seen in some of these patients, may raise the question of whether such cases should be considered more closely related to a MDS/MPN rather than to a true MPN, the nosological attribution of CEL, NOS and HES in the current WHO classification scheme. Clinically, patients with abnormal BM often showed features suggestive of a myeloid neoplasm, more frequently having constitutional symptoms, organomegaly, higher lactate dehydrogenase concentration, higher white blood cell and absolute eosinophil counts, abnormal platelet count and anemia. In contrast, they had fewer symptoms related to eosinophil activation, such as allergy, hypersensitivity, arthritis, muscle aches, and gastrointestinal, pulmonary, and cardiac-related symptoms. There were 18 (45%) disease-associated deaths in patients with abnormal BM morphology, including three which occurred due to progression to acute myeloid leukemia; many deaths were due to complications of BM failure. In contrast, there were only eight (8%) deaths in patients with normal BM morphology, and the causes of deaths were mainly cardiac or pulmonary complications. Patients with abnormal BM had a significantly inferior median survival compared to that of patients without significantly abnormal BM morphology. The median survival of 48.1 months appeared to be better than the previously reported 15-22 months7,20 in patients with CEL, NOS. However, the previous studies only included cases with an abnormal karyotype and/or increased blasts. A similar significance of abnormal BM morphology for survival was also observed in HES patients with a normal karyotype and no increased BM or PB blasts, who otherwise would be diagnosed as having idiopathic HES. These findings were underscored in the multivariable analysis, which showed that abnormal BM morphology, but not abnormal BM karyotype, was an independent prognostic marker when other factors were co-analyzed. In this study, we were also able to correlate mutational data with morphology and clinical data. Of 76 patients studied, 21 (27.6%) were found to have mutations. Similarly to what we found previously,7 mutations frehaematologica | 2017; 102(8)

quently involved genes involved in DNA methylation and chromatin modification, such as ASXL1, TET2, and DNMT3A. Although, most of these mutations are also frequently reported to occur in normal aging individuals,9,10 making it challenging to apply mutation data in the establishment of a clonal hematopoietic stem cell neoplasm, mutations involving at least one gene (60% versus 16%) as well as two or more genes were significantly more frequent in patients with abnormal BM morphology. The caveat was that the patients with abnormal BM morphology were significantly older (64.0 versus 47.7 years). Interestingly, it has been shown recently that in chronic myelomonocytic leukemia, age-related somatic mutations through successive acquisition convert a myelomonocytic biased hematopoiesis into overt leukemia.24 We noted that in our patients mutations more specific for a myeloid neoplasm (TP53, EZH2, SETBP1, NRAS, CSF3R, JAK2) were only found in subjects with abnormal BM morphology. Similar findings have been reported in MDS, in that certain specific mutations, the number of mutations and VAF, are predictive of MDS evolvement in cytopenic patients with clonal hematopoiesis of undetermined potential.9,25 Based on our findings of differences in mutation frequency and affected genes associated with BM morphology, we recommend that abnormal BM morphology should prompt NGS study using a myeloid mutation panel to try to establish evidence of clonality, and identify a neoplastic hematopoietic disease. In summary, we found that most patients with CEL, NOS and about 20% of those with idiopathic HES have abnormal BM morphology, while the remainder have unremarkable BM morphology with the exception of increased eosinophils. The abnormal BM findings in these cases are similar to those seen in MDS, MDS/MPN and/or BCR-ABL1-negative MPN. Isolated cytological abnormality of eosinophils is not entirely specific for a neoplastic process, but its presence should prompt careful assessment of BM morphology and appropriate molecular genetic testing. We found that abnormal BM morphology correlates with clinical presentations typically associated with a myeloid neoplasm, such as constitutional symptoms, splenomegaly, high lactate dehydrogenase concentration, anemia and abnormal platelet counts and less commonly with symptoms associated with an eosinophil 1359

S.A. Wang et al.

activation syndrome, such as allergy, respiratory or gastrointestinal symptoms, or cardiac involvement. Abnormal BM morphology is significantly correlated with abnormal karyotype and the presence of myeloid neoplasm-related mutations, and is highly associated with inferior patientsâ&#x20AC;&#x2122; outcome. The prognostic significance is independent of the effect of abnormal karyotypes, muta-

References 9. 1. Gotlib J. World Health Organizationdefined eosinophilic disorders: 2015 update on diagnosis, risk stratification, and management. Am J Hematol. 2015; 90(11):10771089. 2. Bain BJ. Eosinophilic leukaemias and the idiopathic hypereosinophilic syndrome. Br J Haematol. 1996;95(1):2-9. 3. Butt NM, Lambert J, Ali S, et al. Guideline for the investigation and management of eosinophilia. Br J Haematol. 2017;176(4): 553-572. 4. Bain BJ GD, Vardiman JW, Horny H-P. Chronic eosinophilic leukemia, not otherwise specified. In: Swerdlow SH, Campo E, Harris NL, et al, ed. WHO Classification of Tumors of Haematopoietic and Lymphoid Tissues: Lyon, France: IARC Press, 2008:5153. 5. 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. 6. Schwaab J, Umbach R, Metzgeroth G, et al. KIT D816V and JAK2 V617F mutations are seen recurrently in hypereosinophilia of unknown significance. Am J Hematol. 2015;90(9):774-777. 7. Wang SA, Tam W, Tsai AG, et al. Targeted next-generation sequencing identifies a subset of idiopathic hypereosinophilic syndrome with features similar to chronic eosinophilic leukemia, not otherwise specified. Mod Pathol. 2016;29(8):854864. 8. Andersen CL, Nielsen HM, Kristensen LS, et al. Whole-exome sequencing and genomewide methylation analyses identify novel disease associated mutations and methylation patterns in idiopathic hypereosinophilic

1360

10.

11.

12.

13.

14. 15.

16.

17.

18.

tions or other risk factors. We conclude that abnormal BM morphology should be regarded as a critical parameter useful for identifying a neoplastic subset of patients considered to have idiopathic HES. The presence of abnormal BM findings should be added to the abnormal karyotype and excess PB and/or BM blasts as one of the defining criteria for diagnosing CEL, NOS.

syndrome. Oncotarget. 2015;6(38):4058840597. Cargo CA, Rowbotham N, Evans PA, et al. Targeted sequencing identifies patients with preclinical MDS at high risk of disease progression. Blood. 2015;126(21):2362-2365. Steensma DP, Bejar R, Jaiswal S, et al. Clonal hematopoiesis of indeterminate potential and its distinction from myelodysplastic syndromes. Blood. 2015; 126(1):9-16. Chusid MJ, Dale DC, West BC, Wolff SM. The hypereosinophilic syndrome: analysis of fourteen cases with review of the literature. Medicine (Baltimore). 1975;54(1):1-27. Flaum MA, Schooley RT, Fauci AS, Gralnick HR. A clinicopathologic correlation of the idiopathic hypereosinophilic syndrome. I. Hematologic manifestations. Blood. 1981;58(5):1012-1020. Kueck BD, Smith RE, Parkin J, Peterson LC, Hanson CA. Eosinophilic leukemia: a myeloproliferative disorder distinct from the hypereosinophilic syndrome. Hematol Pathol. 1991;5(4):195-205. Weller PF, Bubley GJ. The idiopathic hypereosinophilic syndrome. Blood. 1994; 83(10):2759-2779. Valent P, Klion AD, Horny HP, et al. Contemporary consensus proposal on criteria and classification of eosinophilic disorders and related syndromes. J Allergy Clin Immunol. 2012;130(3):607-612.e609. Thiele J, Kvasnicka HM, Facchetti F, Franco V, van der Walt J, Orazi A. European consensus on grading bone marrow fibrosis and assessment of cellularity. Haematologica. 2005;90(8):1128-1132. Yamamoto A, Kojima T, Aoki T, Sasai M, Taniuchi S, Kobayashi Y. Eosinophil hypersegmentation is a possible marker to monitor the disease activity of atopic dermatitis. Allergology International. 2001;50(4): 325â&#x20AC;&#x201C; 330. Xu X. Nuclear hypersegmentation of neu-

19.

20.

21.

22.

23.

24.

25.

trophils, eosinophils, and basophils due to hydroxycarbamide (hydroxyurea). Blood. 2014;124(9):1392. Orazi A. Histopathology in the diagnosis and classification of acute myeloid leukemia, myelodysplastic syndromes, and myelodysplastic/myeloproliferative diseases. Pathobiology. 2007;74(2):97-114. Helbig G, Soja A, Bartkowska-Chrobok A, Kyrcz-Krzemien S. Chronic eosinophilic leukemia-not otherwise specified has a poor prognosis with unresponsiveness to conventional treatment and high risk of acute transformation. Am J Hematol. 2012;87(6):643645. Pardanani A, Brockman SR, Paternoster SF, et al. FIP1L1-PDGFRA fusion: prevalence and clinicopathologic correlates in 89 consecutive patients with moderate to severe eosinophilia. Blood. 2004;104(10):30383045. Walz C, Metzgeroth G, Haferlach C, et al. Characterization of three new imatinibresponsive fusion genes in chronic myeloproliferative disorders generated by disruption of the platelet-derived growth factor receptor beta gene. Haematologica. 2007;92(2):163-169. Schwaab J, Jawhar M, Naumann N, et al. Diagnostic challenges in the work up of hypereosinophilia: pitfalls in bone marrow core biopsy interpretation. Ann Hematol. 2016;95(4):557-562. Mason CC, Khorashad JS, Tantravahi SK, et al. Age-related mutations and chronic myelomonocytic leukemia. Leukemia. 2016;30(4):906-913. Fernandez-Pol S, Ma L, Ohgami RS, Arber DA. Significance of myelodysplastic syndrome-associated somatic variants in the evaluation of patients with pancytopenia and idiopathic cytopenias of undetermined significance. Mod Pathol. 2016;29(9):9961003.

haematologica | 2017; 102(8)

ARTICLE

Chronic Myeloid Leukemia

Single cell immune profiling by mass cytometry of newly diagnosed chronic phase chronic myeloid leukemia treated with nilotinib Stein-Erik Gullaksen,1 Jørn Skavland,1 Sonia Gavasso,2,3 Vinko Tosevski,4 Krzysztof Warzocha,5 Claudia Dumrese,6 Augustin Ferrant,7 Tobias Gedde-Dahl,8 Andrzej Hellmann,9 Jeroen Janssen,10 Boris Labar,11 Alois Lang,12 Waleed Majeed,13 Georgi Mihaylov,14 Jesper Stentoft,15 Leif Stenke,16 Josef Thaler,17 Noortje Thielen,10 Gregor Verhoef,18 Jaroslava Voglova,19 Gert Ossenkoppele,10 Andreas Hochhaus,20 Henrik Hjorth-Hansen,21,22 Satu Mustjoki,23,24 Sieghart Sopper,25 Francis Giles,26 Kimmo Porkka,23 Dominik Wolf25,27 and Bjørn Tore Gjertsen1,28

Centre of Cancer Biomarkers CCBIO, Department of Clinical Science, Precision Oncology Research Group, University of Bergen, Norway; 2Department of Clinical Medicine, University of Bergen, Norway; 3Neuroimmunology Lab, Haukeland University Hospital, Bergen, Norway; 4Mass Cytometry Facility, University of Zurich, Switzerland; 5 Department of Hematology, Institute of Hematology and Transfusion Medicine, Warsaw, Poland; 6Flow Cytometry Facility, University of Zurich, Switzerland; 7Hematology Department, Cliniques Universitaires St Luc, Brussels, Belgium; 8Department of Medicine, Oslo University Hospital, Norway; 9Department of Hematology, Medical University of Gdańsk, Poland; 10Department of Hematology, VU University Medical Center, Amsterdam, the Netherlands; 11Department of Hematology, University Hospital Center Rebro, Zagreb, Croatia; 12Internal Medicine, Hospital Feldkirch, Austria; 13 Department of Hemato-Oncology, Stavanger University Hospital, Norway; 14Clinic for Hematology, University Hospital Sofia, Bulgaria; 15Hematology Unit, Aarhus University Hospital, Denmark; 16Department of Medicine, Karolinska University Hospital, Stockholm, Sweden; 17Department of Internal Medicine IV, Wels-Grieskirchen Hospital, Wels, Austria; 18Department of Hematology, University Hospital Leuven, Belgium; 194th Department of Internal Medicine – Hematology, University Hospital Hradec Kralove, Czech Republic; 20Department of Hematology and Medical Oncology, Universitätsklinikum Jena, Germany; 21Department of Hematology, St Olavs Hospital, Trondheim, Norway; 22IKM, NTNU, Trondheim, Norway; 23Hematology Research Unit Helsinki, University of Helsinki and Helsinki University Hospital Comprehensive Cancer Center, Department of Hematology, Finland; 24Department of Clinical Chemistry, University of Helsinki, Finland; 25Department of Hematology and Oncology, Innsbruck Medical University and Tyrolean Cancer Research Institute, Innsbruck, Austria; 26NMDTI, Robert H. Lurie Comprehensive Cancer Center of Northwestern University, Chicago, IL, USA; 27Medical Clinic 3, Oncology, Hematology and Rheumatology, University Hospital Bonn (UKB), Germany and 28Department of Internal Medicine, Haukeland University Hospital, Bergen, Norway

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1361-1367

ABSTRACT

onitoring of single cell signal transduction in leukemic cellular subsets has been proposed to provide deeper understanding of disease biology and prognosis, but has so far not been tested in a clinical trial of targeted therapy. We developed a complete mass cytometry analysis pipeline for characterization of intracellular signal transduction patterns in the major leukocyte subsets of chronic phase chronic myeloid leukemia. Changes in phosphorylated Bcr-Abl1 and the signaling pathways involved were readily identifiable in peripheral blood single cells already within three hours of the patient receiving oral nilotinib. The signal transduction profiles of healthy donors were clearly distinct from those of the patients at diagnosis. Furthermore, using principal component analysis, we could show that phosphorylated transcription factors STAT3 (Y705) and CREB (S133) within seven days reflected BCR-ABL1IS at three and six months. Analyses of peripheral blood cells longitudinally collected from patients in the ENEST1st clinical trial showed that single cell mass cytometry appears to be highly suitable for future investigations addressing tyrosine kinase inhibitor dosing and effect. (clinicaltrials.gov identifier: 01061177) haematologica | 2017; 102(8)

Correspondence: bjorn.gjertsen@med.uib.no

Received: February 19, 2017. Accepted: May 8, 2017. Pre-published: May 18, 2017. doi:10.3324/haematol.2017.167080 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1361 ©2017 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.

1361

S.-E. Gullaksen et al.

Introduction Chronic myeloid leukemia (CML) is a hematologic stem cell disorder driven by transcription of the fusion protein Bcr-Abl1, a constitutively active tyrosine kinase.1 The deregulated kinase activity is sufficient to maintain the CML phenotype2,3 and leads to activation of several pathways, including JAK/STAT,4 PI3K/Akt5 and Ras/MAPK.6 Accordingly, specific tyrosine kinase inhibitors (TKIs) targeting Bcr-Abl1 have revolutionized CML treatment. The majority of patients display excellent cytogenetic and molecular responses and remain free of disease progression.7-9 However, for many CML patients, a cure will probably not be achieved by TKIs alone because of the persistence of leukemic stem cells that may cause relapse.10,11 Furthermore, a significant number of patients are switched from one TKI to another because of insufficient efficacy or adverse effects.12 Early identification of suboptimal responders to TKI could facilitate optimal dosing and choice of TKI thereby limiting adverse events and securing optimal molecular responses.13 This could be accomplished by the direct assessment of Bcr-Abl1 kinase activity in the leukemic cells early after start of TKI therapy. Pharmacodynamic analysis of TKI effect on Bcr-Abl1 tyrosine kinase activity and subsequent downstream signal transduction in CML patients has been technically challenging and thus rarely exploited in clinical trials.14,15 However, recent advances in single cell analysis of intracellular signal transduction show promise both in diagnostics and in therapeutic leukemia monitoring.16-18 Cytometry by Time of Flight (CyTOF) allows single cell characterization of immunophenotypes and intracellular signaling status in healthy individuals19,20 and patients with hematologic malignancies.21 The technique enables measurement of more than 40 parameters per cell, thereby dramatically increasing the dimensionality of acquired data compared to conventional flow cytometry. This is achieved by exploiting the resolution of mass spectrometry and stable rare-earth isotope-conjugated antibodies to perform highly multiplexed assays on single cells. This has permitted the dynamic signal transduction events within the hematologic hierarchy in acute myeloid leukemia to be evaluated.22-24 A panel of antibodies targeting more than 30 cell-surface and intracellular phospho-specific epitopes covering the major signaling molecules in Bcr-Abl1 oncogenic signaling (pBCR Y177, pAbl Y245, pCRKL Y207, pSTAT1 Y701, pSTAT3 Y705, pSTAT5 Y694, pCREB S133, pERK 1/2 T202/Y204, pS6 ribosomal protein S235/36 and S240/44) were developed. By analyzing longitudinally collected samples we show that both TKI-mediated changes in signal transduction and immunophenotypes is measurable, and propose that a 1-tube-based analysis by mass cytometry is a feasible biomarker strategy in clinical trials of leukemia.

Methods Patients Patients were eligible for the present study if they were enrolled in the international phase IIIb ENEST1st study.25 Patients were enrolled when they had newly diagnosed chronic phase CML, were 18 years of age or older, and had a World Health Organization (WHO) performance status no higher than 2. The 1362

demographic and clinical parameters of all patients included in this study (n=17) are shown in Online Supplementary Tables S1 and S2. Treatment consisted of nilotinib 300 mg BID. A peripheral blood (PB) sample was collected before the first nilotinib dose, and three hours, seven days and 28 days after the start of treatment. The study was performed in accordance with the Declaration of Helsinki, and all patients provided written informed consent. The study was approved by the local institutional review boards of all participating centers and is registered at clinicaltrials.gov identifier: 01061177. PB and bone marrow (BM) were collected from healthy individuals after written informed consent (University of Bergen, Norway, local ethical committee approval 2012/1045).

Materials The primary material was fixed and erythrocytes lysed by Lyse/Fix buffer (BD Phosflow) in local hospital laboratories and immediately frozen in saline for long-term storage and shipping at -80°C. The longitudinal PB samples from each patient (before first dose of nilotinib, and after 3 hours, 7 days, and 28 days) were barcoded for identification, pooled and stained with the antibody panels (Online Supplementary Table S3) according to the manufacturer's recommendation (MaxPar Phospho-Protein Staining Protocol). In the preliminary dataset #1, a protocol for barcoding was adapted from Zunder et al.26 (Online Supplementary Figure S1) and barcoded reagents were a kind gift from Prof. Bernd Bodenmiller, University of Zurich, Switzerland. The acquisition of samples was performed on the CyTOF2 mass cytometer (Fluidigm) at the Mass Cytometry Facility, University of Zurich, Switzerland. In dataset #2, a commercially available barcoding kit was used (Fluidigm). Acquisition of this dataset was performed using the Helios mass cytometer (Fluidigm) at the Flow Cytometry Core Facility, University of Bergen, Norway.

Statistical analysis Friedman non-parametric test with Dunn’s multiple comparison was used to find statistically significant changes in phosphorylation of intracellular targets (before, and after 3 hours and 7 days of treatment). Wilcoxon matched-pairs signed rank test was used to identify statistically significant changes in cell abundances before and after initiation of nilotinib therapy. Prism software (GraphPad Software Inc.) was used to calculate P-values and plot the graphs. P≤0.05 was considered statistically significant.

Results The immunophenotype and signal transduction in CD34+cells of CML patients It has been proposed that both number of hematopoietic stem and progenitor cells (CD34+CD38low and CD34+CD38high, respectively), and their signaling response patterns to TKIs are important for response prediction in CML.27 Therefore, we performed an initial analysis of the PB CD34+ compartment before and after receiving the first dose (3 hours and 7 days) of nilotinib in 17 patients enrolled in the ENEST1st study. The 3 samples from each patient were barcoded, pooled and stained with a validated panel of antibodies (Online Supplementary Figures S2 and S3). The high dimensional single-cell data for each patient were clustered using the SPADE algorithm,28 where manual annotation identified several major hematologic cell subsets, including the CD34+ cells. This population was further gated manually into CD34+CD38low and CD34+CD38high cells. Due to inter-patient variation in CD38 expression range, the gating was tailored to each haematologica | 2017; 102(8)

Single cell profiling of CML by mass cytometry

longitudinal patient sample set. Only patients for whom a minimum of 100 cells could be identified as CD34+CD38low and CD34+CD38high for all samples were included in the final analysis (n=10). To characterize the debulking of the tumor load at the single cell level and over time, all CD34+ cells were analyzed using viSNE algorithm.21 This allowed single cells from different samples to be graphed in a unifying 2-dimensional plot, where its (x,y) position contains information about cell surface marker expression (see Figure 1A for 2 representative patients' sample sets). CD34+CD38low cells resided in the lower part of the viSNE plot, while CD34+CD38high cells resided in the middle/top right. It has been suggested that CD25 can be used to identify leukemic stem cells in CML.29 In this pool of CD34+ cells collected from PB, CD34+CD25+ cells are strongly aggregated towards the very bottom of the viSNE plot, in the CD34+CD38low region. During seven days of TKI treatment, we saw a statistically significant (P≤0.01) depletion of CD34+ cells from PB leukocytes (Figure 1B). Furthermore, after gating of CD34+CD25+ double positive cells, we saw a statistically significant depletion of CD34+CD25+ cells compared to

Mass cytometry identified hematologic remission in clinical trial samples To validate our methodological approach, we analyzed a second cohort of patients (n=8) enrolled in the ENEST1st clinical trial. From each patient, 3 longitudinally collected samples (before, and after 3 hours and day 7 of TKI treat-

CD34+ cells at seven days of therapy (Figure 1C). The 85th percentile metal intensity of pAbl Y245, pCRKL Y207, pSTAT3 Y705 and pSTAT5 Y694 was calculated for the manually gated CD34+CD38low and CD34+CD38high single cell populations. Comparing the changes of phosphorylation level across patients over time, we found a statistically significant increase in pSTAT3 Y705 in the CD34+CD38high population (P≤0.05) between samples collected before and after seven days of TKI therapy (Figure 1D). Based on these observations in the CML progenitor compartment, we decided to examine more highly differentiated CML cells based on a more extensive panel of antibodies characterizing the myeloid and lymphoid lineages in greater detail (Online Supplementary Figures S4 and S5)

Figure 1. Nilotinib dosing altered signal transduction of the CD34+ cell population in chronic phase chronic myeloid leukemia (CML) patients analyzed by mass cytometry. Patient peripheral blood (PB) was collected at trial inclusion (t=0; before dosing), and after three hours (t=3h) and at day 7 (t=7d) of nilotinib (300 mg BID). (A) The CD34+ subset identified by the SPADE algorithm was further analyzed by the viSNE algorithm.21 The CD34+ cells from 2 representative patients at the three time points are shown. For each patient, the range of expression of each cell surface marker (CD34, CD38 and CD25) is color-coded from minimum to maximum expression of the three longitudinal samples. (B) The ratio of CD34+ cells and the total PB counts was calculated for all patients (n=10) in all three longitudinal samples. We found a statistically significant decrease of CD34+ cells in the PB between t=0 and day 7, and between after three hours and day 7 (Friedman non-parametric with Dunn’s multiple comparison; **P≤0.01). (C) For all patients, CD34+CD25+ cells were identified on the viSNE plots (see gating scheme). The ratio between CD34+CD25+ and all CD34+ cells was calculated showing a statistically significant decrease in CD34+CD25+ cells between diagnosis and day 7, and between after three hours and day 7 (Friedman non-parametric with Dunn’s multiple comparison; **P≤0.01). (D) Intracellular signaling transduction targets of Bcr-Abl1. The 85th percentile metal intensities of pAbl Y245, pCRKL Y207, pSTAT3 Y705, pSTAT5 Y694 and pCREB S133 were calculated and data of the patient cohort (n=10) for the CD34+CD38low and CD34+CD38high populations box plotted as a function of time. We observed a statistically significant change of pSTAT3 Y705 in the CD34+CD38high population (Friedman non-parametric with Dunn’s multiple comparison, *P≤0.05) between samples collected at diagnosis and day 7.

haematologica | 2017; 102(8)

1363

S.-E. Gullaksen et al. A

Figure 2. High-resolution single cell immune profiles of healthy and patient leukocytes. Longitudinally collected samples (before, after 3 hours and day 7 on nilotinib) from 8 patients in the ENEST1st trial, together with 4 healthy peripheral blood (PB) and bone marrow (BM) samples, were barcoded using the 20-plex metal barcoding kit (Fluidigm). (A) Data from the longitudinal samples from each patient, and the 4 healthy PB and BM samples, were pooled and clustered using the SPADE algorithm28 and manually annotated to identify cellular subsets. The SPADE tree analysis of patient 4702_0004 is shown. The size of each node represents the number of cells clustered and the expression of pSTAT3 Y705 is color-coded. The red bubble highlights the mature neutrophil population. (B and C) The relative abundance of the major PB populations identified in samples collected at diagnosis and after seven days of tyrosine kinase inhibitor (TKI) therapy is shown for the patient cohort (n=8, error bars showing standard error of mean, SEM), with the appropriate subpopulation in the healthy samples shown in a lighter shade. CD34+ cells could only be identified in 7 out of 8 patients. Wilcoxon matched-pairs rank test was used to identify statistically significant changes from before and after seven days of TKI therapy, where P≤0.05 was considered statistically significant.

ment) were barcoded, along with 4 healthy PB and 4 healthy BM samples, using the commercially available 20plex metal barcoding kit (Fluidigm). Each longitudinal set of 3 samples from a patient, and healthy PB and BM, was analyzed using the SPADE clustering algorithm,28 and the major leukocyte subsets were manually identified and annotated (Figure 2A). The relative abundance of a selection of cellular subsets quantified in longitudinal samples collected before and after seven days of TKI therapy is shown in Figure 2B and C. The abundance of the corresponding healthy cell subset is shown in a lighter shade, identified either from the healthy PB or BM samples (n=4 each). Similar to previous results (Online Supplementary Figure S3), already at day 7 of TKI therapy we observed a statistically significant change in the relative abundance of several cellular subsets. We measured a statistically significant reduction of myelocytes and progenitor cells (CD34+CD38low), and an expansion of mature neutrophils, CD8+ T cells, and activated T cells, (Wilcoxon matched pairs signed rank test P>0.05 for all).

The direct measurement of phosphorylated Bcr-Abl1 in single cells from patients treated with nilotinib The phosphorylation level of several key signaling mol1364

ecules in the Bcr-Abl1 signal transduction network was characterized in each patient, as a function of time and nilotinib treatment (before, and after 3 hours and 7 days of TKI therapy; n=8), and in the healthy PB and BM (n=4 each) (Figure 3A). In general, these data support our initial findings (Online Supplementary Figure 3C), reproducing the clearly distinct signaling status observed in healthy PB and CML at diagnosis, including the attenuated level of pSTAT3 Y694 phosphorylation in the CML neutrophils. In the larger patient cohort (n=8), we are able to measure a statistically significant decrease in pCRKL Y207 as a function of Bcr-Abl1 inhibition in the basophils (between before and after 7 days on TKI; P≤0.05), and an increase in pSTAT5 Y694 in the CD4 T helper cells (between diagnosis and after 3 hours of TKI; P≤0.01). We also measured a clear trend of pAbl Y245 downregulation in the CD34+CD38– population as a function of TKI-therapy (simple Wilcoxon test after 7 days of TKI P=0.0313). The seemingly ubiquitous regulation of pCREB S133, initial decrease and subsequent increase, reached statistical significance in the neutrophils (between before and after 7 days, and between diagnosis and after 3 hours of TKI; P≤0.05 for both) but was observed in several cellular subsets. In the hematopoietic stem and progenitor comparthaematologica | 2017; 102(8)

Single cell profiling of CML by mass cytometry

Figure 3. Single cell signaling profiles correlate to BCR-ABL1IS molecular response. The phosphorylation level (75th percentile of metal intensity) of the intracellular Bcr-Abl1 signaling network (n=10) was measured in the longitudinal samples (before, after 3 hours and after 7 days of nilotinib) in the patient cohort (n=8), together with 4 healthy peripheral blood (PB) and bone marrow (BM) samples. (A) We observed distinctly different states of signal transduction in the patient samples compared to the healthy controls. (B) The Lin-CD34+CD38low and Lin–CD34+CD38high cell populations were manually gated from the CD34+ population identified by SPADE, and the intracellular signal transduction measured. CD34+ cells could only be identified in 7 out of 8 patients. The same strategy was followed for healthy CD34+ cells. For both (A and B) statistical significance was determined using Friedman non-parametric with Dunn’s multiple comparison test, where P≤0.05 was considered statistically significant (*P≤0.05, **P≤0.01). (C) The BCR-ABL1IS of each patient during tyrosine kinase inhibitor (TKI) therapy. The patient cohort was divided into two groups based on the BCR-ABL1IS at three and six months (red and green). (D) An unsupervised PCA analysis of the signal transduction arcsinh fold change compared samples before and after seven days of nilotinib treatment in the mature neutrophils (Neutro) and myelocytes (Myelo) was performed using Unscramble software (CAMO Software). The categorical color-coding from (C) was superimposed on the resulting PCA plot, and the dashed line was drawn manually.

ment (Figure 3B), we reproduced a statistically significant increase in pSTAT3 Y694 (between after 3 hours of TKI and after 7 days of TKI; P≤0.05) similar to what can be seen in the CD34+CD38high progenitor cell population (Figure 1D).

TKI-mediated changes in single cell signal transduction reflects BCR-ABL1IS at 3 and 6 months Finally, we examined whether the observed TKI-induced changes in signal transduction networks could provide prognostic information. We split the patients according to their BCR-ABL1IS at three and six months into two response groups (Figure 3C, low in green and high in red). The arcsinh fold change of each signal transduction epitope (n=10) between diagnosis and after seven days of TKI treatment was calculated for the neutrophils and myelocytes for all patients (n=8). We performed an unsupervised multivariate principle component analysis (PCA) of the fold change data using Unscrambler X (CAMO Software), and then manually overlaid the response group affiliation (Figure 3D). The PCA analysis revealed underlying differences in signaling status between the response groups, driven mainly by the rates of downregulation of pCREB S133 and upregulation of pSTAT3 Y705.

Discussion Mass cytometry enabled the measurement of more than 30 intracellular and cell surface markers on each single cell, with an additional six channels reserved for metal barcodhaematologica | 2017; 102(8)

ing for sample multiplexing.30 The convenient single tube labeling of multiplexed samples combined with semiautomatic analysis makes the technique highly efficient. We were able to recapitulate the expected hematologic response, and could also demonstrate the debulking of leukemic CD34+ cells in the PB by immunophenotyping as early as one week after start of therapy. Importantly, by simultaneously probing key intracellular phosphorylation targets of the Bcr-Abl1 signaling network,2,3 we monitored changes in signal transduction of individual cell types for each patient undergoing TKI therapy. Unsupervised principle component analysis of these early changes in signal transduction allowed patients to be identified according to their BCR-ABL1IS, indicating a possible future prognostic impact of this approach. The proportion of leukemic stem cells (as determined by FISH) in the BM of chronic phase CML have been shown to have a prognostic significance.26 In the PB of our patients, we were able to monitor the therapy-dependent debulking of CD34+ cells (Figure 1B), and both CD34+CD38low and CD34+CD38high cells (data not shown), at day 7 of therapy. Interestingly, at day 7 we saw a statistically significant depletion of CD34+CD25+, a putative CML stem cell subset27 compared to all CD34+ cells (Figure 1C). This could indicate an increased sensitivity to TKI or alternatively an increased bone marrow homing of this cell subset.29 Early relative increase of lymphocytes in the PB of CML patients treated with TKI has also previously been shown to have prognostic importance.31 Our immune profiling of patient leukocytes allowed a highresolution picture of PB subpopulations, enabling the 1365

S.-E. Gullaksen et al.

simultaneous monitoring of the size of all major hematopoietic lineages. Here, the analysis of healthy PB and BM side-by-side with primary material clearly delineates cell populations dominating in CML compared to healthy individuals (Figure 2). Interestingly, we were able to measure statistically significant changes in abundances for several cellular subpopulations, including an increase in PB CD8+ cytotoxic T cells already after seven days of TKI therapy. We believe that a particular strength of our single cell analysis is the direct monitoring of nilotinib-induced changes in signal transduction. This includes the primary target kinase for nilotinib, Abl1, as well as its immediate substrates like pCRKL Y20732 and pivotal Abl1-affected signal transduction pathways in myeloid neoplasias, e.g. the STAT3/5 pathway.16,17,33 We established a metal barcoded pool of stimulated cell lines for functional validation of our panel of phospho-specific antibodies (Online Supplementary Figure S5). Indeed, we could measure a statistically significant reduction of pCRKL Y207 in the basophils as a function of nilotinib therapy in our patents (n=8) (Figure 3A). Compared to healthy controls, we could measure a clearly attenuated pSTAT3 Y705 in mature neutrophils (Figure 3A and Online Supplementary Figure S3C) and in the CD34+CD38high progenitor compartment (Figures 1C and 3C) collected before treatment initiation, which subsequently increased during TKI treatment. A similar increase in pSTAT3 Y705 has previously been reported in both mature neutrophils and CD34+ cells.34,35 In BCR-ABL1 oncogene-addicted CML cells, it has been shown that growth factor receptor signaling is dampened through MEK/ERK-dependent negative feedback of the BCR-ABL1 signal.36 This may explain the attenuated basal level of phosphorylation of pSTAT3. Finally, we observe a decrease of pCREB S133 as early as three hours after first TKI dosing in several cellular subsets followed by an increase after seven days. Interestingly, we see that, compared to non-CML cell lines, K562 cells express high levels of pCREB S133 at baseline that was reduced by TKI treatment (Online Supplementary Figure S5C). Using a principle component analysis to analyze changes in signal transduction between diagnosis and after seven days of TKI treatment, we could discriminate patient groups according to their BCR-ABL1IS at three and six months (Figure 3D). The PCA analysis was driven mainly by mutual regulation of pCREB S133 and pSTAT3 Y705, and underscores a potential prognostic application for this assay.

References 1. Nowell PC, Hungerford DA. A minute chromosome in human granulocytic leukemia. Science. 1960;32:1497-1501. 2. Deininger MW, Goldman JM, Melo JV. The molecular biology of chronic myeloid leukemia. Blood. 2000;96(10):3343-3356. 3. Cilloni D, Saglio G. Molecular pathways: BCR-ABL. Clin Cancer Res. 2012;18(4): 930-937. 4. Chai SK, Nichols GL, Rothman P. Constitutive activation of JAKs and STATs in BCR-Abl-expressing cell lines and peripheral blood cells derived from leukemic patients. J Immunol. 1997;159 (10):4720-4728.

1366

In summary, here we demonstrate the first human study of signal transduction in cancer cells from nilotinibtreated CML patients in a prospective clinical trial, demonstrating significant phosphorylation of CREB and STAT3 in chronic phase CML patients early after start on therapy. We propose that mass cytometry, enabling simultaneous measurement of more than 30 parameters on single cells, may provide a highly detailed characterization of CML that allows comprehensive 1-tube-based monitoring of the disease in patients during TKI therapy. The method outlined here will allow early changes in individual patient’s immunophenotype and signal transduction as a function of TKI therapy to be monitored in clinical trials. Furthermore, this approach may shed light on the relation between the kinetics of TKI-modulated signal transduction and clinical response. We anticipate this technology to have high clinical impact by providing cancer cell signal transduction profiles in patients close to real time after start of TKI therapy. Additional work on standardization of antibody panels is a prerequisite if mass cytometry is to be established as a future tool to guide signal transductiontargeted therapy. Acknowledgments The authors would like to thank personnel at the following ENEST1st study centers for participating in this sub-study: VU University Medical Center, Amsterdam, the Netherlands; Cliniques Universitaires St Luc, Brussels, Belgium; University of Helsinki and Comprehensive Cancer Center, Helsinki University Hospital, Finland; Cliniques Universitaires St Luc, Brussels, Belgium; University Hospital Bonn (UKB), Bonn, Germany; Wels-Grieskirchen Hospital, Wels, Austria; Hospital Feldkirch, Austria; University Hospital, Aarhus, Denmark; Oslo University Hospital, Oslo, Norway; St Olavs Hospital Trondheim, Norway; Stavanger University Hospital, Norway; Haukeland University Hospital, Bergen; Norway; MTZ Clinical Research, Warsaw, Poland; University Hospital Leuven, Belgium; University Hospital Center Rebro, Zagreb, Croatia; Karolinska University Hospital, Stockholm, Sweden; Medical University of Gdańsk, Poland; University Hospital Sofia, Bulgaria. Funding This study was supported by The Research Council of Norway (Petromaks program grant #220759), Helse Vest health trust and the Norwegian Cancer Society with Solveig & Ole Lunds Legacy. Novartis is acknowledged for financially supporting data and sample collection for this sub-study.

5. Ly C, Arechiga AF, Melo JV, Walsh CM, Ong ST. Bcr-Abl kinase modulates the translation regulators ribosomal protein S6 and 4E-BP1 in chronic myelogenous leukemia cells via the mammalian target of rapamycin. Cancer Res. 2003;63(18):57165722. 6. Cortez D, Reuther G, Pendergast AM. The Bcr-Abl tyrosine kinase activates mitogenic signaling pathways and stimulates G1-to-S phase transition in hematopoietic cells. Oncogene. 1997;15(19):2333-2342. 7. Hughes TP, Hochhaus A, Branford S, et al. Long-term prognostic significance of early molecular response to imatinib in newly diagnosed chronic myeloid leukemia: an analysis from the International

10.

11.

Randomized Study of Interferon and STI571 (IRIS). Blood. 2010;116(19):37583765. Rosti G, Palandri F, Castagnetti F, et al. Nilotinib for the frontline treatment of Ph(+) chronic myeloid leukemia. Blood. 2009;114(24):4933-4938. Cortes JE, Jones D, O'Brien S, et al. Results of dasatinib therapy in patients with early chronic-phase chronic myeloid leukemia. J Clin Oncol. 2010;28(3):398-404. Graham SM, Jørgensen HG, Allan E, et al. Primitive, quiescent, Philadelphia-positive stem cells from patients with chronic myeloid leukemia are insensitive to STI571 in vitro. Blood. 2002;99(1):319-325. Jorgensen HG, Allan EK, Jordanides NE,

haematologica | 2017; 102(8)

Single cell profiling of CML by mass cytometry

12.

13.

14. 15.

16.

17.

18.

19.

Mountford JC, Holyoake TL. Nilotinib exerts equipotent antiproliferative effects to imatinib and does not induce apoptosis in CD34+ CML cells. Blood. 2007; 109(9):4016-4019. Hochhaus A, O'Brien SG, Guilhot F, et al. Six-year follow-up of patients receiving imatinib for the first-line treatment of chronic myeloid leukemia. Leukemia. 2009;23(6):1054-1061. QuintĂĄs-Cardama A, Kantarjian H, Jones D, et al. Delayed achievement of cytogenetic and molecular response is associated with increased risk of progression among patients with chronic myeloid leukemia in early chronic phase receiving high-dose or standard-dose imatinib therapy. Blood. 2009;113(25):6315-6321. Iliuk AB, Tao WA. Is phosphoproteomics ready for clinical research? Clin Chim Acta. 2013;420:23-27. Baca Q, Cosma A, Nolan G, Gaudilliere B. The road ahead: Implementing mass cytometry in clinical studies, one cell at a time. Cytometry B Clin Cytom. 2017; 92(1):10-11. Skavland J, JĂ¸rgensen KM, Hadziavdic K, et al. Specific cellular signal-transduction responses to in vivo combination therapy with ATRA, valproic acid and theophylline in acute myeloid leukemia. Blood Cancer J. 2011;1(2):e4. Kotecha N, Flores NJ, Irish JM, et al. Singlecell profiling identifies aberrant STAT5 activation in myeloid malignancies with specific clinical and biologic correlates. Cancer Cell. 2008;14(4):335-343. Jalkanen SE, Lahesmaa-Korpinen AM, Heckman CA, et al. Phosphoprotein profiling predicts response to tyrosine kinase inhibitor therapy in chronic myeloid leukemia patients. Exp Hematol. 2012;40(9):705-714.e03. Bendall SC, Simonds EF, Qiu P, et al. Singlecell mass cytometry of differential immune

haematologica | 2017; 102(8)

20.

21.

22.

23.

24.

25.

26.

27.

and drug responses across a human hematopoietic continuum. Science. 2011; 332(6030):687-696. Bodenmiller B, Zunder ER, Finck R, et al. Multiplexed mass cytometry profiling of cellular states perturbed by small-molecule regulators. Nat Biotechnol. 2012;30(9):858867. Amir el-AD, Davis KL, Tadmor MD, et al. viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leukemia. Nat Biotechnol. 2013;31(6):545-552. Levine JH, Simonds EF, Bendall SC, et al. Data-Driven Phenotypic Dissection of AML Reveals Progenitor-like Cells that Correlate with Prognosis. Cell. 2015;162(1):184-197. Behbehani GK, Samusik N, Bjornson ZB, Fantl WJ, Medeiros BC, Nolan GP. Mass Cytometric Functional Profiling of Acute Myeloid Leukemia Defines Cell-Cycle and Immunophenotypic Properties That Correlate with Known Responses to Therapy. Cancer Discov. 2015;5(9):9881003. Han L, Qiu P, Zeng Z, et al. Single-cell mass cytometry reveals intracellular survival/proliferative signaling in FLT3ITD-mutated AML stem/progenitor cells. Cytometry A. 2015;87(4):346-356. Hochhaus A, Rosti G, Cross NC, et al. Frontline nilotinib in patients with chronic myeloid leukemia in chronic phase: results from the European ENEST1st study. Leukemia. 2016;30(1):57-64. Zunder ER, Finck R, Behbehani GK, et al. Palladium-based mass tag cell barcoding with a doublet-filtering scheme and singlecell deconvolution algorithm. Nat Protoc. 2015;10(2):316-333. Mustjoki S, Richter J, Barbany G, et al. Impact of malignant stem cell burden on therapy outcome in newly diagnosed chronic myeloid leukemia patients.

Leukemia. 2013;27(7):1520-1526. 28. Qiu P, Simonds EF, Bendall SC, et al. Extracting a cellular hierarchy from highdimensional cytometry data with SPADE. Nat Biotechnol. 2011;29(10):886-891. 29. Kobayashi CI, Takubo K, Kobayashi H, et al. The IL-2/CD25 axis maintains distinct subsets of chronic myeloid leukemia-initiating cells. Blood. 2014;123(16):2540-2549. 30. Gavasso S, Gullaksen SE, Skavland J, Gjertsen BT. Single-cell proteomics: potential implications for cancer diagnostics. Expert Rev Mol Diagn. 2016;16(5):579-589. 31. Kumagai T, Matsuki E, Inokuchi K, et al. Relative increase in lymphocytes from as early as 1 month predicts improved response to dasatinib in chronic-phase chronic myelogenous leukemia. Int J Hematol. 2014;99(1):41-52. 32. Nichols GL, Raines MA, Vera JC, Lacomis L, Tempst P, Golde DW. Identification of CRKL as the constitutively phosphorylated 39-kD tyrosine phosphoprotein in chronic myelogenous leukemia cells. Blood. 1994; 84(9):2912-2918. 33. Irish JM, Hovland R, Krutzik PO, et al. Single cell profiling of potentiated phospho-protein networks in cancer cells. Cell. 2004;118(2):217-228. 34. Jalkanen SE, Vakkila J, Kreutzman A, Nieminen JK, Porkka K, Mustjoki S. Poor cytokine-induced phosphorylation in chronic myeloid leukemia patients at diagnosis is effectively reversed by tyrosine kinase inhibitor therapy. Exp Hematol. 2011;39(1):102-113.e1. 35. Eiring AM, Page BD, Kraft IL, et al. Combined STAT3 and BCR-ABL1 inhibition induces synthetic lethality in therapyresistant chronic myeloid leukemia. Leukemia. 2015;29(3):586-597. 36. Asmussen J, Lasater EA, Tajon C, et al. MEK-dependent negative feedback underlies BCR-ABL-mediated oncogene addiction. Cancer Discov. 2014;4(2):200-215.

1367

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Chronic Myeloid Leukemia

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1368-1377

Natural killer-cell counts are associated with molecular relapse-free survival after imatinib discontinuation in chronic myeloid leukemia: the IMMUNOSTIM study

Delphine Rea,1,2,3* Guylaine Henry,4 Zena Khaznadar,1,5 Gabriel Etienne,3,6 François Guilhot,3,7 Franck Nicolini,3,8 Joelle Guilhot,3,7 Philippe Rousselot,3,9 Françoise Huguet,3,10 Laurence Legros,3,11 Martine Gardembas,3,12 Viviane Dubruille,3,13 Agnès Guerci-Bresler,3,14 Aude Charbonnier,3,15 Frédéric Maloisel,16 Jean-Christophe Ianotto,17 Bruno Villemagne,18 François-Xavier Mahon,3,6 Hélène Moins-Teisserenc,1,4,5 Nicolas Dulphy1,4,5* and Antoine Toubert1,4,5

INSERM UMRS-1160, Paris; 2Service d’Hématologie Adulte, Hôpital Saint-Louis, Paris; France Intergroupe des Leucémies Myéloïdes Chroniques (Fi-LMC), Institut Bergonié, Bordeaux; 4Laboratoire d’Immunologie et Histocompatibilité, Hôpital Saint-Louis, Paris; 5 Institut Universitaire d'Hématologie, Université Paris Diderot-Paris 7; 6Service d'Oncologie Médicale, Institut Bergonié, Bordeaux; 7INSERM CIC 1402, CHU de Poitiers; 8 Service d'Hématologie Clinique, CHU Lyon Sud, Pierre Bénite; 9Service d’Hématologie Oncologie et INSERM UMR-1173, Centre Hospitalier de Versailles, Le Chesnay; 10Service d'Hématologie, IUCT Oncopole, Toulouse; 11Service d'Hématologie Clinique, Hôpital de l'Archet, CHU de Nice; 12Service des Maladies du Sang, CHRU Angers; 13Service d'Hématologie Clinique, Hôpital Hôtel Dieu, Nantes; 14Service d’Hématologie, CHU Brabois, Vandoeuvre les Nancy; 15Service d’Onco-Hématologie, Institut Paoli Calmettes, Marseille; 16Groupe Oncologie-Maladies du Sang, Clinique Sainte Anne, Strasbourg; 17 Service Hématologie Clinique, Hôpital Morvan, CHRU de Brest and 18Service Médecine Onco-hématologie, CH de la Roche sur Yon, France 1 3

*DR and ND contributed equally to this work.

ABSTRACT

Correspondence: delphine.rea@aphp.fr or nicolas.dulphy@univ-paris-diderot.fr Received: January 21, 2017. Accepted: May 8, 2017. Pre-published: May 18, 2017. doi:10.3324/haematol.2017.165001 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1368 ©2017 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.

1368

espite persistence of leukemic stem cells, patients with chronic myeloid leukemia who achieve and maintain deep molecular responses may successfully stop the tyrosine kinase inhibitor imatinib. However, questions remain unanswered regarding the biological basis of molecular relapse after imatinib cessation. In IMMUNOSTIM, we monitored 51 patients from the French Stop IMatinib trial for peripheral blood T cells and natural killer cells. Molecular relapse-free survival at 24 months was 45.1% (95% CI: 31.44%-58.75%). At the time of imatinib discontinuation, non-relapsing patients had significantly higher numbers of natural killer cells of the cytotoxic CD56dim subset than had relapsing patients, while CD56bright natural killer cells, T cells and their subsets did not differ significantly. Furthermore, the CD56dim natural killer-cell count was an independent prognostic factor of molecular-relapse free survival in a multivariate analysis. However, expression of natural killer-cell activating receptors, BCR-ABL1+ leukemia cell line K562-specific degranulation and cytokine-induced interferon-gamma secretion were decreased in nonrelapsing and relapsing patients as compared with healthy individuals. After imatinib cessation, the natural killer-cell count increased significantly and stayed higher in non-relapsing patients than in relapsing patients, while receptor expression and functional properties remained unchanged. Altogether, our results suggest that natural killer cells may play a role in controlling leukemia-initiating cells at the origin of relapse after imatinib cessation, provided that these cells are numerous enough to compensate for their functional defects. Further research will decipher mechanisms underlying functional differences between natural killer cells from patients and healthy individuals and evaluate the potential interest of immunostimulatory approaches in tyrosine kinase inhibitor discontinuation strategies. (ClinicalTrial.gov Identifier NCT00478985) haematologica | 2017; 102(8)

NK cells and outcome after stopping imatinib

Introduction Chronic myeloid leukemia (CML) is a myeloproliferative neoplasia caused by the fusion of the BCR and ABL1 genes, as the result of the acquired reciprocal t(9;22)(q34;q11) translocation. In the early 2000s, imatinib, the first ATP-competitive inhibitor of the BCR-ABL1 oncoprotein, revolutionized the management of CML, providing most patients a dramatic progression-free survival benefit.1 Since then, newer generations of tyrosine kinase inhibitors (TKI) have been developed in order to overcome some of the drawbacks of imatinib, but imatinib remains one of the key initial therapies for newly diagnosed patients.2 When imatinib treatment is addressed appropriately, life expectancy of adult patients diagnosed with chronicphase CML (CP-CML) is close to that of the general population.3.4 However, the current recommendation is to administer treatment lifelong because of the inability of imatinib and other TKI to eliminate quiescent leukemic stem cells.5-8 This recommendation represents a substantial challenge with respect to long-term safety, quality of life and economic burden. Therefore in the past few years, clinical trials have investigated the feasibility of discontinuing imatinib treatment in patients with sustained deep molecular responses. In the pioneering STIM trial, patients on imatinib therapy for a minimum of 3 years in whom BCR-ABL1 transcripts were undetectable for at least 2 years had a probability of maintaining deep molecular responses without any treatment of about 40%, challenging the statement that TKI may never be stopped.9 These findings were rapidly corroborated by the independent TWISTER trial.10 However, a definitive cure remains uncertain in patients who do not relapse. Indeed, serial assessments with reverse transcriptase quantitative polymerase chain reaction (RT-qPCR) showed that peripheral blood BCR-ABL1 transcripts could be detected in patients who successfully stopped imatinib, albeit in low amounts.9 The use of genomic DNA-based PCR as a monitoring tool revealed that patients continued to harbor the BCR-ABL1 gene after discontinuation of imatinib, even when the corresponding transcripts were undetectable.11 In patients who had been off TKI therapy for several years, BCR-ABL1 transcripts could be amplified in CD34+ cell-derived colony-forming cells and long-term cultureinitiating cells despite undetectable residual disease in the peripheral blood.8 Altogether, these results indicate that a reservoir of primitive leukemic cells persists in most if not all TKI-treated patients regardless of outcome after treatment discontinuation. There is great clinical interest in trying to identify patients who are more likely to succeed in discontinuing imatinib in order to minimize potential risks of a leukemic rebound and to avoid undesirable drug-withdrawal symptoms.12 So far, the search for clinical variables predictive of outcome has been challenging but factors such as the Sokal score, duration of therapy, depth of molecular response and duration of deep molecular response have provided some insights into the probability of successful imatinib discontinuation in several studies.9,13,14 However, biological factors directing the fate of residual leukemic cells once TKI pressure is released are unclear. Given the susceptibility of CML to adaptive and innate immune cellular attack, an efficient autologous anti-CML response might help to control the leukemic load beyond cessation haematologica | 2017; 102(8)

of TKI treatment.15,16 We designed and conducted an ancillary biological study within the STIM trial, named IMMUNOSTIM, with the goal of analyzing peripheral blood T cells and natural killer (NK) cells and investigated whether immune parameters were associated with molecular relapse-free survival.

Methods Patients IMMUNOSTIM is a sub-study of the STIM trial approved by French health authorities (NCT00478985).9 Written informed consent was given in agreement with the Declaration of Helsinki. Imatinib was stopped after ≥3 years of therapy and ≥2 years of undetectable BCR-ABL1 transcripts. Stringent monitoring by RTqPCR was performed after imatinib discontinuation to detect a molecular relapse.9 The assay sensitivity was ≥4.5 log. Consecutively detectable peripheral blood BCR-ABL1 transcripts showing a ≥1 log increase or loss of a major molecular response [BCR-ABL1/ABL1 internationally standardized (IS) ratio ≤0.1%] defined molecular relapse and triggered imatinib resumption. In IMMUNOSTIM, heparinized blood was collected at baseline, bimonthly for 6 months then every 6 months until 24 months unless imatinib was resumed. Healthy donors were recruited through the Paris Saint-Louis Blood Donation Center and gave informed consent. Experiments were performed in a centralized fashion, allowing ≤48 h from blood collection to processing.

Immunophenotyping Patients’ whole blood cell counts were determined using a Sysmex XS 1000i analyzer. T cells and NK cells were quantified by dual-platform flow cytometry using monoclonal antibodies recognizing CD3, CD4, CD127, CD25, CD8, CD45RA, CCR7, CD27, CD56, CD16 and NKG2D (Online Supplementary Methods). Peripheral blood mononuclear cells (PBMC) were purified with Ficoll density-gradient centrifugation and cryopreserved in liquid nitrogen. NK-cell receptor expression was studied by flow cytometry using thawed PBMC stained with monoclonal antibodies recognizing CD56, CD16, CD3, DNAM-1, KIR2D, NKp46, NKp30, NKG2A, CD94 and CD57 (Online Supplementary Methods).

Natural killer-cell functional assays To assess degranulation, PBMC were thawed, maintained in 10% fetal calf serum-RPMI-1640 overnight at 37°C and incubated with or without the HLA class I-deficient BCR-ABL1+ leukemia cell line K562 at a 1:1 effector to target ratio for 18 h at 37°C with a CD107a monoclonal antibody. Thereafter, PBMC were stained with monoclonal antibodies recognizing CD137, CD56, CD3 and CD16. CD137+ and CD107a+ NK cells were detected by flow cytometry (Online Supplementary Methods). To study interferon (IFN)-γ production, thawed PBMC were activated overnight in medium supplemented with interleukin (IL)-12 and IL-18 (10 ng/mL each). The following day, brefeldin A was added for 4 h. IFN-γ–producing CD3-CD56brightCD16-/low NK cells were detected by flow cytometry as previously described (Online Supplementary Methods).17

Statistical analyses A Mann-Whitney U-test, a Kruskal-Wallis test and a Wilcoxon matched-pairs signed ranked test were used to compare, respectively, quantitative variables from two independent groups, more than two groups with a Dunn test for multiple comparisons, and quantitative variables following imatinib discontinuation. Molecular relapse-free survival was estimated with the Kaplan1369

D. Rea et al.

Meier method.9 Clinical variables and NK cells were assessed as potential prognostic factors for molecular relapse-free survival. Quantitative factors were categorized into two groups with cutoffs set at the median. All variables were assessed by univariate analysis using the Kaplan-Meier method and the two-tailed logrank test. A backward stepwise multivariate Cox proportional analysis was performed to determine the influence of variables potentially associated with outcome in univariate analysis (P≤0.10). In the final model, P values <0.05 were considered statistically significant. Quantitative variables were categorized into two groups with cut-offs set at the median. Hazard ratios and 95% confidence intervals (CI) were estimated from the Cox regression analysis.

Results Patients’ characteristics Fifty-one of the 100 patients enrolled in the STIM trial who agreed to participate to IMMUNOSTIM and from whom blood samples were received and processed within 48 h of collection were included; their baseline characteristics are detailed in Table 1. Thirty-three patients (64.7%) were females and the Sokal score at diagnosis of CP-CML was low in 28 (54.9%). Single-agent imatinib was given as first-line TKI therapy in all patients, either soon after the diagnosis of CML (n=24) or after resistance or intolerance to IFN-α (n=27). None of the patients had a history of allogeneic stem cell transplantation. The median duration of imatinib treatment was 57 months (range; 36-94). The median age at imatinib discontinuation was 63 years (range; 39-81) and the median follow-up after imatinib discontinuation was 63 months (range; 47-72). After treatment cessation, 28 patients (54.9%) experienced a protocol-defined molecular relapse and all restarted imatinib treatment. The median BCR-ABL1/ABL1 IS ratio was 0.015% (range; 0.0004-0.201) at first detection of molecular relapse, 0.056% (range; 0.006-1.54) at confirmation of molecular relapse and 0.165% (range; 0.024-4.49) at imatinib resumption (Figure 1A). The median time from imatinib discontinuation to first detection of molecular relapse was 2 months (range, 1-20) and all molecular relapses but

one occurred before 6 months. The median time until imatinib resumption was 4 months (range, 3-29). At 24 months, molecular relapse-free survival was 45.1% (95% CI: 31.44%-58.75%) (Figure 1B). These findings were consistent with those of the entire STIM study population.18 The median follow-up for non-relapsing patients was 63 months (range, 48-72).

Leukocyte, T-cell and natural kill-cell counts at imatinib discontinuation Leukocyte and lymphocyte counts at the time of imatinib discontinuation did not differ significantly between non-relapsing and relapsing patients (Online Supplementary Table S1). When CD3+ T cells, CD4+ and CD8+ T cells, naïve, central memory, effector memory CD4+ and CD8+ subsets and CD4+CD25+CD127low/- regulatory T cells were analyzed, no significant differences were observed between non-relapsing and relapsing patients (Online Supplementary Table S1). In contrast, although CD3-CD56+ NK cells were within the range established in our laboratory using samples from healthy donors (data not shown), non-relapsing patients had significantly higher frequencies

Table 1. Baseline characteristics of the patients (n=51).

Parameters

Results (n=51)*

Median age (range) 63 years (39-81) Female gender n, (%) 33 (64.7) Sokal score at diagnosis n, (%) Low 28 (54.9) Intermediate 18 (35.3) High 4 (7.8) Unknown 1 (2) IFN-α intolerance or resistance prior to imatinib n, (%) 27 (52.9) Median duration of imatinib treatment (range) 57 months (36-94) Daily dose of imatinib n, (%) 400 mg 44 (86.3) <400 mg 4 (7.8) Unknown 3 (5.9) Median follow-up (range) 63 months (47-72) *70 patients were enrolled in IMMUNOSTIM but only 51 were included. The 19 other patients were excluded because of missing baseline samples (n=4) or arrival of samples more than 48 h after blood was drawn (n=15).

1370

Figure 1. Evolution of BCR-ABL1 transcripts in relapsing patients and molecular relapse-free survival. (A) Scatter dot plots represent BCR-ABL1 IS % for each individual relapsing patient at first detection of relapse, relapse confirmation and at imatinib reintroduction; median values (horizontal bars) are also shown. (B) Kaplan-Meier estimate of molecular relapse-free survival defined as the time interval between imatinib discontinuation and first occurrence of molecular relapse or death, whichever came first. Data were censored at last molecular assessment for patients who were alive and had not relapsed.

haematologica | 2017; 102(8)

NK cells and outcome after stopping imatinib

and counts of CD3-CD56+ NK cells than had relapsing patients, with a median count of 233/mm3 (range, 70-727) in the former and 145/mm3 (range, 67-450) in the latter (P=0.014) (Figure 2A,B). This finding was not explained by differences in age or imatinib doses between the two groups (data not shown). In addition, we did not find any association between CD3-CD56+ NK-cell counts and duration of imatinib therapy (data not shown). CD3-CD56+ NK cells were thus subtyped based upon cell-surface density of the adhesion molecule CD56 and the FcRÎłIII receptor CD16. Counts of the cytotoxic CD3-CD56dimCD16+ population (hereafter named CD56dim) were significantly greater in non-relapsing patients [median: 216/mm3 (range, 55723)] than in relapsing patients [median: 139/mm3 (range, 58-438)] (P=0.011) and accounted for higher total NK-cell numbers in the former (Figure 2C). Indeed, counts of the cytokine-secreting CD3-CD56brightCD16-/low fraction (hereafter named CD56bright) did not differ between the two groups of patients (Figure 2D).

Baseline prognostic factors for molecular relapse In the STIM trial, the Sokal risk group and duration of imatinib treatment were associated with outcome following imatinib discontinuation.9,18 Here, clinical and biological variables and NK-cell counts at baseline were analyzed as potential prognostic factors for molecular relapse-free survival using univariate analysis. Patients with a low

Sokal score had a significantly higher molecular relapsefree survival rate than patients with an intermediate or high Sokal score (P=0.017) (Figure 3A). There was a trend toward a longer duration of imatinib treatment and higher estimated molecular relapse-free survival rate (P=0.078) (Figure 3B). Notably, patients with CD56dim NK-cell counts higher than the median (>162/mm3) at baseline had a significantly higher molecular relapse-free survival rate than those with a lower CD56dim NK-cell count (P=0.0008) (Figure 3C). No significant association with age, sex, or prior IFN-Îą therapy was found (Figure 3D-F). After multivariable analysis, CD56dim NK-cell count was identified as an independent prognostic factor for molecular-relapsefree survival (Table 2). These findings led us to analyze NK cells further in the setting of immunoprofiling and functional experiments.

Natural killer-cell receptors and maturation marker at imatinib discontinuation NK-cell receptors play a key role in recognizing targets and transducing activating or inhibitory signals upon binding to their ligands, thereby controlling cell function.19 We thus examined the expression of a large panel of NK-cell receptors and that of the carbohydrate antigen CD57, a marker linked to NK-cell terminal differentiation.20,21 No statistically significant differences were found on NK cells between non-relapsing and relapsing patients with respect

Figure 2. Natural killer cells at imatinib discontinuation in non-relapsing and relapsing patients. (A) Flow cytometry dot plot from a representative patient showing CD3-CD56+ NK cells in the upper left quadrant after lymphocyte gating using the side and forward scatter display. (B) Scatter dot plots represent CD3-CD56+ NK-cell counts for each individual non-relapsing and relapsing patient; median values (horizontal bars) are also shown. (C) Scatter dot plots represent CD3-CD56dim NK-cell counts for each individual non-relapsing and relapsing patient; median values (horizontal bars) are also shown. (D) Scatter dot plots represent CD3-CD56bright NK-cell counts for each individual non-relapsing and relapsing patient; median values (horizontal bars) are also shown. P values (by the Mann-Whitney U-test) are indicated for each panel.

haematologica | 2017; 102(8)

1371

D. Rea et al.

to expression of the activating receptors CD16, NKG2D, NKp46, NKp30 and DNAM-1, the maturation marker CD57, the activating and inhibitory KIR2D isoforms and the C-type lectin inhibitory receptor NKG2A/CD94 (data not shown). However, when we compared NK-cell receptor expression profile of patients with that of healthy individuals, significant alterations were found. The proportions

of NK cells expressing the activating receptors NKp46 and DNAM-1 were significantly lower in the CD56dim and CD56bright subsets of non-relapsing and relapsing patients (Figure 4). In addition, the CD56bright subset of non-relapsing and relapsing patients showed significantly increased KIR2D and CD57 and decreased NKG2A expression as compared to the expression in healthy donors (Figure 4).

Figure 3. Molecular relapse-free survival after discontinuation of imatinib according to clinico-biological factors. (A) Sokal risk group, (B) imatinib treatment duration, (C) CD56dim NK-cell counts at baseline, (D) age, (E) sex, and (F) prior exposure to IFN-Îą. For each survival plot, a corresponding log-rank P value is shown.

1372

haematologica | 2017; 102(8)

NK cells and outcome after stopping imatinib

Natural killer-cell function at imatinib discontinuation The capacity of NK cells to degranulate cytolytic vesicles was measured through the delocalization of the lysosomal-associated membrane protein-1 CD107a onto the cell surface after stimulation with the NK-sensitive BCRABL1+ cell line K562. This assay is considered a general indicator of NK activity and particularly of tumor cell lysis.22 Raw quantification of CD107a in the presence of K562 indicated a comparable degranulation in non-relapsing and relapsing patients and in healthy donors (Figure 5A). However, the propensity of NK cells for surface CD107a expression in the absence of K562 revealed significantly higher levels of spontaneous degranulation in both groups of patients (Figure 5A). Consequently, K562-specific degranulation of NK cells from non-relapsing and relapsing patients was significantly weaker than that of NK cells from healthy donors (Figure 5B). This finding was corroborated by a lower induction of the tumor necrosis factor receptor CD137 (also known as 4-1BB) upon K562 encounter in patients, a marker upregulated following NKcell activation (Figure 5C). We also investigated the ability of the CD56bright NK-cell subset to secrete IFN-γ after stimulation by IL-12 and IL-18. We found that IFN-γ production was comparable in non-relapsing and relapsing patients and significantly reduced compared to that in healthy donors (Figure 5D).

Natural killer cells after imatinib discontinuation Several studies support the notion that TKI exert off-target effects on NK cells and other cells of the immune system.23-26 We thus wondered whether cell counts, receptor expression and functional features assessed at baseline could have been influenced at least in part by imatinib. To address this point, experiments were repeated after cessation of imatinib. Because most relapsing patients resumed imatinib within a short time frame (64.3% within 3 months and 85.7% within 6 months), follow-up samples obtained within 6 months after imatinib discontinuation were used, prior to imatinib reintroduction. After imatinib was stopped, median leukocyte and lymphocyte counts increased in non-relapsing and relapsing patients from 4810/mm3 (range, 2250-7860) to 5950/mm3 (range, 376010440) (P<0.0001) and 1310/mm3 (range, 720-2610) to 1520/mm3 (range, 670-3060), respectively (P<0.0001). A rise in NK cells also occurred in both groups of patients, from a median value of 179/mm3 (range, 67-727) at baseline to 205/mm3 (range, 74-736) after imatinib discontinuation (P=0.0011). This rise was observed within both the CD56dim subset [median count 164/mm3 (range, 55-723) at baseline and 195/mm3 (range, 62-723) after imatinib discontinuation (P=0.0032)] and the CD56bright fraction [median count 7/mm3 (range, 2-31) at baseline and 10/mm3 (range, 4-37) after imatinib discontinuation (P=0.0008)]. Of note, CD3+ T cells also increased from a median value of 889/mm3 (range, 315-1949) at baseline to 990/mm3 (range, 265-1817) after imatinib discontinuation (P=0.0445). Importantly, the comparison between groups after imatinib discontinuation showed that NK cells of the CD56dim subset remained significantly higher in nonrelapsing patients, with a median value of 249/mm3 (range, 85-723), than in relapsing patients, who had a median value of 148/mm3 (range, 62-442) (P=0.0179) (Table 3). Immunophenotypic analyses revealed no significant modification in the expression of NKp46, NKp30, DNAM-1 and KIR2D after imatinib discontinuation in haematologica | 2017; 102(8)

either group of patients (Online Supplementary Figure S1AD). A significant reduction in NKG2A+ NK cells was observed in non-relapsing patients but not in relapsing patients (Online Supplementary Figure S1E). The proportion of CD57+ NK cells was significantly decreased in relapsing patients (Online Supplementary Figure S1F). NK-cell degranulation and activation and IFN-γ production did not show any significant enhancement (Online Supplementary Figure S1G-I).

Discussion The current principles of CML management rely on the use of life-saving TKI through the induction of an optimal response to prevent progression to blast crisis, followed by maintenance of the optimal response by means of lifelong TKI treatment.7 However, recent demonstrations that imatinib can be successfully stopped in a substantial proportion of patients with deep and durable molecular responses is currently modifying this view and avoiding lifelong dependency on TKI treatment may become the ultimate goal.27-29 While serial molecular monitoring is essential to detect a relapse rapidly after an attempt to discontinue TKI, in order to trigger medical intervention and avoid any negative impact on the patient’s outcome, our ability to predict relapses accurately remains limited and determinants of relapse are unknown.30 Here, we demonstrated that patients free from molecular relapse after imatinib discontinuation within the STIM trial had higher numbers of circulating NK cells than had patients who relapsed. In addition to CML-related variables, such as the Sokal score and imatinib treatment duration, a greater load of CD56dim NK cells at the time of imatinib discontinuation was associated with a greater chance of treatment cessation being successful in univariate analysis. Although IFN-α has the ability to activate NK cells, no association was found between past exposure to IFN-α and molecular relapse-free survival, in agreement with what was previously found in the single-agent imatinib discontinuation studies STIM and TWISTER.9,10 Multivariable analysis enabled us to show that the amount of CD56dim NK cells was the only independent prognostic factor for molecular relapse-free survival in this population of patients. Two recent and independent studies are in line with our results. Indeed in the DADI dasatinib discontinuation trial and in the EUROSKI imatinib cessation immunological sub-study, greatest peripheral NK-cell burden at the time TKI therapy was stopped was associated with a higher probability of treatment-free remission.31,32 Given the importance of NK cells in immune

Table 2. Potential prognostic factors of molecular relapse-free survival: multivariable Cox proportional analysis.

Variable Sokal score* (n=50) Imatinib duration (n=51) CD56dim NK-cell counts (n=51)

Molecular relapse-free survival over time (n=51) Hazard ratio (95% CI) P value 0.474 (0.219-1.024) 0.573 (0.256-1.286) 0.292 (0.122-0.699)

0.0576 0.177 0.0057

*The Sokal score was categorized as low versus intermediate and high. Quantitative variables were categorized into two groups with cut-offs set at the median. P<0.05 was considered statistically significant.

1373

D. Rea et al.

surveillance against malignancies, these results and ours led us to hypothesize that NK cells and especially the differentiated and cytotoxic CD56dim subset20 might contribute to prevent overt CML relapse originating from residual leukemic cells, in a direct or indirect fashion, after imatinib discontinuation. The involvement of NK cells in immune surveillance against CML is also supported by

other lines of evidence. BCR-ABL1+ CD34+ cells from CML patients express NK activating receptor ligands.33-37 In addition, BCR-ABL1+ CD34+ cells from CML patients are sensitive to NK-cell-mediated lysis in vitro. Indeed, Sconocchia and colleagues34 found that IL-2-stimulated NK cells from allogeneic HLA-matched siblings inhibited BCR-ABL1+ CD34+-derived colony-forming unit â&#x20AC;&#x201C; granulocyte-

Figure 4. Expression of natural killer cell receptors in CD56dim and CD56bright natural killer cell subsets. Scatter dot plots for healthy donors (black dots, n=44), non-relapsing patients (filled gray dots, n=19) and relapsing patients (empty black dots, n=27) with median values (horizontal bars) are shown. Overall P value (by the Kruskall-Wallis test) is indicated for each panel.

1374

haematologica | 2017; 102(8)

NK cells and outcome after stopping imatinib

macrophage growth in colony formation assays. Yong and colleagues38 reported that primitive quiescent BCR-ABL1+ CD34+ stem cells were sensitive to lysis by NK cells from HLA-identical siblings, although to a lesser extent than cycling CD34+ leukemic cells. Cervantes and colleagues39 observed that IL-2-activated NK cells were capable of suppressing autologous BCR-ABL1+ CD34+ cell-derived colony-forming cells and long-term culture-initiating cells. A further indication of a protective role of NK cells against CML is that early recovery of NK cells was able to predict a positive clinical outcome for patients undergoing T-celldepleted allogeneic transplantation from HLA-identical

siblings.40 Moreover during treatment with the NK-cellstimulating cytokine, IFN-α, an association between the achievement of a complete cytogenetic response and NKcell cytolytic activity was described.41,42 Finally, cross-talk between NK cells and other immune cells, such as dendritic cells, may potentiate anti-CML adaptive immune responses.43,44 Despite striking quantitative differences, we found that the immunophenotypic and functional features of NK cells from non-relapsing and relapsing patients at the time of imatinib discontinuation were comparable. In addition, NK cells from both groups of patients differed phenotypically and functionally from those of healthy individuals. Expression of the activating receptors NKp46 and DNAM1 was decreased. Overall degranulation upon encounter of K562 cells was preserved, as found in the immunological EUROSKI sub-study but contrary to EUROSKI, we were able to show lower activation marker induction and higher basal CD107a mobilization in patients, suggesting poorer K562-specific degranulation. These differences between the immunological EUROSKI sub-study and our study remain unexplained. A high level of spontaneous degranulation was described in other pathological settings, such as in human immunodeficiency virus infection, and was linked to chronic inflammation but its significance in wellcontrolled CML remains unclear.45 Unfortunately, it was not possible to assess degranulation toward autologous leukemic cells because it is not routine practice to collect and store CD34+ BCR-ABL1+ cells at CML diagnosis, it is challenging to find and isolate very rare residual leukemic stem cells at imatinib discontinuation and this was not planned in our study. Finally, we found that IFN-γ production in response to cytokine stimulation in the CD56bright subset was lower in patients irrespective of their relapse status. Although NK cells are not derived from the leukemic clone in CP-CML, deficient NK-cell function has been described by several groups.25,46,47 In newly diagnosed CP-CML prior to any treatment, NK-cell proliferation in response to the K562 leukemic cell line and IL-2 is reduced, as is degranulation triggered by K562 alone.25

Table 3. Leukocytes, lymphocytes and natural killer-cell subsets after imatinib discontinuation.

Non-relapsing (n=23) D

Figure 5. Natural killer-cell function. (A) NK-cell degranulation assay against K562 or medium control. (B) Target-specific degranulation estimated by the CD107a ratio. (C) NK-cell activation estimated by CD137 expression in the presence of K562 or medium control. (D) Production of IFN-γ upon stimulation with IL-12 and IL-18. Scatter dot plots for healthy donors (black dots, n=43), nonrelapsing patients (filled gray dots, n=15) and relapsing patients (empty black dots, n=25) at baseline with median values (horizontal bars) are shown. Overall P value (by the Kruskall-Wallis test) is indicated for each panel.

haematologica | 2017; 102(8)

Leukocytes /mm3 Median 6360 Range 3760-10440 Lymphocytes /mm3 Median 1580 Range 1140-3060 CD3-CD56+ NK cells /mm3 Median 260 Range 99-736 CD56dim NK cells /mm3 Median 249 Range 85-723 CD56bright NK cells /mm3 Median 11 Range 5-37

Relapsing (n=26) 5465 3920-7100 1375 670-2500

P value* 0.072

0.0134

155 74-452

0.0205

148 62-442

0.0179

9 4-27

0.4455

*The Mann-Whitney U test was used to compare variables from non-relapsing and relapsing patients with a level of significance of 0.05. Median values and range (minmax) are shown.

1375

D. Rea et al.

These alterations are partially restored by imatinib, thereby suggesting that the disease itself has a deleterious impact on NK cells which persists beyond the imatinibinduced remission.25 In our study, imatinib discontinuation was accompanied by a rapid and significant increase in leukocytes and lymphocytes, as also observed by Park and colleagues.48 In addition, a significant increase in NK cells occurred. This finding is concordant with the inhibitory impact of imatinib on NK-cell expansion in vitro, although such an impact was described at higher concentrations than the trough plasma concentrations usually expected in patients treated with imatinib at a dose of 400 mg daily.25 It is also consistent with the fact that patients in deep molecular response who have successfully stopped imatinib have higher NK-cell counts than those with a comparable level of response but who do not stop the drug.49 However, it is important to note that in our study, non-relapsing patients maintained higher counts of NK cells than relapsing patients after imatinib cessation and that the imatinib dose prior to discontinuation was similar in non-relapsing and relapsing patients. Although imatinib plasma levels were not measured prior to discontinuation, this argues against a different drug exposure as the sole explanation for a lower NK-cell burden in relapsing patients. Other factors involved in NK-cell peripheral expansion, such as stromal cell factors or the immunoregulatory cytokine IL-15, need to be explored in future research studies.50 In our experimental setting, discontinuation of imatinib did not result in the disappearance of NK-cell differences with healthy individuals, ruling out a major deleterious effect of the drug on NK-cell function. Expression of the activating receptors NKp46 and DNAM-1 remained unchanged. A significant decrease in NKG2A+ NK cells in non-relapsing patients was observed but there was no gain in cytolytic and CD56bright-IFN-γ secretion capacities. Our study was not designed to decipher the underlying mechanisms of NK-cell defects in CML but several hypotheses can be made, including a negative impact of residual leukemic stem cells or their microenvironment on immune cell function or a weaker activation of the NK-cell repertoire due to HLA class I/KIR variation as compared to that in healthy individuals.36 In the immunological

References 1. Druker BJ, Guilhot F, O'Brien SG, et al; IRIS Investigators. Five-year follow-up of patients receiving imatinib for chronic myeloid leukemia. N Engl J Med. 2006;355(23):2408-2417. 2. Kalmanti L, Saussele S, Lauseker M, et al. Safety and efficacy of imatinib in CML over a period of 10 years: data from the randomized CML-study IV. Leukemia. 2015;29(5): 1123-1132. 3. Gambacorti-Passerini C, Antolini L, Mahon FX, et al. Multicenter independent assessment of outcomes in chronic myeloid leukemia patients treated with imatinib. J Natl Cancer Inst. 2011;103(7):553-561. 4. Sasaki K, Strom SS, O'Brien S, et al. Relative survival in patients with chronic-phase chronic myeloid leukaemia in the tyrosine-

1376

EUROSKI sub-study, individual KIR genes and haplotypes were not associated with imatinib discontinuation outcome but, unfortunately, comparisons with the general population were not performed.32 To conclude, we provide evidence that NK cells are associated with outcome after imatinib discontinuation in CP-CML patients in deep molecular response. Beyond functional aspects, larger amounts of cytotoxic CD56dim NK cells delivered to the leukemic reservoir in non-relapsing patients may play an important role in controlling residual CML-initiating cells and their progeny soon after cessation of imatinib treatment while a reduced CD56dim compartment, and thus a lower effector to target cell ratio in relapsing patients, may leave less chance to counterattack in an efficient manner. Of course, our results do not preclude the role of additional biological aspects in dictating patients’ outcome, such as the involvement of other immunological factors or functional heterogeneity of the residual BCR-ABL1+ hematopoietic stem-cell reservoir. Thus integration of NK-cell counts into the TKI discontinuation decision-making process is not reasonable at this stage. In addition, NK-cell alterations in patients suggest that the use of agents that stimulate NK-cell expansion or function, such as the IL-15 cytokine, or antibodies modulating NK-cell receptors may increase the chance of successfully TKI discontinuation, an aspect that may be addressed in future studies. Acknowledgments The authors would like to thank the patients and the personnel from clinical centers involved in this study and Ms. Nicole Biscay, Valérie Hubert, Eliane Melgire and Véronique Delasse from the Cell Therapy Unit, Hôpital Saint-Louis, Paris, France for expert logistic and technical assistance. Funding This work was supported by research grants from the Association pour la Recherche sur le Cancer (ZK, #DOC20100600956), Association LMC France (#R13102HH), Assistance Publique-Hôpitaux de Paris (Translational Research Grant in Biology 2010 #RTB10002) and the French Ministry of Health/Institut National du Cancer (Programme Hospitalier de Recherche 2006).

kinase inhibitor era: analysis of patient data from six prospective clinical trials. Lancet Haematol. 2015;2(5):e186-e193. Corbin AS, Agarwal A, Loriaux M, Cortes J, Deininger MW, Druker BJ. Human chronic myeloid leukemia stem cells are insensitive to imatinib despite inhibition of BCR-ABL activity. J Clin Invest. 2011;121(1):396-409. Hamilton A, Helgason GV, Schemionek M, et al. Chronic myeloid leukemia stem cells are not dependent on Bcr-Abl kinase activity for their survival. Blood. 2012;119(6):15011510. Baccarani M, Deininger MW, Rosti G, et al. European LeukemiaNet recommendations for the management of chronic myeloid leukemia: 2013. Blood. 2013;122(6):872-884. Chomel JC, Bonnet ML, Sorel N, et al. Leukemic stem cell persistence in chronic myeloid leukemia patients in deep molecular response induced by tyrosine kinase

inhibitors and the impact of therapy discontinuation. Oncotarget. 2016;7(23):3529335301. 9. Mahon FX, Réa D, Guilhot J, et al; Intergroupe Français des Leucémies Myéloïdes Chroniques. Discontinuation of imatinib in patients with chronic myeloid leukaemia who have maintained complete molecular remission for at least 2 years: the prospective, multicentre Stop Imatinib (STIM) trial. Lancet Oncol. 2010;11(11): 1029-1035. 10. Ross DM, Branford S, Seymour JF, et al. Safety and efficacy of imatinib cessation for CML patients with stable undetectable minimal residual disease: results from the TWISTER study. Blood. 2013;122(4):515522. 11. Ross DM, Branford S, Seymour JF, et al. Patients with chronic myeloid leukemia who maintain a complete molecular

haematologica | 2017; 102(8)

NK cells and outcome after stopping imatinib

12.

13.

14.

15. 16.

17.

18.

19. 20.

21.

22.

23.

24.

response after stopping imatinib treatment have evidence of persistent leukemia by DNA PCR. Leukemia. 2010;24(10):17191724. Richter J, Söderlund S, Lübking A, et al. Musculoskeletal pain in patients with chronic myeloid leukemia after discontinuation of imatinib: a tyrosine kinase inhibitor withdrawal syndrome. J Clin Oncol. 2014;32 (25):2821-2823. Mori S, Vagge E, le Coutre P, et al. Age and dPCR can predict relapse in CML patients who discontinued imatinib: the ISAV study. Am J Hematol. 2015;90(10):910-914. Lee SE, Choi SY, Song HY, et al. Imatinib withdrawal syndrome and longer duration of imatinib have a close association with a lower molecular relapse after treatment discontinuation: the KID study. Haematologica. 2016;101(6):717-723. Rohon P. Biological therapy and the immune system in patients with chronic myeloid leukemia. Int J Hematol. 2012;96(1):1-9. Barrett AJ, Ito S. The role of stem cell transplantation for chronic myelogenous leukemia in the 21st century. Blood. 2015;125(21):3230-3235. Khaznadar Z, Henry G, Setterblad N, et al. Acute myeloid leukemia impairs natural killer cells through the formation of a deficient cytotoxic immunological synapse. Eur J Immunol. 2014;44(10):3068-3080. Etienne G, Guilhot J, Rea D, et al, on behalf of the Intergroupe Français des Leucémies Myéloïdes Chroniques (Fi-LMC). Long-term follow-up of the French Stop Imatinib Study (STIM1) in chronic myeloid leukemia patients. J Clin Oncol. 2017;35(3):298-305. Carrillo-Bustamante P, Ke mir C, de Boer RJ. The evolution of natural killer cell receptors. Immunogenetics. 2016;68(1):3-18. Björkström NK, Riese P, Heuts F, et al. Expression patterns of NKG2A, KIR, and CD57 define a process of CD56dim NK-cell differentiation uncoupled from NK-cell education. Blood. 2010;116(19):3853-3864. Nielsen CM, White MJ, Goodier MR, Riley EM. Functional significance of CD57 expression on human NK cells and relevance to disease. Front Immunol. 2013;4(422):1-8. Alter G, Malenfant JM, Altfeld M. CD107a as a functional marker for the identification of natural killer cell activity. J Immunol Methods. 2004;294(1-2):15-22. Cebo C, Da Rocha S, Wittnebel S, et al. The decreased susceptibility of Bcr/Abl targets to NK cell-mediated lysis in response to imatinib mesylate involves modulation of NKG2D ligands, GM1 expression, and synapse formation. J Immunol. 2006;176(2):864-872. Salih J, Hilpert J, Placke T, et al. The BCR/ABL-inhibitors imatinib, nilotinib and dasatinib differentially affect NK cell reactivity. Int J Cancer. 2010;127(9):2119-2128.

haematologica | 2017; 102(8)

25. Chen CI, Koschmieder S, Kerstiens L, et al. NK cells are dysfunctional in human chronic myelogenous leukemia before and on imatinib treatment and in BCR-ABL-positive mice. Leukemia. 2012;26(3):465-474. 26. Zitvogel L, Rusakiewicz S, Routy B, Ayyoub M, Kroemer G. Immunological off-target effects of imatinib. Nat Rev Clin Oncol. 2016;13(7):431-446. 27. Rea D, Rousselot P, Guilhot J, Guilhot F, Mahon FX. Curing chronic myeloid leukemia. Curr Hematol Malig Rep. 2012;7 (2):103-108. 28. Hughes TP, Ross DM. Moving treatmentfree remission into mainstream clinical practice in CML. Blood. 2016;128(1):17-23. 29. Saußele S, Richter J, Hochhaus A, Mahon FX. The concept of treatment-free remission in chronic myeloid leukemia. Leukemia. 2016;30(8): 1638-1647. 30. Deininger MW. Molecular monitoring in CML and the prospects for treatment-free remissions. Hematology Am Soc Hematol Educ Program. 2015;(2015):257-263. 31. Imagawa J, Tanaka H, Okada M, et al; DADI Trial Group. Discontinuation of dasatinib in patients with chronic myeloid leukaemia who have maintained deep molecular response for longer than 1 year (DADI trial): a multicentre phase 2 trial. Lancet Haematol. 2015;2(12):e528-e535. 32. Ilander M, Olsson-Strömberg U, Schlums H, et al. Increased proportion of mature NK cells is associated with successful imatinib discontinuation in chronic myeloid leukemia. Leukemia. 2017;31(5):1108-1116. 33. Salih HR, Antropius H, Gieseke F, et al. Functional expression and release of ligands for the activating immunoreceptor NKG2D in leukemia. Blood. 2003;102(4):1389-1396. 34. Sconocchia G, Lau M, Provenzano M, et al. The antileukemia effect of HLA-matched NK and NK-T cells in chronic myelogenous leukemia involves NKG2D-target-cell interactions. Blood. 2005;106(10):3666-3672. 35. Boissel N, Rea D, Tieng V, et al. BCR/ABL oncogene directly controls MHC class I chain-related molecule A expression in chronic myelogenous leukemia. J Immunol. 2006;176(8):5108-5116. 36. Verheyden S, Demanet C. NK cell receptors and their ligands in leukemia. Leukemia. 2008;22(2):249-257. 37. Hilpert J, Grosse-Hovest L, Grünebach F, et al. Comprehensive analysis of NKG2D ligand expression and release in leukemia: implications for NKG2D-mediated NK cell responses. J Immunol. 2012;189(3):1360-1371. 38. Yong AS, Keyvanfar K, Hensel N, et al. Primitive quiescent CD34+ cells in chronic myeloid leukemia are targeted by in vitro expanded natural killer cells, which are functionally enhanced by bortezomib. Blood. 2009;113(4):875-882. 39. Cervantes F, Pierson BA, McGlave PB,

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

Verfaillie CM, Miller JS. Autologous activated natural killer cells suppress primitive chronic myelogenous leukemia progenitors in long-term culture. Blood. 1996;87(6): 2476-2485. Savani BN, Rezvani K, Mielke S, et al. Factors associated with early molecular remission after T cell-depleted allogeneic stem cell transplantation for chronic myelogenous leukemia. Blood. 2006;107(4):16881695. Pawelec G, Da Silva P, Max H, et al. Relative roles of natural killer- and T cell-mediated anti-leukemia effects in chronic myelogenous leukemia patients treated with interferon-alpha. Leuk Lymphoma. 1995;18(56):471-478. Meseri A, Delwail V, Brizard A, et al. Endogenous lymphokine activated killer cell activity and cytogenetic response in chronic myelogenous leukaemia treated with alphainterferon. Br J Haematol. 1993;83(2):218222. Degli-Esposti MA, Smyth MJ. Close encounters of different kinds: dendritic cells and NK cells take centre stage. Nat Rev Immunol. 2005;5(2):112-124. Walzer T, Dalod M, Robbins SH, Zitvogel L, Vivier E. Natural-killer cells and dendritic cells: "l'union fait la force". Blood. 2005;106(7):2252-2258. Lichtfuss GF, Cheng WJ, Farsakoglu Y, et al. Virologically suppressed HIV patients show activation of NK cells and persistent innate immune activation. J Immunol. 2012;189(3):1491-1499. Takahashi N, Miura I, Saitoh K, Miura AB. Lineage involvement of stem cells bearing the philadelphia chromosome in chronic myeloid leukemia in the chronic phase as shown by a combination of fluorescenceactivated cell sorting and fluorescence in situ hybridization. Blood. 1998;92(12):47584763. Mellqvist UH, Hansson M, Brune M, Dahlgren C, Hermodsson S, Hellstrand K. Natural killer cell dysfunction and apoptosis induced by chronic myelogenous leukemia cells: role of reactive oxygen species and regulation by histamine. Blood. 2000;96(5): 1961-1968. Park JS, Lee SE, Jeong SH, et al. Change of health-related profiles after Imatinib cessation in chronic phase chronic myeloid leukemia patients. Leuk Lymphoma. 2016;57(2):341-347. Ohyashiki K, Katagiri S, Tauchi T, et al. Increased natural killer cells and decreased CD3(+)CD8(+)CD62L(+) T cells in CML patients who sustained complete molecular remission after discontinuation of imatinib. Br J Haematol. 2012;157(2):254-256. Fehniger TA, Caligiuri MA. Interleukin 15: biology and relevance to human disease. Blood. 2001;97(1):14-32.

1377

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Acute Myeloid Leukemia

Ferrata Storti Foundation

FLT3 and FLT3-ITD phosphorylate and inactivate the cyclin-dependent kinase inhibitor p27Kip1 in acute myeloid leukemia Ines Peschel,1 Silvio R. Podmirseg,1 Martin Taschler,1 Justus Duyster,2 Katharina S. Götze,3 Heinz Sill,4 David Nachbaur,5 Heidelinde Jäkel1 and Ludger Hengst1

Haematologica 2017 Volume 102(8):1378-1390

Division of Medical Biochemistry, Biocenter, Medical University of Innsbruck, Austria; Department of Hematology, Oncology and Stem Cell Transplantation, University Medical Center Freiburg, Germany; 3Department of Internal Medicine III, Klinikum Rechts der Isar, Technical University of Munich, Germany; 4Division of Hematology, Department of Internal Medicine, Medical University of Graz, Austria and 5Department of Internal Medicine V, Medical University of Innsbruck, Austria

1 2

ABSTRACT

Correspondence: heidelinde.jaekel@i-med.ac.at

Received: November 25, 2016. Accepted: May 8, 2017. Pre-published: May 18, 2017.

27Kip1 (p27) can prevent cell proliferation by inactivating cyclindependent kinases. This function is impaired upon phosphorylation of p27 at tyrosine residue 88. We observed that FLT3 and FLT3-ITD can directly bind and selectively phosphorylate p27 on this residue. Inhibition of FLT3-ITD in cell lines strongly reduced p27 tyrosine 88 phosphorylation and resulted in increased p27 levels and cell cycle arrest. Subsequent analysis revealed the presence of tyrosine 88 phosphorylated p27 in primary patient samples. Inhibition of FLT3 kinase activity with AC220 significantly reduced p27 tyrosine 88 phosphorylation in cells isolated from FLT3 wild type expressing acute myeloid leukemia (AML) patients. In FLT3-ITD positive AML patients, p27 tyrosine 88 phosphorylation was reduced in 5 out of 9 subjects, but, surprisingly, was increased in 4 patients. This indicated that other tyrosine kinases such as Src family kinases might contribute to p27 tyrosine 88 phosphorylation in FLT3-ITD positive AML cells. In fact, incubation with the Src family kinase inhibitor dasatinib could decrease p27 tyrosine 88 phosphorylation in these patient samples, indicating that p27 phosphorylated on tyrosine 88 may be a therapeutic marker for the treatment of AML patients with tyrosine kinase inhibitors.

doi:10.3324/haematol.2016.160101 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1378 ©2017 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.

1378

Introduction Cell proliferation and cell cycle progression are tightly regulated by the sequential activation and inactivation of specific cyclin-dependent kinases (CDKs).2 Binding of the CDK inhibitor p27Kip1 (p27) can regulate CDK activity and thereby control cell cycle progression from G0/G1 phase to S phase. p27 regulates not only CDK activity, but also transcription and cell motility.2,3 p27 levels are elevated in non-proliferating cells and decline when cells progress towards S phase.4 Whereas p27 mRNA levels are frequently not altered during the cell cycle, protein levels of p27 can fluctuate dramatically.2,4 The rapid elimination of p27 at the G1/S transition is triggered through ubiquitin-dependent proteasomal degradation by the SCFSkp2 E3 ligase complex.5 Cyclin-dependent kinase inactivation by p27 involves the insertion of a 310-helix of the inhibitor into the catalytic cleft of the kinase, thereby blocking access of ATP.6 Interestingly, phosphorylation of p27 on residue tyrosine 88 (pY88) leads to the ejection of the inhibitory 310-helix from the catalytic cleft, permitting access of ATP7 and partial activation of p27-bound CDK complexes.7-11 The partially active cyclin-CDK2 can now phosphorylate substrates, including the bound p27 on T187.7 T187-phosphorylation is a prerequisite for p27 ubiquitination by SCFSkp2, haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

initiating its proteasomal degradation.5 This mechanism directly couples mitogen-induced activation of tyrosine kinases to cell cycle control, but can also be used during oncogenic transformation of cancer cells.12 The non-receptor tyrosine kinases JAK2, Abl, BCR-Abl, Lyn, Yes, Src, and Brk can phosphorylate p27 on Y88 and likely employ this mechanism to inactivate p27 and to promote cell proliferation.7,8,11,13 The Fms-like tyrosine kinase 3 (FLT3) is a member of the class III subfamily of receptor tyrosine kinases and is activated by FLT3 ligand (FL).14 FLT3 is expressed in early hematopoietic progenitor cells in the bone marrow.14 High FLT3 levels have been detected in acute myeloid leukemia (AML),15,16 where activating FLT3 mutations can be found in approximately 30% of the patients.14,17 In fact, the most common mutation in AML is the internal tandem duplication (ITD) in the juxtamembrane domain of FLT3 with a 20-27% occurrence. FLT3-ITD serves as a prognostic marker since it positively correlates with higher blast counts, increased relapse rate, and worse overall survival.17-19 Several activating point mutations in the tyrosine kinase domain (TKD) have also been identified.14 Acute myeloid leukemia cells show increased proliferation and survival, as well as impaired hematopoietic differentiation.14 FLT3-ITD or FLT3 activation confers proliferative and survival advantages to cells14,20 by activating Src family tyrosine kinases (SFKs), the PI3K/Akt-, mitogenactivated protein kinase (MAPK) pathways, and, in the case of FLT3-ITD, also Stat5.20 Identifying the downstream targets of FLT3 and FLT3ITD is essential to understanding the mechanisms through which they promote leukemia development. In the present study, we identified p27 as a novel direct substrate of FLT3 and FLT3-ITD. FLT3 inhibitor treatment efficiently reduced pY88-p27 in FLT3-ITD expressing cell lines and increased p27 protein levels. Analysis of cells from AML patients demonstrates for the first time that p27 is phosphorylated on Y88 in primary patient material. This uncovers a novel pathway with which FLT3 can promote hyperproliferation of AML cells.

Methods Cell lines and primary cells Cells were incubated at 37Â°C with 5% CO2 in DMEM (293T, U2OS) or RPMI (MV4;11, U937, Ba/F3, 32D) medium including 10% FCS. Primary blast cells were obtained from bone marrow aspirates or peripheral blood of AML patients. Written informed consent was obtained from all individuals in accordance with the Declaration of Helsinki. The use of human material was approved by the ethics committees of the Medical University of Innsbruck (AN2014-0362 344/4.22 345/4.4 346/4.1), Graz (27-372 14/15), and the Technical University of Munich (5689/13, 349/13, 276/15). Mononuclear cells were purified with Biocoll Separating Solution (Biochrom, Berlin, Germany), frozen in media containing 10% DMSO or immediately cultured in RPMI medium supplemented with 20% FCS for two hours (h) before treatment with the FLT3 inhibitor AC220.

Transient transfection, Western blot and immunoprecipitation 293T cells were analyzed 40 h after transfection using the calcium phosphate method.21 Protein extracts were prepared with 0.5% NP-40 containing lysis buffer and analyzed by Western blot haematologica | 2017; 102(8)

as described.13 Bands were visualized by ECL- (GE-Healthcare) or Odyssey infrared fluorescence detection (LI-COR Biosciences) as indicated. Immunoprecipitation was performed as described13 using antibodies covalently coupled to protein A-agarose (Immunosorb A; Medicargo, Sweden). The antibodies are specified in the Online Supplementary Methods.

FACS Cell cycle analyses were performed using bromodeoxyuridine (BrdU) and propidium iodide (PI) double staining, and analyzed as described previously.22

Protein interaction analysis Pull down assays with recombinant isolated proteins were performed as described previously.13 Pull down experiments using His-p27 domains bound to AffiGel10 are described in the Online Supplementary Methods.

Kinase assay In vitro phosphorylation was performed with Flag-tagged catalytically active cytoplasmic domain of FLT3-ITD (amino acids 564-1006) immunoprecipitated from transfected 293T cells or recombinant FLT3-ITD fused to GST-His6 (ProQinase) and Hisp27, His-p21 and His-p57 as substrates, as described previously.13

Immunoflourescence U2OS cells transfected with Flag-FLT3-ITD, Flag-FLT3, p27 or p27-Y88F were analyzed as described.13 Briefly, cells were fixed with 4% para-formaldehyde, permeabilized with 0.025% saponine, and blocked with 0.5% gelatine. A Leica DMi8 inverted widefield microscope was used for imaging.

Statistical analysis P-values were determined by Wilcoxon test. ***P<0.001, **P<0.01, *P<0.05. n.s.: not significant.

Results FLT3 and FLT3-ITD induce p27 phosphorylation, and inhibition of FLT3-ITD increases p27 expression MV4;11 leukemia-derived cells express constitutively active FLT3-ITD.23 AC220 (Quizartinib) is a potent small molecule inhibitor of FLT3/FLT3-ITD24 currently under clinical phase III investigation (clinicaltrials.gov identifier: 02039726). Inhibition of FLT3-ITD with AC220 for 18 h resulted in a clear increase of p27 in MV4;11 cells (Figure 1A and Online Supplementary Figure S1). As expected, AC220 also inhibited cell proliferation, induced an accumulation of cells in G0/G1 phase of the cell cycle, and increased the proportion of apoptotic cells (Figure 1B and C and Online Supplementary Figures S1-S3). To exclude potential cell-type specific effects of AC220, we also inhibited FLT3-ITD in a murine 32D cell line which had been stably transfected to express FLT3-ITD (32D-FLT3-ITD) and which proliferates independently of cytokines in the presence of this oncogene. Inhibition with AC220 led to a similar increase in p27, cell cycle arrest, and increased apoptosis (Figure 1D-F and Online Supplementary Figures S4 and S5). Inactivation of FLT3-ITD was confirmed by inhibited autophosphorylation on Y589/591 (Figure 1A and D, pY-FLT3). Since FLT3 undergoes glycosylation in the Golgi, immature (130 kDa) and mature forms (160 kDa) were detected.16,25 Whereas active FLT3-ITD predominantly accumulates in its underglycosylated form, inhibition of FLT3-ITD has 1379

I. Peschel et al.

been described to increase the ratio of the mature over the immature form, and to increase its abundance in MV4;11 cells25 (Figure 1A and Online Supplementary Figure S1). To investigate if FLT3-ITD or FLT3 can induce tyrosine phosphorylation of p27, we expressed the proteins in 293T cells. Transfection of FLT3-ITD or FLT3 induced a strong phosphorylation of over-expressed p27 on tyrosine residues (Figure 1G, pY-p27). Using a monoclonal antibody specific for p27 phosphorylated at tyrosine residue 88 (pY88-p27), we observed that phosphorylation on this site was strongly induced by FLT3 and FLT3-ITD (Figure 1G, pY88-p27). Phosphorylation on Y88 converts p27 into a substrate of the bound CDK2/cyclin complex, leading to subsequent phosphorylation of p27 on T187 and its proteasomal degradation.7 Consistent with this observation, CDK2-cyclin E-dependent T187 phosphorylation of p27 was increased upon FLT3-ITD overexpression and resulted in decreased p27 levels (Figure 1H, left panel).

Importantly, the p27-Y88F mutant failed to show the increase in p27-T187 phosphorylation as well as the decrease of p27 levels (Figure 1H, right panel), supporting the model that Y88 phosphorylation mediates the FLT3ITD-induced decline of p27.

p27 is directly and selectively phosphorylated on tyrosine 88 by FLT3 and FLT3-ITD To investigate the individual contribution of each of the three tyrosine residues present in p27 (Y74, Y88 and Y89), we expressed p27 proteins with one or several tyrosine residues mutated to phenylalanine together with FLT3ITD. Tyrosine phosphorylation of all p27 mutant proteins lacking Y88 was dramatically reduced, and the mutant lacking only Y88 was not significantly phosphorylated on the remaining tyrosines (Figure 2A). This demonstrates that FLT3-ITD results in a highly selective phosphorylation of p27 on residue Y88 in vivo.

Table 1. Characteristics of acute myeloid leukemia patients.

Patient Age Sex FAB n. (years)

WBC Specimen Disease status

Cytogenetics FLT3 mut. NPM1 Other Blasts Blasts (Ratio) mut. mutations PB (%) BM (%)

66.9

De novo

46,XX

Neg

Pos

91.6

De novo

46,XX

Neg

Pos

2.8

De novo

Neg

46.5

49.5

73 72

F M

M1 NA

33 50

De novo Relapse

46,XX,del(7) (q22)/46,XX 46,XX NA

Neg

4 5

BM/ PB PB BM

Neg Neg

NA NA

80 96.5

80 97

160

De novo

46,XX

MLL-PTD

De novo t(6;9) (p23;q34)

ITD (4.32)

t(6;9)

85.5

M4/5

40.8

De novo

46,XY

ITD (0.259)

Neg

80.9

De novo

46,XY

ITD

Pos

6.3

De novo

46,XX

ITD

Pos

13.7

t-AML

ITD

Pos

139

De novo

45~46,XX,+4 [cp15]/46,XX 46,XX

DNMT3A, TET2 NA

ITD

Neg

De novo

46,XY

Pos

139

De novo

46, XX

ITD (0.76) ITD (0.55)

pos

ITD Neg (0.894))

Therapy

CR/ relapse

3+7; CR; 2x relapse; LFU auto HSCT 3+7: refractory; 2x relapse; re-induction; CR deceased 3+7 relapse; deceased primary refractory deceased FLAG, Vidaza relapse; deceased RATIFY study CR w/o Midostaurin; CR; allo HSCT 3+7+Midostaurin; relapse; HiDACI: PR; deceased allo HSCT 3+7: CR; t-MDS; t-AML allo HSCT 2nd allo HSCT; relapse; LFU 3+7; CR allo HSCT 3+7 Relapse; allo HSCT IdaFLAG; CR Best supportive LFU care 3+7: primary allo HSCT; CR refractory salvage HAM; 3+7 CR Midostaurin 7+3: RD; IdaFLAG; CR allo HSCT

n.: number: mut: mutation; PB: peripheral blood; BM: bone marrow; Neg: negative; Pos: positive; F: female; M: male; FAB: French-American-British classification; WBC: white blood cell counts (103 cells per microliter); ITD: internal tandem duplication; NA: not available; t-MDS: therapy-related myelodysplastic syndrome; t-AML: therapy-related AML; LFU: lost to follow up; HSCT: hematopoietic stem cell transplantation; auto HSCT: autologous HSCT; allo HSCT: allogeneic HSCT; CR: complete remission; PR: partial remission; RD: residual disease; 3+7: anthracycline+AraC; HAM: high-dose AraC+Mitoxantron; HiDACI: high-dose AraC+idarubicin; FLAG: fludarabine, AraC; G-CSF: granulocyte-colony stimulating factor; Ida: idarubicin; RATIFY clinical trial: induction (daunorubicin+AraC), consolidation (high-dose AraC) + Midostaurin. All patient samples were subjected to Western blot analyses.

1380

haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

p27 can be a direct substrate of FLT3/FLT3-ITD or its phosphorylation might be indirect, e.g. through Src family kinases (SFKs) activated by FLT3 kinases. To investigate if FLT3-ITD can phosphorylate recombinant purified p27 in vitro, p27 was incubated with the catalytically active cyto-

plasmic domain of FLT3-ITD. While wild-type p27 was efficiently phosphorylated, mutation of Y88 of p27 to phenylalanine strongly decreased the phosphorylation (Figure 2B and Online Supplementary Figure S6). No reduction of phosphorylation was observed when Y89 and/or

Figure 1. FLT3 and FLT3-ITD induce tyrosine phosphorylation of p27. Inhibition of FLT3-ITD with AC220 results in upregulation of p27 and cell cycle arrest in G1 phase of the cell cycle. Incubation of MV4;11 (A-C) and 32D-FLT3-ITD (D-F) cells in presence or absence of 100 nM AC220 for 18 hours (h). For Western blot and FACS analyses, one representative experiment out of three independent experiments is shown. (A and D) Western blot analysis: AC220 treatment increases p27 expression. pY589/Y591-FLT3 monitors kinase inhibition and Î˛-tubulin serves as loading control. (B and E) Cell cycle phase distribution was determined by flow cytometry using propidium iodide (PI) and BrdU staining. (C and F) Cell proliferation is decreased in presence of AC220. Cells were seeded in a 12-well plate and cell numbers of viable (Trypan-blue negative) MV4;11 (C) and 32D-FLT3-ITD (F) cells were determined after 18, 24 and 42 h as indicated. (G) Expression of FLT3 or FLT3-ITD and HA-p27 leads to p27-Y88 phosphorylation. Transfected 293T cell lysates were heated [65Â°C, 10 minutes (min)] to precipitate the majority of proteins. The heat stable p27 and its tyrosine phosphorylation were simultaneously determined by Odyssey infrared imaging. The monoclonal 4G10 mouse antibody was used to detect overall p27 tyrosine-phosphorylation and a specific mouse monoclonal anti-pY88-p27 antibody to detect pY88-p27. Polyclonal rabbit anti-p27 antibodies were used to detect p27. FLT3 and FLT3-ITD were detected in non-heated lysates. (H) FLT3-ITD expression results in increased p27-T187 phosphorylation and decreased p27 levels. 293T cells were transfected with cyclin-E and CDK2 together with HA-p27, HA-p27-Y88F and Flag-FLT3-ITD as indicated. p27 phosphorylated on T187, p27 and FLT3 were detected with specific antibodies by Western blot analyses. ctr: control (vehicle-treated cells).

haematologica | 2017; 102(8)

1381

I. Peschel et al. A

Figure 2. FLT3 phosphorylates p27 selectively on Y88 in vivo and in vitro and Y88-phosphorylation of p27 is Src family tyrosine kinase (SFK) independent. (A) FLT3-ITD phosphorylates p27 on Y88. In a similar approach as described in Figure 1G, HA-p27 and HA-p27 mutants where tyrosines were replaced by phenylalanine were expressed together with FLT3-ITD in 293T cells. p27 was detected simultaneously with overall tyrosine phosphorylation in heat-treated extracts. Co-migration indicates that the tyrosine-phosphorylated protein is p27 (pY-p27). (B) FLT3-ITD phosphorylates p27 on Y88 in vitro. Kinase assays were performed with recombinant His-p27 or His-p27 mutant proteins as indicated. The Flag-tagged cytoplasmic domain of FLT3-ITD was expressed in 293T cells and precipitated with anti-Flag antibodies. p27 protein or mutants were incubated with the sepharose-A bound kinase and Îł-32P-ATP. Incorporation of Îł-32P-ATP in p27 was detected by autoradiography (top panel) following SDS-PAGE and Coomassie brilliant blue staining (bottom panel). (C) FLT3-ITD selectively phosphorylates p27. In vitro kinase assay was performed as described above (B) using the cytoplasmic domain of recombinant FLT3-ITD isolated from insect cells as kinase and recombinant His-p21, His-p27 and His-p57 as substrates. (D) The receptor tyrosine kinase inhibitor SU5614, but not the SFK inhibitor PP2, decreases pY88-p27 after FLT3 or FLT3-ITD expression. 293T cells were treated with SU5614 (1 mM), PP2 (10 mM), or vehicle (DMSO) for 3 hours (h) following transient tansfection of HA-p27 together with FLT3 (top panel), FLT3-ITD (middle panel), or Src (bottom panel). Western blot analysis was performed as described in Figure 1G. (E) p27-Y88 phosphorylation occurs independently of activated SFK members. 32D-FLT3-ITD cells were treated with SU5614 (1 mM), PP2 (10 mM), or vehicle (DMSO) for 2 h. pY88-p27 was immunoprecipitated with pY88-p27-specific antibody and analyzed by Western blot with HRP-coupled anti-p27 antibodies. pY694-Stat5 serves as a control for SU5614 and pY416-SFK for PP2 treatment. Actin was used as loading control. One representative experiment of three independent experiments is shown. ctr: control (vehicle-treated cells).

1382

haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

Figure 3. FLT3 and FLT3-ITD directly bind to p27. (A) FLT3-ITD interacts directly with p27 in vitro. GST and the GST-tagged cytoplasmic domain of FLT3-ITD were isolated from E.coli, bound to glutathione-sepharose, and incubated with recombinant p27. Eluates were analyzed by Western blot using anti-GST and anti-p27 antibodies. (B) HA-p27 co-precipitates with Flag-FLT3-ITD, and vice versa after expression in 293T cells. Protein expression and Y88-phosphorylation of p27 (Input, left panel) as well as p27-co-immunoprecipitates (right panel) and FLT3-co-immunoprecipitates (lower panel) were analyzed by Western blot. (C and D) Endogenous p27 co-precipitates with FLT3-ITD in 32D-FLT3-ITD cells (C) and with endogenous FLT3-ITD in MV4;11 cells (D). Immunoprecipitates were detected with anti-FLT3 antibodies and co-immunoprecipitates with anti-p27 antibodies in Western blots. Normal rabbit IgG was used in control IPs (AB ctr). Coupled pY88-p27 antibodies were loaded to exclude unspecific signals of the antibody (IP ctr.). (E) Schematic representation of FLT3-ITD and its domains used in (A, F and G). (F) Recombinant His-p27 was bound to Affi-Gel10 and incubated with cell lysates from 293T cells transfected with the respective Flag-FLT3 domains. Eluates were analyzed by Western blot using anti-Flag-antibodies. Bovine serum albumin (BSA) coupled to Affi-Gel as well as Affi-Gel alone incubated with Flag-FLT3-ITD served as negative controls. (G) Recombinant full-length His-p27 and its C-terminal (C-term) and N-terminal (N-term) domains were bound to Affi-Gel 10 and incubated with cell lysates from 293T cells transfected with the Flag-FLT3 tyrosine kinase domain (Flag-FLT3-KD). Eluates were analyzed by Western blot using Flag-antibody. BSA coupled to Affi-Gel as well as Affi-Gel alone, incubated with Flag-FLT3-ITD, served as negative controls. One representative experiment of three independent experiments is shown. IP: immunoprecipitation; cytFLT3-ITD: cytoplasmic domain; KD: kinase domain; TKD1: tyrosine kinase domain 1.

haematologica | 2017; 102(8)

1383

I. Peschel et al.

Figure 4. p27 partially co-localizes with FLT3 and FLT3-ITD. Immunofluorescence analysis of U2OS cells transfected with p27 and FlagFLT3-ITD (panels 1-8) or Flag-FLT3 WT (panels 9-16) with a Leica DMi8 inverted widefield microscope. Panels 1-3 and 9-11 show the localization of pY88-p27 (red) and FLT3-ITD or FLT3 WT (green), respectively. Panels 5-7 and 13-15 show the localization of p27 (green) and FLT3-ITD or FLT3 WT (red). As a specificity control for pY88p27 antibodies, U2OS cells were transfected with a phosphodeficient p27 mutant (p27-Y88F) and FLT3-ITD (panels 17-20). DNA was stained with DAPI (blue). Scale bar represents 10 mm.

Y74 were mutated. A mutant preserving Y88 as single tyrosine residue (p27-Y74,89F) was as efficiently phosphorylated as wild-type p27 (Figure 2B and Online Supplementary Figure S6), supporting the hypothesis that Y88 is selectively and directly phosphorylated by FLT3 in vitro. When p27 phosphorylation by recombinant FLT3ITD was compared to p21 and p57, an efficient phosphorylation was only observed for p27 (Figure 2C), suggesting that p27 is the principal FLT3 target of the CIP/KIP family members. FLT3/FLT3-ITD activates SFK members such as Src, Lyn and Fyn,26-28 and Src and Lyn can phosphorylate p27.7,8 Phosphorylation of p27 by Src, however, also includes Y74,8 but phosphorylation at this tyrosine was not detected in transfection assays (Figure 2A, p27-Y88,89F), suggesting that Src is not the main tyrosine kinase that contributes to FLT3-ITD-induced p27 phosphorylation. To assess the contribution of SFKs to p27 phosphorylation in cell lines, we used SU5614 to inactivate FLT3,29 and PP2 as SFK inhibitor. In 293T cells over-expressing p27 and FLT3 or FLT3-ITD, p27-Y88 phosphorylation was inhibited by SU5614, and remained unchanged by PP2 treatment (Figure 2D, top and middle panels), indicating that SFK activity is not required for FLT3-induced p27 phosphorylation. As expected, PP2 treatment abolished p27-Y88 phosphorylation in cells transfected with p27 and Src (Figure 2D, lower panel). 1384

Importantly, Y88 phosphorylation of endogenous p27 in 32D-FLT3-ITD cells was also strongly reduced by SU5614, whereas treatment with the SFK inhibitor PP2 did not inhibit p27 phosphorylation (Figure 2E). Inhibition of FLT3ITD by SU5614 as well as inhibition of SFK by PP2 was confirmed by the phosphorylation status of Y694-Stat5 and Y416-SFK, respectively, demonstrating that SFKs remain active in SU5614-treated cells (Figure 2E). These observations are consistent with the hypothesis that p27 is a direct substrate of FLT3 and FLT3-ITD.

p27 is in a complex with FLT3 and FLT3-ITD If p27 is a substrate of FLT3/FLT3-ITD, both proteins might associate in a stable complex. To investigate if p27 can bind to FLT3/FLT3-ITD, we initially performed pull down experiments with the GST-tagged cytoplasmic domain of FLT3-ITD (cytFLT3-ITD) and His-tagged p27. To exclude the presence of potential bridging factors, the proteins were expressed in E.coli. GST-cytFLT3-ITD, but not GST, co-precipitated p27 (Figure 3A), demonstrating a stable and direct interaction in vitro. The interaction of p27 with FLT3-ITD was confirmed in 293T cells in which HAp27 was co-immunoprecipitated with Flag-FLT3-ITD, and vice versa (Figure 3B). HA-p27 co-precipitated also with Flag-FLT3 (Online Supplementary Figure S7). Importantly, a stable complex of endogenous p27 and endogenous FLT3haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

Figure 5. FLT3 ligand induces p27-Y88 phosphorylation in U937 cells. pY88-p27 is decreased by FLT3 inhibitors in FLT3-ITD positive acute myeloid leukemia (AML) cell lines. (A) FLT3 ligand (FL) induces p27-Y88-phosphorylation in U937 cells. U937 were serum-starved for 6 hours (H) (starved) before stimulation with 50 ng/ml rhFL for the indicated time points. pY88-p27 was immunoprecipitated with specific antibodies and detected in Western blots using HRP-coupled anti-p27 antibodies. Further, anti-FLT3, p27 and GAPDH antibodies were used. Mouse IgG served as isoptype control for immunoprecipitation (AB ctr.). (B and C) pY88-p27 is decreased and p27 levels are increased by FLT3 inhibitors. Western blot analyses are shown. (B) MV4;11 cells were incubated with SU5614 (1 ÂľM) or vehicle for 1 h. (C) Ba/F3 cells were incubated with AC220 (50 nM) or vehicle for 1 h. p27, FLT3 and pY88-p27 were detected as described above. pY589/Y591-FLT3 and pY694-Stat5 served as controls for inhibitor treatment and actin as loading control. Skp2 immunoprecipitates revealed no change in Skp2 expression. Mouse IgG was used as isoptype control for immunoprecipitation (AB ctr.). (D) p27 is phosphorylated on Y88 in FLT3-D835Y cells. Total p27 correlates inversely with pY88-p27 in FLT3-D835Y and FLT3-ITD cells. Analyses were performed as described in (B and C). One representative experiment of three independent experiments is shown.

ITD was precipitated from MV4;11 cells (Figure 3D). Similarly, endogenous p27 co-precipitated with FLT3-ITD in 32D-FLT3-ITD cells (Figure 3C). These data demonstrate that p27 can form a stable complex with FLT3 and FLT3-ITD, and support the hypothesis that p27 is a direct substrate of these kinases. To identify essential domains for the p27-FLT3 interaction, we performed pull down experiments using fragments of both proteins. FLT3 can be subdivided in an extracellular domain, a transmembrane portion, and a cytoplasmic region containing the split kinase domain30 (Figure 3E). Cell lysates expressing different regions of the cytoplasmic domain were incubated with recombinant His-tagged p27. p27 bound to the FLT3 kinase domain (KD) and to its N-terminal lobe (TKD1) (Figure 3F). Coimmunoprecipitation experiments with transfected proteins confirmed binding of the kinase domain and the TKD1 domain to p27 (Online Supplementary Figure S8). On the other hand, full-length p27 and its N-terminal kinaseinhibitory domain precipitated FLT3-KD efficiently (Figure 3G). Interestingly, this mapping analysis indicates that the kinase domain of FLT3 interacts with the CDK-inhibitory domain of p27. In order to be a significant substrate of FLT3/FLT3-ITD, p27 and the kinase should at least partially co-localize. FLT3 localizes to the plasma membrane and can be restrained in ER membranes upon incomplete glycosylahaematologica | 2017; 102(8)

tion.25 p27 is predominantly a nuclear protein that shuttles between nucleus and cytoplasm.31 To interact with the cytoplasmic domain of the FLT3 kinases, p27 needs to be, at least partly, exported to the cytoplasm. To determine the subcellular localization of both proteins, U2OS cells were transfected with p27 and FLT3-ITD (Figure 4, panels 1-8) or p27 and FLT3 (Figure 4, panels 9-16). Immunoflourescence analyses revealed that FLT3-ITD and FLT3 are mainly localized in extended structures surrounding the nucleus, potentially representing ER membranes (Figure 4, panels 2,6 and 10,14, respectively). With p27 accumulated in the nucleus, however, a portion of the protein was detected in the cytoplasm (Figure 4, panels 5,13). This extranuclear fraction of p27 partially co-localized with FLT3-ITD/FLT3 (Figure 4, panels 7,15). Interestingly, when Y88-phosphorylated p27 was analyzed, the degree of co-localization with FLT3-ITD and FLT3 was significantly enhanced (Figure 4, panels 3,11). These observations support the hypothesis that p27 interacts with FLT3ITD and FLT3.

Stimulation of FLT3 with its ligand FL induces p27 phosphorylation. Inhibition of FLT3-ITD in AML cell lines reduces p27-Y88 phosphorylation and restores p27 protein levels. To investigate the potential physiological role of p27Y88 phosphorylation by FLT3, we investigated if endoge1385

I. Peschel et al.

nous p27 tyrosine phosphorylation follows FL-induced activation of endogenous FLT3 in U937 cells. pY88-p27 was low in asynchronously proliferating cells and remained low upon serum starvation (Figure 5A). FLT3 autophosphorylation is known to occur 5-15 min after FL

addition.32 Interestingly, 5 min after activation by FL p27Y88 phosphorylation was already strongly induced (Figure 5A), suggesting that p27 phosphorylation is an immediate event following FLT3 activation. p27-Y88 phosphorylation slowly declined within 90 min (Figure 5A). FLT3 protein

1386

Figure 6. pY88-p27 is detected in primary FLT3 wild-type (WT) acute myeloid leukemia (AML) cells and reduced upon FLT3 inhibition. (A) Western blot of pY88-p27 immunoprecipitates, p27 and actin in FLT3 primary AML cells. FLT3 WT expressing primary cells were cultured overnight in RPMI medium containing 20% FBS and equal amounts of protein extracts were analyzed by immunoblotting. Cells from patient #B were stimulated with FL (100 ng/mL) for 20 minutes (min). (B-D) AC220 treatment of FLT3 expressing primary AML cells down-regulates pY88-p27. Primary AML cells were cultured overnight (frozen samples) or for 2 hours (h) (fresh samples) in medium containing 20% FBS before incubation with AC220 (100 nM), dasatinib (100 nM) or vehicle for 2 h. Equal total protein amounts as determined by DC protein quantification assay were used for pY88-p27 immunoprecipitation with specific antibodies and subjected to Western blot analysis using HRP-coupled anti-p27 antibodies. (B) Immunoblot analysis of one representative patient (#1) expressing FLT3 WT. p27, FLT3, and the loading control actin were detected. pY589/Y591FLT3, pY694-Stat5, and pT202/Y204-Erk served as controls for FLT3 inhibition. (C) Quantitative analysis (top panel) of Western blots (bottom panel) determining pY88-p27 levels in all 5 FLT3 WT AML patient cells that were treated with or without AC220. Sample #1 is also shown in Figure 6B. Western blots were quantified by densitometry using ImageJ software and normalized for actin. (D) Mean decrease of pY88-p27 levels of 5 FLT3 WT AML patients upon AC220 treatment (2 h). Wilcoxon test, *P=0.043.

haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

started to decline 60 min after activation (Figure 5A), probably following internalization.33 In asynchronous FLT3-ITD-positive MV4;11 cells, p27 was strongly phosphorylated on Y88 (Figure 5B). The FLT3 inhibitor SU5614 abolished p27-Y88 phosphorylation to undetectable levels, suggesting that this phosphorylation depends on FLT3-ITD kinase activity (Figure 5B). The efficiency of SU5614 was monitored by reduction of pY589/Y591-FLT3 and pY694-Stat5 (Figure 5B). Similar results were observed in Ba/F3-FLT3-ITD cells treated with AC220 (Figure 5C). Y88 phosphorylation of p27 can initiate its SCFSkp2-mediated proteasomal degradation.7,8,13 Consistent with our earlier findings,7,13 long-term inhibition of FLT3-ITD led to a robust accumulation of p27 (Figure 1A and F) which could be the cause or consequence of the cell cycle arrest. To exclude cell cycle positioning effects, we analyzed p27 after very short periods of inhibition. In MV4;11 or Ba/F3-FLT3-ITD cells, an initial increase in p27 protein could already be detected 60 min after FLT3-ITD inhibition (Figure 5B and C). When p27 phosphorylation was compared in 32D cells stably expressing either FLT3-ITD or FLT3-TKD (D835Y), higher p27-Y88 phosphorylation and lower p27 levels were found in FLT3-D835Y expressing cells (Figure 5D). These observations are consistent with the previous finding that pY88-p27 can initiate its SCFSkp2-triggered degradation, a mechanism that leads to CDK activation and cell proliferation.7,8,13

p27 is phosphorylated on Y88 in primary AML cells The p27-Y88 phosphorylation and its physiological consequences have been studied in vitro and in cell lines.711,13,34,35 However, the presence, abundance or regulation of this modification have not yet been established in primary patient material. We went on to investigate the detectability of pY88-p27 in primary blast cells obtained from AML patients. Immunoprecipitation analysis of cell extracts from AML patients positive for FLT3 wild type (WT) revealed the presence of pY88-p27 and p27 at levels comparable to those in FLT3-ITD positive MV4;11 cells (Figure 6A and Online Supplementary Figure S9). To gain more insight into the p27-Y88 phosphorylation in primary AML patient samples, we investigated the response to FLT3 inhibition by AC220. (We did not test the effect of other FLT3 inhibitors.) Primary blast cells from 14 AML patients, including 5 FLT3 WT and 9 FLT3ITD positive patients, were analyzed (Table 1). All samples included in this analysis were positive for FLT3 or FLT3-ITD expression and FLT3 autophosphorylation, and responded to AC220 as monitored by reduced pY589/Y591-FLT3, pY694-Stat5, or pT202/Y204-Erk (Figures 6B, 7A and D, Online Supplementary Figures S10 and S11, and data not shown). Inhibitor treatment of FLT3 WT positive AML cells for 2 h resulted in a significant reduction of p27-Y88 phosphorylation to an average 59% (Figure 6C and D). All individual samples of this group responded to AC220 with a decline in p27-Y88 phosphorylation (Figure 6C). In FLT3-ITD-expressing AML cells, AC220 treatment decreased pY88-p27 levels in 5 out of 9 patient samples (Figure 7B). Surprisingly, in 4 patients, pY88-p27 levels were even increased in the presence of FLT3 inhibitor (Figure 7C), although FLT3-ITD kinase activity was inhibited, as monitored by marker protein phosphorylation (reduction of pY589/Y591-FLT3, pY694Stat5, or pT202/Y204-Erk) (Figure 7D, Online haematologica | 2017; 102(8)

Supplementary Figure S11, and data not shown). The impaired ability of AC220 to prevent p27-Y88 phosphorylation in FLT3-ITD expressing cells could be attributed to other activated tyrosine kinases present in these cells. Members of the SFK family can be activated in AML cells.26-28,36,37 Therefore, we investigated the activity of SFKs in the patient with the strongest increase of pY88-p27 upon AC220 incubation (#13, 63% increase). In this case, the SFK autophosphorylation did not decrease upon AC220 incubation (Figure 7D). SFKs can also phosphorylate p27 on Y88.7,8 To investigate if increased SFK activity might contribute to p27-Y88 phosphorylation, we treated these cells with AC220, dasatinib or both inhibitors (Figure 7D). pY88-p27 was increased 1.7-fold by AC220 alone, despite efficient inhibition of FLT3 autophosphorylation (Figure 7D). Interestingly, dasatinib treatment alone reduced pY88-p27 to 38% of vehicle-treated cells and the combination of dasatinib and AC220 further reduced pY88-p27 to 23%. Phosphorylation of marker proteins Stat5, ERK and SFK was also efficiently reduced following dasatinib treatment (Figure 7D), suggesting that, in this patient, increased activity of SFKs (or other desatinib-sensitive kinases) significantly contribute to p27-Y88 phosphorylation. Subsequent analyses of patient samples #11 and #14, where AC220 treatment also increased pY88-p27 levels, revealed that dasatinib treatment decreased p27Y88 phosphorylation status of patient sample #14 but had no effect on sample #11 (Online Supplementary Figure S11). These data demonstrate that pY88-p27 can be detected in primary human patient material and that p27 is targeted by FLT3 and FLT3-ITD in primary AML cells. Some AML patients express additional p27-phosphorylating tyrosine kinases that can be inhibited by dasatinib or combinations of tyrosine kinase inhibitors.

Discussion p27 is a substrate of several non-receptor tyrosine kinases.7-11,13 Here we report that the receptor tyrosine kinase FLT3 and its constitutively active counterpart FLT3-ITD can bind and selectively phosphorylate p27on Y88. p27 can shuttle between the nucleus and cytoplasm.31 Increased cytoplasmic localization of p27 results from mitogen stimulation and involves Akt dependent phosphorylation events.2,38 Since FLT3/FLT3-ITD signaling activates the PI3K/Akt pathway, it is likely that enhanced Akt activity reinforces the nuclear export of p27. Binding of p27 to FLT3 involves the N-terminal domain of p27 and TKD1 of the tyrosine kinase domain of FLT3. The N-terminal domain of p27 uses a folding-on-binding mechanism for cyclin/CDK binding, and inserts a 310-helix surrounding Y88 into the catalytic cleft of the kinase.6,39 Since the kinase domain and especially TKD1 of FLT3 bind to the CDK-inhibitory domain of p27, it will be interesting to determine if this binding follows a similar folding-on-binding mechanism of p27 on FLT3. This may explain the enhanced binding by the entire kinase domain and might also position Y88 for selective phosphorylation. The precise mechanism and dynamics of p27 binding to the FLT3 tyrosine kinase domain still need to be determined. Whereas p27 levels were low in cells expressing active FLT3-ITD, inhibition of FLT3-ITD led to increased p27 levels and cell cycle arrest in G0/G1-phase. To avoid cell cycle 1387

I. Peschel et al.

position effects, we analyzed p27 after very short inhibition of FLT3. Already one hour following FLT3 inhibition, we observed a reproducible increase in p27. Degradation induced by Y88 phosphorylation requires CDK-dependent phosphorylation of the inhibitor and expression of the cell cycle-regulated ubiquitin E3 ligase SCFSkp2.5,7 Since proliferating cells express the SCFSkp2 E3 ligase, it is likely that FLT3-ITD-mediated phosphorylation of p27 con-

tributes to the reduction of p27 levels in MV4;11 and proliferating AML cells. SCFSkp2-dependent proteasomal degradation of T187-phosphorylated p27 may occur in the cytoplasm or in the nucleus. Since Skp2 translocates to the cytoplasm upon Akt phosphorylation, it may also support cytoplasmic p27 ubiquitination. On the other hand, p27 was phosphorylated on Y88 but did not decline upon FL stimulation of serum starved U937 cells. Lack of p27

Figure 7. AC220 regulates pY88-p27 in primary FLT3-ITD acute myeloid leukemia (AML) cells. Experiments with FLT3-ITD positive AML cells were performed as described in Figure 6. (A) Western blot analysis of a representative FLT3-ITD patient (#6) of the group of patients, where pY88-p27 decreased upon AC220 treatment. pY589/Y591-FLT3 and pY694-Stat5, which is a direct substrate of FLT3-ITD, serve as controls for FLT3 inhibition. p27, FLT3 and Î˛-tubulin were monitored as loading control. (B) Quantitative analysis (top panel) of Western blots (bottom panel) determining pY88-p27 levels in 5 FLT3-ITD AML patients where AC220 treatment for 2 hours (h) decreased pY88-p27 levels. (C) Western blots (bottom panel) and its quantification (top panel) of pY88-p27 levels from 4 FLT3-ITD patients where AC220 treatment increased pY88-p27 levels. (D) Dasatinib treatment reduces pY88-p27 in an FLT3-ITD positive patient sample where AC220 increased pY88-p27. FLT3ITD AML cells (#13) were treated for 2 h with AC220, dasatinib, the combination of both inhibitors, or vehicle. p27, FLT3, and the loading control Î˛-tubulin were detected. pY589/Y591-FLT3 (pY-FLT3), pY694-Stat5 (pY-Stat5), and pT202/Y204-Erk (pY-Erk) serve as controls for FLT3 inhibition. pY416-SFK (pY-SFK) is a control for dasatinib-mediated Src inhibition.

1388

haematologica | 2017; 102(8)

Phosphorylation of p27 by FLT3

degradation is likely caused by low Skp2 expression in starved cells.40 Interestingly, decreased p27 levels have been associated with poor prognostic outcome in AML patients.41-43 In contrast to these observations, Haferlach et al. reported that low CDKN1B levels correlates with a better prognosis in AML patients.44 However, Haferlach and co-workers analyzed only CDKN1B mRNA expression and p27 mRNA frequently does not correlate with its protein expression.8 In addition, Y88 phosphorylation and cytoplasmic localization can inactivate the CDK inhibitor function and convert the p27 tumor suppressor into a cytoplasmic oncogene.45 The molecular consequences of p27-Y88 phosphorylation have been extensively described.7-11,13,34,35 In contrast to data in cell culture systems, confirmation of the presence of pY88-p27 in primary patient material has been lacking. We found that pY88-p27 could readily be detected in primary AML blast cells from patients expressing either FLT3 WT or FLT3-ITD. Levels of pY88-p27 are comparable to levels detected in MV4;11 cells (Figure 6A and Online Supplementary Figures S9 and S10). This is the first time that pY88-p27 has been detected in primary patient material. FLT3 inhibition with AC220 significantly reduced pY88p27 in all FLT3 WT patient samples. These data indicate that p27 is a substrate of FLT3 in primary AML cells. In contrast to experiments with leukemic cell lines, p27 protein levels were not increased following AC220 treatment. This is most likely caused by the cell culture conditions, which do not support proliferation of the primary cells.42 This would prevent Skp2 expression, and the E3 ligase is required for p27 degradation following p27 tyrosine phosphorylation. The reduction of pY88-p27 by AC220 in FLT3 WT AML cells reflects an increased CDK inhibitory activity of p27 and possibly increased p27 levels in proliferating cells. These findings support the suggestion that FLT3 WT AML patients may benefit from FLT3 inhibitor treatment.46 Of note, a phase II study is currently investigating the effect of AC220 monotherapy in FLT3 WT AML patients (clinicaltrials.gov identifier: 00989261). Interestingly, we observed that pY88-p27 was also decreased by dasatinib treatment in the single FLT3 WT AML patient investigated (Figure 6B), indicating that FLT3 WT patients might also benefit from multikinase inhibitor therapy. AC220 is under investigation in a phase III trial for patients expressing FLT3-ITD (clinicaltrials.gov identifier:

References 1. Malumbres M, Barbacid M. Cell cycle, CDKs and cancer: a changing paradigm. Nat Rev Cancer. 2009;9(3):153-166. 2. Chu IM, Hengst L, Slingerland JM. The Cdk inhibitor p27 in human cancer: prognostic potential and relevance to anticancer therapy. Nat Rev Cancer. 2008;8(4):253-267. 3. Sharma SS, Pledger WJ. The non-canonical functions of p27(Kip1) in normal and tumor biology. Cell Cycle. 2016;15(9):1189-1201. 4. Hengst L, Reed SI. Translational control of p27Kip1 accumulation during the cell cycle. Science. 1996;271(5257):1861-1864. 5. Carrano AC, Eytan E, Hershko A, Pagano M. SKP2 is required for ubiquitin-mediated

haematologica | 2017; 102(8)

8. 9.

02039726). Approximately 30% of FLT3-ITD patients do not respond to FLT3 inhibitor treatment.47 Surprisingly, we observed both up- and downregulation of pY88-p27 following AC220 treatment of individual FLT3-ITD patient samples. The increase of pY88-p27 upon AC220 treatment in some patient samples raised the question as to whether other tyrosine kinases were responsible for p27-Y88 phosphorylation in these patients. We found that SFKs might contribute to this phosphorylation since dasatinib treatment efficiently reduced pY88-p27 in 2 out of 3 FLT3-ITD positive patients with increased pY88-p27 levels after AC220 treatment (Figure 7D and Online Supplementary Figure S11, patient #14). Dasatinib treatment further reduced pY88-p27 in one out of 2 FLT3-ITD patients where AC220 already reduced pY88-p27 levels (Online Supplementary Figure S10, patient #10). SFKs can be activated in primary AML cells36,37 and Src, Yes and Lyn can phosphorylate p27.7,8 Of note, dasatinib can also inhibit other tyrosine kinases including Abl, which might cause p27 phosphorylation. It will be important to explore if p27Y88 phosphorylation can serve as a therapeutic marker in AML and act as an indicator for treatment decisions involving tyrosine kinase inhibitors. Similarly, it should be determined if inhibition of p27-Y88 phosphorylation correlates with therapeutic outcome in AML patients undergoing therapy with tyrosine kinase inhibitors. Interestingly, small molecule CDK inhibitors are currently being investigated for use in AML therapy48 (clinicaltrials.gov identifiers: 02310243 and 00278330), and recent investigations indicate a synergistic effect of CDK4/6 kinase inhibitor and FLT3 inhibitors to impair the survival of leukemic FLT3-ITD positive cells.49 Preventing p27-Y88 phosphorylation by combinations of tyrosine kinase inhibitors might be an alternative approach to restore CDK inactivation in AML. Acknowledgments The authors would like to thank Caroline Linhart for performing statistical analyses of the ethics proposal, Jonathan Vosper for proofreading and all members of the Hengst lab for support and stimulating discussions. Funding This work has been supported by the Austrian Science Fund (FWF) Grant P24031 (LH) and the Ă&#x2013;sterreichische Krebshilfe Tirol (Ă&#x2013;KH-KG) (IP).

degradation of the CDK inhibitor p27. Nat Cell Biol. 1999;1(4):193-199. Russo AA, Jeffrey PD, Patten AK, Massague J, Pavletich NP. Crystal structure of the p27Kip1 cyclin-dependent-kinase inhibitor bound to the cyclin A-Cdk2 complex. Nature. 1996;382(6589):325-331. Grimmler M, Wang Y, Mund T, et al. Cdkinhibitory activity and stability of p27Kip1 are directly regulated by oncogenic tyrosine kinases. Cell. 2007;128(2):269-280. Chu I, Sun J, Arnaout A, et al. p27 phosphorylation by Src regulates inhibition of cyclin E-Cdk2. Cell. 2007;128(2):281-294. James MK, Ray A, Leznova D, Blain SW. Differential modification of p27Kip1 controls its cyclin D-cdk4 inhibitory activity.

Mol Cell Biol. 2008;28(1):498-510. 10. Larrea MD, Liang J, Da Silva T, et al. Phosphorylation of p27Kip1 regulates assembly and activation of cyclin D1-Cdk4. Mol Cell Biol. 2008;28(20):6462-6472. 11. Patel P, Asbach B, Shteyn E, et al. Brk/Protein tyrosine kinase 6 phosphorylates p27KIP1, regulating the activity of cyclin D-cyclin-dependent kinase 4. Mol Cell Biol. 2015;35(9):1506-1522. 12. Jakel H, Peschel I, Kunze C, Weinl C, Hengst L. Regulation of p27 (Kip1) by mitogen-induced tyrosine phosphorylation. Cell Cycle. 2012;11(10):1910-1917. 13. Jakel H, Weinl C, Hengst L. Phosphorylation of p27Kip1 by JAK2 directly links cytokine receptor signaling to cell cycle control.

1389

I. Peschel et al. Oncogene. 30(32):3502-3512. 14. Gilliland DG, Griffin JD. The roles of FLT3 in hematopoiesis and leukemia. Blood. 2002;100(5):1532-1542. 15. Birg F, Courcoul M, Rosnet O, et al. Expression of the FMS/KIT-like gene FLT3 in human acute leukemias of the myeloid and lymphoid lineages. Blood. 1992; 80(10):2584-2593. 16. Carow CE, Levenstein M, Kaufmann SH, et al. Expression of the hematopoietic growth factor receptor FLT3 (STK-1/Flk2) in human leukemias. Blood. 1996; 87(3):1089-1096. 17. Meshinchi S, Appelbaum FR. Structural and functional alterations of FLT3 in acute myeloid leukemia. Clin Cancer Res. 2009; 15(13):4263-4269. 18. Schnittger S, Schoch C, Dugas M, et al. Analysis of FLT3 length mutations in 1003 patients with acute myeloid leukemia: correlation to cytogenetics, FAB subtype, and prognosis in the AMLCG study and usefulness as a marker for the detection of minimal residual disease. Blood. 2002;100(1):59-66. 19. Thiede C, Steudel C, Mohr B, et al. Analysis of FLT3-activating mutations in 979 patients with acute myelogenous leukemia: association with FAB subtypes and identification of subgroups with poor prognosis. Blood. 2002;99(12):4326-4335. 20. Takahashi S. Downstream molecular pathways of FLT3 in the pathogenesis of acute myeloid leukemia: biology and therapeutic implications. J Hematol Oncol. 2011;4:13. 21. Graham FL, van der Eb AJ. A new technique for the assay of infectivity of human adenovirus 5 DNA. Virology. 1973; 52(2):456-467. 22. Vosper J, Masuccio A, Kullmann M, Ploner C, Geley S, Hengst L. Statin-induced depletion of geranylgeranyl pyrophosphate inhibits cell proliferation by a novel pathway of Skp2 degradation. Oncotarget. 2015;6(5):2889-2902. 23. Quentmeier H, Reinhardt J, Zaborski M, Drexler HG. FLT3 mutations in acute myeloid leukemia cell lines. Leukemia. 2003;17(1):120-124. 24. Zarrinkar PP, Gunawardane RN, Cramer MD, et al. AC220 is a uniquely potent and selective inhibitor of FLT3 for the treatment of acute myeloid leukemia (AML). Blood. 2009;114(14):2984-2992. 25. Schmidt-Arras D, Bohmer SA, Koch S, et al. Anchoring of FLT3 in the endoplasmic reticulum alters signaling quality. Blood. 2009;113(15):3568-3576. 26. Chougule RA, Kazi JU, Ronnstrand L. FYN

1390

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

expression potentiates FLT3-ITD induced STAT5 signaling in acute myeloid leukemia. Oncotarget. 2016;7(9):99649974. Leischner H, Albers C, Grundler R, et al. SRC is a signaling mediator in FLT3-ITDbut not in FLT3-TKD-positive AML. Blood. 119(17):4026-4033. Robinson LJ, Xue J, Corey SJ. Src family tyrosine kinases are activated by Flt3 and are involved in the proliferative effects of leukemia-associated Flt3 mutations. Exp Hematol. 2005;33(4):469-479. Spiekermann K, Dirschinger RJ, Schwab R, et al. The protein tyrosine kinase inhibitor SU5614 inhibits FLT3 and induces growth arrest and apoptosis in AML-derived cell lines expressing a constitutively activated FLT3. Blood. 2003;101(4):1494-1504. Rosnet O, Marchetto S, deLapeyriere O, Birnbaum D. Murine Flt3, a gene encoding a novel tyrosine kinase receptor of the PDGFR/CSF1R family. Oncogene. 1991;6(9):1641-1650. Connor MK, Kotchetkov R, Cariou S, et al. CRM1/Ran-mediated nuclear export of p27(Kip1) involves a nuclear export signal and links p27 export and proteolysis. Mol Biol Cell. 2003;14(1):201-213. Grafone T, Palmisano M, Nicci C, Storti S. An overview on the role of FLT3-tyrosine kinase receptor in acute myeloid leukemia: biology and treatment. Oncol Rev. 2012;6(1):e8. Turner AM, Lin NL, Issarachai S, Lyman SD, Broudy VC. FLT3 receptor expression on the surface of normal and malignant human hematopoietic cells. Blood. 1996; 88(9):3383-3390. Kardinal C, Dangers M, Kardinal A, et al. Tyrosine phosphorylation modulates binding preference to cyclin-dependent kinases and subcellular localization of p27Kip1 in the acute promyelocytic leukemia cell line NB4. Blood. 2006;107(3):1133-1140. Ou L, Ferreira AM, Otieno S, Xiao L, Bashford D, Kriwacki RW. Incomplete folding upon binding mediates Cdk4/cyclin D complex activation by tyrosine phosphorylation of inhibitor p27 protein. J Biol Chem. 2011;286(34):30142-30151. Dos Santos C, Demur C, Bardet V, PradeHoudellier N, Payrastre B, Recher C. A critical role for Lyn in acute myeloid leukemia. Blood. 2008;111(4):2269-2279. Ozawa Y, Williams AH, Estes ML, et al. Src family kinases promote AML cell survival through activation of signal transducers and

38.

39.

40.

41.

42.

43.

44.

45. 46.

47.

48. 49.

activators of transcription (STAT). Leuk Res. 2008;32(6):893-903. Shin I, Yakes FM, Rojo F, et al. PKB/Akt mediates cell-cycle progression by phosphorylation of p27(Kip1) at threonine 157 and modulation of its cellular localization. Nat Med. 2002;8(10):1145-1152. Lacy ER, Filippov I, Lewis WS, et al. p27 binds cyclin-CDK complexes through a sequential mechanism involving bindinginduced protein folding. Nat Struct Mol Biol. 2004;11(4):358-364. Wei W, Ayad NG, Wan Y, Zhang GJ, Kirschner MW, Kaelin WG Jr. Degradation of the SCF component Skp2 in cell-cycle phase G1 by the anaphase-promoting complex. Nature. 2004;428(6979):194-198. Abdel Alim A, Youssef SR, Farouk HM. The prognostic potential of bone marrow cyclin E and P27, in Egyptian patients with acute myeloid leukemia. Egypt J Immunol. 2010;17(1):9-18. Radosevic N, Delmer A, Tang R, Marie JP, Ajchenbaum-Cymbalista F. Cell cycle regulatory protein expression in fresh acute myeloid leukemia cells and after drug exposure. Leukemia. 2001;15(4):559-566. Yokozawa T, Towatari M, Iida H, et al. Prognostic significance of the cell cycle inhibitor p27Kip1 in acute myeloid leukemia. Leukemia. 2000;14(1):28-33. Haferlach C, Bacher U, Kohlmann A, et al. CDKN1B, encoding the cyclin-dependent kinase inhibitor 1B (p27), is located in the minimally deleted region of 12p abnormalities in myeloid malignancies and its low expression is a favorable prognostic marker in acute myeloid leukemia. Haematologica. 2011;96(6):829-836. Besson A, Dowdy SF, Roberts JM. CDK inhibitors: cell cycle regulators and beyond. Dev Cell. 2008;14(2):159-169. Ozeki K, Kiyoi H, Hirose Y, et al. Biologic and clinical significance of the FLT3 transcript level in acute myeloid leukemia. Blood. 2004;103(5):1901-1908. Kindler T, Lipka DB, Fischer T. FLT3 as a therapeutic target in AML: still challenging after all these years. Blood. 2010;116(24): 5089-5102. Bose P, Grant S. Rational Combinations of Targeted Agents in AML. J Clin Med. 2015;4(4):634-664. Uras IZ, Walter GJ, Scheicher R, et al. Palbociclib treatment of FLT3-ITD+ AML cells uncovers a kinase-dependent transcriptional regulation of FLT3 and PIM1 by CDK6. Blood. 2016;127(23):2890-2902.

haematologica | 2017; 102(8)

ARTICLE

Acute Myeloid Leukemia

Prognostic and biologic significance of long non-coding RNA profiling in younger adults with cytogenetically normal acute myeloid leukemia

Dimitrios Papaioannou,1 Deedra Nicolet,1,2 Stefano Volinia,3 Krzysztof MrĂłzek,1 Pearlly Yan,1 Ralf Bundschuh,4 Andrew J. Carroll,5 Jessica Kohlschmidt,1,2 William Blum,1 Bayard L. Powell,6 Geoffrey L. Uy,7 Jonathan E. Kolitz,8 Eunice S. Wang,9 Ann-Kathrin Eisfeld,1 Shelley J. Orwick,1 David M. Lucas,1 Michael A. Caligiuri,1 Richard M. Stone,10 John C. Byrd,1 Ramiro Garzon1* and Clara D. Bloomfield1* *RG and CDB contributed equally to this work.

1 The Ohio State University Comprehensive Cancer Center, Columbus, OH, USA; 2The Alliance for Clinical Trials in Oncology Statistics and Data Center, Mayo Clinic, Rochester, MN, USA; 3Department of Morphology, Surgery and Experimental Medicine, University of Ferrara, Italy; 4Department of Physics, Department of Chemistry & Biochemistry, Division of Hematology, Department of Internal Medicine, Center for RNA Biology, The Ohio State University, Columbus, OH, USA; 5Department of Genetics, University of Alabama at Birmingham, AL, USA; 6The Comprehensive Cancer Center of Wake Forest University, Winston-Salem, NC, USA; 7Siteman Cancer Center, Washington University School of Medicine, St. Louis, MO, USA; 8Hofstra North Shore-Long Island Jewish School of Medicine, Lake Success, NY, USA; 9Roswell Park Cancer Institute, Buffalo, NY, USA and 10Dana-Farber Cancer Institute, Harvard University, Boston, MA, USA

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1391-1400

ABSTRACT

ong non-coding ribonucleic acids (RNAs) are a novel class of RNA molecules, which are increasingly recognized as important molecular players in solid and hematologic malignancies. Herein we investigated whether long non-coding RNA expression is associated with clinical and molecular features, as well as outcome of younger adults (aged <60 years) with de novo cytogenetically normal acute myeloid leukemia. Whole transcriptome profiling was performed in a training (n=263) and a validation set (n=114). Using the training set, we identified 24 long non-coding RNAs associated with event-free survival. Linear combination of the weighted expression values of these transcripts yielded a prognostic score. In the validation set, patients with high scores had shorter disease-free (P<0.001), overall (P=0.002) and event-free survival (P<0.001) than patients with low scores. In multivariable analyses, long non-coding RNA score status was an independent prognostic marker for disease-free (P=0.01) and event-free survival (P=0.002), and showed a trend for overall survival (P=0.06). Among multiple molecular alterations tested, which are prognostic in cytogenetically normal acute myeloid leukemia, only double CEBPA mutations, NPM1 mutations and FLT3-ITD associated with distinct long non-coding RNA signatures. Correlation of the long non-coding RNA scores with messenger RNA and microRNA expression identified enrichment of genes involved in lymphocyte/leukocyte activation, inflammation and apoptosis in patients with high scores. We conclude that long noncoding RNA profiling provides meaningful prognostic information in younger adults with cytogenetically normal acute myeloid leukemia. In addition, expression of prognostic long non-coding RNAs associates with oncogenic molecular pathways in this disease. clinicaltrials.gov Identifier: 00048958 (CALGB-8461), 00899223 (CALGB-9665), and 00900224 (CALGB-20202). Introduction

Correspondence: ramiro.garzon@osumc.edu or clara.bloomfield@osumc.edu Received: February 7, 2017. Accepted: May 2, 2017. Pre-published: May 4, 2017. doi:10.3324/haematol.2017.166215 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1391 ÂŠ2017 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.

Acute myeloid leukemia (AML) is a highly heterogeneous disease with regard to genetic abnormalities and clinical course.1 The prognosis of adult AML is generally haematologica | 2017; 102(8)

1391

D. Papaioannou et al.

poor. Only 40% of younger adult (aged <60 years) and 10% of older (aged ≥60 years) AML patients achieve longterm survival.1 Currently, chromosomal aberrations2-4 and recurrent gene mutations5-8 are considered the most reliable and reproducible prognostic markers in AML, and are used in the clinic to identify patients at high risk of death and to guide treatment. Aberrant levels of messenger RNA (mRNA)9-11 and microRNA (miR) transcripts12,13 also have prognostic significance, and efforts have been made to incorporate gene-expression profiling into prognostic algorithms.14-16 Long non-coding RNAs (lncRNAs) are a novel class of RNA molecules that are longer than 200 nucleotides, have no protein coding potential and are either located within the intergenic stretches of the genome or overlap (in sense or antisense direction) protein coding genes.17,18 These transcripts regulate key cellular functions, such as chromosome dosage compensation,19 imprinting,20 cell cycle progression,21 and differentiation.22 In cancer, individual lncRNAs have been shown to play an important role in malignant transformation.23-25 Despite the growing understanding of the biologic significance of deregulated lncRNA expression in malignant diseases, the value of these molecules as potential biomarkers in the clinical setting has not been extensively studied.26,27 With regard to cytogenetically normal AML (CN-AML), the prognostic and biologic significance of lncRNAs in younger adult patients remains unknown. Therefore, we analyzed, using whole transcriptome sequencing (RNA-seq), the lncRNA profiles of younger adults with de novo CN-AML, who were comprehensively characterized with regard to molecular abnormalities and outcome. Herein, we show that lncRNA profiling provides independent prognostic information in these patients. We also show that expression levels of prognostic lncRNAs correlate with distinct mRNA and miR signatures, and provide insights into the leukemogenic pathways that these lncRNAs potentially regulate.

Methods

mutations that have been reported to associate with clinical outcome of CN-AML patients (i.e., mutations in the ASXL1, DNMT3A [R882 and non-R882], IDH1, IDH2 [R140 and R172], NPM1, RUNX1, TET2 or WT1 genes, and FLT3-tyrosine kinase domain [FLT3-TKD] mutations), as described previously.26,29 A variant allele frequency of ≥10% was used as the cutoff to distinguish between mutated versus wild-type alleles of these genes. The presence of mutations in the CEBPA gene and FLT3-internal tandem duplications (FLT3-ITD) were evaluated using Sanger sequencing30 and fragment analysis,31 as described previously. Since only double CEBPA mutations are favorable prognostic markers in CN-AML,32 we considered only this genotype as mutated.

Transcriptome analyses RNA samples of all studied patients (n=377) were analyzed with total RNA sequencing (after depletion of ribosomal and mitochondrial RNA) using the Illumina HiSeq 2500 platform. Due to RNA quality restrictions, a subset of 300 patients could be additionally analyzed with small RNA sequencing, for profiling of miR expression. Further details are provided in the Online Supplementary Appendix. To determine the expression status of patients (i.e., high versus low expressers) with regard to prognostic expression markers (e.g., BAALC), the median values of normalized RNA sequencing reads were used as the cutoff.

Statistical analyses Clinical endpoint definitions are given in the Online Supplementary Appendix. Baseline demographic, clinical, and molecular features were compared between patients with low and those with high lncRNA scores (later on referred to as favorable and unfavorable, see below), and between the training and validation sets using the Wilcoxon rank-sum and Fisher’s exact tests for continuous and categorical variables, respectively.33 The estimated probabilities of disease-free (DFS), overall (OS) and event-free survival (EFS) were calculated using the Kaplan–Meier method, and the log-rank test evaluated differences between survival distributions.34 Cox proportional hazard models were used to calculate hazard ratios (HR) for DFS, OS and EFS.33 Multivariable proportional hazards models were constructed using a backward selection procedure. All statistical analyses were performed by The Alliance Statistics and Data Center.

Patients and treatment Pretreatment bone marrow (BM) or blood samples were obtained from a training (n=263) and a validation set (n=114) of younger adult patients (aged 17-59 years) with de novo CN-AML, who received intensive cytarabine/anthracycline-based first-line therapy on Cancer and Leukemia Group B (CALGB)/Alliance for Clinical Trials in Oncology (Alliance) trials and were alive 30 days after initiation of treatment. Per protocol, no patient received allogeneic stem cell transplantation in first complete remission (CR). Details regarding treatment protocols are provided in the Online Supplementary Appendix. All patients provided written informed consent, and all study protocols were in accordance with the Declaration of Helsinki and approved by institutional review boards at each center.

Cytogenetic and molecular analyses Cytogenetic analyses were performed in CALGB/Allianceapproved institutional laboratories and results confirmed by central karyotype review.28 The diagnosis of normal karyotype was based on at least 20 metaphase cells analyzed in BM specimens subjected to short-term (24- or 48-hour) unstimulated cultures. Targeted amplicon sequencing using the MiSeq platform (Illumina) was used to analyze DNA samples for presence of gene 1392

Results Global expression of lncRNAs To investigate the role of lncRNAs in AML, we first identified all known lncRNAs which were present in the transcriptomes of the younger CN-AML patients who were studied (n=377). After exclusion of contaminating ribosomal RNA molecules, we identified 22,166 non-coding RNA transcripts. According to the GENCODE v22 database,35 23% of these transcripts were categorized as processed pseudogenes, 21% as intergenic/intervening lncRNAs, 21% as antisense lncRNAs, 4% as sense intronic/overlapping lncRNAs and 31% were classified as other transcripts (e.g., as unitary pseudogenes, unprocessed pseudogenes etc.; Figure 1).

Generation of a prognostic lncRNA score in the training set To assess the prognostic significance of lncRNA expression in younger adults with CN-AML, we performed exploratory analysis in a training set (n=263) of younger CN-AML patients and used a separate patient cohort to haematologica | 2017; 102(8)

lncRNA expression in younger adults with CN-AML

validate our findings (validation set, n=114). Comparison of clinical and molecular characteristics at diagnosis between the training and validation sets showed that they were relatively similar, with the exceptions that patients in the training set had higher percentages of blood blasts

(P=0.03), were more frequently FLT3-TKD-positive (P=0.02), and had higher ERG (P=0.01) and BAALC (P=0.002) expression levels (Online Supplementary Table S1). We first identified all lncRNAs that were highly associated with EFS (P<10-6) in the training set by univariable

Table 1. Comparison of clinical and molecular characteristics by favorable and unfavorable long non-coding RNA (lncRNA) score in the validation set of younger adults with cytogenetically normal acute myeloid leukemia.

Characteristic

Age, years Median Range Sex, n. (%) Male Female Race, n. (%) White Non-white Hemoglobin (g/dL) Median Range Platelet count (x109/L) Median Range WBC count (x109/L) Median Range Blood blasts, % Median Range Bone marrow blasts, % Median Range Extramedullary involvement, n. (%) Autologous HCT in 1st CR, n. (%) NPM1, n. (%) Mutated Wild-type FLT3-ITD, n. (%) Present Absent CEBPA, n. (%) Double Mutated Wild-type FLT3-TKD, n. (%) Present Absent WT1, n. (%) Mutated Wild-type TET2, n. (%) Mutated Wild-type IDH1, n. (%) Mutated Wild-type IDH2, n. (%) Mutated R140 R172 Wild-type ASXL1, n. (%) Mutated Wild-type haematologica | 2017; 102(8)

Favorable lncRNA Score (n=57)

Unfavorable lncRNA Score (n=57)

44 18-59

47 18-59

28 (49) 29 (51)

29 (51) 28 (49)

51 (91) 5 (9)

50 (89) 6 (11)

9.1 4.2-25.1

8.8 4.8-13.4

52 10-271

55 8-433

24.9 0.9-475.0

45.7 2.2-295.0

45 0-90

63 0-97

63 21-91 15 (28) 33 (65)

68 18-95 18 (32) 23 (48)

37 (65) 20 (35)

15 (27) 40 (73)

30 (54) 26 (46)

8 (15) 46 (85)

6 (12) 45 (88)

4 (7) 51 (93)

1 (2) 54 (98)

4 (7) 52 (93)

10 (19) 44 (81)

6 (11) 50 (89)

3 (6) 51 (94)

4 (7) 52 (93)

3 (5) 52 (95)

7 (13) 4 3 49 (88)

6 (11) 6 0 49 (89)

2 (4) 54 (96)

1 (2) 51 (98)

0.44 1.00 1.00 0.66 0.49 0.009 0.06 0.25 0.68 0.11 1.00 0.007

0.78 0.36

0.09 0.49 1.00 1.00

1.00

1393

D. Papaioannou et al. Characteristic

DNMT3A, n. (%) Mutated R882 Non-R882 Wild-type RUNX1, n. (%) Mutated Wild-type ELN Risk Category,* n. (%) Favorable Intermediate Adverse ERG expression group,† n. (%) High Low BAALC expression group,† n. (%) High Low MN1 expression group,† n. (%) High Low miR-181a expression group,† n. (%) High Low miR-3151, n. (%) Expressed Not expressed miR-155 expression group,† n. (%) High Low

Favorable lncRNA Score (n=57)

Unfavorable lncRNA Score (n=57)

20 (36) 14 6 36 (64)

23 (43) 20 3 31 (57)

3 (5) 53 (95)

2 (4) 52 (96)

37 (71) 11 (21) 4 (8)

23 (43) 20 (38) 10 (19)

22 (39) 35 (61)

23 (41) 33 (59)

19 (36) 34 (64)

21 (39) 33 (61)

18 (33) 37 (67)

29 (52) 27 (48)

24 (50) 24 (50)

18 (40) 27 (60)

8 (17) 40 (83)

4 (9) 41 (91)

16 (33) 32 (67)

31 (69) 14 (31)

0.56

1.00

0.02

0.85

0.84 0.06

0.41 0.36

<0.001

*Among patients with cytogenetically normal acute myeloid leukemia (CN-AML), the ELN favorable risk category comprises patients with double-mutated CEBPA and patients with mutated NPM1 without FLT3-ITD or with FLT3-ITDlow. The ELN intermediate risk category includes patients with wild-type NPM1 without FLT3-ITD or wild-type NPM1 and FLT3-ITDlow or mutated NPM1 and FLT3-ITDhigh. The ELN adverse risk category comprises patients with wild-type NPM1 and FLT3-ITDhigh and/or mutated RUNX1 (if it does not cooccur with a favorable AML subtype) and/or mutated ASXL1 (if does not co-occur with a favorable AML subtype) and/or mutated TP53. FLT3-ITDlow is defined by a FLT3-ITD/FLT3 wild-type allelic ratio of less than 0.5 and FLT3-ITDhigh is defined by a FLT3-ITD/FLT3 wild-type allelic ratio of equal to or more than 0.5.1 †The median expression value was used as the cut point. WBC: white blood cell; HCT: hematopoietic cell transplant; CR: complete remission; ELN: European LeukemiaNet; FLT3-ITD: internal tandem duplication of the FLT3 gene; FLT3-TKD: tyrosine kinase domain mutation in the FLT3 gene; lncRNA: long non-coding ribonucleic acid; miR: microRNA.

Cox analysis (Figure 2). EFS was used because it comprehensively evaluates the lncRNAs that are associated with response to chemotherapy, probability of relapse and probability of survival. We detected 24 lncRNAs associated with EFS (P<10-6; Online Supplementary Table S2). Next, we derived a prognostic lncRNA score by linear combination of the weighted expression values of these 24 lncRNAs. The median value of the lncRNA score was used to dichotomize the training set of patients. Patients with low lncRNA scores (n=132) had longer DFS (P<0.001), OS (P<0.001) and EFS (P<0.001) than patients with high lncRNA scores (n=131). We therefore classified low lncRNA scores as “favorable” and high as “unfavorable” (Online Supplementary Table S3 and Online Supplementary Figure S1).

Association of lncRNA score with patient characteristics and clinical outcome in the training set With regard to clinical and molecular characteristics, patients with favorable lncRNA scores in the training set were more likely to present with higher hemoglobin lev1394

els (P=0.02), lower white blood cell (WBC) counts (P<0.001), and lower percentages of BM blasts (P=0.02). They were also less likely to harbor FLT3-ITD (P<0.001), DNMT3A (P=0.01) and RUNX1 (P=0.009) mutations and more likely to harbor double CEBPA mutations (P<0.001). Patients with favorable lncRNA scores in the training set differed with regard to their distribution in the Risk Categories of the European LeukemiaNet (ELN) classification of AML,1 when compared with patients with unfavorable lncRNA scores (P<0.001); patients with favorable lncRNA scores were more frequently classified as favorable and less frequently as intermediate or adverse risk than those with unfavorable lncRNA scores (Online Supplementary Table S4). Favorable lncRNA score status also associated with high expression of miR-181a (P<0.001) and low expression of miR-155 (P=0.03, Online Supplementary Table S4). Association of a favorable lncRNA score with longer DFS, OS and EFS remained significant in multivariable analyses (P<0.001 for all 3 end points, Online Supplementary Table S5), after adjusting for other co-variates. haematologica | 2017; 102(8)

lncRNA expression in younger adults with CN-AML

Figure 1. Distribution of the 22,166 detected non-coding RNA transcripts among different classes of non-coding RNA molecules. Annotation of transcripts was performed according to the GENCODE v22 database. lncRNA indicates long non-coding RNA and lincRNA denotes long intergenic/intervening non-coding RNA. *Other refers to: microRNAs, miscellaneous non-coding RNAs, unprocessed pseudogenes, small RNAs, translated unprocessed pseudogenes, processed transcripts, small nucleolar RNAs, transcribed processed pseudogenes, T-cell receptor pseudogenes, immunoglobulin genes, immunoglobulin pseudogenes, unitary pseudogenes, small cajal body specific RNAs, polymorphic pseudogenes, 3-prime overlapping non-coding RNAs, transcribed unitary pseudogenes and macro lncRNAs. lncRNA: long non-coding ribonucleic acid.

Association of lncRNA score with patient characteristics and clinical outcome in the validation set We used the median value of the lncRNA score, as calculated in the training set to divide the validation set into favorable and unfavorable lncRNA score groups (Figure 2). Patients with favorable lncRNA scores (n=57) were less likely to present with higher WBC counts at the time of diagnosis (P=0.009) or to harbor FLT3-ITD (P=0.007). lncRNA score status also associated with significantly different distribution of the patients in the Risk Categories of the ELN guidelines (P=0.02).1 Patients with favorable lncRNA scores were more likely to belong to the favorable and less likely to belong to the intermediate or adverse risk category. Patients with favorable lncRNA scores in the validation set were less likely to be miR-155 high-expressers (P<0.001) than patients with unfavorable lncRNA scores (n=57; Table 1). Patients with favorable lncRNA scores had longer DFS than those with unfavorable lncRNA scores (P<0.001; Figure 3A). Five years after diagnosis, 51% of patients with favorable lncRNA scores remained alive and leukemia-free, in contrast to only 17% of those with unfavorable lncRNA scores. Favorable lncRNA score status also associated with longer OS (P=0.002, 5-year rates, 52% versus 26%; Figure 3B) and longer EFS (P<0.001, 5year rates, 46% versus 16%; Figure 3C, Online Supplementary Table S6). The prognostic value of the lncRNA score in the validation set remained significant when it was analyzed as a continuous variable. Increasingly favorable lncRNA scores associated with longer DFS (P<0.001), OS (P=0.007) and EFS (P=0.002). In multivariable analyses, favorable lncRNA score status was an independent marker for longer DFS (P=0.01), after adjusting for miR-155 expression status, and EFS (P=0.002), after adjusting for the presence of FLT3-ITD (Table 2). With regard to OS, patients with a favorable lncRNA score had a trend for longer survival (P=0.06), after adjustment for FLT3-ITD and MN1 expression status. haematologica | 2017; 102(8)

Associations of recurrent gene mutations with lncRNA expression We evaluated if recurrent prognostic gene mutations in CN-AML associated with distinct expression patterns of lncRNAs in younger adults with CN-AML. For this purpose, mutation-related lncRNA signatures were derived in the training set using stringent criteria (for details see Methods and the Online Supplementary Appendix). Double-mutated CEBPA showed the strongest association with lncRNA expression; 82 lncRNAs were upregulated and 186 lncRNAs were downregulated in patients who harbored double-mutated CEBPA (Figure 4A, Online Supplementary Table S7). Among the CEBPA mutationrelated lncRNAs, NEAT1 was significantly underexpressed in the group of patients with CEBPA mutations. This lncRNA has been involved in myeloid differentiation of acute promyelocytic leukemia cells after all-trans retinoic acid treatment.36 Mutations in the NPM1 gene also strongly associated with a lncRNA signature, which comprised 35 transcripts upregulated and 37 transcripts downregulated in patients harboring NPM1 mutations (Figure 4B, Online Supplementary Table S8). Thirty-three of the 35 lncRNAs overexpressed in patients with NPM1 mutations, were downregulated in patients with CEBPA mutations. This finding is consistent with the observation that double CEBPA and NPM1 mutations rarely co-occur in CN-AML. NPM1 mutations were positively associated with lncRNAs embedded within the HOX gene loci (HOXAAS3, HOXB-AS3) and other lncRNAs implicated in myelopoiesis (EGOT137) or carcinogenesis (e.g., PCAT1838 and LUCAT139). The FLT3-ITD-related lncRNA signature consisted of 26 transcripts, 19 of which were upregulated and 7 downregulated in patients with this mutation (Figure 4C, Online Supplementary Table S9). The host gene of miR-155 (MIR155HG) was among the lncRNAs overexpressed in FLT3-ITD-positive patients. High MIR155HG expression independently associates with poor outcome in CN1395

D. Papaioannou et al.

Figure 2. Overview of the study design.

AML.40 The WT1-AS lncRNA was also highly expressed among FLT3-ITD-positive patients; it has been reported to post-translationally regulate the protein levels of WT1.41 To assess the capacity of gene mutation-related lncRNA signatures to detect their corresponding molecular alterations in CN-AML patients, we applied these signatures to the validation set. The mutated CEBPA-related signature showed the highest level of accuracy (specificity and sensitivity of mutated CEBPA detection: ≥93% and ≥98%, respectively), followed by the mutated NPM1-related (sensitivity: ≥80%, specificity ≥73%) and the FLT3-ITDrelated signatures (sensitivity ≥70%, specificity: ≥76%). The remaining prognostic gene mutations that were tested either did not associate with differential expression of lncRNAs (i.e., TET2 mutations) or generated signatures that failed to reliably detect the mutational status of patients in the validation set (e.g., DNMT3A, WT1 mutations).

Biologic implications of the lncRNA score To gain biologic insights into the molecular pathways that may be affected by differences in the lncRNA score, we examined the correlation between the lncRNA score and the mRNA/miRNA expression in 300 younger CNAML patients who had available mRNA and miRNA profiling data. We identified 410 mRNA transcripts whose expression levels correlated with the lncRNA score, 172 of which correlated positively and 238 negatively with unfavorable lncRNA scores (Figure 5A, Online Supplementary Table S10). Among highly expressed genes in patients with unfavorable lncRNA scores, putative oncogenes and key mediators of the oncogenic AP-1 pathway such as ATF3, FOS, FOSB, JUN, and MAFF were identified. With regard to hematopoiesis, the AP-1 pathway has been shown to regulate proliferation of erythroleukemia cells,42 to mediate 1396

monocyte/macrophage differentiation of myeloid cells43 and to co-regulate miR-155 expression in stimulated macrophages.44 Genes that regulate immune responses (e.g., IL1B, IRF7, CD80) and genes that mediate immune evasion (e.g., IER3, LILRB4) were also highly expressed in patients with unfavorable lncRNA profiles. Finally, oncogenes promoting proliferation of malignant cells (e.g., RET, ETS2, PLK2, NEK6, PLK3 and SRC) were found to be overexpressed in patients with unfavorable lncRNA scores. Gene ontology analysis revealed that genes involved in lymphocyte/leukocyte activation, inflammation, response to wounding and regulation of apoptosis were enriched in the subset of patients with unfavorable lncRNA scores (Figure 5B, Online Supplementary Table S11). Among mRNA molecules downregulated in patients with unfavorable lncRNA scores, we detected transcripts with reported tumor-suppressive function (APC, JADE1, BRMS1L, and ING3). Gene ontology analysis showed that genes that participate in the regulation of transcription, the regulation of RNA metabolic processes and DNA binding were underexpressed in the group of patients with unfavorable lncRNA scores (Figure 5C, Online Supplementary Table S11). With regard to miR expression, 10 miRs were found to correlate positively (miR-660, miR-502, miR-532-5p, miR155, miR-500a-3p, miR-500a-5p, miR-532-3p, miR-362, miR-339 and miR-23a) and 4 miRs to correlate negatively (miR-192, miR-625, miR-100 and miR-194) with unfavorable lncRNA scores (Online Supplementary Table S12). Among the 10 miRs that positively correlated with unfavorable lncRNA scores, 7 were located in close proximity on chromosome X; miR-660, miR-502, miR-532-5p, miR500a-3p, miR-500a-5p, miR-532-3p and miR-362 are all imbedded in intron 3 of the CLCN5 gene. This miR cluster mediates an anti-apoptotic effect in chronic lymphocytic leukemia cells.45 miR-155, which also positively correlated haematologica | 2017; 102(8)

lncRNA expression in younger adults with CN-AML

these patients.26 Since CN-AML in younger adults differs with regard to clinical features, associated molecular abnormalities and outcome from that in older patients, we investigated the prognostic value and biologic implications of lncRNA expression in a total of 377 CN-AML adult patients younger than 60 years. First, we identified 24 lncRNAs highly correlated with EFS. Similarly to the previously reported older CN-AML patients,26 only a small number of these prognostic lncRNAs associated with prognostic gene mutations: MIR155HG was upregulated in patients who harbor FLT3ITD, AC006129.2 was upregulated in patients with double CEBPA mutations, whereas AL122127.25, RP11-946L16.2,

with unfavorable lncRNA scores, is an established adverse prognosticator in CN-AML40 and has been implicated in leukemogenesis of FLT3-ITD-positive AML.46

Discussion Over the past 5 years, lncRNAs have emerged as new players in cancer biology and biomarker discovery.47 Our group has previously reported that distinctive lncRNA signatures are associated with prognostic gene mutations in older CN-AML patients, and that expression levels of a small group of lncRNAs have prognostic significance in

Table 2. Multivariable analyses for outcome in the validation set of younger adults with cytogenetically normal acute myeloid leukemia.

Variables in final models lncRNA score, favorable versus unfavorable miR-155, high versus low* FLT3-ITD, present versus absent MN1, high versus low*

Disease-free survival HR (95% CI)

Overall survival HR (95% CI)

Event-free survival HR (95% CI)

0.46 (0.26-0.83) 1.81 (1.01-3.24) -

0.01 0.05 -

0.6 (0.35-1.03) 1.96 (1.17-3.29) 1.92 (1.16-3.17)

0.06 0.01 0.01

0.48 (0.30-0.77) 2.15 (1.36-3.41) -

0.002 0.001 -

Hazard ratios greater than (less than) 1.0 indicate higher (lower) risk for relapse or death (disease-free survival) or death (overall survival) or for failure to achieve complete remission, relapse or death (event-free survival) for the first category listed. Variables considered for model inclusion were: lncRNA score status (favorable versus unfavorable), age (as a continuous variable, in 10-year increments), sex (male versus female), race (white versus non-white), white blood cell count (as a continuous variable, in 50-unit increments), hemoglobin (as a continuous variable, in 1-unit increments), platelet count (as a continuous variable, in 50-unit increments), extramedullary involvement (present versus absent), ASXL1 mutations (mutated versus wild-type), CEBPA mutations (double-mutated versus single-mutated or wild-type), DNMT3A mutations (mutated versus wild-type), FLT3-ITD (present versus absent), FLT3-TKD (present versus absent), IDH1 mutations (mutated versus wild-type), IDH2 mutations (mutated versus wild-type), NPM1 mutations (mutated versus wild-type), RUNX1 mutations (mutated versus wild-type), TET2 mutations (mutated versus wild-type), WT1 mutations (mutated versus wild-type), BAALC expression levels (high versus low), ERG expression levels (high versus low), MN1 expression levels (high versus low), miR-181a expression levels (high versus low), miR3151 (expressed versus not expressed), and miR-155 expression levels (high versus low). * The median expression value was used as the cut point. HR: hazard ratio; CI: confidence interval; lncRNA: long non-coding RNA; FLT3-ITD: internal tandem duplication of the FLT3 gene.

P<0.001

P=0.002

C P<0.001

Figure 3. Outcomes of younger adult patients with cytogenetically normal acute myeloid leukemia with favorable and unfavorable long non-coding RNA (lncRNA) scores in the validation set. (A) Disease-free survival, (B) overall survival and (C) event-free survival. The lncRNA score of each individual patient was computed as a weighted score of 24 prognostic lncRNAs.

haematologica | 2017; 102(8)

1397

D. Papaioannou et al. A

Figure 4. Long non-coding RNA (lncRNA) expression signatures associated with prognostic gene mutations in younger adult patients with cytogenetically normal acute myeloid leukemia. Heat maps for (A) double CEBPA, (B) NPM1 and (C) FLT3-ITD mutation-related lncRNA signatures are presented. The lncRNA signatures were derived in the training set of the studied cohort. Expression values of the lncRNA transcripts are represented by color, with green indicating expression less than and red indicating expression greater than the median value for the lncRNA transcript. Gray color indicates lack of detectable expression. Rows represent lncRNA transcripts, and columns represent patients. Patients are grouped by the gene mutational status (i.e., mutated [mut] versus wild-type [wt]). For a full list of the lncRNAs that associated with prognostic gene mutations see the Online Supplementary Appendix.

SDHAP3 and SENC3 were downregulated in patients with double CEBPA mutations. Of the 24 prognostic lncRNA genes, only MIR155HG has previously been associated with the clinical outcome of CN-AML patients.40,48 Linear combination of the weighted expression values of lncRNA transcripts yielded a prognostic score, which strongly associated with DFS, EFS and OS duration in younger adult CN-AML patients. Favorable lncRNA score status was an independent marker for longer DFS and EFS (and also showed a strong trend towards significance for longer OS). We were intrigued to find no overlap between the 48 prognostic lncRNAs that we previously identified in older CN-AML patients26 and the 24 transcripts reported herein in younger patients. This finding could be interpreted as an additional biologic difference between CN-AML of younger and that of older patients, similar to the agedependent difference in frequency of some recurrent prognostic gene mutations (e.g., mutations in the ASXL1 and RUNX1 genes).1 We also examined the associations between recurrent prognostic gene mutations and lncRNA expression, and found double CEBPA and NPM1 mutations and FLT3-ITD to associate with distinct lncRNA signatures. We identified several lncRNAs that were commonly associated with these gene mutations in both younger and older CN-AML patients26 (e.g., the HOX-loci embedded lncRNAs in the NPM1 mutation-related lncRNA signature, WT1-AS in the FLT3-ITD-related signature, etc.). On the other hand, such gene mutations as RUNX1 and ASXL1 that are more frequent in older CN-AML patients and were found to associate with differential expression of lncRNAs26 could not be tested in younger CN-AML patients, because too few 1398

younger patients harbored these mutations. Of note, mutations in the TET2 gene showed no correlation with differential expression of lncRNA molecules in either older26 or younger CN-AML patients, despite their impact on the epigenome49 and adequate numbers of patients in both studied cohorts. To gain insights into biologic pathways affected by differences in the lncRNA score, we investigated correlations between mRNA and miR expression signatures and lncRNA scores. In concordance with the adverse outcome that unfavorable lncRNA scores bestow, a number of previously described oncogenes and oncomiRs were found overexpressed in patients with unfavorable lncRNA score status. Similarly, genes with reported tumor-suppressive activity were found downregulated in this patient group. Only a small fraction of these transcripts have been reported in gene mutation-related mRNA signatures or other prognostic gene-expression signatures.14-16 These findings indicate that, in addition to being independent of prognostic mutations, the differential expression of prognostic lncRNAs may regulate distinct molecular pathways in CN-AML. Notably, 5 important mediators of the AP-1 pathway (ATF3, FOS, FOSB, JUN, and MAFF) were found upregulated in patients with unfavorable lncRNA scores. The high number of cell cycle regulators and proliferation-inducing kinases that were also upregulated in this patient group is consistent with aberrant activation of the AP-1 pathway. In this work, we used whole transcriptome sequencing techniques to identify and measure the expression of prognostic lncRNA molecules in younger adults with CNAML. While this technology is becoming rapidly cheaper and widely available, it will most likely continue to serve haematologica | 2017; 102(8)

lncRNA expression in younger adults with CN-AML

Figure 5. Messenger RNA (mRNA) transcripts which associate with the long non-coding RNA (lncRNA) score in younger adults with cytogenetically normal acute myeloid leukemia (CN-AML). (A) Heat map of the gene-expression signature associated with the lncRNA score. Rows represent protein-coding genes and columns represent patients. Patients are grouped by lncRNA score: favorable on the left and unfavorable on the right. The lncRNA score of each individual patient was computed as a weighted score of 24 prognostic lncRNAs. Expression values of the lncRNA transcripts are represented by color: green: expression less than median value; red: expression greater than median value; gray: lack of detectable expression. Top 5 gene ontology terms that positively (B) or negatively (C) correlate with unfavorable lncRNA scores in younger patients with CN-AML are displayed. Gene ontology terms in (B) and (C) are ranked according to fold enrichment.

as a research tool rather than a diagnostic test to guide patient treatment. Despite this, alternative techniques for measuring RNA transcripts in a clinically applicable manner are available and are used to risk stratify patients with certain solid malignancies.50 Similar assays could be developed in order to obtain targeted measurements of prognostic lncRNAs in a fast and clinically meaningful manner. The potential of such assays to improve risk stratification of AML patients should be evaluated in future prospective clinical studies. Acknowledgments The authors would like to thank: Donna Bucci and Wacharaphon Vongchucherd of The Alliance NCTN Biorepository and Biospecimen Resource for sample processing and storage services, Karl Kroll for technical support, and Lisa J.

References 1. Dรถhner H, Estey E, Grimwade D, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood. 2017;129(4):424-447. 2. Byrd JC, Mrรณzek K, Dodge RK, et al. Pretreatment cytogenetic abnormalities are predictive of induction success, cumulative incidence of relapse, and overall survival in

haematologica | 2017; 102(8)

Sterling and Christine Finks of The Ohio State University, Comprehensive Cancer Center, Columbus, OH for data management. Funding This work was supported by the National Cancer Institute of the National Institutes of Health under Award Numbers CA180821 and CA180882 (to the Alliance for Clinical Trials in Oncology), CA077658, CA180850, CA180861, CA140158, CA16058, and CA197734. This work was also supported in part by the Leukemia Clinical Research Foundation, D Warren Brown Foundation, and the Pelotonia Fellowship Program. The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health.

adult patients with de novo acute myeloid leukemia: results from Cancer and Leukemia Group B (CALGB 8461). Blood. 2002;100(13):4325-4336. 3. Grimwade D, Hills RK, Moorman AV, et al. Refinement of cytogenetic classification in acute myeloid leukemia: determination of prognostic significance of rare recurring chromosomal abnormalities among 5876 younger adult patients treated in the United Kingdom Medical Research Council trials. Blood. 2010;116(3):354-365.

4. Mrรณzek K, Heerema NA, Bloomfield CD. Cytogenetics in acute leukemia. Blood Rev. 2004;18(2):115-136. 5. Papaemmanuil E, Gerstung M, Bullinger L, et al. Genomic classification and prognosis in acute myeloid leukemia. N Engl J Med. 2016;374(23):2209-2221 6. Metzeler KH, Herold T, RothenbergThurley M, et al. Spectrum and prognostic relevance of driver gene mutations in acute myeloid leukemia. Blood. 2016;128(5):686698.

1399

D. Papaioannou et al. 7. Patel JP, Gönen M, Figueroa ME, et al. Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med. 2012;366(12):1079-1089. 8. Döhner H, Weisdorf DJ, Bloomfield CD. Acute myeloid leukemia. N Engl J Med. 2015;373(12):1136-1152. 9. Baldus CD, Tanner SM, Ruppert AS, et al. BAALC expression predicts clinical outcome of de novo acute myeloid leukemia patients with normal cytogenetics: a Cancer and Leukemia Group B study. Blood. 2003;102(5):1613-1618. 10. Marcucci G, Baldus CD, Ruppert AS, et al. Overexpression of the ETS-related gene, ERG, predicts a worse outcome in acute myeloid leukemia with normal karyotype: a Cancer and Leukemia Group B study. J Clin Oncol. 2005;23(36):9234-9242. 11. Schwind S, Marcucci G, Kohlschmidt J, et al. Low expression of MN1 associates with better treatment response in older patients with de novo cytogenetically normal acute myeloid leukemia. Blood. 2011; 118(15): 4188-4198. 12. Garzon R, Volinia S, Liu C-G, et al. MicroRNA signatures associated with cytogenetics and prognosis in acute myeloid leukemia. Blood. 2008;111(6):3183-3189. 13. Marcucci G, Radmacher MD, Maharry K, et al. MicroRNA expression in cytogenetically normal acute myeloid leukemia. N Engl J Med. 2008;358(18):1919-1928. 14. Bullinger L, Döhner K, Bair E, et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med. 2004; 350(16):1605-1616. 15. Metzeler KH, Hummel M, Bloomfield CD, et al. An 86-probe-set gene-expression signature predicts survival in cytogenetically normal acute myeloid leukemia. Blood. 2008;112(10):4193-4201. 16. Li Z, Herold T, He C, et al. Identification of a 24-gene prognostic signature that improves the European LeukemiaNet risk classification of acute myeloid leukemia: an international collaborative study. J Clin Oncol. 2013;31(9):1172-1181. 17. Gibb EA, Brown CJ, Lam WL. The functional role of long non-coding RNA in human carcinomas. Mol Cancer. 2011;10:38. 18. Ponting CP, Oliver PL, Reik W. Evolution and functions of long noncoding RNAs. Cell. 2009;136(4):629-641. 19. Zhao J, Sun BK, Erwin JA, Song J-J, Lee JT. Polycomb proteins targeted by a short repeat RNA to the mouse X chromosome. Science. 2008;322(5902):750-756. 20. Bartolomei MS, Zemel S, Tilghman SM. Parental imprinting of the mouse H19 gene. Nature. 1991;351(6322):153-155. 21. Hung T, Wang Y, Lin MF, et al. Extensive and coordinated transcription of noncoding RNAs within cell-cycle promoters. Nat Genet. 2011;43(7):621-629. 22. Ulitsky I, Shkumatava A, Jan CH, Sive H, Bartel DP. Conserved function of lincRNAs in vertebrate embryonic development

1400

23. 24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34. 35.

36.

despite rapid sequence evolution. Cell. 2011;147(7):1537-1550. Yildirim E, Kirby JE, Brown DE, et al. Xist RNA is a potent suppressor of hematologic cancer in mice. Cell. 2013;152(4):727-742. Gupta RA, Shah N, Wang KC, et al. Long non-coding RNA HOTAIR reprograms chromatin state to promote cancer metastasis. Nature. 2010;464(7291):1071-1076. Trimarchi T, Bilal E, Ntziachristos P, et al. Genome-wide mapping and characterization of Notch-regulated long noncoding RNAs in acute leukemia. Cell. 2014;158(3): 593-606. Garzon R, Volinia S, Papaioannou D, et al. Expression and prognostic impact of lncRNAs in acute myeloid leukemia. Proc Natl Acad Sci USA. 2014;111(52):1867918684. Zhou M, Guo M, He DA, et al. A potential signature of eight long non-coding RNAs predicts survival in patients with non-small cell lung cancer. J Transl Med. 2015;17 (13)231. Mrózek K, Carroll AJ, Maharry K, et al. Central review of cytogenetics is necessary for cooperative group correlative and clinical studies of adult acute leukemia: the Cancer and Leukemia Group B experience. Int J Oncol. 2008;33(2):239-244. Kroll KW, Eisfeld AK, Lozanski G, Bloomfield CD, Byrd JC, Blachly JS. MuCor: mutation aggregation and correlation. Bioinformatics. 2016; 32(10):1557-1558. Marcucci G, Maharry K, Radmacher MD, et al. Prognostic significance of, and gene and microRNA expression signatures associated with, CEBPA mutations in cytogenetically normal acute myeloid leukemia with high-risk molecular features: a Cancer and Leukemia Group B study. J Clin Oncol. 2008;26(31):5078-5087. Whitman SP, Archer KJ, Feng L, et al. Absence of the wild-type allele predicts poor prognosis in adult de novo acute myeloid leukemia with normal cytogenetics and the internal tandem duplication of FLT3: a Cancer and Leukemia Group B study. Cancer Res. 2001;61(19):7233-7239. Wouters BJ, Löwenberg B, ErpelinckVerschueren CAJ, van Putten WLJ, Valk PJM, Delwel R. Double CEBPA mutations, but not single CEBPA mutations, define a subgroup of acute myeloid leukemia with a distinctive gene expression profile that is uniquely associated with a favorable outcome. Blood. 2009;113(13):3088-3091. Vittinghoff E, Glidden DV, Shiboski SC, McCulloch CE. Regression methods in biostatistics: linear, logistic, survival and repeated measures models. New York, NY: Springer, 2005. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481. Harrow J, Frankish A, Gonzalez JM, et al: GENCODE: the reference human genome annotation for The ENCODE Project. Genome Res. 2012; 22(9):1760-1774. Zeng C, Xu Y, Xu L, et al. Inhibition of long

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47. 48.

49.

50.

non-coding RNA NEAT1 impairs myeloid differentiation in acute promyelocytic leukemia cells. BMC Cancer. 2014;14:693. Wagner LA, Christensen CJ, Dunn DM, et al. EGO, a novel, noncoding RNA gene, regulates eosinophil granule protein transcript expression. Blood. 2007;109(12): 5191-5198. Crea F, Watahiki A, Quagliata L, et al. Identification of a long non-coding RNA as a novel biomarker and potential therapeutic target for metastatic prostate cancer. Oncotarget. 2014;5(3):764-774. Thai P, Statt S, Chen CH, Liang E, Campbell C, Wu R. Characterization of a novel long noncoding RNA, SCAL1, induced by cigarette smoke and elevated in lung cancer cell lines. Am J Respir Cell Mol Biol. 2013;49(2):204-211. Marcucci G, Maharry KS, Metzeler KH, et al. Clinical role of microRNAs in cytogenetically normal acute myeloid leukemia: miR-155 upregulation independently identifies high-risk patients. J Clin Oncol. 2013; 31(17):2086-2093. Moorwood K, Charles AK, Salpekar A, Wallace JI, Brown KW, Malik K. Antisense WT1 transcription parallels sense mRNA and protein expression in fetal kidney and can elevate protein levels in vitro. J Pathol. 1998;185(4):352-359. Smith MJ, Prochownik EV. Inhibition of c-jun causes reversible proliferative arrest and withdrawal from the cell cycle. Blood. 1992;79(8):2107-2115. Gaynor R, Simon K, Koeffler P. Expression of c-jun during macrophage differentiation of HL-60 cells. Blood. 1991;77(12):26182623. O'Connell RM, Taganov KD, Boldin MP, Cheng G, Baltimore D. MicroRNA-155 is induced during the macrophage inflammatory response. Proc Natl Acad Sci USA. 2007;104(5):1604-1609. Ruiz-Lafuente N, Alcaraz-García MJ, Sebastián-Ruiz S, et al. IL-4 up-regulates miR-21 and the miRNAs hosted in the CLCN5 gene in chronic lymphocytic leukemia. PLoS One. 2015;10(4):e0124936. Gerloff D, Grundler R, Wurm AA, et al. NFκB/STAT5/miR-155 network targets PU.1 in FLT3-ITD-driven acute myeloid leukemia. Leukemia. 2015;29(3):535-547. Huarte M. The emerging role of lncRNAs in cancer. Nat Med. 2015;21(11):1253-1261. Marcucci G, Yan P, Maharry K, et al. Epigenetics meets genetics in acute myeloid leukemia: clinical impact of a novel seven-gene score. J Clin Oncol. 2014;32(6):548-556. Ko M, Huang Y, Jankowska AM, et al. Impaired hydroxylation of 5-methylcytosine in myeloid cancers with mutant TET2. Nature. 2010;468(7325):839-843. Liu MC, Pitcher BN, Mardis ER, et al. PAM50 gene signatures and breast cancer prognosis with adjuvant anthracycline- and taxane-based chemotherapy: correlative analysis of C9741 (Alliance). NPJ Breast Cancer. 2016;(2):15023.

haematologica | 2017; 102(8)

ARTICLE

Chronic Lymphocytic Leukemia

Targeted activation of the SHP-1/PP2A signaling axis elicits apoptosis of chronic lymphocytic leukemia cells

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Elena Tibaldi,1 Mario Angelo Pagano,2 Federica Frezzato,3,4 Valentina Trimarco,3,4 Monica Facco,3,4 Giuseppe Zagotto,2 Giovanni Ribaudo,2 Valeria Pavan,2 Luciana Bordin,1 Andrea Visentin,3,4 Francesca Zonta,5 Gianpietro Semenzato,3,4 Anna Maria Brunati1§ and Livio Trentin3,4§

Department of Molecular Medicine, University of Padua; 2Department of Pharmaceutical and Pharmacological Sciences, University of Padua; 3Department of Medicine, University of Padua; 4Venetian Institute of Molecular Medicine (VIMM), Centro di Eccellenza per la Ricerca Biomedica, Padua and 5Department of Biomedical Sciences, University of Padua, Italy 1

Haematologica 2017 Volume 102(8):1401-1412

Co-senior authors

ABSTRACT

yn, a member of the Src family of kinases, is a key factor in the dysregulation of survival and apoptotic pathways of malignant B cells in chronic lymphocytic leukemia. One of the effects of Lyn’s action is spatial and functional segregation of the tyrosine phosphatase SHP-1 into two pools, one beneath the plasma membrane in an active state promoting pro-survival signals, the other in the cytosol in an inhibited conformation and unable to counter the elevated level of cytosolic tyrosine phosphorylation. We herein show that SHP-1 activity can be elicited directly by nintedanib, an agent also known as a triple angiokinase inhibitor, circumventing the phospho-S591-dependent inhibition of the phosphatase, leading to the dephosphorylation of pro-apoptotic players such as procaspase-8 and serine/threonine phosphatase 2A, eventually triggering apoptosis. Furthermore, the activation of PP2A by using MP0766, a novel FTY720 analog, stimulated SHP-1 activity via dephosphorylation of phospho-S591, which unveiled the existence of a positive feedback signaling loop involving the two phosphatases. In addition to providing further insights into the molecular basis of this disease, our findings indicate that the PP2A/SHP-1 axis may emerge as an attractive, novel target for the development of alternative strategies in the treatment of chronic lymphocytic leukemia. Introduction Reversible protein phosphorylation is the fundamental post-translational modification by which virtually all cellular events are regulated, enabling cells to respond properly to intra- and extra-cellular cues. The crucial players involved in this dynamic process are protein kinases and protein phosphatases, which are placed at the different levels of cellular signaling, and, albeit traditionally considered as functionally opposed to one another, not rarely cooperate to finely orchestrate and appropriately drive signal transduction.1-3 The significance of both classes of enzymes for a cell’s life and fate is mirrored by the effects of their dysregulation, whether related to altered expression or activity, which frequently underlies the onset and progression of a plethora of diseases.4-7 B-cell chronic lymphocytic leukemia (CLL), the most common leukemia in the western world,8-10 is no exception to this paradigm, with a high level of intracellular phosphorylation being mediated by the abnormal activity of several kinases downstream of the B-cell receptor, including Lyn, Syk, Btk, PI3K, and Akt.11 This condition, along with the aberrant expression of anti-apoptotic molecules12 and the supportive microenvironment,13 leads to the derangement of cell signaling and contributes to the growth and survival of leukemia cells. haematologica | 2017; 102(8)

Correspondence: annamaria.brunati@unipd.it or mario.pagano@unipd.it Received: September 1, 2016 Accepted: June 14, 2017. Pre-published: June 15, 2017. doi:10.3324/haematol.2016.155747 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1401 ©2017 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.

1401

E. Tibaldi et al.

Accumulating evidence suggests that the abnormal signal transduction is also supported by the lack of a proper counterbalance mediated by a number of phosphatases, whose expression or activity is altered in CLL cells. For instance, the expression of PTEN,14 PTPROt,15 PHLPP116,17 and SHIP118 is significantly decreased, leaving tonic prosurvival signaling intact, whereas PTPN22, which acts as a positive regulator of anti-apoptotic signals by hampering the negative regulation of B-cell receptor-dependent signaling pathways, is overexpressed.19 By contrast, protein phosphatase 2A (PP2A)20 and the Src homology 2 domaincontaining phosphatase 1 (SHP-1)21 are expressed in CLL at levels comparable to those in normal B cells, but are functionally dysregulated by a variety of mechanisms, which are chiefly mediated by the Src family kinase, Lyn.22,23 In normal B cells, this tyrosine kinase is central to propagating signals initiated by engagement of the B-cell receptor through phosphorylation of the immunoreceptor tyrosine-based activation motifs of the B-cell receptor itself, promoting the formation of a multiprotein “signalosome”.24 It also possesses the unique ability to phosphorylate immunoreceptor tyrosine inhibitory motifs of inhibitory cell surface co-receptors, including FcγRIIb1, CD22, CD72 and CD5, and eventually to negatively regulate B-cell receptor signaling.25,26 In CLL cells, Lyn is overexpressed and constitutively active, distributed between the plasma membrane and an aberrant cytosolic multiprotein complex,27,28 exerting a crucial anti-apoptotic role by phosphorylating and thereby modifying the functional status of, a variety of protein targets.22,29 As to PP2A, the phosphorylation of its catalytic subunit (PP2Ac) by Lyn stabilizes the interaction of PP2Ac itself with SET, an endogenous PP2A inhibitor overexpressed in CLL cells,30 resulting in blockade of the phosphatase activity31,32 and a persistently high level of phosphorylation of factors implicated in CLL cell survival.22 On the other hand, Lyn induces the spatial segregation of SHP-1 into two pools, one being associated with the inhibitory coreceptor CD5 in an active form triggering membranederived anti-apoptotic signals, the other being located in the cytosol in an inactive conformation.30 Importantly, this latter condition is thought to be promoted by phosphorylation of the C-terminal tail at Ser591, whereas PP2A seems to play a role in the reactivation of SHP-1.22 The landscape of cell signaling inhibitors approved for the treatment of CLL has expanded rapidly and several agents with novel mechanisms of action (inhibitors of BTK, PI3K and Bcl2) have been introduced into routine clinical practice with promising results documented also in patients with relapsed/refractory disease.33-36 Because CLL is characterized by a high level of Lyn-dependent tyrosine phosphorylation in the cytosol,27-29 we wondered whether compounds capable of directly or indirectly driving the activation of SHP-1 could counter the pervasive action of Lyn and induce cell demise. Here we demonstrate that nintedanib, a small molecule tyrosine kinase inhibitor approved for the treatment of pulmonary fibrosis and lung adenocarcinoma,37 and MP0766, a novel fingolimod analog designed and synthesized in our laboratories, generate a positive feedback signaling loop involving SHP-1 and PP2A, which dismantles the oncogenic machinery supported by the pro-survival action of the aberrant cytosolic form of Lyn. These molecules, which specifically target and activate molecular players other than those traditionally considered compo1402

nents of the CLL signalosome, may represent new weapons for the treatment of CLL patients.

Methods Ethics statement Written informed consent was obtained from all patients, prior to sample collection, according to the Declaration of Helsinki. The ethical approval for our study was granted by the local ethical committee of “Regione Veneto on Chronic Lymphocytic Leukemia” (3259/AO/14).

Patients, cell separation and culture conditions B cells from 37 untreated patients with CLL were purified and cultured as previously described,27 and subjected to the treatments described throughout the text. The patients’ relevant features are reported in Online Supplementary Table S1. Information concerning reagents, cell lysis and subcellular fractionation, SHP-1 activity assays immunoprecipitation of SHP-1, [32P]-Band 3 preparation, the PP2A activity assay, the Casp8 activity assay, cell transfection, co-culture conditions, apoptosis assays, western blotting, and the statistical analysis is detailed in the Online Supplementary Data.

Results Nintedanib directly activates SHP-1 in the cytosol of chronic lymphocytic leukemia cells We previously demonstrated that SHP-1 is present in CLL cells in two forms, one bound to the plasma membrane receptor CD5 in an active state, and the other in the cytosol in an inhibited conformation.23 As shown in Figure 1A, as well as in Online Supplementary Figure S1A, the plasma membrane-enriched fraction (particulate) and the cytosol of CLL cells were separated from total lysates and immunoblotted with anti-pY536-SHP-1 and anti-pS591SHP-1 antibodies, a positive response to which is indicative of either activation or inhibition of SHP-1, respectively. As expected, the distribution of SHP-1 in the two cellular compartments paralleled the phosphorylation state, these characteristics proving independent of the diverse biological and clinical features of the single patients (Online Supplementary Table S1). Moreover, to establish how the phosphorylation state affected SHP-1 activity, SHP-1 was immunoprecipitated from the cytosolic and particulate fractions and its activity was measured by using [32P]-Band 3 as a substrate.38 The phosphatase activity of the cytosolic fraction of SHP-1 was negligible as compared to that of the particulate (Figure 1B and Online Supplementary Figure S1B), underscoring that the catalytic activity of SHP-1 can be profoundly changed by phosphorylation at different residues. This finding raised the issue of whether activating the cytosolic pool of SHP-1 might represent a means of countering the elevated level of tyrosine phosphorylation of CLL cells in this compartment, which was previously shown to promote anti-apoptotic mechanisms in this disease.22 We, therefore, first performed in vitro phosphatase activity assays on the cytosolic pool of SHP-1 in the presence of increasing concentrations of nintedanib, a receptor tyrosine kinase inhibitor recently shown to trigger SHP-1 activity39,40 despite the inhibitory S591 phosphorylation.41 First, SHP-1 was immunoprecipihaematologica | 2017; 102(8)

SHP-1/PP2A as a new druggable axis in CLL

tated from the cytosolic fraction of CLL cells in the absence or presence of serine/threonine phosphatase inhibitors, so that the inhibitory phosphorylation at S591 could be removed or preserved, respectively (Figure 1C, inset), and the effect of nintedanib on the activity of the

non-phosphorylated and phosphorylated forms of SHP-1 could be evaluated. Figure 1C shows that nintedanib was capable of activating the phosphorylated, and inhibited, form of SHP-1 (Ip2-SHP-1, right-hand panel), as expected,41 whereas the non-phosphorylated form was not influ-

Figure 1. In vitro effect of nintedanib on the differently phosphorylated forms of SHP-1 pulled down from chronic lymphocytic leukemia cells. (A) Phosphorylation state of SHP-1 in total cell lysates, particulate and cytosol of CLL cells of patients belonging to various clinical and biological subtypes. Anti-LDH (cytosolic marker), anti-PMCA (plasma membrane marker). (B) Tyrosine phosphatase activity of SHP-1 immunoprecipitated from particulate and cytosol of CLL of patients #2, #17, and #36 measured as [32P] released from in vitro [32P]-Band 3. (C) Tyrosine phosphatase activity of SHP-1 immunoprecipitated in the absence (Ip1-SHP-1) and presence (Ip2-SHP-1) of serine/threonine phosphatase inhibitors from the cytosol of CLL cells of 15 patients as determined in vitro in the presence of increasing concentrations of nintedanib supplemented without (solid circles) or with 25 mM PTP I-I (open circles). Data are mean Âą SD of three experiments performed in triplicate (*Pâ&#x2030;¤0.01). LDH: lactate dehydrogenase; PMCA: plasma membrane Ca2+ ATPase; Wb: western blot; IP: immunoprecipitation.

haematologica | 2017; 102(8)

1403

E. Tibaldi et al.

enced by this compound, the phosphatase being already active (Ip1-SHP-1, left-hand panel). These results are concordant with the hypothesis that this drug causes a change in the inhibited conformation of SHP-1 induced by phosphorylation at S591. Similarly, we analyzed the activity of SHP-1 immunoprecipitated from the cytosolic and particulate fractions of CLL cells treated with increasing concentrations of nintedanib, considering that SHP-1 in these

compartments is differently phosphorylated and active. As shown in Figure 2A, SHP-1 reached full activation at a concentration as high as 15 mM nintedanib (left-hand panel), as determined by the dephosphorylation of [32P]Band 3, whereas SHP-1 from the particulate was unaffected (right-hand panel). After incubating CLL cells with 15 mM nintedanib, we evaluated the phosphorylation status of SHP-1 from the cytosolic and particulate fractions by

Figure 2. Effect of nintedanib on the activity of the differently phosphorylated forms of SHP-1 inside chronic lymphocytic leukemia cells. (A) Phosphatase activity of SHP-1 immunoprecipitated from the cytosol (left) and particulate (right) of CLL cells cultured in the absence or presence of increasing concentrations of nintedanib for 1 h measured as [32P] released from [32P]-Band 3. Data are mean ± SD of three experiments performed in triplicate. (B) Expression and phosphorylation state of SHP-1 of CLL cells cultured in the presence of 15 mM nintedanib over time. (C) Phosphatase activity of SHP-1 immunoprecipitated from the cytosol (left) or particulate (right) of the CLL cells described in (B) measured as [32P] released from [32P]-Band 3. (D) Densitometric analysis of western blots probed with anti-pS591 or antipY536 antibody (arbitrary units, open circles, left- and right-hand panel, respectively) and phosphatase activity from (C) normalized as percentage (solid circles, left and right panel, respectively) and plotted as line graphs. Data are mean ± SD of three experiments performed in triplicate (*P≤ 0.01), n=16. Wb: western blot; IP: immunoprecipitation; incub: incubation.

1404

haematologica | 2017; 102(8)

SHP-1/PP2A as a new druggable axis in CLL

western blot analysis with anti-pS591-SHP-1 and antipY536-SHP-1 antibodies, respectively, and performed phosphatase activity assays on SHP-1 immunoprecipitates at each time interval. Nintedanib did not affect either the phosphorylation status (Figure 2B, right-hand panel) or the catalytic activity (Figure 2C, right-hand panel) of the SHP1 pool of the particulate, whereas SHP-1 in the cytosol was readily activated by nintedanib, reaching maximal efficacy already at the earliest incubation times by circumventing the inhibitory phosphorylation at S591 (Figure 2B,C, left-hand panel). The data obtained from western blot analysis and phosphor imaging were quantitated as arbitrary units and are illustrated in Figure 2D for clarity’s sake.

Nintedanib induces activation of caspase-8 and PP2A by decreasing tyrosine phosphorylation in chronic lymphocytic leukemia cells To verify whether the inhibition of SHP-1 contributed to the elevated tyrosine phosphorylation of CLL cells, which we earlier demonstrated to be dependent on the delocalized and constitutively active HSP90-bound form of Lyn27,28 (Online Supplementary Figure S2), freshly isolated CLL cells underwent SHP-1 knockdown (Figure 3A) and then incubated for 1 h in the presence of increasing concentrations of nintedanib (0-30 mM). After cell lysis, western blotting with anti-phosphotyrosine antibody showed that nintedanib caused a dramatic reduction in tyrosine phosphorylation at a concentration of 10 mM (Figure 3B and Online Supplementary S3B, left-hand panel), which is significantly higher than the nanomolar range reported to

inhibit the receptor tyrosine kinases that nintedanib is known to target.39,40 Importantly, the effect of nintedanib was abrogated by the genetic inhibition of the phosphatase itself (Figure 3B and Online Supplementary S3B, right-hand panel). On the other hand, tyrosine kinase activities were only affected to a limited extent by nintedanib, as assessed by the level of phosphorylation of poly-Glu-Tyr or cdc2[6-20], peptide substrates used to determine global (including that of receptor tyrosine kinases) or Src family kinase-specific tyrosine kinase activity, respectively.42,43 In fact, as shown by the histograms in Figure 3C and Online Supplementary Figure S3C, both types of activities were affected at most by 30% of the control only at high concentrations of nintedanib (over 10 mM). Notably, the slight decrease in tyrosine kinase activity paralleled the activation state of Lyn, as monitored by western blot analysis with anti-pTyr-396-Lyn antibody (Figure 3D and Online Supplementary Figure S3D), whereas the phosphorylation of CD5, to which the pool of SHP-1 associated with the plasma membrane is bound,23 was unchanged (Figure 3E and Online Supplementary Figure S3E). Moreover, stimulation with 20 mg/mL anti-IgM antibody at different time points, which only weakly increased the level of phosphorylation of CLL cells compared to that in the resting state,23,27 did not affect the decrease in tyrosine phosphorylation caused by pre-incubation with 15 mM nintedanib for 1 h (Online Supplementary Figure S4), again underscoring the importance of the activation of the cytosolic pool of SHP-1 in dampening phospho-tyrosine-mediated signals. Additionally, we wanted to evaluate whether

E B

haematologica | 2017; 102(8)

Figure 3. Effect of nintedanib on the tyrosine phosphorylation of chronic lymphocytic leukemia cells. (A) Expression of SHP-1 in CLL cells of patient #18 transfected with scrambled and SHP-1 siRNA. (B) Phosphorylation pattern of CLL cells transfected with scrambled or SHP-1 siRNA and cultured in the absence or presence of increasing concentrations of nintedanib for 1 h. (C) Global tyrosine kinase activity and specific Src activity in CLL cells cultured as in (B) determined by [32P] incorporation into the nonspecific random polymer polyGlu4Tyr (top panel) or the specific peptide substrate cdc2(6-20) (bottom panel). (D) Activation state of Lyn in CLL cells cultured as in (B) determined by western blotting with anti-pTyr396-Lyn antibody. (E) Phosphorylation state of CD5 as assessed by western blotting with anti-pTyr antibody of CD5 immunoprecipitates from total cell lysates of CLL cells treated as in (B). Data are mean ± SD of three experiments performed in triplicate (*P≤0.05). Wb: western blotting; IP: immunoprecipitation.

1405

E. Tibaldi et al.

nintedanib could affect CLL cell viability, considering that compounds decreasing the tyrosine phosphorylation in CLL cells also induced apoptosis.27,28 Freshly isolated CLL cells incubated with increasing concentrations of nintedanib (0-24 mM) for 24 and 48 h exhibited a marked level of apoptosis, as determined by annexin Vâ&#x20AC;&#x201C;propidium iodide flow cytometry and pooled densitometric analysis (Figure 4A, left-hand panel). Moreover, the degree of variability observed among the different subsets of patients as spontaneous apoptosis was attenuated after treatment with nintedanib (Online Supplementary Figure S5). Notably, the concentrations of nintedanib able to induce apoptosis in CLL cells did not alter the survival of normal B cells (Online Supplementary Figure S6). Mechanistically, PARP cleavage suggested that activation of the apoptotic pathway was the major route leading to cell death (Figure 4A, right-hand panel), which was further confirmed by the use of the pan-caspase inhibitor z-VADfmk44 (Online Supplementary Figure S7A) and by experiments exploring alternative mechanisms possibly contributing to cell demise, such as necroptosis or autophagy. In fact, survival of CLL cells did not vary if treatment with nintedanib followed pre-incubation with necrostatin-1, an inhibitor of necroptosis (Online Supplementary Figure S7B), and the protein levels of p62/SQSTM1, the degradation of which is indicative of autophagy,45,46 remained stable for the entire duration of the experiments performed in the presence of nintedanib (Online Supplementary Figure S7C). In the latter case, the positive control was provided by supplementing the incubation medium of CLL cells with the mTOR inhibitor everolimus, which caused a decrease in the protein level of p62/SQSTM1 (Online Supplementary Figure S7D).47 Similar results were obtained by co-culturing CLL cells with bone marrow mesenchymal stromal cells, which had been cultured in monolayer and grown to confluence (approximately 1x105 mesenchymal stromal cells per well), in a 20:1 ratio, to mimic the tissue microenviroment in which CLL cells proliferate48 (Online Supplementary Figure S8). To further investigate the contribution of SHP-1 to nintedanib-induced apoptosis, we used genetic and pharmacological inhibition of SHP-1 itself. Figure 4B (left-hand panel) shows that the rate of apoptosis after 24 h of treatment with nintedanib (bar 4) was dramatically reduced by the silencing of SHP-1 (bar 8) to levels comparable to those reached by knocking down the phosphatase in the absence of nintedanib (bar 6). At the molecular level, nintedanib induced caspase-dependent apoptosis, as witnessed by the cleavage of caspase-3 and PARP (Figure 4B, right-hand panel, lane 4), the latter event being negligible when silencing SHP-1, irrespective of the presence of nintedanib (lanes 6 and 8). These findings were consistent with our previous data demonstrating that SHP-1 knockdown brings about caspase-independent apoptosis by targeting the plasma membrane pool of the phosphatase, which is catalytically active and orchestrates anti-apoptotic signals.21,23 The pharmacological inhibition of SHP-1 mediated by PTP-I-I provided results overlapping those obtained by using SHP-1 short interfering RNA (Figure 4C). These observations led us to hypothesize that the dephosphorylation of specific SHP-1 substrates induced apoptosis following treatment with nintedanib in CLL cells. We, therefore, focused on two factors that we had previously explored and the activity of which we found to be inhibited in CLL via phosphorylation by the aberrant 1406

cytosolic form of Lyn, procaspase-8 (procasp8)29 and PP2A.22 After incubating freshly isolated CLL cells with increasing concentrations of nintedanib, we performed western blot analysis with antibodies directed against the phosphorylated form of specific inhibitory residues of these two proteins, Y380 of procasp8 and Y307 of PP2Ac. Both tyrosines were phosphorylated when SHP-1 was not silenced and nintedanib was not added to the incubation medium (Figure 5A,C, left-hand panels, 0 mM), the level of phosphorylation gradually declining as nintedanib concentration increased (1-20 mM). Moreover, total lysates from the same samples were assayed for the activity of the two enzymes in vitro (see the Online Supplementary Data for details on the commercial kits employed), which allowed us to conclude that the activation of both enzymes depended on the action of nintedanib exerted on SHP-1 (Figure 5B,D, left-hand panels). Importantly, dephosphorylation and activation of these factors were blocked, albeit in the presence of nintedanib, by silencing SHP-1 in CLL cells (Figure 5, right-hand panels), indicating that procasp8 and PP2Ac were substrates for SHP-1 and effectors of a SHP-1- dependent pro-apoptotic pathway.

Apoptosis of chronic lymphocytic leukemia cells can be induced by indirect activators of SHP-1 The data collected thus far confirmed that nintedanib could trigger SHP-1 by circumventing the phosphorylation at the inhibitory residue S591. Interestingly, we recently demonstrated that pS591 could be dephosphorylated by PP2A, the activity of which was impaired by phosphorylation at Y307 as well as by the interaction with its endogenous inhibitor SET.22 In this scenario, the restoration of PP2A activity by a fingolimod analog devoid of immunosuppressive action, the so-called MP07-66 [(2,2diethoxyethyl{[4-(hexyloxy)phenyl]methyl})amine], and the subsequent dephosphorylation of PP2A substrates, was shown to trigger apoptosis.22 These findings led us to conjecture that MP07-66 could be used as an indirect activator of SHP-1 via PP2A, and exploited to potentiate the action of nintedanib on SHP-1 itself to reinforce the apoptotic response of CLL cells. As shown in Figure 6A (lefthand panel), incubation with increasing concentrations of MP07-66 (0-24 mM) for 24 and 48 h brought about a marked level of apoptosis in CLL cells, as determined by annexin Vâ&#x20AC;&#x201C;propidium iodide flow cytometry. As in the case of nintedanib (Figure 4), MP07-66 evoked apoptosis in a caspase-dependent manner, as shown by the use of the pan-caspase inhibitor zVADfmk44 (Figure 6A, Online Supplementary Figure S9A,B), with no evidence of necroptosis, the number of viable cells remaining unchanged following pre-incubation with necrostatin-1 prior to MP0766 treatment, or autophagy, as indicated by the stable level of expression of p62/SQSTM145 (Online Supplementary Figure S9C) for the duration of the experiments in the presence of PP2A activator. Under these conditions, the survival of normal B cells was not altered (Online Supplementary Figure S10). To evaluate whether SHP-1 was implicated in these events, aliquots from the cytosol for each time interval were immunoblotted with anti-pS591SHP-1 antibody and tested for phosphatase activity in the presence of [32P]-Band 3 as a substrate. The elevation of activity of SHP-1 was concomitant with the dephosphorylation thereof as the concentration of MP07-66 increased (Figure 6B,C) and the results were highly similar for all the samples tested (Online Supplementary Figure S11A,B). haematologica | 2017; 102(8)

SHP-1/PP2A as a new druggable axis in CLL

Figure 4. Effect of nintedanib on the viability of chronic lymphocytic leukemia cells. Apoptosis of CLL cells from ten patients belonging to the various clinical and biological subtypes cultured (A) in the presence of increasing concentrations of nintedanib for 24 and 48 and assessed h by annexin V–propidium iodide flow cytometry, (B) in the absence (bars 1-2, 5-6) or presence (bars 3-4, 7-8) of 15 mM nintedanib for 0 or 24 h after transfection with either scrambled siRNA or SHP-1-siRNA (white and gray bars, respectively, left-hand panel) analyzed as in (A), and (C) pre-incubated without (1-4) or with 25 mM PTP-I-I (5-8) for 1 h (white and gray bars, respectively, left-hand panel) and then cultured in the absence (bars 1-2, 5-6) or presence (bars 3-4, 7-8) of 15 mM nintedanib for 0 or 24 h analyzed as in (A). Expression of SHP-1 in CLL cells of patient #34 after transfection with either scrambled siRNA or SHP-1-siRNA by western blotting analysis and pooled densitometric analysis (arbitrary units) of the Wb bands of ten patients is represented by the histograms in the inset of (B). Data are mean percentage of early and late apoptosis ± SD from three separate experiments performed in triplicate (left-hand panel, *P≤0.01). Western blotting analysis with anti-caspase 3 and anti-PARP antibodies monitored caspase-dependent apoptosis; anti-b-actin antibody was used as a loading control (right-hand panels). Wb: western blot.

haematologica | 2017; 102(8)

1407

E. Tibaldi et al.

Further evidence for the role of SHP-1 in mediating apoptosis of CLL cells upon treatment with MP07-66 was provided by western blotting with antibodies against pY380procasp8 and pY307-PP2Ac, which revealed the dephosphorylation of the inhibitory residues (Figure 6D, lane 6) at 6 h. As expected, this event was blocked by the specific SHP-1 inhibitor PTP-I-I (Figure 6D, lane 8) as well as by okadaic acid, a phosphatase inhibitor that is highly selective toward PP2A in the low nanomolar range49 (lane 7), which confirmed that the activation of PP2A drove the dephosphorylation of SHP-1 S591, thereby promoting global tyrosine dephosphorylation and cell death. Again, these results were similar for all the samples tested, as reported in the pooled densitometric analysis shown in Online Supplementary Figure S11C. All these observations were consistent with the hypothesis that SHP-1 and PP2A form a signaling axis wherein the single phosphatases, when stimulated, can

activate one another, and that such a process can be amplified by a combination of molecules activating both simultaneously.

MP07-66 potentiates the pro-apoptotic effect of nintedanib Since our data suggest that the activation of either PP2A or SHP-1 triggered by specific small molecules caused stimulation of each other’s activity, thereby evoking a positive feedback signaling loop promoting apoptosis, we wondered whether the combination of nintedanib and MP07-66 could result in a more robust apoptotic response of CLL cells. We, therefore, incubated freshly isolated CLL cells with 15 mM nintedanib and 8 mM MP07-66 at different times and monitored apoptosis by annexin V–propidium iodide flow cytometry (Figure 7A). Nintedanib proved moderately effective at inducing apoptosis of CLL cells after 6 and 12 h of incubation, its efficacy being large-

1408

Figure 5. Effect of nintedanib on the phosphorylation state and activity of procaspase 8 and PP2Ac. (A) Western blotting (Wb) analysis with anti-pY380-procasp8 antibody, stripped and reprobed with anti-procasp8 antibody and with anti-b-actin antibody as a loading control, of total cell lysates of CLL cells from ten patients belonging to the various clinical and biological subtypes transfected by nucleofection with either scrambled siRNA (left-hand panels) or SHP-1-siRNA (right-hand panels) and cultured in the presence of increasing concentrations of nintedanib for 1 h. Densitometric analysis (arbitrary units) of the pY380-procasp8 and pro-casp8 bands is represented as histograms. (B) In vitro casp8 activity on cell lysates from CLL cells treated as in (A) as described in the Methods section. Compared with the effect of nintedanib, changes due to siRNA were statistically significant (*P≤0.01). (C) Wb analysis with anti-pY307-PP2Ac antibody, stripped and reprobed with anti-PP2Ac antibody and with anti-b-actin antibody as a loading control, of total cell lysates of CLL cells from ten patients treated as in (A). Densitometric analysis (arbitrary units) of the pY307-PP2Ac and PP2Ac bands is represented as histograms. Data are mean ± SD from four experiments performed in triplicate. (D) In vitro PP2A activity on cell lysates from CLL cells treated as in (A) by using a specific phosphopeptide as a substrate, as described in the Methods section of the Online Supplementary Data. Compared with the effect of nintedanib, changes due to siRNA were statistically significant (*P≤0.01).

haematologica | 2017; 102(8)

SHP-1/PP2A as a new druggable axis in CLL

ly improved by the concomitant presence of MP07-66, which itself exhibited a pro-apoptotic activity overlapping that of nintedanib when used as a single agent. Furthermore, the variability exhibited by the different subsets of patients as spontaneous apoptosis was attenuated after treatment with the combination of the two agents (Online Supplementary Figure S12). Similarly, co-cultures of CLL cells with bone marrow mesenchymal stromal cells were treated as above with overlapping results (Online Supplementary Figure S13). Moreover, to explore how PP2A activation took part in the apoptotic process induced by MP07-66, CLL cells were incubated with 15 mM nintedanib for 6 h in the absence or presence of okadaic acid. As expected, 5 nM okadaic acid drastically reduced the rate of apoptosis of CLL cells treated with MP07-66 or the combination with nintedanib, but only to a lesser extent with nintedanib alone (Figure 7B, left-hand panel). Moreover, the PP2A activity assay performed using a commercial PP2A assay kit on cell lysates of CLL cells treated as above showed a trend similar to that observed for the apoptotic rate (right-hand panel). Overall, these data corroborate the hypothesis that the inhibition of PP2A is cen-

tral to CLL cell viability and that its activation is facilitated by the supportive action of SHP-1, as demonstrated by the effect produced by the simultaneous use of the respective activators.

Discussion In this study, we show that nintedanib induces caspasedependent apoptosis in CLL cells via dephosphorylation of pro-apoptotic key players such as procasp8 and PP2A by directly activating the cytosolic pool of the tyrosine phosphatase SHP-1. SHP-1 is a tyrosine phosphatase that negatively regulates signaling in cells of hematopoietic lineage, having a key role in modulating the response to antigens and contributing to the development of tolerance to self-antigens. In CLL, SHP-1 undergoes multiple regulatory mechanisms leading to both spatial and functional segregation, which seem to be crucial in supporting the cancer phenotype.23 Phosphorylation of different residues, especially at the Cterminus, significantly changes the activation status and

B Figure 6. Effect of MP 07-66 on the chronic lymphocytic leukemia cell survival. (A) Apoptosis of CLL cells from ten patients belonging to the various clinical and biological subtypes cultured in the presence of increasing concentrations of MP07-66 for 24 and 48 h analyzed by annexin V– propidium iodide flow cytometry (left-hand panel). Data are mean percentages of early and late apoptosis ± SD from three separate experiments performed in triplicate (*P≤0.01). Western blotting (Wb) analysis of total cell lysate of CLL cells with anti-caspase 3 and anti-PARP antibodies monitored caspase-dependent apoptosis; anti-β-actin antibody was used as a loading control (right-hand panel). (B) Tyrosine phosphatase activity of SHP-1 immunoprecipitated from the cytosol of CLL cells of patient #28 cultured in the presence of increasing concentrations of MP0766 for 1 h and measured as [32P] released from in vitro [32P]-Band 3. Data are expressed as mean ± SD from one experiment performed in triplicate (*P≤ 0.01). (C) Expression of SHP-1 and phosphorylation state of pS591 of CLL cells of patient #28 cultured in the presence of increasing concentrations of MP07-66. Data are expressed as mean ± SD from one experiment performed in triplicate (*P≤0.01). (D) Expression and phosphorylation state of procasp8 and PP2Ac in the CLL cells of patient #28 cultured in the absence (lanes 1 and 5) and presence of 8 mM MP07-66 supplemented with 5 nM okadaic acid (OA) (lanes 3 and 7) or 25 mM PTP-I-I (lanes 4 and 8) for 0 and 6 h.

haematologica | 2017; 102(8)

1409

E. Tibaldi et al.

localization of SHP-1, phospho-Y536 being typical of the activated pool bound to the plasma membrane co-receptor CD5, and phospho-S591 characterizing the inhibited pool of SHP-1 in the cytosol. This latter form appears to be one of the factors that sustains the aberrant Lyn-dependent tyrosine phosphorylation of countless proteins in the cytosol of CLL cells, which is ultimately key to the antiapoptotic signaling network in this disease.22,27,29 Here, we demonstrate that nintedanib, a small molecule known to act as a triple angiokinase inhibitor in the low nanomolar range,39,40 activates the cytosolic fraction of SHP-1 by circumventing pS591-dependent inhibition, as recently

described in another cancer model.41 Significantly, in addition to leaving SHP-1 at the plasma membrane unaffected, nintedanib only marginally modifies tyrosine kinase activities in CLL cells even at micromolar concentrations. On the other hand, a dramatic drop in tyrosine phosphorylation occurs as a result of SHP-1 activation in the cytosol with consequent caspase-dependent apoptosis, suggesting that the massive tyrosine phosphorylation in CLL cells directly affects the function of factors that counteract the oncogenic machinery. Notably, although genetic or pharmacological inhibition of SHP-1 can prevent the caspasedependent apoptosis evoked by nintedanib, again sup-

D Figure 7. Effect of the combined action of nintedanib and MP07-66 on chronic lymphocytic leukemia cell survival. (A) Apoptosis of CLL cells from ten patients belonging to the various clinical and biological subtypes cultured in the absence and presence of 15 mM nintedanib, 8 mM MP0766 or both over time analyzed by annexin V–propidium iodide flow cytometry. (B) Apoptosis of CLL cells from ten patients as described in (A) supplemented without or with 5 nM okadaic acid (OA) (right panels) for 6 h. Analysis by annexin V– propidium iodide flow cytometry is expressed as mean percentages of early and late apoptosis ± SD from three separate experiments performed in triplicate (*P≤0.01). (C) PP2A phosphatase activity performed on the total lysates of CLL cells treated as in (B). (D) Working model of the positive feedback signling loop triggered by the combination of nintedanib and MP07-66, which counters the high level of cytosolic tyrosine phosphorylation, the crucial factor sustaining the oncogenic machinery in CLL cells, mediated by the aberrant form of HSP90-bound Lyn.

1410

haematologica | 2017; 102(8)

SHP-1/PP2A as a new druggable axis in CLL

porting the hypothesis that the action of this drug is mediated by SHP-1, such treatments can still induce caspaseindependent apoptosis, which is in line with the role of the CD5-bound form of SHP-1 as a pro-survival agent in CLL.23 Our findings indicate that the distribution of the two forms of SHP-1 is central to their differentiated function, at the plasma membrane in an active form taking part in a signalosome that orchestrates survival signals, and in the cytosol in an inhibited conformation. This condition renders SHP-1 unable to dephosphorylate cytosolic Lyn targets endowed with pro-apoptotic potential, such as procasp8 and PP2A. Procasp8 occurs as an inactive homodimer in CLL cells, the trigger for dimerization being the phosphorylation of Y380 mediated by Lyn.29 Here, nintedanib, via direct activation of SHP-1 in the cytosol, induces dephosphorylation, autocatalysis and activation of procasp8, which explains the caspase-dependent apoptosis observed. As to PP2A, Lyn-mediated phosphorylation at Y307 of the catalytic subunit stabilizes its interaction with its physiological inhibitor SET, hampering the activity of the phosphatase.22 This results in the persistent serine/threonine phosphorylation of PP2A substrates, including the cytosolic pool of SHP-1, and propagates prosurvival and anti-apoptotic signals.22 Conversely, nintedanib-activated SHP-1 dephosphorylates PP2A, facilitating the disruption of the PP2A/SET complex, with acti-

References 1. Johnson LN. The regulation of protein phosphorylation. Biochem Soc Trans. 2009;37(4): 627-641. 2. Lim WA, Pawson T. Phosphotyrosine signalling: evolving a new cellular communication system. Cell. 2010;142(5):661-667. 3. Tonks NK. Protein tyrosine phosphatases. from house keeping enzymes to master regulators of signal transduction. FEBS J. 280(2):346-378. 4. Cohen P. Protein kinases - the major drug targets of the twenty-first century? Nat Rev Drug Discov. 2002;1(4):309-315. 5. Lahiry P, Torkamani A, Schork NJ, Hegele RA. Kinase mutations in human diseases: interpreting genotype-phenotype relationships. Nat Rev Genet. 2010;11(1):60-74. 6. Julien SG, Dube N, Hardy S, Tremblay ML. Inside the human cancer tyrosine phosphatome. Nat Rev Cancer. 2011;1(1):35-49. 7. Khatri A, Wang J, Pendergast AM. Multifunctional Abl kinases in health and disease. J Cell Sci. 2016;129(1): 9-16. 8. Dighiero G, Hamblin TJ. Chronic lymphocytic leukaemia. Lancet. 2008;371(9617): 1017-1029. 9. Burger JA, Chiorazzi N. B cell receptor signalling in chronic lymphocytic leukemia. Trends Immunol. 2013;34(12):592-601. 10. Hallek M. Chronic lymphocytic leukemia: 2015 Update on diagnosis, risk stratification, and treatment. Am J Hematol. 2015;90(5):446-460. 11. Zhang S, Kipps TJ. The pathogenesis of chronic lymphocytic leukemia. Annu Rev Pathol. 2014;9:103-118. 12. Del Gaizo Moore V, Brown JR, Certo M,

haematologica | 2017; 102(8)

13.

14.

15.

16.

17.

18.

19.

vation of PP2A itself and dephosphorylation of SHP-1 (Figure 7D), eventually triggering the apoptotic response. This latter event is magnified by the combination of nintedanib with the FTY720 analog MP07-66 in that this latter compound interferes directly with the interaction between PP2A and SET,22 further aiding in the reactivation of PP2A. In this scenario, nintedanib and MP07-66, direct activators of SHP-1 and PP2A, respectively, appear to initiate a positive feedback signaling loop which opposes Lynmediated oncogenic signaling, thus promoting the dephosphorylation of crucial players in the deranged signaling architecture of CLL, switching off anti-apoptotic signals and unleashing cell death. In conclusion, our findings indicate that phosphatase activators may represent a new weapon against this form of leukemia, especially in the light of the heterogeneity and the unavoidable progression of the disease as well as the resistance to the front-line drugs currently in use, not to mention the adverse effects recently reported for the most promising second-line drugs.50 Acknowledgments This work was supported by grants from A.I.R.C. (Milan) to GS and LT by MIUR with funds to LT; by Regione Veneto on Chronic Lymphocytic Leukemia; by AIRC Regional Project with Fondazione CARIPARO and CARIVERONA and by FIRB (Rome).

Love TM, Novina CD, Letai A. Chronic lymphocytic leukemia requires BCL2 to sequester prodeath BIM, explaining sensitivity to BCL2 antagonist ABT-737. J Clin Invest. 2007;117(1):112-121. Ten Hacken E, Burger JA. Microenvironment interactions and B-cell receptor signalling in chronic lymphocytic leukemia: implications for disease pathogenesis and treatment. Biochim Biophys Acta. 2016;1863(3):401-413. Zou ZJ, Fan L, Wang L, et al. miR-26a and miR-214 down-regulate expression of the PTEN gene in chronic lymphocytic leukemia, but not PTEN mutation or promoter methylation. Oncotarget. 2015;6(2): 1276-1285. Motiwala T, Majumder S, Kutay H, et al. Methylation and silencing of protein tyrosine phosphatase receptor type O in chronic lymphocytic leukemia. Clin Cancer Res. 2007;13(11):3174-3181. Suljagic M, Laurenti L, Tarnani M, Alam M, Malek SN, Efremov DG. Reduced expression of the tumor suppressor PHLPP1 enhances the antiapoptotic B-cell receptor signal in chronic lymphocytic leukemia Bcells. Leukemia. 2010;24(12):2063-2071. O'Hayre M, Niederst M, Fecteau JF, et al. Mechanisms and consequences of the loss of PHLPP1 phosphatase in chronic lymphocytic leukemia (CLL). Leukemia. 2012;26(7):1689-1692. Cui B, Chen L, Zhang S, et al. MicroRNA155 influences B-cell receptor signalling and associates with aggressive disease in chronic lymphocytic leukemia. Blood. 2014;124(4):546-554. Negro R, Gobessi S, Longo PG, et al. Overexpression of the autoimmunity-asso-

20.

21.

22.

23.

24.

25.

26.

27.

ciated phosphatase PTPN22 promotes survival of antigen-stimulated CLL cells by selectively activating AKT. Blood. 2012;119 (26):6278-6287. Perrotti D, Neviani P. Protein phosphatase 2A (PP2A), a druggable tumor suppressor in Ph1(+) leukemias. Cancer Metastasis Rev. 2008;27(2):159-168. Tsui FW, Martin A, Wang J, Tsui HW. Investigations into the regulation and function of the SH2 domain-containing proteintyrosine phosphatase, SHP-1. Immunol Res. 2006;35(1-2):127-136. Zonta F, Pagano MA, Trentin L, et al. Lyn sustains oncogenic signalling in chronic lymphocytic leukemia by strengthening SETmediated inhibition of PP2A. Blood. 2015;125(24):3747-3755. Tibaldi E, Brunati AM, Zonta F, et al. Lynmediated SHP-1 recruitment to CD5 contributes to resistance to apoptosis of B-cell chronic lymphocytic leukemia cells. Leukemia. 2011;25(11):1768-1781. Pierce SK, Liu W. The tipping points in the initiation of B cell signalling: how small changes make big differences. Nat Rev Immunol. 2010;10(11):767-777. DeFranco AL, Chan VW, Lowell CA. Positive and negative roles of the tyrosine kinase Lyn in B cell function. Semin Immunol. 1998; 10(4):299â&#x20AC;&#x201C;307. Xu Y, Harder KW, Huntington ND, Hibbs ML, Tarlinton DM. Lyn tyrosine kinase: accentuating the positive and the negative. Immunity. 2005;22(1):9â&#x20AC;&#x201C;18. Contri A, Brunati AM, Trentin L, et al. Chronic lymphocytic leukemia B cells contain anomalous Lyn tyrosine kinase, a putative contribution to defective apoptosis. J Clin Invest. 2005;115(2):369-378.

1411

E. Tibaldi et al. 28. Trentin L, Frasson M, Donella-Deana A, et al. Geldanamycin-induced Lyn dissociation from aberrant Hsp90-stabilized cytosolic complex is an early event in apoptotic mechanisms in B-chronic lymphocytic leukemia. Blood. 2008;112(12):4665-4674. 29. Zonta F, Pagano MA, Trentin L, et al. Lynmediated procaspase 8 dimerization blocks apoptotic signalling in B-cell chronic lymphocytic leukemia. Blood. 2014;123(6): 875-883. 30. Christensen DJ, Chen Y, Oddo J, et al. SET oncoprotein overexpression in B-cell chronic lymphocytic leukemia and non-Hodgkin lymphoma: a predictor of aggressive disease and a new treatment target. Blood. 2011;118(15):4150-4158. 31. Oaks J, Ogretmen B. Regulation of PP2A by sphingolipid metabolism and signalling. Front Oncol. 2015;4:388. 32. Saddoughi SA, Gencer S, Peterson YK, et al. Sphingosine analogue drug FTY720 targets I2PP2A/SET and mediates lung tumour suppression via activation of PP2A-RIPK1dependent necroptosis. EMBO Mol Med. 2013(1);5:105-121. 33. Woyach JA, Furman RR, Liu TM, et al. Resistance mechanisms for the Bruton's tyrosine kinase inhibitor ibrutinib. N Engl J Med. 2014;370(24):2286-2294. 34. 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. 35. Burger JA, Tedeschi A, Barr PM, et al. Ibrutinib as initial therapy for patients with

1412

36.

37. 38.

39.

40.

41.

42.

43.

chronic lymphocytic leukemia. N Engl J Med. 2015;373(25):2425-2437. Roberts AW, Davids MS, Pagel JM, et al. Targeting BCL2 with venetoclax in relapsed chronic lymphocytic leukemia. N Engl J Med. 2016;374(4):311-322. McCormack PL. Nintedanib: first global approval. Drugs. 2015;75(1):129-139. Tibaldi E, Zonta F, Bordin L, et al. The tyrosine phosphatase SHP-1 inhibits proliferation of activated hepatic stellate cells by impairing PDGF receptor signalling. Biochim Biophys Acta. 2014;1843(2):288-298. Hilberg F, Roth GJ, Krssak M, et al. BIBF 1120: triple angiokinase inhibitor with sustained receptor blockade and good antitumor efficacy. Cancer Res. 2008;68(12):47744782. Roth GJ, Heckel A, Colbatzky F, et al. Design, synthesis, and evaluation of indolinones as triple angiokinase inhibitors and the discovery of a highly specific 6methoxycarbonyl-substituted indolinone (BIBF 1120). J Med Chem. 2009;52(14):44664480. Tai WT, Shiau CW, Li YS, et al. Nintedanib (BIBF-1120) inhibits hepatocellular carcinoma growth independent of angiokinase activity. J Hepatol. 2014;61(1):89-97. Gringeri E, Carraro A, Tibaldi E, et al. Lynmediated mitochondrial tyrosine phosphorylation is required to preserve mitochondrial integrity in early liver regeneration. Biochem J. 2009;425(2):401-412. Tibaldi E, Venerando A, Zonta F, et al. Interaction between the SH3 domain of Src

44.

45.

46.

47.

48.

49.

50.

family kinases and the proline-rich motif of HTLV-1 p13: a novel mechanism underlying delivery of Src family kinases to mitochondria. Biochem J. 2011;439(3):505-516. Pietkiewicz S, Schmidt JH, Lavrik IN. Quantification of apoptosis and necroptosis at the single cell level by a combination of imaging flow cytometry with classical annexin V/propidium iodide staining. J Immunol Methods. 2015;423:99-103. Degterev A, Hitomi J, Germscheid M, et al. Identification of RIP1 kinase as a specific cellular target of necrostatins. Nat Chem Biol. 2008;4(5):313-321. Katsuragi Y, Ichimura Y, Komatsu M. p62/SQSTM1 functions as a signaling hub and an autophagy adaptor. FEBS J. 2015;282(24):4672-4678. Calimeri T, Ferreri AJ. m-TOR inhibitors and their potential role in haematological malignancies. Br J Haematol. 2017;177(5): 684-702. Kurtova AV, Balakrishnan K, Chen R, et al. Diverse marrow stromal cells protect CLL cells from spontaneous and drug-induced apoptosis: development of a reliable and reproducible system to assess stromal cell adhesion-mediated drug resistance. Blood. 2009;114(20):4441-4450. Swingle M, Ni L, Honkanen RE. Small-molecule inhibitors of ser/thr protein phosphatases: specificity, use and common forms of abuse. Methods Mol Biol. 2007;365:2338. Brown JR. Idelalisib has CLL on the run! Blood. 2015;126(25):2656-2657.

haematologica | 2017; 102(8)

ARTICLE

Non-Hodgkin Lymphoma

Prognostic relevance of CD163 and CD8 combined with EZH2 and gain of chromosome 18 in follicular lymphoma: a study by the Lunenburg Lymphoma Biomarker Consortium Wendy B.C. Stevens,1* Matias Mendeville,2* Robert Redd,3 Andrew J. Clear,4 Reno Bladergroen,2 Maria Calaminici,4 Andreas Rosenwald,5 Eva Hoster,6 Wolfgang Hiddemann,6 Philippe Gaulard,7 Luc Xerri,8 Gilles Salles,9 Wolfram Klapper,10 Michael Pfreundschuh,11 Andrew Jack,12 Randy D. Gascoyne,13 Yasodha Natkunam,14 Ranjana Advani,15 Eva Kimby,16 Birgitta Sander,17 Laurie H. Sehn,13 Anton Hagenbeek,18 John Raemaekers,1 John Gribben,4 Marie José Kersten,18 Bauke Ylstra,2 Edie Weller3 and Daphne de Jong2

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1413-1423

*WBCS and MM contributed equally to this work

1 Department of Hematology, Radboudumc, Nijmegen, the Netherlands; 2Department of Pathology, VU University Medical Center, Amsterdam, the Netherlands; 3Department of Biostatistics and Computational Biology, Dana-Farber Cancer Institute, Boston, MA, USA; 4 Centre for Haemato-Oncology, Barts Cancer Institute, University of London, UK; 5Institute of Pathology, Comprehensive Cancer Center Mainfranken, University of Würzburg, Germany; 6Department of Medicine III, University Hospital Grosshadern, Munich, Germany; 7 Department of Pathology and Inserm U955, Hôpital Henri Mondor, University Paris-Est, Créteil, France; 8Département de Biopathologie, Institut Paoli-Calmettes, Marseille, France; 9 Service d'Hématologie,Hospices Civils de Lyon & Université Claude Bernard Lyon-1, UMR CNRS 5239, France; 10Institute of Pathology, University of Schleswig-Holstein, Kiel, Germany; 11Medical Clinic I, Saarland University Hospital, Homburg, Germany; 12 Haematological Malignancy Diagnostic Service, St James's University Hospital, Leeds, UK; 13 Department of Pathology & Medical Oncology, Centre for Lymphoid Cancer, British Columbia Cancer Agency, University of British Columbia, Vancouver, Canada; 14Department of Pathology, Stanford University School of Medicine, CA, USA; 15Department of Hematology, Stanford University School of Medicine, CA, USA; 16Department of Medicine, Division of Hematology, Karolinska Institute, Stockholm, Sweden; 17Department of Laboratory Medicine, Division of Pathology, Karolinska Institute and Karolinska University Hospital, Stockholm, Sweden and 18Department of Hematology, Academic Medical Center, Amsterdam, the Netherlands

ABSTRACT

n follicular lymphoma, studies addressing the prognostic value of microenvironment-related immunohistochemical markers and tumor cell-related genetic markers have yielded conflicting results, precluding implementation in practice. Therefore, the Lunenburg Lymphoma Biomarker Consortium performed a validation study evaluating published markers. To maximize sensitivity, an end of spectrum design was applied for 122 uniformly immunochemotherapy-treated follicular lymphoma patients retrieved from international trials and registries. The criteria were: early failure, progression or lymphoma-related death <2 years versus long remission, response duration of >5 years. Immunohistochemical staining for T cells and macrophages was performed on tissue microarrays from initial biopsies and scored with a validated computer-assisted protocol. Shallow whole-genome and deep targeted sequencing was performed on the same samples. The 96/122 cases with complete molecular and immunohistochemical data were included in the analysis. EZH2 wild-type (P=0.006), gain of chromosome 18 (P=0.002), low percentages of CD8+ cells (P=0.011) and CD163+ areas (P=0.038) were associated with early failure. No significant differences in other markers were observed, thereby refuting previous claims of their prognostic significance. Using an optimized study design, this Lunenburg Lymphoma Biomarker Consortium study substantiates wild-type EZH2 status, gain of chromosome 18, low percentages of CD8+ cells and CD163+ area as predictors of early failure to immunochemotherapy in follicular lymphoma treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP [-like]), while refuting the prognostic impact of various other markers. haematologica | 2017; 102(8)

Correspondence: wendy.stevens@radboudumc.nl

Received: January 31, 2017. Accepted: April 11, 2017. Pre-published: April 14, 2017. doi:10.3324/haematol.2017.165415 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1413 ©2017 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.

1413

W.B.C. Stevens et al.

Introduction The disease course of follicular lymphoma (FL) is characterized by multiple relapses with variable remission durations, which tend to get shorter after each line of treatment.1-7 Approximately 15% of patients die within the first few years, largely due to histological transformation or refractory disease. In contrast, the majority of patients show prolonged survival without relapse and a substantial number of patients never require treatment. Currently, clinical factors captured in the Follicular Lymphoma International Prognostic Index (FLIPI)8,9 and the adjusted FLIPI210 are the primary prognostic tools utilized to predict disease progression. FLIPI is based on easily available clinical data designed to offer an accurate yet simple prognostic index. However, the biological behavior of FL is likely determined by a more complex interaction between tumor genetics, microenvironment and patient characteristics. Several gene expression profiling studies have underlined the major influence of characteristics of the nonmalignant tumor microenvironment on prognosis in FL.11,12 However, evaluation by immunohistochemistry (IHC) of T-cell and macrophage populations for their prognostic translational impact have produced valuable yet conflicting results.13-25 Contradictory results are likely caused by multiple factors, including variable patient selection, heterogeneity of treatments across cohorts,25 insufficient statistical power due to underrepresentation of poor outcome patients, technical issues in IHC staining and inter-observer variability in scoring. Previous IHC-based validation studies of microenvironment cell populations by the Lunenburg Lymphoma Biomarker Consortium (LLBC), in both diffuse large B-cell Lymphoma (DLBCL) and FL, have highlighted the poor reproducibility of manual scoring even by experienced pathologists, and advocated computer-assisted scoring as being the more reliable technique.26,27 Recently, the mutational spectrum of FL tumor cells and their prognostic value have been reported.28-30 Most notably, Pastore et al. demonstrated the value of combining mutation status with clinical FLIPI and performance status to improve upon the prognostic value of FLIPI alone (45-55%) to 64-72% for M7-FLIPI, with a comparable negative prognostic value.31 This method may also be valuable for very high-risk FL cohorts, as reported by the same group.32 The objective of the study herein by the LLBC consortium was to critically assess, in a homogeneously treated patient cohort, whether the previously implicated microenvironmental and molecular markers of the tumor have clinically relevant prognostic value. We hypothesized that the microenvironmental and molecular markers would be most prominent when we compared tissue samples of patients with an extremely poor prognosis (i.e., progression or death within 2 years, a well-established criterion for poor prognosis in FL)33 with those with a very favorable prognosis (a response to first-line treatment lasting >5 years). The LLBC gene panel incorporated the molecular markers which were frequently mutated, and were at that time published, as well as markers that were rarely mutated such as FAS and MYD88,28-30 since with our study design we hypothesized the ability to also reveal the prognostic value of rare mutations. To avoid interlaboratory technical variations and interpretation, all assays were uniformly processed in a single 1414

laboratory and IHC was scored using computer-assisted technology based on a previous LLBC FL validation study,27 and molecular analysis was performed using established next-generation sequencing (NGS) procedures for mutation and copy number analyses.

Methods Patient selection for an end of spectrum design Tumor specimens of patients with FL, histologic grades 1, 2 and 3A were retrieved from the randomized Lymphoma Study Association (LYSA) FL2000 study,4,6 the German Low-Grade Lymphoma Study Group (GLSG) GLSG2000 study,3 and the St Bartholomew’s Hospital Registry, London, UK. The patients required treatment and the inclusion criteria in the trials were comparable. All patients were treated with R-CHOP or R-CHOPlike regimens, with or without interferon-α (IFN) maintenance for 2 years. Patients were selected upon the following criteria for an end of spectrum design: (1) early failure (EF), defined as no remission or progression, or lymphoma-related death within 2 years after start of first-line treatment, or (2) long remission (LR), defined as a complete or partial remission lasting >5 years after starting first-line treatment. Patients that fell in between these criteria were not included in this study. The availability of complete clinical information at diagnosis, follow-up data until relapse, progression or at least 5 years post-treatment if the patient was still in remission, and availability of formalin-fixed paraffin-embedded (FFPE) diagnostic biopsy samples were prerequisites for inclusion. Detailed clinical information on demographic parameters, staging procedures, treatment regimens and outcome were collected by the involved data centers (LYSA, GLSG and St Bartholomew’s Hospital).

Microenvironment analysis on tissue microarrays using immunohistochemistry and automated image analysis scoring Tissue microarrays (TMAs) were constructed centrally according to LLBC validated protocols27 at the Department of Pathology, Würzburg, Germany, from the biopsy part identified by the pathologist (AR), using duplicate cores of 1mm diameter. Three µm slides were stained for CD3, CD4, CD8, CD68, CD163, FOXP3, PD1, and P53 (Online Supplementary Table S1) according to standard procedures at the Bartholomew’s Pathology Research Laboratory, London, UK. A computerized system with automated scanning microscope and computerized image analysis (Ariol SL-8, Leica Microsystems, Wetzlar, Germany) was used for scoring as described in the LLBC validation study.27 Macrophages and all T-cell populations were scored for the whole core, and in the intrafollicular and interfollicular areas separately, as described by Wahlin et al.22 Color and shape class defined positive and negative nucleated cells (T-cell classes), or positive and negative areas (macrophage classes).16 Cores with less than 50% scorable core surface (non-representative areas or tissue artifacts) were excluded and average scores of duplicates were used when available. Using the Ariol software algorithm, CD3, CD4, CD8, PD1, FOXP3 and p53 positive nucleated cells were scored as the percentage of all nucleated cells. For CD68 and CD163 the positive area versus the whole area was scored to accommodate the large size and long cytoplasmic extensions of macrophages as an optimal determination of cell numbers. The typical perifollicular infiltration pattern of FOXP3 was scored manually by 3 observers,27 and was considered positive if at least 2 out of the 3 pathologists (DdJ, AR and MC) identified and counted at least one rim of haematologica | 2017; 102(8)

CD8, CD163, EZH2, chromosome 18, FL prognostic biomarkers densely packed positive cells at the periphery of a follicle.14 All histopathological assessments were carried out without insight into the patient’s clinical data and outcome.

Gene mutation and copy number analysis using NGS DNA was extracted from FFPE cores with a QIAamp DNA FFPE Tissue Kit (QIAGEN, Hilden, Germany) and quantified using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Carlsbad, CA, USA). 250ng DNA from each patient sample was sheared on a Covaris S2 (Covaris Inc., Woburn, MA, USA), with settings adjusted to DNA from FFPE tissue.34 NGS libraries were prepared using KAPA Library Preparation kits (KAPA Biosystems, Inc., Wilmington, MA, USA). In short, uniquely 8-bp indexed adapters (Roche NimbleGen, Madison, WI, USA.) were ligated to the FFPEextracted DNA, followed by size selection of fragments in the range of 150 to 400bp. One aliquot of this library was subjected to shallow wholegenome sequencing (WGS) for genome-wide copy number analysis,34 and another aliquot was subjected to hybrid capture target enrichment (Roche NimbleGen, Madison, WI, USA) for mutation analysis. Eight libraries were equimolarly pooled per capture. The hybrid capture panel covers 122 exons (~50.000 base pairs) of 11 frequently mutated genes (KMT2D, CREBBP, MEF2B, EZH2, EP300, BCL2, FAS, TNFRSF14, CARD11, TNFAIP3 and MYD88) in the FL-LLBC-NGS target enrichment panel (Follicular Lunenburg Lymphoma Biomarker Consortium NGS-panel, Online Supplementary Table S2). All libraries were sequenced on a HiSeq 2000 (Illumina, San Diego, CA, USA); 50bp single-end for shallow WGS and 125bp paired-end for mutation analysis. All sequence lanes were multiplexed with up to 24 barcoded sample libraries. Shallow WGS data was analyzed using the Bioconductor package QDNAseq (v1.5.1).34 For gene mutation analysis, variant calling was performed by VarScan 2 (v2.3.7)35 using very strict criteria, which excluded any synonymous mutations, any intronic mutations with predicted low impact and all germline variants reported in the Single Nucleotide Polymorphism database (dbSNP build142). For prognostic analysis, only non-silent mutations (missense, nonsense, in-frame or frame-shift insertions and deletions) were included. For detailed laboratory data analysis procedures see the Online Supplementary Methods section.

Ethical Committee statement The study and protocols to obtain human archival tissues and patient data were approved by the local ethical committee of the VU University Medical Center, Amsterdam (FWA00017598) for all collaborating centers and comply with the Code for Proper Secondary Use of Human Tissue in The Netherlands.

Statistics Univariate and multivariate analyses were used to evaluate the distribution of the biomarkers for the 2 cohorts. For microenvironment analysis, biomarkers were included in the analysis using the scoring categories as defined above, and the average of the biomarker score from 2 cores was used in the analysis. For mutation analysis, the genes were included in the analysis as mutated or wild-type. To correct for multiple comparisons, the BenjaminiHochberg method was used, and P-values of less than 0.05 were considered significant. Patient characteristics were summarized with descriptive statistics. The Fisher’s exact test was used to test for association between pairs of categorical variables, and univariate and multivariable logistic regression was used for the binary outcome of cohorts. Odds ratios (OR) and 95% confidence intervals (CI) were reported. The Wilcoxon rank-sum test was used to assess a location shift in the distribution of continuous variables haematologica | 2017; 102(8)

between 2 groups. In a secondary analysis, optimal cut points for 8 continuous markers were determined using recursive partitioning models for a binary outcome. To evaluate agreement among the 3 pathologists for the FOXP3 patterns the free-marginal κ statistics of Brennan and Prediger are reported with a bootstrap confidence CI.36,37 Analyses were performed using SAS Software version 9.3 (SAS Institute Inc., Cary, NC, USA; 2005) and R version 3.3.0 (R Core Team, Vienna, Austria; 2016).38

Results Patient selection and immunohistochemical biomarker assessment A total of 122 patients fulfilled the selection criteria and had biopsy material available that met the input requirements for both IHC and gene mutation analysis (EF, n=49 and LR, n=73). In 105 cases a complete set of IHC markers was available (Online Supplementary Table S3). For NGS analysis, DNA of sufficient quality and with sufficient NGS read depth could be retrieved for 111 cases, resulting in the complete molecular data for gene mutation analysis and copy number profiles. For 96 of the 122 cases (79%) both IHC and molecular data could be generated that met our quality criteria and was included for downstream analysis. Table 1 shows the clinical characteristics of the 96 patients (Online Supplementary Table S4 provides the characteristics for all 122 patients, with marginal statistical differences for prognostic subgroup representation). The FLIPI high-risk category was overrepresented in the EF cohort (P=0.009), underpinning the validity of the selection criteria for this end of spectrum design.

Impact of microenvironment T-cell and macrophage cell populations by cohort The distribution of IHC markers per end of spectrum prognostic subgroup analyzed in the whole core is shown in Figure 1 and the Online Supplementary Table S5 (Online Supplementary Table S6 shows the distribution of IHC markers of the 105 cases). Statistically significant differences were found in the median percentage of CD8+ nucleated cells (median EF vs. LR is 7.9% vs. 8.6%, P=0.011) and CD163+ macrophages area (median EF vs. LR is 3.6% vs. 5.2%, P=0.038). In logistic regression analyses, the estimated OR for CD8+ cells are 3.9 (95% CI: [1.5-12.1], P=0.01) for a decrease of 10% CD8+ cells. For the CD163+ area the odds are 2.0 (95% CI: [1.1- 4.4], P=0.04) for a decrease of 10% CD163+ area (Online Supplementary Table S7). Adjusting for other IHC markers without the FLIPI score in a multivariable model, only the CD8+ T cells retained significance (OR=4.5, 95% CI: [1.1-21.2] P=0.04), however, with the FLIPI score included they no longer reached significance. No significant differences were found for the other markers (CD3, CD4, CD68, CD163, PD1, FOXP3 and p53) between the two cohorts (Online Supplementary Table S7). Since CD163 was binary scored as positive or negative, a high overall score for the 2 prognostic subgroups might be caused by a higher cell density and/or by larger individual cells. Visual assessment by two pathologists (DdJ, MC) was performed and confirmed that the higher percentage of positive areas was due to higher cell density. A secondary analysis using recursive partitioning was performed to evaluate markers that could separate the 2 prognostic sub1415

W.B.C. Stevens et al.

groups and to determine the optimal cut points for the identified markers for the whole core. The optimal cut points were 12.6% and 6.3% for CD8+ and CD163+, respectively. The percentage of patients with low levels of both CD8+ T cells and CD163+ area (defined as lower than the optimal cut points) was 79% (95% CI 64 – 91%) vs. 39% (95% CI 16 -52%) for EF (n=39) vs. LR (n=57) (P<0.001). Similarly, results were obtained if the upper quartiles were used instead of the optimal cut points (12.2% (n=39) vs. 8.4% (n=57) for EF [79% (95% CI: 64 – 91%)] vs. LR [46% (95% CI: 32 – 59%)], P=0.001).

Impact of spatial distribution and perifollicular pattern of FOXP3 by prognostic subgroup The spatial distribution of T-cell populations and

macrophages has been claimed to be of greater influence on prognosis than the overall numbers of infiltrating cells and therefore intrafollicular and interfollicular populations were assessed separately.14-17,19,22,39 However, in this series, we could not validate this claim for most T-cell or macrophage classes. Except in the univariate analysis of the interfollicular population, differences in CD8+ cells and CD163+ area were significant; CD8+ cells with an OR 3.6 (95% CI: [1.2-6.4], P=0.03) and CD163+ area 1.9 (95% CI: [1.1-3.6], P=0.03). Only CD8+ interfollicular populations have a minor influence in multivariate analysis excluding the FLIPI score (OR 3.7, 95% CI [1.2-15.5], P<0.01) (Online Supplementary Tables S8 and S9). The perifollicular pattern of FOXP3+ T cells was scored manually by 3 pathologists (DdJ, AR, BS). Agreement

Table 1. Clinical characteristics of patients with all immunohistochemical and molecular markers available.

Group Barts GLSG LYSA Age at diagnosis Median (range) < 60 Sex Female Grade Grade 1, 2 Grade 3A Missing Stage Stage I-II Stage III-IV Missing B-symptoms Absent Present Missing ECOG PS 0 1 2 3 Missing FLIPI risk categories Low Intermediate High Missing First-line therapy R-CHOP R-CHVP-I

Total n = 96

Early failure n = 39

Long remission n = 57

8 (8%) 80 (83%) 8 (8%)

6 (15%) 29 (74%) 4 (10%)

2 (4%) 51 (89%) 4 (7%)

58 (27-75) 50 (52%)

61 (27-75) 18 (46%)

58 (32-69) 32 (56%)

47 (49%)

17 (44%)

30 (53%)

75 (78%) 6 (6%) 15 (16%)

28 (72%) 3 (8%) 8 (21%)

47 (82%) 3 (5%) 7 (12%)

2 24%) 94 (87%) 1 (1%)

1 (3%) 38 (97%) 0

1 (2%) 55 (96%) 1 (2%)

59 (61%) 35 (36%) 2 (2%)

23 (59%) 16 (41%) 0

36 (63%) 19 (33%) 2 (4%)

32 (33%) 58 (60%) 2 (2%) 1 (1%) 3 (2%)

13 (33%) 21 (54%) 2 (5%) 1 (3%) 2 (5%)

19 (33%) 37 (65%) 0 (0%) 0 (0%) 1 (2%)

10 (10%) 35 (36%) 46 (48%) 5 (5%)

1 (3%) 11 (28%) 26 (67%) 1 (3%)

9 (16%) 24 (42%) 20 (36%) 4 (7%)

87 (91%) 9 (9%)

34 (87%) 5 (13%)

53 (93%) 4 (7%)

Barts: Bartholomew’s Hospital Registry London, UK; GLSG: German low-grade Lymphoma Study Group; LYSA: the Lymphoma Study Association; ECOG: Eastern Cooperative Oncology Group; PS: performance score; FLIPI: follicular lymphoma international prognostic index; R-CHOP: rituximab, cyclophosphamide, adriamycin, vincristine and prednisone; R-CHVP-I: rituximab, cyclophosphamide, adriamycin, etoposide, prednisolone and interferon-α2a.

1416

haematologica | 2017; 102(8)

CD8, CD163, EZH2, chromosome 18, FL prognostic biomarkers between pathologists reached levels similar to the validation study (Brennan-Prediger estimate of 0.78 and 0.83 for cores 1 and 2 in the study herein versus 0.85 for the validation study).29 The perifollicular pattern did not differ significantly (P=0.46) between the 2 prognostic subgroups, 10/39 (26%) in the EF subgroup and 11/57 (19%) in the LR subgroup (Online Supplementary Table S10).

Frequency distribution of chromosomal copy numbers and gene mutations in FL Shallow WGS resulted in high quality genome-wide copy number aberration (CNA) plots for all cases. The most common aberrations, detected in at least 10% of patients, included complete or partial gains of chromosomes 1q, 2, 7, 8, 12 and 18 and losses of 1p, 6q and 10q

Figure 1. Boxplots per immuunhistochemical marker. For CD4, CD8, CD3, FOXP3, PD1 and P53 they show the percentage of positive nucleated cells of all nucleated cells, and for CD163 and CD68 they show the percentage of positive cell area of the total cell area. Early failure n=39, long remission n=57.

haematologica | 2017; 102(8)

1417

W.B.C. Stevens et al. Table 2. Distribution of gene mutation status by cohort (n=96).

BCL2 Mutated Unmutated KMT2D Mutated Unmutated CREBBP Mutated Unmutated TNFRSF14 Mutated Unmutated MEF2B Mutated Unmutated EZH2 Mutated Unmutated TNFAIP3 Mutated Unmutated EP300 Mutated Unmutated CARD11 Mutated Unmutated FAS Mutated Unmutated MYD88 Mutated Unmutated

Total n=96 (%)

Early failure n=39 (%)

Long remission n=57 (%)

OR [95% CI]

P (unadjusted)

P (adjusted)

88 (92) 8 (8)

37 (95) 2 (5)

51 (89) 6 (11)

0.46 [0.04 - 2.78]

0,47

0,9

68 (71) 28 (29)

28 (72) 11 (28)

40 (70) 17 (30)

0.93 [0.34 - 2.47]

> 0.99

64 (67) 32 (33)

29 (74) 10 (26)

35 (61) 22 (39)

0.55 [0.20 - 1.45]

0,27

0,6

29 (30) 67 (70)

10 (26) 29 (74)

19 (33) 38 (67)

1.44 [0.54 - 4.04]

0,5

0,9

10 (10) 86 (90)

4 (10) 35 (90)

6 (11) 51 (89)

1.03 [0.22 - 5.33]

> 0.99

17 (18) 79 (82)

1 (3) 38 (97)

16 (28) 41 (72)

14.53 [2.06 - 635.92]

< 0.001

0,006

7 (7) 89 (93)

1 (3) 38 (97)

6 (11) 51 (89)

4.41 [0.50 - 210.74]

0,23

0,6

7 (7) 89 (93)

2 (5) 37 (95)

5 (9) 52 (91)

1.77 [0.27 - 19.52]

0,7

> 0.99

8 (8) 88 (92)

3 (8) 36 (92)

5 (9) 52 (91)

1.15 [0.21 - 7.89]

> 0.99

2 (2) 94 (98)

2 (5) 37 (95)

0 (0) 57 (100)

0.00 [0.00 - 3.62]

0,16

0,6

2 (2) 94 (98)

1 (3) 38 (97)

1 (2) 56 (98)

0.68 [0.01 - 54.63]

> 0.99

OR: odds ratio; CI: confidence interval.

(Figure 2 and Online Supplementary Table S11). The landscape of CNAs showed notable enrichment of FL-related genes, including focal losses of TNFRSF14 (1p36.32), TNFAIP3 (6p23.3) and FAS and PTEN (10q23.31) and focal gains that harbor FL-related oncogenes, like BCL11A and REL (2p16.1). Genes included in the FL-LLBC-NGS target enrichment panel showed non-synonymous mutations in at least 2 FL cases (Figure 3 and Online Supplementary Table S12). BCL2, a known target for aberrant somatic hypermutation (aSHM) in FL, was most frequently mutated (88/96, 92%) with 0 to 78 mutations per case. Chromatin modifying genes KMT2D (71%) and CREBBP (67%) were mutated with high frequency, and epigenetic modifiers EZH2 (18%), MEF2B (10%), EP300 (7%) at lower rates with non-silent mutations in 1-4 of these chromatin modifying genes in 90% of FL patients, consistent with the critical role of epigenetic deregulation in the majority of FL. No patterns of co-occurrence or mutual exclusivity were 1418

observed. Non-silent mutations were found in TNFRSF14 (30%), TNFAIP3 (7%), CARD11 (8%), FAS (2%) and MYD88 (2%) (Table 2, Online Supplementary Table S13 shows the distribution of mutations of the 111 cases).

Impact of chromosomal copy numbers and gene mutations by prognostic subgroup The distribution of CNAs and gene mutations by prognostic subgroup are shown in Figures 2 and 3. Statistically significant differences after multiple testing correction were only found for gain of chromosomal region 18p11.32-q21.33 (gain EF vs. LR 49% vs. 12%, P<0.001), with an estimated OR in logistic regression analysis of 0.15 (95% CI: [0.05-0.44]) and for EZH2 mutation status (unmutated EF vs. LR 90% vs. 72%, P<0.001) (Table 2). The odds ratio for unmutated EZH2 was 14.53 (95% CI:[2.06-635.92]). No significant impact was found for various markers, previously implicated as (borderline) prognostic such as CREBBP (OR 0.55, 95% CI [0.20-1.45], haematologica | 2017; 102(8)

Losses and gains

CD8, CD163, EZH2, chromosome 18, FL prognostic biomarkers

P-value and FDR

P value FDR

Figure 2. Distribution and significance of copy number gains and losses by subgroup. Top panel: Percentages of gains (top; green) and losses (bottom; red) in early failure (EF, n=39) and long remission (LR, n=57) per chromosomal region. X-axis: chromosomal regions, ordered by genomic coordinates of chromosomes 1 to 22. Y-axis: percentage of cases showing CNAs. Vertical dotted lines: boundaries between chromosomes. Bottom panel: statistical significance of differences in frequencies of gains (top) and losses (bottom) between cohorts. X-axis: chromosomal regions, ordered by genomic coordinates of chromosomes 1 to 22. Y-axis: P-value (blue) and false discovery rate (FDR; yellow). Horizontal dotted lines show the threshold of significance for P (0.05) and FDR (0.1), based on a Wilcoxon rank-sum test with 10 000 permutations.

Table 3. Odds ratio (OR) (95% Cl) for a 10% change in the immunohistochemical markers and absent or present molecular markers in univariate analysis, and multivariate analysis without and with the FLIPI.

Univariate OR (95% Cl) %CD8 %CD163 CHR 18 EZH2 FLIPI high

3.86 (1.48, 12.13) 2.01 (1.11, 4.37) 0.15 (0.05, 0.39) 14.83 (2.82, 274.07) 0.28 (0.11, 0.66)

Multivariable without FLIPI OR (95% Cl)

0.01 0.04 <0.001 0.011

3.15 (1.03, 11.87) 1.54 (0.75, 3.83) 0.27 (0.09, 0.79) 13.76 (2.35, 264.94)

0.005

P 0.064 0.29 0.019 0.017

Multivariable with FLIPI OR (95% Cl) 2.58 (0.83, 10.02) 1.43 (0.69, 3.52) 0.24 (0.07, 0.75) 9.99 (1.69, 192.73) 0.36 (0.12, 0.98)

P 0.13 0.38 0.018 0.036

CI: confidence interval; FLIPI: follicular lymphoma international prognostic index.

P=0.6), EP300 (OR 1.77, 95% CI [0.27-19.52], P>0.99), CARD11 (OR 1.15, 95% CI [0.21-7.89], P>0.99) and MEF2B (OR 1.03, 95% CI [0.22-5.33], P>0.99).

Integrated modeling of immunohistochemical and molecular analysis of the prognostic subgroups Correlation analysis was performed for molecular and IHC markers, showing significant correlation between mutated CREBBP status and higher level infiltrates of PD1 positive T cells (P<0.005), but not with CD4 and CD8 positive T cells (Figure 4A). TNFRSF14 mutation status showed significant correlation with lower CD4 and CD8 positive T-cell infiltrates (P=0.037 and P=0.030) (Figure 4B). Microenvironmental populations were not significantly differentially distributed for other molecular markers, including EZH2 (data not shown). haematologica | 2017; 102(8)

In a multivariable model combining the 4 markers, CD8, CD163, gain chromosome 18 and EZH2 mutation, that are statistically significant in the univariate analysis, only the gain of chromosome 18 (OR 0.27 (95% CI: [0.09,0.79], P=0.019)) and EZH2 (OR 13.76 (95% CI: [2.53,264.94], P=0.017)) retained significance. After incorporating the FLIPI score into the model, gain of chromosome 18 (P=0.018) and EZH2 (P=0.036) (Table 3) retained significance.

Discussion This LLBC study is, as far as we know, the first to comprehensively explore the combined prognostic impact of microenvironment T-cell and macrophage infiltration and 1419

W.B.C. Stevens et al.

Figure 3. Distribution of mutations in 11 genes by subgroup. Each column represents an individual case. Genes are clustered based on functional category and mutations are color-coded based on effect prediction. Mutation frequencies for each gene by cohort are shown in the bar graph on the right. EF: early failure; LR: long remission.

tumor genetics of FL patients with extremely poor outcome (EF) versus those with a prolonged remission (LR). We show poor outcome to be characterized by a lower number of CD8+ T cells, a smaller CD163+ macrophage area (indicative of fewer macrophages), wild-type EZH2 and a copy number gain of chromosome 18. These observations, in part, confirm previous studies. The gain of chromosome 18, despite its statistically strong prognostic value, has not been reported previously. Equally important, cellular densities of various other cell populations, such as PD1+ T follicular helper (TFH) cells and FOXP3+ T regulatory (Treg) cells, previously claimed to predict clinical outcome, were not confirmed in our study.16,22,24,39-41 This study was specifically designed to verify the impact of previously published IHC and molecular markers in FL in the rituximab-chemotherapy era. Therefore, we implemented a dedicated study design to reduce noise such that all cases were retrieved from clinical trials and registries which guaranteed complete clinical information at presentation and detailed treatment information. This allowed us to make a homogeneously treated patient selection, with subgroups at the extreme ends of the prognostic spectrum, the EF and the LR subgroups. By balancing inclusion of the rare EF subgroup, we maximized the sensitivity to observe clinically relevant differences in the microenvironment and mutations, while allowing an overall relatively small patient cohort. To reduce inter-observer variability, the validated quantitative computerized IHC scoring method of the TMA was implemented, which has previously been shown to be more reproducible than any semi-quantitative method.27 For microenvironment populations, our results regarding the CD163+ area validate those of Kridel et al., showing that a higher CD163+ pixel count or CD163+ area are independent predictors of prolonged progression-free survival (PFS) in patients treated with R-CHOP (P=0.0110.030), while the CD68+ macrophages population did not have a significant impact by pixel count or area.42 Previous reports provided conflicting data for CD68+ macrophages, suggesting correlation with either adverse17,39,43,44 or favorable outcome.45 Differences in scoring methods and varia1420

tions in treatment characteristics may explain these discrepancies. This LLBC study also showed that a lower percentage of CD8+ cytotoxic T cells in the whole TMA core and interfollicular areas was associated with EF after treatment with rituximab-chemotherapy. This confirms findings in several series using computerized scoring or flow cytometry of cell suspensions, both in the rituximab era and prerituximab era.15,22,46 However, inconsistent results were obtained in studies using manual scoring,14,18,47 corroborating the need for a reliable scoring method of CD8+ T cells. None of the other T-cell markers, including CD3, CD4, PD1 and FOXP3 demonstrated a significant prognostic impact. Debates mostly focus on PD1+ cells, FOXP3+ cells and perifollicular patterns of FOXP3 positive T cells, with conflicting results being published. However, the positive prognostic value of PD1+ cells was seen in studies in which only some of the patients received rituximab,16,22 whereas in studies in which the majority of patients were treated with rituximab-chemotherapy, either no effect or a negative effect was reported.39,41 Frequencies of mutations and distribution of alterations in hotspots and functional domains of the tested genes are largely in line with those previously reported.28,31,48-50 Almost all FLs are reported to have mutations in 1 or more histone-modifying gene, such as KMT2D (reported as 7689%), CREBBP (reported as 33-68%), MEF2B (reported as 15%), EP300 (reported as 9-15%) and EZH2 (reported as 7-26%).49 Histone-modifying genes exert their function largely indirectly via co-activation of various transcription factors, and as such their specific role in B-cell oncogenesis and immunological functions is difficult to predict.51 Gene expression analysis has suggested that mutated CREBBP may downregulate major histocompatibility complex (MHC) class II genes, resulting in impaired T-cell activation and possibly lower T-cell levels in tumor samples.49 This hypothesis could not be supported in the study herein, which showed no association of CREBBP mutation status and total numbers of T cells or specific subsets, apart from an association of CREBBP mutated status with higher numbers of PD1 positive T cells. It should be noted, haematologica | 2017; 102(8)

CD8, CD163, EZH2, chromosome 18, FL prognostic biomarkers

Figure 4. Correlation molecular markers and IHC markers. A: CREBBP B: TNFRSF14. Blue bars are mutated, gray bars are wild-type. On the X-axis the number of cases per IHC marker, on the Y-axis the percentages of positive cells or area.

however, that the differences, albeit statistically significant, take place in a very narrow dynamic range. KMT2D, which was mutated in 70% of both EF and LR cases, showed a strong correlation to both CD8 and CD163 with high levels of both markers in wild-type cases. KMT2D is known to decrease apoptosis and increase B-cell proliferation both directly and indirectly in germinal center B cells.52 Conditional mouse models indicate a role in plasma cell and germinal center differentiation.53 Regulatory alterations, impacting on immunological interactions with T cells or macrophages have not been described, however. In line with this, it is less remarkable that the mutation status of TNFRSF14, a protein that is directly involved in immune response regulation, does correlate with CD4+ and CD8+ T-cell infiltration in tumor samples. By shallow WGS, copy number profiles of all tumors were studied in both prognostic groups. Of all previously haematologica | 2017; 102(8)

published numerical alterations in FL, only gain of chromosome 18 stood out as an independent prognostic marker. Our results confirm EZH2 as a strong prognostic marker in FL, with wild-type gene status associated with poor disease outcome (EF),31,32 but other markers, such as EP300 and TNFRSF14, which have been implicated to have a significant, though minor, impact on prognosis, were not substantiated in our study.31,54 This is likely due to selection bias and relative underrepresentation of poor prognosis patients in previous series for which our study design was specifically optimized. This LLBC study design precludes integration of a complete multifactorial prognostic model such as the M7-FLIPI index, however, the prognostic trend of EZH2 as reported by Pastore et al. follows the same direction as in our study, where statistical significance is reached and lack of significance of 4 other markers is confirmed. 1421

W.B.C. Stevens et al.

The pertinent question is whether these findings on the prognostic value of microenvironment populations and genomic alterations can be translated to application in daily clinical practice. The mutations in EZH2 are largely clustered in codon Y646 (Online Supplementary Figure S1) and are therefore technically very easily amenable to simple polymerase chain reaction (PCR) techniques,55 while chromosomal gains can be monitored using fluorescence in situ hybridization (FISH) or NGS.56,57 This signifies that EZH2 mutation status and chromosome 18 gain have a high potential for clinical implementation, in contrast to the microenvironment population markers CD8 and CD163. With current techniques, even if optimized, the absolute quantitative differences are too small between these 2 extreme cohorts to become a powerful and clinically useful tool for the scores around the cut point. At best the prediction of the extremes can be used for this purpose. In conclusion, the literature with regard to IHC prognostic markers in FL has produced highly conflicting results, and concerning mutation analysis, only very limited data are currently available on prognostic impact. By making use of a specialized study design and a homogeneously rituximab-chemotherapy treated group of patients, we can now confirm that lower percentages of CD8+ T cells, CD163+ M2 macrophage areas, EZH2 wild-type status and gain of chromosome 18 in the initial tumor biopsy specimen predict a poor prognosis in FL for this treatment cohort. Of equal importance, in the study herein we could not substantiate the previously reported claims on the prognostic impact of the other most commonly mutated genes, such as TNFRSF14 and EP300, of T-cell populations and classes of macrophages, as well as a perifollicular distribution of FOXP3+ T cells for patients treated with R-

References 1. Swenson WT, Wooldridge JE, Lynch CF, Forman-Hoffman VL, Chrischilles E, Link BK. Improved survival of follicular lymphoma patients in the United States. J Clin Oncol. 2005;23(22):5019-5026. 2. Marcus R, Imrie K, Belch A, et al. CVP chemotherapy plus rituximab compared with CVP as first-line treatment for advanced follicular lymphoma. Blood. 2005;105(4):1417-1423. 3. Hiddemann W, Kneba M, Dreyling M, et al. Frontline therapy with rituximab added to the combination of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) significantly improves the outcome for patients with advanced-stage follicular lymphoma compared with therapy with CHOP alone: results of a prospective randomized study of the German LowGrade Lymphoma Study Group. Blood. 2005;106(12):3725-3732. 4. Salles G, Mounier N, de Guibert S, et al. Rituximab combined with chemotherapy and interferon in follicular lymphoma patients: results of the GELA-GOELAMS FL2000 study. Blood. 2008;112(13):48244831. 5. Salles G, Seymour JF, Offner F, et al. Rituximab maintenance for 2 years in patients with high tumour burden follicular lymphoma responding to rituximab plus

1422

7. 8. 9.

10.

11.

CHOP (-like).14,16,18,22,24,31,39-41,54 Moreover, the study herein provides further insight into the relationship between gene mutation status and the most relevant microenvironmental populations in FL. Contributors The Lunenburg Lymphoma Biomarker Consortium (LLBC) is a collaboration of 10 international lymphoma research groups, each represented by a clinical investigator and one or more hematopathologists and supported by a team of statisticians. Foundation of the LLBC was made possible with a grant from the Van Vlissingen Lymphoma Foundation. EORTC Lymphoma group: Daphne de Jong, John Raemaekers. HOVON Lymphoma group: Daphne de Jong, Marie-JosĂŠ Kersten, Anton Hagenbeek. LYSA: Philippe Gaulard, Gilles Salles, Luc Xerri. British Columbia Cancer Agency: Randy D. Gascoyne, Laurie Sehn. ECOG: Randy D. Gascoyne. GLSG: Andreas Rosenwald, Wolfram Klapper, Michael Pfreundschuh, Wolfgang Hiddemann, Eva Hoster. NLG: Birgitta Sander, Eva Kimby. Barts Cancer Institute: Andrew J. Clear, Maria Calaminici, John Gribben. Leeds Registry: Andrew Jack Stanford: Yasodha Natkunam, Ranjana Advani. Dana-Farber Cancer Institute, Boston, USA: Edie Weller, Robert Redd. Acknowledgments The authors would like to thank Daoud Sie for his advice and providing genomics and compute infrastructure Funding This study was supported by unrestricted grants from: Genentech/Roche, GlaxoSmithKline, Pfizer Pharma, Teva, Pharmaceuticals/Cephalon and Millenium Pharmaceuticals Inc, Celgene.

chemotherapy (PRIMA): a phase 3, randomised controlled trial. Lancet. 2011;377 (9759):42-51. Bachy E, Houot R, Morschhauser F, et al. Long-term follow up of the FL2000 study comparing CHVP-interferon to CHVPinterferon plus rituximab in follicular lymphoma. Haematologica. 2013;98(7):11071114. Freedman A. Follicular lymphoma: 2011 update on diagnosis and management. Am J Hematol. 2011;86(9):768-775. Solal-Celigny P, Roy P, Colombat P, et al. Follicular lymphoma international prognostic index. Blood. 2004;104(5):1258-1265. Montoto S, Lopez-Guillermo A, Altes A, et al. Predictive value of Follicular Lymphoma International Prognostic Index (FLIPI) in patients with follicular lymphoma at first progression. Ann Oncol. 2004;15(10):14841489. Federico M, Bellei M, Marcheselli L, et al. Follicular lymphoma international prognostic index 2: a new prognostic index for follicular lymphoma developed by the international follicular lymphoma prognostic factor project. J Clin Oncol. 2009;27(27):4555-4562. Dave SS, Wright G, Tan B, et al. Prediction of survival in follicular lymphoma based on molecular features of tumor-infiltrating immune cells. N Engl J Med. 2004;351(21):2159-2169.

12. Glas AM, Kersten MJ, Delahaye LJ, et al. Gene expression profiling in follicular lymphoma to assess clinical aggressiveness and to guide the choice of treatment. Blood. 2005;105(1):301-307. 13. Ott G, Katzenberger T, Lohr A, et al. Cytomorphologic, immunohistochemical, and cytogenetic profiles of follicular lymphoma: 2 types of follicular lymphoma grade 3. Blood. 2002;99(10):3806-3812. 14. Lee AM, Clear AJ, Calaminici M, et al. Number of CD4+ cells and location of forkhead box protein P3-positive cells in diagnostic follicular lymphoma tissue microarrays correlates with outcome. J Clin Oncol. 2006;24(31):5052-5059. 15. Alvaro T, Lejeune M, Camacho FI, et al. The presence of STAT1-positive tumorassociated macrophages and their relation to outcome in patients with follicular lymphoma. Haematologica. 2006;91(12):16051612. 16. Carreras J, Lopez-Guillermo A, Fox BC, et al. High numbers of tumor-infiltrating FOXP3-positive regulatory T cells are associated with improved overall survival in follicular lymphoma. Blood. 2006;108(9): 2957-2964. 17. Canioni D, Salles G, Mounier N, et al. High numbers of tumor-associated macrophages have an adverse prognostic value that can be circumvented by rituximab in patients with follicular lymphoma enrolled onto the

haematologica | 2017; 102(8)

CD8, CD163, EZH2, chromosome 18, FL prognostic biomarkers

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

GELA-GOELAMS FL-2000 trial. J Clin Oncol. 2008;26(3):440-446. Glas AM, Knoops L, Delahaye L, et al. Gene-expression and immunohistochemical study of specific T-cell subsets and accessory cell types in the transformation and prognosis of follicular lymphoma. J Clin Oncol. 2007;25(4):390-398. Carreras J, Lopez-Guillermo A, Roncador G, et al. High numbers of tumor-infiltrating programmed cell death 1-positive regulatory lymphocytes are associated with improved overall survival in follicular lymphoma. J Clin Oncol. 2009;27(9):14701476. Clear AJ, Lee AM, Calaminici M, et al. Increased angiogenic sprouting in poor prognosis FL is associated with elevated numbers of CD163+ macrophages within the immediate sprouting microenvironment. Blood. 2010;115(24):5053-5056. Klapper W, Hoster E, Rolver L, et al. Tumor sclerosis but not cell proliferation or malignancy grade is a prognostic marker in advanced-stage follicular lymphoma: the German Low Grade Lymphoma Study Group. J Clin Oncol. 2007;25(22):33303336. Wahlin BE, Aggarwal M, Montes-Moreno S, et al. A unifying microenvironment model in follicular lymphoma: outcome is predicted by programmed death-1--positive, regulatory, cytotoxic, and helper T cells and macrophages. Clin Cancer Res. 2010;16(2):637-650. Wahlin BE, Sundstrom C, Holte H, et al. T cells in tumors and blood predict outcome in follicular lymphoma treated with rituximab. Clin Cancer Res. 2011;17(12):41364144. Farinha P, Al-Tourah A, Gill K, Klasa R, Connors JM, Gascoyne RD. The architectural pattern of FOXP3-positive T cells in follicular lymphoma is an independent predictor of survival and histologic transformation. Blood. 2010;115(2):289-295. de Jong D, Koster A, Hagenbeek A, et al. Impact of the tumor microenvironment on prognosis in follicular lymphoma is dependent on specific treatment protocols. Haematologica. 2009;94(1):70-77. de Jong D, Rosenwald A, Chhanabhai M, et al. Immunohistochemical prognostic markers in diffuse large B-cell lymphoma: validation of tissue microarray as a prerequisite for broad clinical applications--a study from the Lunenburg Lymphoma Biomarker Consortium. J Clin Oncol. 2007;25(7):805812. Sander B, de Jong D, Rosenwald A, et al. The reliability of immunohistochemical analysis of the tumor microenvironment in follicular lymphoma: a validation study from the Lunenburg Lymphoma Biomarker Consortium. Haematologica. 2014;99(4): 715-725. Pasqualucci L, Khiabanian H, Fangazio M, et al. Genetics of follicular lymphoma transformation. Cell Rep. 2014;6(1):130140. Okosun J, Bodor C, Wang J, et al. Integrated genomic analysis identifies recurrent mutations and evolution patterns driving the initiation and progression of follicular lym-

haematologica | 2017; 102(8)

phoma. Nat Genet. 2014;46(2):176-181. 30. Green MR, Gentles AJ, Nair RV, et al. Hierarchy in somatic mutations arising during genomic evolution and progression of follicular lymphoma. Blood. 2013;121(9): 1604-1611. 31. Pastore A, Jurinovic V, Kridel R, et al. Integration of gene mutations in risk prognostication for patients receiving first-line immunochemotherapy for follicular lymphoma: a retrospective analysis of a prospective clinical trial and validation in a population-based registry. Lancet Oncol. 2015;16(9):1111-1122. 32. Jurinovic V, Kridel R, Staiger AM, et al. Clinicogenetic risk models predict early progression of follicular lymphoma after first-line immunochemotherapy. Blood. 2016;128(8):1112-1120. 33. Casulo C, Byrtek M, Dawson KL, et al. Early Relapse of Follicular Lymphoma After Rituximab Plus Cyclophosphamide, Doxorubicin, Vincristine, and Prednisone Defines Patients at High Risk for Death: An Analysis From the National LymphoCare Study. J Clin Oncol. 2015;33(23):2516-2522. 34. Scheinin I, Sie D, Bengtsson H, et al. DNA copy number analysis of fresh and formalin-fixed specimens by shallow wholegenome sequencing with identification and exclusion of problematic regions in the genome assembly. Genome Res. 2014;24(12):2022-2032. 35. Koboldt DC, Zhang Q, Larson DE, et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome research. 2012;22(3):568-576. 36. Brennan RL, Prediger DJ. Coefficient kappa - some uses, misuses, and alternatives. Educ Psychol Meas. 1981;41(3):687-699. 37. Mackay A, Weigelt B, Grigoriadis A, et al. Microarray-based class discovery for molecular classification of breast cancer: analysis of interobserver agreement. J Natl Cancer Inst. 2011;103(8):662-673. 38. R Core Team R. A language and environment for statistical computing. R Foundation for Statistical Computering, Vienna, Austria 2013 [cited; Available from: http://www.R-project.org. Last accessed December 2016. 39. Richendollar BG, Pohlman B, Elson P, Hsi ED. Follicular programmed death 1-positive lymphocytes in the tumor microenvironment are an independent prognostic factor in follicular lymphoma. Hum Pathol. 2011;42(4):552-557. 40. Koch K, Hoster E, Unterhalt M, et al. The composition of the microenvironment in follicular lymphoma is associated with the stage of the disease. Hum Pathol. 2012;43(12):2274-2281. 41. Takahashi H, Tomita N, Sakata S, et al. Prognostic significance of programmed cell death-1-positive cells in follicular lymphoma patients may alter in the rituximab era. Eur J Haematol. 2013;90(4):286-290. 42. Kridel R, Xerri L, Gelas-Dore B, et al. The prognostic impact of CD163-positive macrophages in follicular lymphoma: a study from the BC Cancer Agency and the Lymphoma Study Association. Clin Cancer Res. 2015;21(15):3428-3435.

43. Coiffier B, Li W, Henitz ED, et al. Prespecified candidate biomarkers identify follicular lymphoma patients who achieved longer progression-free survival with bortezomib-rituximab versus rituximab. Clin Cancer Res. 2013;19(9):2551-2561. 44. Farinha P, Masoudi H, Skinnider BF, et al. Analysis of multiple biomarkers shows that lymphoma-associated macrophage (LAM) content is an independent predictor of survival in follicular lymphoma (FL). Blood. 2005;106(6):2169-2174. 45. Taskinen M, Karjalainen-Lindsberg ML, Nyman H, Eerola LM, Leppa S. A high tumor-associated macrophage content predicts favorable outcome in follicular lymphoma patients treated with rituximab and cyclophosphamide-doxorubicin-vincristine-prednisone. Clin Cancer Res. 2007;13(19):5784-5789. 46. Wahlin BE, Sander B, Christensson B, et al. Entourage: the immune microenvironment following follicular lymphoma. Blood Cancer J. 2012;2(1):e52. 47. Saifi M, Maran A, Raynaud P, et al. High ratio of interfollicular CD8/FOXP3-positive regulatory T cells is associated with a high FLIPI index and poor overall survival in follicular lymphoma. Exp Ther Med. 2010;1(6):933-938. 48. Morin RD, Mendez-Lago M, Mungall AJ, et al. Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature. 2011;476(7360):298-303. 49. Green MR, Kihira S, Liu CL, et al. Mutations in early follicular lymphoma progenitors are associated with suppressed antigen presentation. Proc Natl Acad Sci USA. 2015;112(10):E1116-1125. 50. Morin RD, Johnson NA, Severson TM, et al. Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat Genet. 2010;42(2):181-185. 51. Jiang Y, Hatzi K, Shaknovich R. Mechanisms of epigenetic deregulation in lymphoid neoplasms. Blood. 2013; 121(21):4271-4279. 52. Zhang J, Dominguez-Sola D, Hussein S, et al. Disruption of KMT2D perturbs germinal center B cell development and promotes lymphomagenesis. Nat Med. 2015; 21(10):1190-1198. 53. Ortega-Molina A, Boss IW, Canela A, et al. The histone lysine methyltransferase KMT2D sustains a gene expression program that represses B cell lymphoma development. Nat Med. 2015;21(10):1199-1208. 54. Cheung KJ, Johnson NA, Affleck JG, et al. Acquired TNFRSF14 mutations in follicular lymphoma are associated with worse prognosis. Cancer Res. 2010;70(22):9166-9174. 55. Bodor C, Grossmann V, Popov N, et al. EZH2 mutations are frequent and represent an early event in follicular lymphoma. Blood. 2013;122(18):3165-3168. 56. Macintyre G, Ylstra B, Brenton JD. Sequencing structural variants in cancer for precision therapeutics. Trends Genet. 2016; 32(9):530-542. 57. Hastings RJ, Bown N, Tibiletti MG, et al. Guidelines for cytogenetic investigations in tumours. Eur J Hum Genet. 2016;24(1):613.

1423

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Plasma Cell Disorders

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1424-1431

Lenalidomide/melphalan/dexamethasone in newly diagnosed patients with immunoglobulin light chain amyloidosis: results of a prospective phase 2 study with long-term follow up Ute Hegenbart,1 Tilmann Bochtler,1,2 Axel Benner,3 Natalia Becker,3 Christoph Kimmich,1 Arnt V. Kristen,4 Jörg Beimler,5 Ernst Hund,6 Markus Zorn,7 Anja Freiberger,8 Marianne Gawlik,1 Hartmut Goldschmidt,1 Dirk Hose,1 Anna Jauch,9 Anthony D. Ho1 and Stefan O. Schönland1,4

Department of Internal Medicine, Division of Hematology/Oncology, University of Heidelberg; 2Clinical Cooperation Unit Molecular Hematology/Oncology, German Cancer Research Center (DKFZ) and Department of Internal Medicine V, University of Heidelberg; 3Division of Biostatistics, German Cancer Research Center, Heidelberg; 4 Division of Cardiology, University of Heidelberg; 5Division of Nephrology, University of Heidelberg; 6Department of Neurology, University of Heidelberg; 7Division of Clinical Chemistry, University of Heidelberg; 8Coordination Center for Clinical Trials, KKS, Heidelberg and 9Institute of Human Genetics, University Heidelberg, Germany 1

ABSTRACT

C Correspondence: ute.hegenbart@med.uni-heidelberg.de

Received: December 24, 2016. Accepted: May 9, 2017. Pre-published: May 18, 2017. doi:10.3324/haematol.2016.163246 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1424

hemotherapy in light chain amyloidosis aims to normalize the involved free light chain in serum, which leads to an improvement, or at least stabilization of organ function in most responding patients. We performed a prospective single center phase 2 trial with 50 untreated patients not eligible for high-dose treatment. The treatment schedule comprised 6 cycles of oral lenalidomide, melphalan and dexamethasone every 4 weeks. After 6 months, complete remission was achieved in 9 patients (18%), very good partial remission in 16 (32%) and partial response in 9 (18%). Overall, organ response was observed in 24 patients (48%). Hematologic and cardiac toxicities were predominant adverse events. Mortality at 3 months was low at 4% (n=2) despite the inclusion of 36% of patients (n=18) with cardiac stage Mayo 3. After a median follow-up of 50 months, median overall and event-free survival were 67.5 months and 25.1 months, respectively. We conclude that the treatment of lenalidomide, melphalan and dexamethasone is very effective in achieving a hematologic remission, organ response and, consecutively, a long survival in transplant ineligible patients with light chain amyloidosis. However, as toxicity and tolerability are the major problems of a 3-drug regimen, a strict surveillance program is necessary and sufficient to avoid severe toxicities. clinicaltrials.gov Identifier: 00883623 (Eudract2008-001405-41).

Introduction ©2017 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.

1424

Immunoglobulin light chain (AL) amyloidosis is a monoclonal plasma cell disorder characterized by the deposition of amyloid fibrils in different tissues. Although the burden of plasma cells in the bone marrow is generally low, the accumulation of amyloid protein leads to progressive and severe end organ failure and, eventually, death.1 The choice of upfront treatment depends on age, performance status and degree of amyloid-related organ dysfunction. High-dose chemotherapy (HDC) followed by autologous stem cell transplantation is very effective and has excellent long-term results, but is reserved for younger patients with nearly normal organ functions.2 In patients not eligible for high-dose chemotherapy, melphalan-dexamethasone (MDex) is considered the standard treatment, and has shown good long-term results in patients without advanced cardiac involvement.3 However, M-Dex is much less effective in patients with advanced cardiac disease when a dose reduction of Dex haematologica | 2017; 102(8)

Len-Mel-Dex is effective in transplant ineligible AL amyloidosis

40 mg (to Dex 20 mg or even less) is necessary.3,4 To improve hematologic remission (HemR) rates, several combination therapies have recently been evaluated. Some of these combination protocols use M-Dex as a backbone and add either bortezomib or lenalidomide.5-8 The combination of bortezomib with M-Dex (B-M-Dex) appears a particularly promising treatment option. A recent publication by Palladini et al. showed a complete remission (CR) rate for B-M-Dex of 42% (HemR 69%) as compared to a CR rate of 19% in a historical M-Dex group (HemR 51%).8 However, the HemR rate was not different when the analysis was restricted to patients without severe cardiac involvement who were taking a full dose of Dex. This again underlines the huge impact of the Dex dosage on the results. Patient recruitment of a randomized clinical trial comparing B-M-Dex with M-Dex has been completed.9 Other combination chemotherapies (lenalidomide plus cyclophosphamide/Dex) have also been evaluated as first-line therapy.10-13 We have performed a prospective phase 2 trial using the combination of lenalidomide, melphalan and dexamethasone (L-M-Dex). As compared to the 3 prior studies also testing this regimen,5-7 the study herein includes the largest patient number with the longest follow up and provides further solid data to support this 3-drug therapy.

Methods Patients Fifty patients with untreated AL amyloidosis could be enrolled. They had to have measurable monoclonal gammopathy (M-spike and/or abnormal free light chain values) as well as symptomatic organ involvement. Diagnosis was made by a congo red positive biopsy, immunohistology to confirm AL type and exclusion of hereditary types when necessary. Patients were not eligible for HDC or autologous stem cell transplantation (center-specific criteria as published by Schönland et al.)14 or refused to undergo it. WHO Performance Status had to be <3 and The New York Heart Association (NYHA) <stage 4. Patients with symptomatic multiple myeloma or a creatinine clearance <40 ml/min were excluded. The patients had to be able to visit the Amyloidosis Clinic once a month.

Study design Patients were enrolled in this investigator initiated trial (IIT) phase 2, single center, open label study combining lenalidomide with M-Ddex (clinicaltrials.gov Identifier: 00883623; Eudract2008001405-41) at the Amyloidosis Center Heidelberg, Germany between 04/2009 and 02/2012. Lenalidomide was supplied by Celgene (München, Germany). The database was closed in 09/2013. Long-term survival, HemR and organ response analysis were performed outside of the protocol in a retrospective fashion with a data cutoff on 01/07/2015. Study treatment consisted of a total of 6 times 4 cycles of lenalidomide 10 mg/day for 21 days, melphalan 0.15 mg/kg/day and dexamethasone 20 mg/day for 4 days each. A prophylactic antibiotic treatment with ciprofloxacin was administered. Prophylactic anti-thrombotic treatment was aspirin 100 mg; lowmolecular weight heparin was applied in patients with a history of deep vein thrombosis, pulmonary embolism or thrombophilic coagulation status. Toxicity was assessed using The National Cancer Institute (NCI) common toxicity criteria (version 3.0). Due to the well described “paradoxical” increase of N-terminal B-type natriuretic peptide (NT-BNP) during lenalidomide treatment this haematologica | 2017; 102(8)

was not recorded as an adverse event or organ progression.15 The primary endpoint of the study was complete remission (CR) after 6 treatment cycles in patients who received at least 3 cycles of chemotherapy. HemR was defined in the protocol as CR or partial remission (PR).16 Very good partial remission (VGPR) was evaluated but not as part of the protocol (it was not yet defined at the time of study initiation); VGPR was retrospectively analyzed in all 42 patients with a dFLC (difference between the involved and uninvolved free light chain) of >50 mg/l.17 For better comparability with other studies we report HemR and CR+VGPR rates after treatment completion (5 or 6 cycles) as an “intention to treat” analysis (ITT, all patients). We defined early mortality as death due to any cause up until 3 months following the start of the treatment. Secondary endpoints were the rate of HemR after end of treatment (5 or 6 cycles) and organ response 6 months after end of treatment, and correlation of cytogenetic results with remission and survival. A retrospective comparison with a historical control group (using the same inclusion and exclusion criteria) treated with MDex either at our center or by local hematologists was performed. Treatment with M-Dex consisted of melphalan 16 mg/m2 day 1 intravenously4 every 4 weeks and Dex 20-40 mg days 1-4 orally, as published previously.4 Evaluation of HemR as well as organ response was performed at our outpatient amyloidosis clinic after every 3 cycles of M-Dex. The study was performed in accordance with the Good Clinical Practice guidelines and the Declaration of Helsinki, and was approved by the Ethics Committee of the University of Heidelberg as well as the competent authority. The Data Monitoring Committee was informed every 6 months about safety data and occurrence of serious adverse events (SAE).

Assessment Baseline assessments and procedures included physical examination, amyloid organ involvement, standard laboratory values as well as serum M-protein analysis, free light chains, NT-proBNP and cardiac troponin T (cTNT)/high-sensitive cTnT (hsTNT). Once a week blood evaluations were performed by the local physician (complete blood counts, creatinine, potassium, bilirubin, C–reactive protein[CRP]). The results were reviewed by the study physician and followed up by a personal phone call with the patient. In the case of any significant clinical problems lenalidomide was withheld until the next evaluation. Bone marrow aspiration and cytogenetic analysis with interphase fluorescence in situ hybridization (iFISH) after CD138+ enrichment was carried out in all patients. Organ involvement and response to therapy were assessed according to the International Consensus Criteria.16 Dose reduction due to hematological toxicity was performed sequentially (first melphalan, second lenalidomide). Dose reduction of dexamethasone was done by 25% in patients with fluid overload in the previous cycle. Mayo cardiac staging was retrospectively applied in all patients using NT-ProBNP and troponin (cTNT or hsTNT).18,19 In addition, organ response was also retrospectively analyzed using this new cardiac criteria.15

Statistical analysis A one-sided exact binomial test was used to test the null hypothesis that the probability of achieving a CR after 6 cycles of L-Mel-Dex is not larger than 16%; this had been the CR rate of patients receiving at least 3 cycles of M-Dex in our institution. Overall survival (OS) was defined as time until death by any cause (failure time) or date of last follow up (censored time). For eventfree survival (EFS) an event was defined as either death by any 1425

U. Hegenbart et al. Table 1. Patient characteristics.

Patient cohort

L-M-Dex

Age, years, median (range) 66.6 (47-75) Sex Female 25 Male 25 Monoclonal light chain, number of patients κ 4 l 46 Absolute involved free light chain concentration, mg/l, median (range): κ 121 (2-591) l 181 (19-3510) dFLC 161 (0-3508) Plasma cell content of the bone marrow, 10 (1-31) median % (range) Number of organs involved, number of patients 1 organ 9 2 organs 22 > 3 organs 19 Dominant organ, number of patients* Heart 26 Liver 2 Gut 1 Kidney 24 Peripheral Nerves 1 Other 6 NT-proBNP, ng/l, median (range) 2784 (54-15066) hsTNT (pg/ml) median (range) 29.5 (0-263) Cardiac involvement, number of patients 36 Mayo staging system,16 number of patients: 1 8 2 24 3 18 Creatinine clearance, ml/min, median (range) 71(41-160) Renal staging system,23 number of patients** 1 11 2 16 3 4 *patients might have several dominant organ involvements, therefore the sum of patients is larger than 50. **only given for patients with renal involvement. NTproBNP: N-terminal prohormone of brain natriuretic; cTNT: cardiac troponin T; hsTNT: high-sensitive cTnT; dFLC: difference between the involved and uninvolved free light chain; L-M-Dex: lenalidomide, melphalan and dexamethasone.

cause, hematological relapse/progression or second-line chemotherapy.16,17 All patients without an event were censored at the date of last contact, defined as date of last visit/response evaluation. Distributions of survival times were estimated by using the method of Kaplan & Meier. 95% confidence intervals (95% CI) were computed using Greenwood's formula for the variance of the Kaplan–Meier estimator. Comparisons of 2 survival curves were performed using the log-rank test. The distribution of follow up times were calculated by the reverse Kaplan-Meier estimate.20 The prognostic impact of treatment in the study cohort and historical M-Dex cohort on EFS and OS was evaluated by the Cox proportional hazards regression model. Estimated hazard ratios (HR) and their corresponding 95% CI were used to express effect sizes. 1426

Table 2. Distribution of the 132 adverse events (AEs) > grade 3 and serious adverse events (SAEs).

Type of AE

Grade 3

Cardiac Hypotension/syncope Thrombosis Neutropenia Lymphopenia Thrombocytopenia Infections Anemia Kidney Liver Gastrointestinal Other

8 5 1 12 3 4 14 1 2 6 6 51

Grade 4

Grade 5

SAE

5 1

1 5 4

The proportional hazards assumption was tested as proposed by Grambsch et al.21 Multivariable Cox proportional hazards regression models were used to adjust effects for additional covariates (Online Supplementary Table S1). For dFLC log2-transformed data were used and HRs are reported according to a 2-fold increase of the original values. The influence of clinical covariates on the CR+VGPR rates was investigated using multivariable logistic regression (Online Supplementary Table S1). Fisher’s exact tests were used for group comparisons of cytogenetic aberrations, response and distributions between the L-M-Dex and historical M-Dex cohorts. Tests are considered to be statistically significant if their corresponding P-value is ≤0.05. All analyses were performed with the statistical software environment R, version 3.1.1.22

Results Patient characteristics Fifty patients with newly diagnosed AL amyloidosis were included. Characteristics of the study cohort are shown in Table 1. Median age at start of L-M-Dex treatment was 66.6 years. Cardiac involvement was present in 36 patients (72%). The median NT-proBNP level was 2.784 ng/l, 6 patients (12%) had >8.500 ng/l. According to the cardiac Mayo 2004 staging system, 8 patients (16%) were stage 1, 24 patients (48%) were stage 2 and 18 patients (36%) were stage 3. The second most affected organ was the kidney (31 patients, 62%). Eleven of these patients were in renal stage 1, 16 patients in stage 2, and 4 patients in stage 3.23

Toxicity All 50 patients were assessed for safety. Thirty-five patients received 6 cycles, and 6 patients completed their 6 cycles without any dose reduction. In total, 136 dose modifications for 43 patients were reported. Lenalidomide dose was reduced or paused in 1 or more cycles in 42 patients and subsequently reintroduced in the next cycle in all of these patients. Melphalan dose was reduced in 20 patients due to hematologic or renal toxicity and the reduced dosage was kept as such in the succeeding cycles. However, all patients received melphalan until the last treatment cycle. One-hundred thirty-two adverse events (AE) of grade >3, according to the Common Terminology haematologica | 2017; 102(8)

Len-Mel-Dex is effective in transplant ineligible AL amyloidosis

Figure 1. Hematologic remission. Distribution of hematologic remission after 3, 6, 9 and 12 months after start of L-M-Dex. pts: patients; SD / PROG: stable disease / progression; PR: partial remission; VGPR: very good partial remission; CR: complete remission.

Criteria for Adverse Events (CTCAE), were recorded in 50 patients including 16 severe AEs (Table 2; at least one AE grade 3 or 4 occurred in all study patients) and all were considered to be treatment-related, as it is often difficult to distinguish between side effects and amyloidosis-related events. Twenty-one hematologic AEs were observed; neutropenia 76%, CTCAE grade 4 in 1 patient and grade 3 in 12 patients, grade 3 thrombocytopenia was reported in 3 cases, anemia grade 3 (<8 g/dl) in 2 cases, and lymphopenia in 3 patients. The most common non-hematologic AE (14 patients) was worsening of cardiac function (e.g., worsening of cardiac failure, atrial fibrillation) or symptoms of autonomic neuropathy. As expected, the median NT-proBNP value increased from baseline (2.784 ng/l, n=50 patients) to 3 months (5.560 ng/l, n=47 patients) and in 38 patients more than 30% compared to baseline, whereas renal function was stable in most patients (baseline median creatinine value 1.0 mg/dl and at 3 months 1.1 mg/dl). The median creatinine level was equal at the start of cycle 2 and 3 (1.1 mg/dl, range 0.5-2.4 and 0.5-3.9 mg/dl, respectively). Eight patients suffered from an infection (1 patient died from sepsis in the first cycle, 4 patients had bronchitis or pneumonia, further AEs classified as infection were fever of unknown origin, gastroenteritis, erysipelas and CRP elevation in several cases (7 patients grade 4). One patient developed acute renal failure and 1 patient a deep vein thrombosis. These latter 2 patients remained within the study following adequate treatment and resolution of the AEs.

Feasibility, hematological remission and organ response Forty-five patients (90%) received 3 cycles and 37 patients (74%) completed the treatment with 5 or 6 cycles. Overall, 253 cycles were administered. Therefore, 45 out haematologica | 2017; 102(8)

of 50 patients were evaluable for the primary endpoint CR rate (1 patient died before the first cycle of L-M-Dex was completed, and 4 additional patients discontinued L-MDex after the first or second cycle due to the deterioration of their performance status mainly related to amyloidosis). After 3 cycles 2 patients achieved CR, 17 patients VGPR and 15 patients PR, respectively. After treatment completion 9 patients achieved CR, 16 VGPR and 9 PR. CR+VGPR rate improved from 38% to 50% from cycle 3 to cycle 6 (ITT analysis, n=50). Five patients received more than 3 but less than 6 cycles: 3 patients received 4 cycles, and 2 received 5 cycles. Two of those with 5 cycles achieved a remission (1 PR and 1 VGPR after 5 cycles) but chose to stop due to moderate toxicity. Interestingly, 3 additional patients achieved a negative immunofixation without further chemotherapy in the follow up at 7 months post start of therapy. With a CR rate of 20% (9/45) among patients with at least 3 cycles, the primary objective of the study was not achieved (P=0.29, one-sided exact binominal test). However, using the 7 month evaluation the primary objective would have been reached with a CR rate of 26.7% (12/45; P=0.05, one-sided exact binominal test). Organ response16 was observed in 8 patients at 6 months after end of treatment. Using the new criteria for organ response,17 10 and 20 patients had an organ response after 12 and 24 months following start of therapy, respectively. Interestingly, another 4 patients developed an organ response later than 24 months without receiving new treatment (overall 24/50, 48%, 95% CI 34%-63%). Organ responses were observed in the heart (decrease of NT-BNP) and kidney in 10 patients (decrease of proteinuria; 1 patient had an organ response in both organs). Three other organ responses were seen in the soft tissue (normalization of factor X), autonomic nerve system and liver. 1427

U. Hegenbart et al.

Survival and Progression Median follow-up time from start of treatment was 50 months (range 1-72). The median duration of CR and VGPR was 39 months (range 1-72) and at 12 months the CR rate was still 20% (Figure 1). Fourteen patients relapsed or progressed with their underlying plasma cell dyscrasia after a median of 25.1 months from start of therapy (95% CI, 20.6–infinity). The duration of organ response in patients with CR/VGPR was 35 months (range 10-67). Twenty patients received second-line therapy (mostly proteasome inhibitors) because they did not respond to L-M-Dex or developed hematological or organ progression. The early death rate was low with 4% (1 death due to septicemia during the first treatment cycle and 1 cardiac death after 3 cycles). Overall, 20 deaths were observed, among them 19 due to progressive disease. The median OS was 67.5 months (95% CI, 49.6-infinity) and median EFS time was 25.1 months (95% CI, 20.6-infinity) (Figure 2). Cox univariate regression showed significant influence of cardiac Mayo 2004 score (3 vs. 1/2) on OS and EFS (P<0.001 in both analyses, Figure 3A,B).

iFISH results We were able to detect at least one cytogenetic abnormality in all 50 study patients. Twelve patients (24%) had gain of 1q21 and translocation t(11;14) was detected in 28 patients (56%). High-risk cytogenetic aberrations (t(4;14), t(14;16), del17p) were only seen in 3 patients and were not further analyzed due to this low number. Gain of 1q21 had no negative influence on outcome. CR+VGPR rate after treatment completion was positively influenced by gain of 1q21 (10/12, 83% vs. 15/38, 39% of patients, P=0.02, Fisher’s exact test). However, there was no significant survival benefit: the HR in the univariate Cox regression of gain 1q21 with no gain as a reference was 0.47 for EFS (95% CI: 0.18-21.24, P=0.13) and 0.66 for OS (95% CI: 0.22-2.01, P=0.47). In patients with translocation t(11;14), the rate of CR+VGPR was significantly lower (32% as compared with 73%, P=0.01, Fisher’s exact test). However, univariate Cox regression revealed no significant adverse influence of t(11;14) on EFS (P=0.09, HR 1.91 [95% CI:0.914.01]) or on OS (P=0.21, HR 1.83 [95% CI: 0.72-4.66]).

Comparison with historical M-Dex We obtained a comparison cohort of 49 consecutive AL patients treated with M-Dex from 2004 to 2009 (median follow up was 87 months). Patient characteristics are shown in the Online Supplementary Table S2. The 2 cohorts were comparable regarding the main clinical characteristics. Thirty-eight patients (78%) completed at least 3 cycles and 22 patients (45%) completed at least 6 M-Dex cycles. Contrary to L-M-Dex, 13 patients received more than 6 cycles of M-Dex. In the L-M-Dex study group, the CR+VGPR rate was higher (25/50, 50%) as compared to the M-Dex group (12/49, 24%, P=0.01, Fisher’s exact test). There was also a longer EFS and OS in the L-M-Dex group (Figure 2) compared to the M-Dex study group (Online Supplementary Figure S1), with a median EFS of 25 vs. 16 months (P=0.005, log-rank test) and a median OS of 67.5 vs. 26.2 months, (P=0.02, log-rank test). In the multivariable regression analysis (Online Supplementary Table S1), L-MDex was again significantly associated with a higher CR+VGPR rate but lost its significance for EFS and OS, 1428

Figure 2. Survival of patients with lenalidomide, melphalan and dexamethasone. Estimated event-free survival (EFS) and overall survival (OS) in the L-M-Dex study group.

whereas the negative prognostic role of Mayo 2004 score 3 was confirmed for all endpoints. However, all these differences were mainly caused by the higher early mortality rate as only 2 L-M-Dex patients (2/50, 4%) died within 3 months following the start of chemotherapy compared to 10 patients (10/49, 20%) in the M-Dex-group (P=0.01, Fisher’s exact test). As a result, the 3-month landmark analyses of EFS and OS no longer reached statistically significant differences (data not shown).

Discussion We report on 50 patients with newly diagnosed AL amyloidosis who received lenalidomide, melphalan and dexamethasone. Poor cardiac function itself was not an exclusion criterion (36% had cardiac Mayo 2004 score 3), but kidney function had to be preserved (creatinine clearance >40 ml/min). To improve tolerability we used 10 mg of lenalidomide instead of the 15 mg proposed by the French phase 1 trial, and 20 mg dexamethasone.5 Treatment resulted in a high rate of CR+VGPR (50%, ITT analysis after therapy completion). Organ response17 was observed in 10 patients 12 months after start of therapy and thereafter increased further in patients with long lasting remission by up to 48%. Sixteen SAEs and 132 AEs grade 3 or 4 occurred in 50 patients, including 1 death due to septicemia during the first treatment cycle and 1 cardiac death after 3 cycles (early mortality rate was only 4%). Most common toxicities were of a hematologic and cardiac type, leading to a (temporary) reduction of at least 1 of the study drugs in 88% of patients. Nevertheless, the targeted number of 6 cycles could be administered in 74% of the patients. Importantly, polyneuropathy neither occurred nor worsened in any of the patients. We imposed a very strict observation with a minimum of 4 weekly visits at our center, weekly evaluations of blood counts and phone calls. In case of symptoms or signs of infections or neutropenia less than 1.0/nl, lenalidomide was immediately stopped. After a median follow up of 50 months the median OS was 67.5 months. Estimated 2- and 4-year OS and EFS were remarkably good with 74/63% and 54/38%, respectively. haematologica | 2017; 102(8)

Len-Mel-Dex is effective in transplant ineligible AL amyloidosis Table 3. Review of L-M-Dex treatment.

Reference

No. of pts % upfront

Single or multicenter recruitment Median follow up

Patient description

Median NT-BNP age (ng/l) (range)

Moreau5

26 100

Multicenter Newly diagnosed 57 yrs 2008-09 Creatinine < 150 umol/l (27-70) 19 months ECOG status < 2

Sanchorawala6

16 69

Single center 2008-11 34 months

No end-stage renal failure SWOG PS < 2

70 yrs not (57-84) reported

Dinner7

25 92

Single center 2009-2012 6 months

No exclusion criteria

67 yrs (52-84)

2200

This study

50 100

Single center Newly diagnosed 2009-2012 Not eligible for HDC 50 months Creatinine clearance > 40 ml/min

67 yrs (47-75)

2900

1100

L-M-Dex dosage Toxicity HemR / CR OS at 6 Number CTCAE (%)* mo and 2 yrs of cycles planned grade > 3, /median number % of patients of cycles received (n) 4 cohorts L 5-20 mg Melphalan 0,18 mg/kg Dex 40 mg, days 1-4 9 / 7 cycles L 10 mg Melphalan 5 mg/m2 Dex 20-40 mg once a week 12 / 6 cycles L 10 mg Melphalan 0,18 mg/kg Dex 40 mg once a week 9 / 3 cycles L 10 mg Melphalan 0,15 mg/kg Dex: 20 mg 6/6

58/23

82/82%

44/6

95/70%

100

58/8

58/58%

100

68/18

86/74%

A 28 days cycle was used in all L-M-Dex studies. Lenalidomide was given days 1-21; Melphalan was given on day 1 â&#x20AC;&#x201C; 4. *HemR rate (PR or better) is reported as an intention-to-treat analysis. No data regarding VGPR were given in the 3 papers. Definitions of events for calculation of EFS were not uniform and are therefore not listed. Definition of OR has changed over time, therefore OR rates are also not listed. CR: complete remission; CTCAE: Common Terminology Criteria for Adverse Events; HDC: High-dose chemotherapy; HemR: hematological Response; L-M-Dex: lenalidomide, melphalan and dexamethasone; mo: months; yrs: years; OS: overall survival; pts: patients; NT-BNP: N-terminal B-type natriuretic peptide; ECOG: Eastern Cooperative Oncology Group; SWOG PS: Southwest Oncology Group performance status.

Comparison with other trials Three other prospective clinical studies with fewer patients and shorter follow up have been published using the same combination. Results are summarized in Table 3 with the focus on inclusion criteria, drug dosages, toxicity, HemR (as an ITT analysis, although time points of evaluation could not be harmonized) and OS (although the 2 USA trials also included relapsed patients). Moreau et al. performed a small phase 1/2 study in 26 patients.5 Patients of all ages could be included. No doselimiting toxicity was observed and the maximum tolerated dosage was defined as lenalidomide 15 mg (together with 40 mg of Dex). The short-term survival (median follow up 19 months) was very good with an estimated 2year survival of 81%. Fifty-eight percent of patients achieved a HemR, 23% a CR and 50% an organ response, and the authors concluded that the dose escalation schedule tended to underestimate response rates. The phase 2 study of Sanchorawala et al. tested the tolerability, HemR and organ response rate in 16 patients (one-third of whom were not treated upfront).6 It was stopped prior to the accrual goal due to toxicities and limited efficacy. The CR rate was lower compared to that reported by Moreau et al.5 at 6%, and only 6 cycles, on average, of the planned 12 could be administered. After discontinuing and publishing the Boston trial, Palumbo and Cavallo advised against this combination treatment in AL patients.24 Subsequently, Dinner et al. published the results of a third trial, and also concluded that this triple combination was toxic and rather ineffective.7 Again, CR was low at 8%. This study included 25 patients, the oldest being 84 years old. The haematologica | 2017; 102(8)

planned Dex dosage was 40 mg. On average, only 3 of the planned 9 cycles could be given, and the OS rate at 6 months was 58%. In our judgment the discrepant results are mostly due to large differences regarding inclusion criteria (age, severity of disease) and dosages of lenalidomide and Dex. The patient cohort of the French trial was younger, had a good performance status and kidney function; some of them were probably also eligible for HDC. Therefore, patients were able to receive a median of 7 of the 9 planned treatment cycles. In the 2 USA trials patients were older, had less strict exclusion criteria and also included patients with relapse or no response after previous chemotherapy, as a consequence of which tolerability and outcome were worse. In our experience, the triple combination chemotherapy was feasible and effective for non-HDC candidates not older than 75 years. Our rigorous surveillance strategy might have avoided a higher toxicity.

Data of patients with Cardiac Mayio 2004 stage 3 The outcome of patients with cardiac Mayo 2004 stage 3 is poor due to the high early mortality.18 In a large retrospective European collaborative study the median OS for Mayo stage 3 patients was 7 months and longer depending on the NT-ProBNP level.25 Patients who died within 3 months had a significantly higher NT-proBNP level (11.794 vs. 7.957 ng/l). Therefore, patients with very high NT-ProBNP levels (e.g., >8.500 ng/l) are mostly excluded from clinical trials. In a recently published multicenter study 60 patients with cardiac stage Mayo 2004 3 were treated with a combination of bortezomib, cyclophos1429

U. Hegenbart et al. A

Figure 3. Survival in relation to the cardiac Mayo Stage. Estimated (A) event-free survival (EFS) and (B) overall survival (OS) in the L-M-Dex study group depending on cardiac Mayo stage 1/2 compared to Mayo stage 3.

phamide and dexamethasone.26 Forty percent of patients died during therapy, the 2-year OS rate was about 50%. In our study, the 2-year OS rate was slightly better than 40% in stage 3 patients (Figure 3B). We summarize that treatment of stage 3 patients is still a challenge and a balancing act between the toxicity of chemotherapy and the goal of >VGPR achievement.

IFISH The prognostic impact of iFISH results was not the same when compared to the patient cohort treated with M-Dex alone.27 In the study herein, gain of 1q21 lost it's negative influence on remission and survival. This might be explained by the addition of lenalidomide. However, these results should be interpreted with caution as the patient number was smaller than in our previous analyses and thus did not allow for a meaningful multivariable analysis.27,28 As of yet, no other published data regarding the influence of iFISH in L-M-Dex exists.29

Comparison with historical cohort of M-DEX To further explore the role of lenalidomide in a triple therapy we used a historical M-Dex cohort. Although the relevant patient characteristics did not differ between LM-Dex and M-Dex, the early mortality rate was higher in the M-Dex cohort (20% vs. 4%) leading to a higher CR+VGPR rate, and longer EFS and OS rates in the L-MDex study cohort. These differences diminished in the 3month landmark analyses. We can only speculate about the causes of the higher early mortality; in our view this might be best explained by the rigid surveillance and the lower Dex dosage we used within the trial. Overall comparisons between the 2 cohorts should therefore be drawn

References 1. Merlini G, Seldin DC, Gertz MA. Amyloidosis: pathogenesis and new therapeutic options. J Clin Oncol. 2011; 29(14):1924-1933.

1430

cautiously given the retrospective nature of the M-Dex analysis and the fact that some of these patients were treated outside of our center.

Conclusion We conclude that the combination treatment of L-M-Dex is effective in patients who are not eligible for HDC. In spite of concerns regarding toxicity raised by prior studies, the LM-Dex regimen could be safely administered in our study. Definitively, a rigid surveillance is needed to immediately modify the dosage of lenalidomide in order to reduce toxicity and mortality. Therefore, this combination therapy is probably best performed by amyloidosis referral centers. Up to now, no direct comparison of Bortezomib-M-Dex versus L-M-Dex as upfront treatment has been performed, thus the best combination of standard chemotherapy with a novel agent remains elusive. In our opinion, principally those AL patients presenting with polyneuropathy, high dFLC levels and who are ineligible for HDM, should be considered for L-M-Dex. Acknowledgments We thank Celgene for their financial support for this investigator-initiated study, and also for supplying lenalidomide at no charge. We would like to acknowledge the excellent collaboration of hematologists from outside hospitals and private practices. We thank our patients for their participation in this trial. We thank the three members of the Data Safety Monitoring Board: M. Kaufmann, MD, Stuttgart. E. Graf, PhD in statistics, and Freiburg, M. Hensel, MD, Mannheim, for their work and advice.

2. Cibeira MT, Sanchorawala V, Seldin DC, et al. Outcome of AL amyloidosis after highdose melphalan and autologous stem cell transplantation: long-term results in a series of 421 patients. Blood. 2011;118(16):43464352. 3. Palladini G, Milani P, Foli A, et al. Oral mel-

phalan and dexamethasone grants extended survival with minimal toxicity in AL amyloidosis: long-term results of a riskadapted approach. Haematologica. 2014;99(4):743-750. 4. Dietrich S, Schonland SO, Benner A, et al. Treatment with intravenous melphalan and

haematologica | 2017; 102(8)

Len-Mel-Dex is effective in transplant ineligible AL amyloidosis

10.

11.

12.

dexamethasone is not able to overcome the poor prognosis of patients with newly diagnosed systemic light chain amyloidosis and severe cardiac involvement. Blood;116(4):522-528. Moreau P, Jaccard A, Benboubker L, et al. Lenalidomide in combination with melphalan and dexamethasone in patients with newly diagnosed AL amyloidosis: a multicenter phase 1/2 dose-escalation study. Blood. 2010;116(23):4777-4782. Sanchorawala V, Patel JM, Sloan JM. Melphalan, lenalidomide and dexamethasone for the treatment of immunoglobulin light chain amyloidosis: results of a phase II trial. Haematologica. 2013;98(5):789-792. Dinner S, Witteles W, Afghahi A, et al. Lenalidomide, melphalan and dexamethasone in a population of patients with immunoglobulin light chain amyloidosis with high rates of advanced cardiac involvement. Haematologica. 2013;98(10):1593-1599. Palladini G, Milani P, Foli A, et al. Melphalan and dexamethasone with or without bortezomib in newly diagnosed AL amyloidosis: a matched case-control study on 174 patients. Leukemia. 2014;28(12):2311-2316. Kastritis E, Leleu X, Arnulf B, et al. A randomized phase III trial of melphalan and dexamethasone (MDex) versus bortezomib, melphalan and dexamethasone (BMDex) for untreated patients with AL amyloidosis. Blood 2016 128:646; abstr. Palladini G, Russo P, Milani P, et al. A phase II trial of cyclophosphamide, lenalidomide and dexamethasone in previously treated patients with AL amyloidosis. Haematologica. 2013;98(3):433-436. Cibeira MT, Oriol A, Lahuerta JJ, Mateos MV, et al. A phase II trial of lenalidomide, dexamethasone and cyclophosphamide for newly diagnosed patients with systemic immunoglobulin light chain amyloidosis. Br J Haematol 2015;170(6):804-813. Kumar SK, Hayman SR, Buadi FK, et al. A.

haematologica | 2017; 102(8)

13.

14.

15.

16.

17.

18.

19.

20.

Lenalidomide, cyclophosphamide, and dexamethasone (CRd) for light-chain amyloidosis: long-term results from a phase 2 trial. Blood. 2012;119(21):4860-4867. Kastritis E, Terpos E, Roussou M, et al. A phase 1/2 study of lenalidomide with lowdose oral cyclophosphamide and low-dose dexamethasone (RdC) in AL amyloidosis. Blood. 2012;119(23):5384-5390. Schonland SO, Dreger P, de Witte T, Hegenbart U. Current status of hematopoietic cell transplantation in the treatment of systemic amyloid light-chain amyloidosis. Bone Marrow Transplant. 2012;47(7):895-905. Tapan U, Seldin DC, Finn KT, et al. Increases in B-type natriuretic peptide (BNP) during treatment with lenalidomide in AL amyloidosis. Blood. 2010;116(23):5071-5072. Gertz MA, Comenzo R, Falk RH, et al. Definition of organ involvement and treatment response in immunoglobulin light chain amyloidosis (AL): a consensus opinion from the 10th International Symposium on Amyloid and Amyloidosis, Tours, France, 18-22 April 2004. Am J Hematol 2005;79(4):319-328. Palladini G, Dispenzieri A, Gertz MA, et al. New criteria for response to treatment in immunoglobulin light chain amyloidosis based on free light chain measurement and cardiac biomarkers: impact on survival outcomes. J Clin Oncol 2012;30(36):45414549. Dispenzieri A, Gertz MA, Kyle RA, et al. Serum cardiac troponins and N-terminal pro-brain natriuretic peptide: a staging system for primary systemic amyloidosis. J Clin Oncol 2004;22(18):3751-3757. Kristen AV, Giannitsis E, Lehrke S, et al. Assessment of disease severity and outcome in patients with systemic light-chain amyloidosis by the high-sensitivity troponin T assay. Blood. 2010;116(14):24552461. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time.

Control Clin Trials 1996;17(4):343-346. 21. Grambsch P, Therneau T. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika. 1994;81:515â&#x20AC;&#x201C;526. 22. R Core Team (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.URL http://www.R-project.org/. 23. Palladini G, Hegenbart U, Milani P, et al. A staging system for renal outcome and early markers of renal response to chemotherapy in AL amyloidosis. Blood. 2014; 124(15):2325-2332. 24. Palumbo A, Cavallo F. Lenalidomide in the treatment of plasma cell dyscrasia: state of the art and perspectives. Haematologica 2013;98(5):660-661. 25. Wechalekar AD, Schonland SO, Kastritis E, et al. A European collaborative study of treatment outcomes in 346 patients with cardiac stage III AL amyloidosis. Blood. 2013;121(17):3420-3427. 26. Jaccard A, Comenzo RL, Hari P, et al. Efficacy of bortezomib, cyclophosphamide and dexamethasone in treatment-naive patients with high-risk cardiac AL amyloidosis (Mayo Clinic stage III). Haematologica. 2014;99(9):1479-1485. 27. Bochtler T, Hegenbart U, Kunz C, et al. Gain of chromosome 1q21 is an independent adverse prognostic factor in light chain amyloidosis patients treated with melphalan/dexamethasone. Amyloid 2014;21(1):917. 28. Bochtler T, Hegenbart U, Kunz C, et al. Translocation t(11;14) is associated with adverse outcome in patients with newly diagnosed AL amyloidosis when treated with bortezomib-based regimens. J Clin Oncol 2015;33(12):1371-1378 29. Muchtar E, Dispenzieri A, Kumar SK, et al. Interphase fluorescence in-situ hybridization (iFISH) in untreated AL amyloidosis has an independent prognostic impact by abnormality type and treatment category. Leukemia. 2016. [Epub ahead of print].

1431

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Plasma Cell Disorders

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1432-1438

Longitudinal fluorescence in situ hybridization reveals cytogenetic evolution in myeloma relapsing after autologous transplantation Maximilian Merz,1 Anna Jauch,2 Thomas Hielscher, 3 Elias K. Mai,1 Anja Seckinger,1 Dirk Hose,1 Uta Bertsch,1 Kai Neben,1 Marc S. Raab,1,4 Hans Salwender,5 Igor W. Blau,6 Hans-Walter Lindemann,7 Ingo Schmidt-Wolf,8 Christof Scheid,9 Mathias Haenel,10 Katja Weisel,11 Hartmut Goldschmidt1,12 and Jens Hillengass1,13

Medizinische Klinik V, University Hospital Heidelberg; 2Institute of Human Genetics, University Heidelberg; 3Division of Biostatistics, German Cancer Research Center (DKFZ), Heidelberg; 4Max-Eder Research Group Experimental Therapies for Hematologic Malignancies, DKFZ, Heidelberg; 5Asklepios Klinik Altona, Hamburg; 6Department of Internal Medicine III, Charité Campus Benjamin Franklin, Berlin; 7Hämatologie/Onkologie, Kath. Krankenhaus Hagen gem. GmbH - St.-Marien-Hospital, Hagen; 8Center for Integrated Oncology, Med. Klinik und Poliklinik III, University of Bonn; 9Department of Internal Medicine I, University of Cologne; 10Klinikum Chemnitz gGmbH; 11University Hospital of Tübingen; 12National Center for Tumor Diseases (NCT), Heidelberg and 13Department of Radiology, German Cancer Research Center DKFZ, Heidelberg, Germany 1

ABSTRACT

T Correspondence: maximilian.merz@med.uni-heidelberg.de

Received: February 28, 2017. Accepted: May 8, 2017. Pre-published: May 11, 2017. doi:10.3324/haematol.2017.168005 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1432 ©2017 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.

1432

o investigate cytogenetic evolution after upfront autologous stem cell transplantation for newly diagnosed myeloma we retrospectively analyzed fluorescence in situ hybridization results of 128 patients with paired bone marrow samples from the time of primary diagnosis and at relapse. High-risk cytogenetic abnormalities (deletion 17p and/or gain 1q21) occurred more frequently after relapse (odds ratio: 6.33; 95% confidence interval: 1.86-33.42; P<0.001). No significant changes were observed for defined IGH translocations [t(4;14); t(11;14); t(14;16)] or hyperdiploid karyotypes between primary diagnosis and relapse. IGH translocations with unknown partners occurred more frequently at relapse. New deletion 17p and/or gain 1q21 were associated with cytogenetic heterogeneity, since some de novo lesions with different copy numbers were present only in subclones. No distinct baseline characteristics were associated with the occurrence of new high-risk cytogenetic abnormalities after progression. Patients who relapsed after novel agent-based induction therapy had an increased risk of developing highrisk aberrations (odds ratio 10.82; 95% confidence interval: 1.65-127.66; P=0.03) compared to those who were treated with conventional chemotherapy. Survival analysis revealed dismal outcomes regardless of whether high-risk aberrations were present at baseline (hazard ratio, 3.53; 95% confidence interval: 1.53-8.14; P=0.003) or developed at relapse only (hazard ratio, 3.06; 95% confidence interval: 1.09-8.59; P=0.03). Our results demonstrate cytogenetic evolution towards highrisk disease after autologous transplantation and underline the importance of repeated genetic testing in relapsed myeloma (EudraCT number of the HD4 trial: 2004-000944-26).

Introduction Multiple myeloma (MM) is a genetically complex and heterogeneous disease.1,2 At the chromosomal level, MM can be subdivided according to the tumor-initiating event into hyperdiploid and non-hyperdiploid myeloma.3 While hyperdiploid MM is defined by gains of odd-numbered chromosomes, non-hyperdiploid myeloma mainly harbors IGH translocations.3 Patients with high-risk cytogenetic abnormalities at haematologica | 2017; 102(8)

Cytogenetic evolution of myeloma

Table 1. Baseline characteristics of patients in the HD4 trial and non-study patients.

HD 4 trial Variable Age(median) Sex ISS

Induction

VGPR after induction VGPR after ASCT

Level

55.5 Male Female I II III PAD TAD VAD Other 14 25

58.5 29 15 20 16 8 22 0 22 0 31.8 56.8

57.0 65.9 34.1 45.5 36.4 18.2 50 0 50 0 32 43

Non-study patients n % 55 29 21 15 21 45 11 21 7 46.4 62.3

65.5 34.5 36.8 26.3 36.8 53.6 13.1 25.0 8.3 46 68

All n

84 44 41 31 29 67 11 43 7 40.7 60.2

65.6 34.4 40.6 30.7 28.7 52.3 8.6 33.6 5.5

ISS: International Staging System; PAD: bortezomib, doxorubicin, dexamethasone; TAD: thalidomide, doxorubicin, dexamethasone; VAD: vincristine, doxorubicin, dexamethasone; VGPR: very good partial remission; ASCT: autologous stem cell transplantation.

primary diagnosis, such as gain of chromosome 1q21, deletion 17p (del17p) and translocation t(4;14), show inferior outcome after high-dose therapy and autologous stem cell transplantation (ASCT), even in the era of novel agents.4,5 Recent studies have demonstrated that disease progression and refractoriness are caused by secondary genetic events.6,7 It has been discussed that systemic treatment might select pre-existing aggressive subclones or cause secondary genetic events and clonal evolution in patients with recurrent disease. Patients without the aforementioned abnormalities at primary diagnosis might, therefore, relapse with cytogenetically defined high-risk disease. However, there are only limited longitudinal data available elucidating cytogenetic changes in relapsed MM after primary therapy.6,8 We therefore performed a retrospective analysis of patients treated with upfront ASCT with an interphase fluorescence in situ hybridization (FISH) analysis of purified plasma cells at primary diagnosis and relapse. We tried to characterize cytogenetic evolution and studied which abnormalities occur upon relapse. We analyzed whether baseline characteristics, treatment or response affect cytogenetic evolution and investigated the prognostic significance of new high-risk abnormalities.

Methods Patients and treatment We identified 128 patients treated with ASCT for newly diagnosed MM with a FISH analysis at initial diagnosis (first FISH) and relapse (second FISH). Forty-four patients were initially enrolled in the prospective HD4 phase III trial of the German-Speaking Myeloma Multicenter Group (GMMG, EudraCT number: 2004000944-26). Results of the GMMG HD4 trial have been published previously.9 In brief, patients were randomly assigned to either the control arm of three cycles of VAD (vincristine, doxorubicin, dexamethasone) followed by tandem-ASCT and thalidomide maintenance for 2 years, or the experimental arm with three cycles of PAD (bortezomib, doxorubicin, dexamethasone) followed by tandem-ASCT and bortezomib maintenance therapy for 2 years. Furthermore we identified 84 patients who were treated at our institution outside the HD4 trial with comparable induction and maintenance therapy regimens before and after ASCT (non-study haematologica | 2017; 102(8)

patients, NSP). Table 1 summarizes baseline characteristics, treatment and response of both populations. The median time to second FISH analysis from progressive disease was 7.4 months (HD4 patients) and 4.4 months (NSP). The study was performed in accordance with the Declaration of Helsinki after informed consent had been obtained. Retrospective analysis was approved by the local ethics committee.

Fluorescence in situ hybridization studies FISH analyses were performed on CD138-purified plasma cells as described previously10 at the Institute of Human Genetics at Heidelberg University Hospital. The following probes were used: gain1q21, gain4p16, gain5q35, gain5p15, del6q21, del8p21, gain9q34, gain11q13, gain11q23, del13q14, gain14q32, gain15q22, del17p13, gain19q13, del22q11, t(4;14), t(11;14), t(14;16) and a probe for IGH splits. The threshold for all aberrations was 10%. If an aberration was found in 10-60% of cells it was defined as subclonal; if it was found in more than 60% it was defined as being the major clone. High-risk cytogenetic abnormalities were defined by the presence of del17p, t(4;14) or gain 1q21.

Statistical analysis An exact McNemar test was used to assess changes between results of FISH assessments at baseline and relapse. A Fisher exact test was used to compare categorical parameters between groups. A multivariable logistic regression model was fitted accounting for baseline karyotype [hyperdiploid or t(11;14)], International Staging System (ISS) stage, time to first progression and to second FISH analysis, age, sex and type of induction therapy to investigate factors associated with the occurrence of high-risk cytogenetic abnormalities at relapse. A Wilcoxon test was used to compare time to progression, which was measured in months and defined as time from the start of chemotherapy to date of progression. Age in years was analyzed as a continuous parameter. The Kaplan-Meier method and log-rank test were used to analyze differences in overall survival times between patients with cytogenetic abnormalities at both FISH assessments and at the first or second FISH only. To analyze effects on overall survival, a multivariable Cox regression model accounting for the same variables as mentioned above was used. P-values <0.05 were considered statistically significant. The last follow-up evaluations were performed in June 2016 (NSP) and November 2015 (HD4). Analyses were carried out with R 3.3 statistical software. 1433

M. Merz et al. A

Figure 1. Clonal evolution in patients with gain of 1q21 and del17p. (A) Distribution of copy numbers of chromosome 1q21. Patients with increasing copy numbers after relapse are represented in red, patients with decreasing copy numbers in blue. (B-D) Examples of clonal evolution in three patients with de novo del17p. Panel B illustrates cause of the disease in a patient with minimal response (MR) after autologous transplantation (TPL). After relapse (PD), the second FISH analysis showed a new del17p present in 60% of analyzed plasma cells and no significant changes of the abnormalities already present at initial diagnosis. The patient represented in panel C achieved a very good partial remission (VGPR). The initially present hyperdiploid clone harboring a del13 and del8 was detected in a smaller subset of analyzed plasma cells with a new, subclonal del17p. In panel D, the patient relapsed after tandem TPL and partial remission with a new del17p and MYC translocation. Compared to the patient in panel C the maternal clone could be detected in a larger proportion of plasma cells after relapse.

Results Differences between primary diagnosis and relapse Table 2 summarizes the frequencies of cytogenetic abnormalities detected at first, second and both FISH examinations. The risk of developing high-risk cytogenetic abnormalities was significantly higher after relapse from ASCT [odds ratio (OR): 6.33; 95% confidence interval (CI): 1.86-33.42; P<0.001]. We found an increased risk of occurrence of del(17p) (OR: 3.4; 95% CI: 1.2-11.79; P=0.02) and gain 1q21 (OR: 16; 95% CI: 2.49-670.96; P<0.001). While none of the patients developed one of the defined IGH translocations [t(4;14), t(11;14) or t(14;16)] after relapse, IGH translocations involving an unknown partner occurred more frequently at second FISH only (OR: 8; 95% CI 1.07-354.98; P=0.039). Hyperdiploidy was observed before and after ASCT in 45 patients (44.6%). Only two patients developed a de novo hyperdiploid karyotype after relapse, while seven patients lost the hyperdiploid karyotype during followup. There were only two patients without detectable cytogenetic abnormalities at primary diagnosis who developed new cytogenetic abnormalities after progression. However, both patients carried a high-risk lesion at relapse (del17p and gain 1q21, respectively). 1434

Clonal evolution in patients with high-risk cytogenetics Analysis of patients with gain 1q21 at first and second FISH (n=48) revealed that most patients had subclones that harbored different copy numbers than the main clone (Figure 1). Five of the 16 patients with de novo gain 1q21 also had subclones with different copy numbers compared to the main clone (Figure 1). In seven of 17 patients with new del17p at relapse, the high-risk cytogenetic abnormality evolved as the major clone (>60% of analyzed cells). In ten patients del17p was detected only in a subclone of cells (10-60%). Figure 1 summarizes changes in copy numbers for patients with gain 1q21 and gives examples of clonal evolution of de novo del17p.

Correlation with baseline characteristics and treatment response We could not identify any baseline cytogenetic abnormalities associated with the occurrence of high-risk disease at relapse. Furthermore, there were no significant differences in high ISS stages: at baseline ISS stage 3 was detected in 29.6% of patients without high-risk cytogenetic abnormalities, in 31.2% of patients with high-risk cytogenetic abnormalities at both time points and in 33.3% of patient with high-risk cytogenetic abnormalities only at second FISH (P=0.75). We also observed no differhaematologica | 2017; 102(8)

Cytogenetic evolution of myeloma

Table 2. Number of patients with aberrations at first and/or second fluorescence in situ hybridization studies.

Only 1st FISH

Not found High-risk Gain1q21 Del17p t(4;14) t(11;14) t(14;16) t(x;14)* Hyperdiploidy Gain5 Gain9q34 Gain11q13 Gain11q23 Gain15q22 Gain19q13 Gain14p16 Del8p21 Del13q14

Only 2nd FISH %

36 58 95 107 97 109 107 47 45 40 65 17 43 47 77 65 50

(29.8) (47.2) (76.6) (88.4) (80.8) (98.2) (88.4) (46.5) (63.4) (38.8) (55.6) (45.9) (42.2) (43.1) (63.1) (59.1) (39.7)

3 1 5 0 1 0 0 7 9 5 5 4 5 14 4 10 9

(2.5) (0.8) (4.0) 0 (0.8) 0 0 (6.9) (12.7) (4.9) (4.3) (10.8) (4.9) (12.8) (3.3) (9.1) (7.1)

19 16 17 0 0 0 8 2 0 4 4 3 4 5 9 11 13

(15.7) (13.0) (13.7) 0 0 0 (6.6) (2.0) 0 (3.9) (3.4) (8.1) (3.9) (4.6) (7.4) (10.0) (10.3)

1st and 2nd FISH n % 63 48 7 14 22 2 6 45 17 54 43 13 50 43 32 24 54

(52.1) (39.0) (5.6) (11.6) (18.3) (1.8) (5.0) (44.6) (23.9) (52.4) (36.8) (35.1) (49.0) (39.4) (26.2) (21.8) (42.9)

*t(x;14) = IGH translocation with unknown partner.

ences in rates of very good partial remission or better after ASCT between the three groups (no high-risk cytogenetic abnormalities: 53.1%, high-risk cytogenetic abnormalities at both FISH: 67.3%, high-risk cytogenetic abnormalities only at second FISH: 57.9%). Rates of very good partial remission or better after ASCT were significantly higher in patients with t(4;14) than in patients without this cytogenetic abnormality (92.3% versus 54.7%; P=0.01). Multivariate analysis revealed that patients who relapsed after ASCT and novel agent-based induction therapy had an increased risk of developing high-risk cytogenetic abnormalities compared to patients not treated with novel agents during induction (OR 10.82; 95% CI: 1.65-127.66; P=0.03).

Prognostic significance When analyzing time to first progression based on cytogenetic abnormalities present at baseline and relapse, we did not observe significant differences. Patients who developed a de novo del17p or gain 1q21 after relapse had similar median times to progression (29.7 and 25.8 months, respectively) as patients with these high-risk cytogenetic abnormalities at both time points (35.5 and 23.7 months, respectively). We observed the same effect when analyzing overall survival in the different groups. Figure 2 shows Kaplan-Meier estimates for overall survival from different landmarks. Multivariate analysis confirmed the observed effects. Patients who developed de novo high-risk cytogenetic abnormalities had the same risk for shorter survival [hazard ratio (HR): 3.06; 95% CI: 1.09-8.59; P=0.03] as patients with high-risk cytogenetic abnormalities already present at baseline (HR: 3.53; 95% CI: 1.53-8.14; P=0.003). Other factors associated with poor outcome in the analyzed cohort were a short time to progression, higher age and higher ISS stage. Patients with a hyperdiploid karyotype at baseline had a better outcome (HR: 0.33; 95% CI: 0.17-0.66; P=0.002). No significant differhaematologica | 2017; 102(8)

ences were observed for patients treated with novel agents during induction or between HD4 patients or NSP. Results from multivariate analysis from different landmarks are summarized in Figure 2.

Discussion The outcome of patients with MM has improved substantially during the last decades as a result of drug development and progress in the understanding of disease biology.11,12 However, even in the era of novel agents some patients with high-risk cytogenetic abnormalities or early relapse after first-line treatment have a dismal outcome.13,14 Clonal heterogeneity and evolution are contributors to disease progression and ultimately refractoriness in MM.6 So far, there are only limited data available that proved clonal evolution in patients relapsing after ASCT for newly diagnosed disease. With our current analysis of 128 patients with FISH data at primary diagnosis and relapse after ASCT we demonstrate that highrisk cytogenetic abnormalities occur more frequently at relapse. This observation was especially due to de novo gains of chromosome 1q and new deletions of chromosome 17p. No changes were observed between primary diagnosis and relapse for defined IGH translocations, including t(4;14). A recent study demonstrated that chromosomal instability is a cornerstone of high-risk myeloma and propagated by bi-allelic inactivation of tumor suppressor genes, such as TP53 located on chromosome 17p.7 Furthermore, gain of chromosome 1q is associated with increased proliferation15 (e.g. compared to hyperdiploid MM) and genomic instability.16 Both cytogenetic abnormalities (del17p and gain 1q21) are considered secondary events in myelomatogenesis1 and associated with a high risk of progression from smoldering to symptomatic MM.17 Our current study confirmed this assumption for the first time in 1435

M. Merz et al. A

Figure 2. Multivariate survival analysis from different landmarks. Kaplan-Meier plots and multivariate analyses for overall survival in patients without high-risk cytogenetic abnormalities (CA) at both time points (black line), high-risk CA only at first (red line), only at second (green line) or at both (blue line) FISH analyses. Different landmarks were used: (A) from start of chemotherapy, (B) from progressive disease, (C) from second FISH analysis.

a large, longitudinally analyzed cohort of patients treated with ASCT. In contrast to the increased rates of the high-risk cytogenetic abnormalities del17p and gain 1q21 after relapse, we did not observe the occurrence of new, defined IGH translocations, including t(4;14) with its adverse implications. This is most likely due to the fact that IGH translocations are considered primary, tumor-initiating events in MM.3 However, a study by the Intergroup Francophone MyĂŠlome found a de novo t(4;14) in 14 out of 268 patients, while in the same study 11 patients lost the t(4;14) after relapse.8 In accordance with our results, real time polymerase chain reaction analysis for IGH-MMSET confirmed the presence of minor t(4;14) subclones in selected cases.8 The authors concluded that t(4;14) can evolve from subclones, present at primary diagnosis, or can be silenced after chemotherapy until next relapse.8 This is in line with our finding, and those of recent studies,9 that patients with t(4;14) show a better response to primary therapy but still suffer from early relapse. To clarify whether clones with high-risk cytogenetic abnormalities remain present as minimal residual disease in patients with serological remission, future longitudinal genome sequencing studies in patients with relapse from minimal residual disease negativity as well as positivity are warranted. Sublclonal evolution was also observed in patients with new del17p and gain 1q21. In a subset of patients, del17p occurred only in less than 60% of analyzed plasma cells, in line with the findings of a previous study.18 Furthermore, patients with gain 1q21 showed massive 1436

clonal heterogeneity after relapse, since many patients had subclones harboring different copy numbers compared to the main clone. Future studies will have to clarify whether the observed clonal heterogeneity in patients with del17p or gain 1q21 might be one of the reasons for the associated dismal outcome. In contrast to the unchanged frequencies of the analyzed, defined IGH translocations [t(4;14), t(11;14) or t(14;16)] after relapse, we observed higher rates of IGH translocations with unknown partners with the IgH breakapart probe. Translocations involving the MYC locus have been associated with disease progression and adverse outcome.19,20 In our cohort, only a small group of patients was tested for MYC translocations at primary diagnosis and relapse, so that we were unable to perform a statistical analysis to confirm the hypothesis that MYC translocations are observed more frequently after relapse. Furthermore, sequencing studies are warranted not only to identify the translocation partners, but also to clarify, whether de novo translocations are caused by class switch recombination or by other mechanisms, as shown previously for t(11;14) and t(14;20).21 Patients with hyperdiploid MM had a favorable outcome in our current analysis and in the majority of patients a hyperdiploid karyotype proved to be stable after relapse. Remarkably, seven patients lost their hyperdiploid karyotype after relapse and gains of odd numbered chromosomes, especially of chromosome 5, were the only cytogenetic abnormalities occurring at lower frequencies after relapse. This might reflect chemosensitivity of the haematologica | 2017; 102(8)

Cytogenetic evolution of myeloma

respective hyperdiploid clones. Since patients with hyperdiploid MM did not have a higher risk of developing highrisk cytogenetic abnormalities, other mechanisms might be responsible for disease progression, e.g. genomic mutations or epigenetic and microenvironmental modifications. However, it must be mentioned that we defined hyperdiploid status based on gains of the aforementioned odd-numbered chromosome loci. To rule out whether abnormalities other than the analyzed trisomies occurred after relapse, we would have had to include whole genome screening methods, such as single-nucleotide polymorphism assays.22 We could not identify other baseline characteristics or cytogenetic abnormalities associated with the occurrence of high-risk cytogenetic abnormalities at relapse and treatment response in patients with new del17p or gain 1q21 did not differ between patients with or without these cytogenetic abnormalities at both time points. Multivariate analysis revealed that patients who relapsed after novel agent-based induction therapy were at higher risk of developing the aforementioned cytogenetic abnormalities. Since the first novel agents were combined with ASCT for the treatment of MM, physicians have debated whether relapses from primary therapy are becoming more aggressive. It has been suggested that the rates of extramedullary disease after novel agent-based treatment are higher23 and early relapse after primary therapy with novel agents is still associated with poor outcome.14 With our current analysis we provide the first evidence that relapse after treatment with novel agents might result in higher rates of high-risk cytogenetic abnormalities. One explanation for this finding might be that effective treatment selects pre-existing aggressive subclones that are below the level of sensitivity of FISH analyses, as proposed in the Intergroup Francophone Myélome study on t(4;14).8 Another explanation might be that chromosomal instability propagates disease progression and causes secondary genetic events.24 Lastly, we investigated the prognostic significance of new cytogenetic abnormalities after relapse. We found that patients with de novo del17p or gain 1q21 had the same dismal outcome as patients with the respective cytogenetic abnormalities detected at both time points. In fact, time to first progression was not significantly shorter whether the respective aberrations were already present at primary diagnosis or not. Again, this could imply that the evolving high-risk clone after relapse might have already been present at a subclonal level at primary diagnosis. There are several potential criticisms of the current

References 1. Manier S, Salem KZ, Park J, Landau DA, Getz G, Ghobrial IM. Genomic complexity of multiple myeloma and its clinical implications. Nat Rev Clin Oncol. 2017;14(2):100113. 2. Landgren O, Rajkumar SV. New developments in diagnosis, prognosis, and assessment of response in multiple myeloma. Clin Cancer Res. 2016;22(22):5428-5433. 3. Morgan GJ, Walker BA, Davies FE. The genetic architecture of multiple myeloma.

haematologica | 2017; 102(8)

study that need to be addressed in the future. First of all, although patients were uniformly treated with ASCT, there were substantial differences in pre-transplant induction therapy in the cohort analyzed. Some patients were treated with proteasome inhibitors or immunomodulatory drugs before and/or after ASCT while others received only conventional chemotherapy. This might have caused different selection pressures on pre-existing clones and – more importantly – might have caused different secondary genetic events, since treatment with immunomodulatory drugs, in particular, has been associated with the occurrence of cytogenetic abnormalities typical of myelodysplastic syndromes.25 This hampers the comparison of study and non-study patients within our analysis. Secondly, our retrospective data included only 128 patients and all patients had relapsed at the time of the second FISH. The findings for some subgroups are, therefore, based on very small numbers and we did not have a validation cohort of patients with a second FISH analysis in remission after ASCT. This weakness could be overcome in future studies including pre-planned bone marrow aspirates with FISH at certain time-points in a prospective clinical trial, e.g. at primary diagnosis, after ASCT before maintenance, at the end of maintenance and relapse. In this way, a longitudinal comparison of changes in the genetic profile of patients in remission and relapse would be possible. Lastly, as mentioned above, we will investigate the partners of the so far unknown IGH translocations observed after relapse in our current study. In summary, our findings underline the importance of FISH analyses at relapse. We demonstrate that cytogenetic analyses need to be repeated, since the risk profile might be worse after treatment than at baseline, especially in the era of novel agents.3 Acknowledgments The authors thank Maria Dörner, Ewelina Nickel, Hendrike Seidt, and Marie-Louise Brygider for technical assistance in the enrichment of CD138-positive plasma cells and Michaela Brough, Michelle Ebentheuer, Stephanie Pschowski-Zuck and Annekathrin Borowski for performing interphase FISH analyses. Funding This work was supported by grants from the German Federal Ministry of Education (BMBF) “CLIOMMICS” (01ZX1309) and “CAMPSIMM” (01ES1103), the Deutsche Forschungsgemeinschaft (SFB/TRR79), the Dietmar Hopp Stiftung “Heidelberger Konzept zur Optimierung der Diagnostik und Therapie des Multiplen Myeloms”, and the 7th EU-framework program "OverMyR”.

Nat Rev Cancer. 2012;12(5):335-348. 4. Moreau P, Cavo M, Sonneveld P, et al. Combination of international scoring system 3, high lactate dehydrogenase, and t(4;14) and/or del(17p) identifies patients with multiple myeloma (MM) treated with front-line autologous stem-cell transplantation at high risk of early MM progressionrelated death. J Clin Oncol. 2014;32(20):2173-2180. 5. 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. 6. Raab MS, Lehners N, Xu J, et al. Spatially divergent clonal evolution in multiple myeloma: overcoming resistance to BRAF inhibition. Blood. 2016;127(17):2155-2157. 7. Weinhold N, Ashby C, Rasche L, et al. Clonal selection and double-hit events involving tumor suppressor genes underlie relapse in myeloma. Blood. 2016;128 (13):1735-1744. 8. Hébraud B, Caillot D, Corre J, et al. The translocation t(4;14) can be present only in

1437

M. Merz et al.

10.

11.

12.

13.

14.

1438

minor subclones in multiple myeloma. Clin Cancer Res. 2013;19(17):4634-4637. Sonneveld P, Schmidt-Wolf IGH, van der Holt B, et al. Bortezomib induction and maintenance treatment in patients with newly diagnosed multiple myeloma: results of the randomized phase III HOVON-65/ GMMG-HD4 trial. J Clin Oncol. 2012;30(24):2946-2955. Neben K, Lokhorst HM, Jauch A, et al. Administration of bortezomib before and after autologous stem cell transplantation improves outcome in multiple myeloma patients with deletion 17p. Blood. 2012;119(4):940-948. Orlowski RZ, Lonial S. Integration of novel agents into the care of patients with multiple myeloma. Clin Cancer Res. 2016; 22(22):5443-5452. RĂśllig C, Knop S, BornhĂ¤user M. Multiple myeloma. Lancet. 2015; 385(9983):21972208. Kaufman GP, Gertz MA, Dispenzieri A, et al. Impact of cytogenetic classification on outcomes following early high-dose therapy in multiple myeloma. Leukemia 2016;30(3):633-639. Majithia N, Rajkumar SV, Lacy MQ, et al. Early relapse following initial therapy for multiple myeloma predicts poor outcomes in the era of novel agents. Leukemia. 2016;

30(11):2208-2213. 15. Hose D, Reme T, Hielscher T, et al. Proliferation is a central independent prognostic factor and target for personalized and risk-adapted treatment in multiple myeloma. Haematologica. 2011;96(1):87-95. 16. Sawyer JR, Tricot G, Lukacs JL, et al. Genomic instability in multiple myeloma: Evidence for jumping segmental duplications of chromosome arm 1q. Genes Chromosomes Cancer. 2005;42(1):95-106. 17. Neben K, Jauch A, Hielscher T, Hillengass J, Lehners N, Seckinger A, 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. 18. An G, Li Z, Tai Y-T, et al. The impact of clone size on the prognostic value of chromosome aberrations by fluorescence in situ hybridization in multiple myeloma. Clin Cancer Res. 2015;21(9):2148-2156. 19. Shou Y, Martelli ML, Gabrea A, et al. Diverse karyotypic abnormalities of the cmyc locus associated with c-myc dysregulation and tumor progression in multiple myeloma. Proc Natl Acad Sci USA. 2000; 97(1):228-233. 20. Weinhold N, Kirn D, Seckinger A, et al. Concomitant gain of 1q21 and MYC

21.

22.

23.

24.

25.

translocation define a poor prognostic subgroup of hyperdiploid multiple myeloma. Haematologica. 2016;101(3):e116-119. Walker BA, Wardell CP, Johnson DC, et al. Characterization of IGH locus breakpoints in multiple myeloma indicates a subset of translocations appear to occur in pregerminal center B cells. Blood. 2013;121(17):34133419. Chretien M-L, Corre J, Lauwers-Cances V, et al. Understanding the role of hyperdiploidy in myeloma prognosis: which trisomies really matter? Blood. 2015;126(25): 2713-2719. Short KD, Rajkumar SV, Larson D, et al. Incidence of extramedullary disease in patients with multiple myeloma in the era of novel therapy, and the activity of pomalidomide on extramedullary myeloma. Leukemia. 2011;25(6):906-908. Ji Z, Zhang L, Peng V, Ren X, McHale CM, Smith MT. A comparison of the cytogenetic alterations and global DNA hypomethylation induced by the benzene metabolite, hydroquinone, with those induced by melphalan and etoposide. Leukemia. 2010;24(5):986-991. Usmani SZ, Sawyer J, Rosenthal A, et al. Risk factors for MDS and acute leukemia following total therapy 2 and 3 for multiple myeloma. Blood. 2013;121(23):4753-4757.

haematologica | 2017; 102(8)

ARTICLE

Plasma Cell Disorders

Rajshekhar Chakraborty,1,2 Eli Muchtar,1 Shaji K. Kumar,1 Dragan Jevremovic,3 Francis K. Buadi,1 David Dingli,1 Angela Dispenzieri,1 Suzanne R. Hayman,1 William J. Hogan,1 Prashant Kapoor,1 Martha Q. Lacy,1 Nelson Leung1 and Morie A. Gertz1

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Haematologica 2017 Volume 102(8):1439-1445

1 Division of Hematology, Mayo Clinic, Rochester; 2Hospitalist Services, Essentia HealthSt. Josephâ&#x20AC;&#x2122;s Medical Center, Brainerd, and 3Department of Laboratory Medicine and Pathology, Division of Hematopathology, Mayo Clinic, Rochester, MN, USA

ABSTRACT

irculating plasma cells at diagnosis, prior to auto-transplant and at relapse have a negative impact on survival in multiple myeloma. However, the impact of kinetics of circulating plasma cells along the course of illness has not been defined. We have analyzed 247 newly diagnosed multiple myeloma patients undergoing early autotransplant who had paired evaluation of circulating plasma cells at diagnosis and pre-transplant by 6-color flow cytometry. A total of 117 patients had no detectable circulating plasma cells at both time points (CPC-/-), 82 had circulating plasma cells at diagnosis followed by complete eradication after induction (CPC+/-) and 48 had circulating plasma cells at transplant, including persistence of cells (CPC+/+; n=45) or emergence of new cells (CPC-/+; n=3) after induction. The rate of posttransplant stringent complete response was 32% in the CPC-/-, 30% in CPC+/- and 12% in CPC+/+ or -/+ groups (P=0.018). At a median follow up of 58 months from transplantation, the median progression-free survival in the 3 respective groups were 30, 24 and 14 months, and the 5-year overall survival rates were 83%, 70% and 43% (P<0.001 for both comparisons). On a multivariate analysis for overall survival, the risk of mortality was higher in CPC +/- (hazard ratio 2.7, 95%CI: 1.35.8; P=0.009) and CPC+/+ or -/+ (hazard ratio 5.7, 95%CI: 2.5-13.1; P<0.001) groups compared to the CPC-/- group. Monitoring for circulating plasma cells before induction therapy and before transplant by 6color flow cytometry is predictive of survival in newly diagnosed myeloma and should be incorporated into clinical trials. Introduction The presence of clonal circulating plasma cells (CPCs) has a negative impact on survival in newly diagnosed and relapsed/refractory multiple myeloma (MM)1-4 despite the use of proteasome inhibitor (PI) and immunomodulator (IMiD)-based therapy. Furthermore, the presence of CPCs prior to autologous stem cell transplantation (ASCT) is a predictor of inferior progression-free survival (PFS) and overall survival (OS), independent of the post-induction depth of response.5-7 However, the impact of changes in CPCs along the course of the disease has not been reported. Since MM is a multi-clonal disease with dynamic evolution along the course of illness, serial assessment of bone marrow (BM) and extramedullary disease burden is important in the treatment decision-making process. Using highly sensitive next-generation flow cytometry, clonal CPCs have been shown to be present in up to 96-100% of patients with newly diagnosed MM (NDMM).8,9 haematologica | 2017; 102(8)

Correspondence: gertz.morie@mayo.edu

Received: February 12, 2017. Accepted: April 28, 2017. Pre-published: May 4, 2017. doi:10.3324/haematol.2017.166629 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1439 ÂŠ2017 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.

1439

R. Chakraborty et al.

Table 1. Baseline clinical characteristics.

Baseline characteristic Median age at transplant (range), years Male (%) ISS stage, n (%) Stage I Stage II Stage III Missing R-ISS stage, n (%) Stage I Stage II Stage III Missing Median number of CPCs (/150,000 total collected events) at diagnosis (n; IQR) Median time from diagnosis to transplant (range), months FISH cytogenetics, n (%) [Data available for 225 patients] Primary cytogenetic abnormalities HRD t(11;14) t(4;14) t(14;16) t(14;20) Deletion(13q) Secondary cytogenetic abnormalities Deletion(17p) Amplification(1q) FISH-defined high-risk* LDH at diagnosis >UNL, % [Data available for 210 patients] LI>1 at diagnosis [data available for 185 patients] Pre-transplant response, n (%) CR VGPR PR MR or SD PD Induction regimen PI-based only, n (%) IMiD-based only, n (%) PI-and IMiD-based, n (%) Stem cell mobilization with cyclophosphamide, n (%) Melphalan dose, n (%) Full Reduced Post-ASCT maintenance, n (%)

Overall (n=247)

CPC -/- (n=117)

CPC +/- (n=82)

CPC +/+ or -/+ (n=48)

61.7 (24.7-76.1) 58.3

61.1 (29.0-75.3) 58.1

63.6 (27.7-76.1) 62.2

61.6 (32.2-72.6) 52.1

0.149 0.529

50 (20.2) 113 (45.7) 63 (25.5) 21 (8.5)

26 (22.2) 57 (48.7) 24 (20.5) 10 (8.6)

16 (19.5) 40 (48.8) 22 (26.8) 4 (4.9)

8 (16.7) 16 (33.3) 17 (35.4) 7 (14.6)

25 (10.1) 155 (62.8) 25 (10.1) 42 (17.0) 3 (0-266)

11 (9.4) 74 (63.3) 8 (6.8) 24 (20.5) -

10 (12.2) 56 (68.3) 8 (9.8) 8 (9.8) 205 (44-778)

4 (8.3) 25 (52.1) 9 (18.8) 10 (20.8) 360 (56-1603)

5.8 (2.3-11.8)

6.1 (2.3-11.8)

5.7 (3.1-10.9)

5.9 (2.4-11.3)

0.268

117 (52.0) 59 (26.2) 18 (8.4) 9 (4.0) 1 (0.4) 94 (41.8)

65 (63.7) 19 (18.6) 5 (4.9) 1 (0.98) 0 (0.0) 37 (36.3)

35 (44.3) 23 (29.1) 9 (11.4) 6 (7.6) 1 (1.3) 37 (46.8)

17 (38.6) 17 (38.6) 5 (11.4) 2 (4.6) 0 (0.0) 20 (45.4)

0.005 0.033 0.204 0.061 0.349 0.308

24 (10.7) 4 (1.8) 45 (20.0) 17.1

14 (13.7) 1 (0.98) 17 (16.7) 10.7

5 (6.3) 1 (1.3) 19 (24.1) 15.3

5 (11.4) 2 (4.6) 9 (20.4) 33.3

0.253 0.381 0.468 0.006

35.7

29.7

43.1

37.9

0.218

34 (13.8) 63 (25.5) 109 (44.1) 32 (13.0) 9 (3.6)

20 (17.1) 27 (23.1) 55 (47.0) 13 (11.1) 2 (1.7)

13 (15.8) 28 (34.2) 35 (42.7) 6 (7.3) 0 (0.0)

1 (2.1) 8 (16.7) 19 (39.6) 13 (27.1) 7 (14.6)

86 (34.8) 95 (38.4) 65 (26.3) 31 (12.6)

35 (29.9) 55 (47.0) 26 (22.2) 8 (6.8)

39 (47.6) 23 (28.1) 20 (24.4) 2 (2.4)

12 (25.0) 17 (35.4) 19 (39.6) 21 (43.8)

217 (87.8) 30 (12.1) 88 (35.6)

103 (88.0) 14 (12.0) 36 (30.8)

72 (87.8) 10 (12.2) 36 (43.9)

42 (87.5) 6 (12.5) 16 (33.3)

0.181

0.113

0.349

<0.001

0.011 0.022 0.074 <0.001 0.995

0.156

ISS: International Staging System; CPC: circulating plasma cells; R-ISS: Revised-International Staging System; ASCT: autologous stem cell transplantation; HRD: hyperdiploidy; FISH: fluorescence in situ hybridization; CR: complete response; VGPR: very good partial response; PR: partial response; MR: minimal response; SD: stable disease; PD: progressive disease; PI: proteasome inhibitor; IMiD: immunomodulators. *High-risk FISH cytogenetics was defined by t(4;14), del(17p), t(14;16), t(14;20) and +1q.

1440

haematologica | 2017; 102(8)

Kinetics of circulating plasma cells in myeloma

Figure 1. Kaplan-Meier curves depicting progression-free survival (PFS) (A) and overall survival (OS) (B) in groups stratified by kinetics of circulating plasma cells (CPCs) before and after induction therapy: CPC-/-, CPC +/- and CPC +/+ or -/+.

Table 2. Post-transplant response and survival.

Groups Post-transplant sCR rate, n (%) Median PFS, months (95% CI) Median OS, months (95% CI) 5-year OS rate, % (95% CI)

CPC-/-

CPC+/-

CPC+/+ or -/+

38 (32.5%) 30.3 (24.5-37.9) NR (NR-NR) 82.9 (74.0-89.2)

25 (30.5%) 24.0 (17.9-27.8) NR (60.4-NR) 69.5 (55.0-80.9)

6 (12.5%) 13.6 (11.7-17.8) 41.6 (23.0-NR) 42.9 (27.7-59.6)

0.018 <0.001 <0.001 <0.001

CPC: circulating plasma cells; sCR: stringent complete response; PFS: progression-free survival; OS: overall survival; NR: not reached; CI: Confidence Interval.

Hence, serial assessment of peripheral blood for detection of CPCs or cell-free DNA could be a feasible approach for detection of minimal residual disease (MRD) outside of the bone marrow. We hypothesized that the reduction of the CPC burden after PI and/or IMiD-based induction therapy would lead to an improved PFS and OS in NDMM patients undergoing upfront ASCT. To test our hypothesis, we retrospectively analyzed NDMM patients undergoing upfront ASCT with available data on paired measurement of CPCs before initiation of induction therapy and before ASCT.

Methods Patients This study was approved by the Mayo Clinic Institutional Review Board and was conducted in accordance with federal regulations and the principles of the Declaration of Helsinki. Informed consent was obtained from all patients for review of their electronic medical records. We have included all consecutive patients who had a serial evaluation for the presence of clonal CPCs at diagnosis (prior to initiating therapy) and prior to stem cell mobilization for ASCT by 6-color multiparameter flow cytometry (MFC). All patients underwent upfront ASCT (<12 months from diagnosis) in Mayo Clinic between January 2007 and May 2015 (the era of PI and IMiD-based induction therapy). Mononuclear cells from peripheral blood samples were isolated by Ficoll gradient and stained with antibodies to CD19, CD38, CD45, CD138 and cytoplasmic kappa and lambda immunoglobuhaematologica | 2017; 102(8)

lin light chains.10 The CPCs were detected by analysis of CD19, CD45, CD38 and CD138. Clonality was assessed by light chain restriction [Kappa: lambda expression ratio of >4:1 (Kappa restricted) or <1:2 (Lambda restricted)]. The target for collection was more than 150,000 cellular events. The clonal CPCs were reported as the number of clonal PCs/150,000 total mononuclear cells. Patients were considered to be negative for clonal CPCs at a sensitivity of 10-4 clonal plasma cells in all tested events. Patients were classified as having high-risk cytogenetics by fluorescent in situ hybridization (FISH) if they had deletion (17p), t(4;14), t(14;20), t(14;16) or +1q at diagnosis or at first presentation to the Mayo Clinic prior to ASCT. The primary end point of this study was best post-transplant response, PFS and OS. Response was determined according to the current International Myeloma Working Group (IMWG) response criteria.11 Best post-transplant response was defined as the best response at any time after ASCT. PFS was defined as time from ASCT to disease progression or death due to any cause. OS was defined as time from ASCT to death due to any cause. Patients who were alive and free of disease were censored at the last follow-up visit. Stem cell mobilization, conditioning and transplant management in the Mayo Clinic have been previously described.12,13

Statistical analysis Two-sided Fisherâ&#x20AC;&#x2122;s exact tests were used to test for differences between categorical variables and two-sided Wilcoxon rank sum tests were used to compare continuous variables. Survival analysis was carried out using the method described by Kaplan and Meier.14 Differences in survival between groups were tested for statistical significance using the two-sided log-rank test. 1441

R. Chakraborty et al.

Univariate analysis using the Cox proportional hazards model was performed with the following variables: age â&#x2030;Ľ65 years at transplant, high-risk cytogenetics by FISH, International Staging System (ISS) stage 3 at diagnosis, CPC kinetics, very good partial response (VGPR) or better at transplant, lactate dehydrogenase (LDH) more than upper normal limit (UNL) at diagnosis, plasma cell labeling index (PCLI) greater than 1 at diagnosis, PI-based induction therapy, IMiD-based induction therapy and PI and IMiD-based induction therapy. Prognostic factors for PFS and OS with P<0.1 in the univariate analysis were studied in a multivariate analysis. The JMP 10.0.0 (SAS Institute Inc., Cary, NC, USA) statistical package was used for all statistical analysis.

Results Baseline characteristics A total of 247 patients had available data on paired evaluation of clonal CPCs at diagnosis and prior to ASCT in our database in the designated time period. The median age at transplant was 62 years (range, 25-76). Clonal CPCs were detected in 127 (51.4%) patients at diagnosis. The median number of clonal CPCs at diagnosis was 247/150,000 total collected events (range 1-88,383). A total of 117 (47.4%) patients did not have detectable CPCs at diagnosis and transplant (CPC-/-). Eighty-two (33.2%) patients had detectable CPCs at diagnosis followed by complete eradication after induction therapy (CPC +/-). Among patients with detectable CPCs at transplant (n=48; 19.4%), persistence of CPCs was seen in 45 patients (CPC +/+) and emergence of new CPCs (CPC-/+) was seen in 3 patients. The median percentage change in the number of CPCs at transplant from that at diagnosis in patients with incomplete resolution after induction therapy was -82% (range, -99% to +122%). There was no significant difference in the median number of CPCs at diagnosis in patients with complete (CPC+/-) or incomplete (CPC+/+) resolution of CPCs after induction therapy (median, 206 vs. 415, respectively; P=0.11). The median number of postinduction CPCs in patients with incomplete resolution of CPCs after induction therapy was 75/150,000 total collected events (range 1-9645). The 3 patients with emergence of new CPCs after induction had a total of 30, 10 and 168 CPCs/150,000 total collected events. For analysis, we divided the patients into 3 groups based on the detection of CPCs before and after induction therapy: CPC-/- (n=117), CPC+/- (n=82) and CPC+/+ or -/+ (n=48). The baseline clinical characteristics in the 3 groups are shown in Table I. There was no significant difference in the ISS stage or revised-ISS stage at presentation in the 3 groups (P=0.181 and 0.113, respectively). Among primary cytogenetic abnormalities, the incidence of hyperdiploidy was significantly higher in patients with CPC-/-, compared to those with CPC+/- and CPC+/+ or -/+ (64%, 44% and 39%, respectively; P=0.005). On the other hand, the incidence of t(11;14) in patients with CPC-/- was 19%, compared to 29% in those with CPC+/and 39% in patients with CPC+/+ or -/+ (P=0.033). There was no significant difference in the incidence of FISHdefined high-risk cytogenetic signatures (including deletion [17p], t[4;14], t[14;16], t[14;20] and/or +1q) in any of the groups (P=0.468). The proportion of patients with an elevated LDH level at diagnosis was 11% in CPC-/-, 15% in CPC+/- and 33% in CPC+/+ or -/+ groups (P=0.006); in other words, patients with elevated LDH had a higher 1442

Figure 2. Kaplan-Meier curves depicting progression-free survival (PFS) in different groups stratified by the receipt of post-transplant maintenance therapy. (A) CPC-/-; (B) CPC+/-; (C) CPC+/+ or -/+.

incidence of having CPCs at diagnosis compared to those with normal LDH (42% vs. 17%). There was no statistically significant correlation between the kinetics of CPCs and LDH level at transplant. The proportion of patients with pre-transplant PCLI of bone marrow plasma cells (BMPCs) greater than 1 was higher in patients in the CPC+/+ or -/+ group (39%), compared to those in CPC+/- (26%) and CPC-/- groups (16%) (P=0.018), suggesting that proliferating plasma cell clones in the bone marrow were associated with the presence of CPCs resistant to clearance after induction therapy. The median time from diagnosis to ASCT was six months, and was similar in the 3 groups. Patients with CPC+/+ or -/+ had a lower incidence of receiving only a PI-based induction regimen (P=0.011) and a trend towards a higher incidence of receiving both PI and IMiD-based induction regimens (P=0.074) prior to ASCT (Table 1). There was no significant difference in the proportion of patients receiving triplet induction regimens in the CPC+/and CPC+/+ or -/+ groups (68% vs. 56%; P=0.170). Very good partial response (VGPR) or better response prior to transplant was achieved in 40% of patients with CPC-/-, 50% in CPC+/- and 19% of patients with CPC+/+ or -/+ groups (P=0.001). The frequency of administration of haematologica | 2017; 102(8)

Kinetics of circulating plasma cells in myeloma

Table 3. Univariate and multivariate analysis for progression-free and overall survival by Cox proportional hazards model.

Variable Age â&#x2030;Ľ65 High-risk cytogenetics by FISH CPC kinetics CPC-/CPC+/CPC+/+ or -/+ â&#x2030;ĽVGPR at transplant ISS stage 3 at diagnosis LDH>UNL at diagnosis LI>1 at diagnosis PI-based induction therapy IMiD-based induction therapy PI- and IMiD-based induction therapy

N.*

247 224 247

247 226 210 185 247 247 247

Progression-free survival Univariate P Multivariate HR (95% CI) HR (95% CI)

0.80 (0.58-1.10) 1.33 (0.89-1.93)

0.178 0.149

NA NA

1 (referent) 1.40 (0.98-1.99) 2.79 (1.87-4.11) 0.66 (0.48-0.90) 1.03 (0.72-1.44) 0.87 (0.54-1.34) 1.53 (1.05-2.20) 1.03 (0.74-1.42) 1.06 (0.78-1.44) 0.92 (0.64-1.28)

0.060 <0.001 0.009 0.869 0.532 0.028 0.846 0.704 0.618

1 (referent) 1.63 (1.08-2.45) 2.88 (1.73-4.68) 0.68 (0.46-0.99) NA NA 1.57 (1.07-2.27) NA NA NA

0.020 <0.001 0.047 NA NA 0.021 NA NA NA

Overall survival Univariate P Multivariate HR (95% CI) HR (95% CI) 0.90 (0.53-1.49) 2.35 (1.31-4.04)

0.902 0.005

NA 2.67 (1.29-5.29)

NA 0.009

1 (referent) 1.82 (0.99-3.35) 0.053 2.68 (1.27-5.84) 4.53 (2.53-8.17) <0.001 5.73 (2.53-13.12) 1.01 (0.61-1.04) 0.973 NA 1.36 (0.80-2.26) 0.249 NA 1.63 (0.86-2.90) 0.125 NA 2.18 (1.22-3.90) 0.009 1.91 (1.03-3.54) 1.29 (0.75-2.14) 0.344 NA 0.59 (0.35-0.97) 0.037 0.79 (0.40-1.49) 1.47 (0.87-2.42) 0.145 NA

0.009 <0.001 NA NA NA 0.039 NA 0.466 NA

NA: not applicable; HR: high-risk; FISH: fluorescence in situ hybridization; CPC: circulating plasma cells; VGPR: very good partial response; sCR: stringent complete response; ISS: International Staging System; LI: Labeling Index; PI: proteasome inhibitors; IMiD: immunomodulators; LDH: lactate dehydrogenase; UNL: upper normal limit. *Indicates number of patients with available data.

post-transplant maintenance therapy was 36%, with no significant difference in the 3 groups (P=0.156).

kinetics of CPCs before and after ASCT is shown in the Online Supplementary Appendix.

Post-transplant response and survival

Impact of maintenance therapy

The median follow up was 58 months (95%CI: 50-64) from ASCT. Patients in CPC-/- and +/- groups had a higher incidence of achieving post-transplant stringent complete response (sCR), compared to those with CPC+/+ or -/+ (32%, 30% and 12%, respectively; P=0.018 for comparison of 3rd with 1st and 2nd groups) (Table 2). The Kaplan-Meier curves for PFS and OS are shown in Figure 1A and B. The median PFS from transplant in the 3 respective groups was 30, 24 and 14 months, respectively, and the 5-year OS rates were 83%, 70% and 43%, respectively (P<0.001 for both comparisons). On a multivariate analysis (MVA) (Table 3), using CPC-/- group as the comparator, PFS and OS was significantly inferior in CPC+/- (P=0.020 and 0.009 for PFS and OS, respectively) and CPC +/+ or -/+ groups (P<0.001 for both PFS and OS). Presence of high-risk cytogenetics by FISH and PCLI greater than 1 at diagnosis also retained their independent prognostic impact for OS on MVA. Data on clonal CPCs at day 100 post transplant was available in 213 out of 247 patients. A total of 10 (4.7%) patients had detectable CPCs on day 100, with the median number of CPCs being 770/150,000 total collected events (range 32-118,285). This included 6 patients with incomplete resolution of pre-transplant CPCs and 4 patients with emergence of new CPCs after transplant. The median PFS in these 10 patients with clonal CPCs on day 100 was 5.3 months (95%CI: 1.1-17.5) and the median OS was 6.6 months (95%CI: 4.9-24.2) from ASCT. The median PFS and OS in patients with complete resolution of pretransplant CPCs by day 100 (n=35) was 16.4 months (95%CI: 12.0-20.5) and 58.4 months (95%CI: 29.0-NR), respectively, the 5-year OS rate being 48% (95%CI: 2967). The Kaplan-Meier curve for PFS and OS stratified by

A total of 88 (36%) patients had received post-transplant maintenance therapy. PI-based maintenance was received by 13%, IMiD-based 20% and both PI and IMiD-based therapy was received by 2% of patients after ASCT. The Kaplan-Meier curves for PFS in the 3 groups (CPC-/-, CPC+/- and CPC+/+ or -/+) stratified by receipt of maintenance therapy are shown in Figure 2. Improvement in PFS with administration of post-transplant maintenance therapy was demonstrated only in the group with complete resolution of CPCs (CPC+/-) after induction therapy (median PFS, 27 vs. 22 months with maintenance vs. no maintenance, respectively; P=0.029). There was no statistically significant difference in OS noted between groups.

haematologica | 2017; 102(8)

Discussion Our study shows that in patients with clonal CPCs at diagnosis, complete eradication of CPCs after induction therapy is associated with a marked improvement in PFS and OS from transplant. However, their survival is still inferior compared to patients with no detectable CPCs at diagnosis and prior to ASCT. Furthermore, in patients with incomplete resolution of CPCs after induction therapy with PIs and/or IMiDs, 5-year OS rate from ASCT is only 43%. The prognostic impact of kinetics of CPCs before induction and before ASCT was independent of other known prognostic variables, including ISS stage, high-risk FISH cytogenetics, LDH at diagnosis and pretransplant response by IMWG criteria. Interestingly, among patients with CPCs at diagnosis, administration of post-transplant maintenance therapy led to PFS improvement only in the group with complete eradication of CPCs 1443

R. Chakraborty et al.

after induction therapy which signifies that resistance of clonal CPCs to induction therapy cannot be overcome by maintenance. Evaluation of CPCs for assessing disease burden in MM is emerging as an attractive modality for non-invasive risk stratification and response assessment. However, it has not been evaluated in prospective studies. The presence of CPCs in newly diagnosed MM has been shown to predict an inferior PFS and OS in various retrospective studies, independent of the cytogenetic risk status at diagnosis.1,3,7,15 However, a uniform cut off for CPCs cannot be ascertained, primarily due to the heterogeneity of flow cytometry techniques used across studies. Using a highly sensitive next-generation flow cytometry method9 with the median limit of detection being 3x10-6, clonal CPCs have been detected in 60% of patients with monoclonal gammopathy of unknown significance, 75% with smoldering MM, 96-100% of patients with newly diagnosed MM, and 85% with relapsed/refractory MM.8,16 The authors in the study by Burgos et al.8 also reported a 100% concordance in FISH-defined cytogenetic abnormalities in 10 paired BMPC and CPC analysis and 75% concordance in gene expression profiling signatures, based on paired data on BMPCs and CPCs in 12 patients. Another study by Lohr et al. on genomic characterization of CPCs in 24 MM patients showed the co-existence of several targeted mutations in BMPCs and CPCs. However, in 3 of 24 patients, the proportion of clonal CPCs harboring TP53R273C, BRAFG469A and NRASG13D was significantly higher compared to that of clonal BMPCs harboring similar mutations.17 Interestingly, in another study in relapsed MM, there was a lack of concordance in the mutational profile in all 6 patients among BMPCs, CPCs and clonal PCs from extramedullary (EM) plasmacytomas. The somatic mutational burden in BMPCs, CPCs and EM PCs was 75%, 77% and 85%, respectively (P=0.07), likely indicating independence from the bone marrow microenvironment in myeloma cells with increasing mutational burden.18 It would be interesting to explore the mutational profile of CPCs at several time points in the course of illness to identify the dynamics of clonal evolution. There was no significant difference in the frequency of FISH-defined high-risk cytogenetic abnormalities between groups in our study. However, patients with CPCs at diagnosis had a higher frequency of harboring t(11;14) compared to those with no CPCs (34 vs. 19%; P=0.033). t(11;14) is known to be more prevalent in patients with primary plasma cell leukemia (PCL) (25-65%) compared to those with newly diagnosed MM (15%).19 Although the frequency of t(11;14) was higher in patients with incomplete resolution of CPCs compared to those with complete eradication after induction therapy (40% vs. 29%, respectively), this difference was not statistically significant (P=0.209). Conversely, the incidence of hyperdiploidy (HRD) was lower in patients with CPCs at diagnosis compared to those without (33% vs. 52%, respectively;

References 1. Gonsalves WI, Rajkumar SV, Gupta V, et al. Quantification of clonal circulating plasma cells in newly diagnosed multiple myeloma: implications for redefining high-risk

1444

P=0.005). HRD as a primary cytogenetic abnormality is extremely rare in primary PCL (0-9%), compared to a prevalence of 50% or more in newly diagnosed MM.19 In our study, 82 out of 127 patients with CPCs at diagnosis had complete resolution after induction therapy with PI and/or IMiDs and this subgroup also achieved PFS benefit from PI/IMiD-based post-transplant maintenance therapy, indicating greater chemosensitivity of the CPC clone in this subgroup. This also highlights the heterogeneity in sensitivity to induction therapy among clonal CPCs present at diagnosis. Patients with incomplete resolution of CPCs after induction therapy, who subsequently achieved complete clearance by day 100 of ASCT still had an inferior OS (5-year OS rate, 48%) compared to those with complete resolution of CPCs after induction (5-year OS rate, 70%). Only 3 of 247 patients in our study had no CPCs at diagnosis followed by emergence of new CPCs after induction therapy. With serial CPC monitoring by higher sensitivity flow cytometry methods, this number might increase and reflect subclonal evolution of PCs developing anchorage independence and growth potential outside of the bone marrow microenvironment under therapeutic and immunological pressure. This study has limitations. The flow cytometry technique employed was not highly sensitive, as demonstrated by the detection of CPCs in only 51% of newly diagnosed patients, compared to 96% with next-generation flow technology.8 With the use of homogeneous high-sensitivity flow cytometry techniques, the impact of quantitative cut offs for decrease in CPCs after induction therapy on survival should be explored, ideally in the setting of clinical trials. Some recent studies have also shown downregulation of CD138 expression on CPCs,20,21 hence CD138-independent strategies should be employed to identify CPCs in MM. Furthermore, there were significant differences in induction regimens received across different groups and the groups were not well balanced in relation to LDH at diagnosis. However, CPC kinetics retained its independent prognostic impact on multivariate analysis. To the best of our knowledge, this is the first study to demonstrate the independent prognostic value of serial assessment of CPC burden on multivariate analysis in newly diagnosed MM patients undergoing early ASCT. Furthermore, it also shows that post-transplant maintenance did not offer any PFS benefit to patients who did not achieve complete eradication of clonal CPCs at the end of induction therapy. Flow cytometry of peripheral blood for CPCs is a non-invasive and cost-effective technique which can be easily employed in resource-limited settings to assess disease burden outside of the bone marrow. Future prospective studies using highly sensitive methods for detection of CPCs should incorporate several time points to explore impact on survival and establish a predictive marker for maintenance therapy. However, the findings presented here provide additional evidence the already established role of CPCs as a prognostic marker in myeloma.

myeloma. Leukemia. 2014;28(10):20602065. 2. Gonsalves WI, Morice WG, Rajkumar V, et al. Quantification of clonal circulating plasma cells in relapsed multiple myeloma. Br J Haematol. 2014;167(4):500-505. 3. An G, Qin X, Acharya C, et al. Multiple

myeloma patients with low proportion of circulating plasma cells had similar survival with primary plasma cell leukemia patients. Ann Hematol. 2015;94(2):257-264. 4. Peceliunas V, Janiulioniene A, Matuzeviciene R, Zvirblis T, Griskevicius L. Circulating plasma cells predict the out-

haematologica | 2017; 102(8)

Kinetics of circulating plasma cells in myeloma

come of relapsed or refractory multiple myeloma. Leuk Lymphoma. 2012; 53(4):641-647. Chakraborty R, Muchtar E, Kumar SK, et al. Risk stratification in myeloma by detection of circulating plasma cells prior to autologous stem cell transplantation in the novel agent era. Blood Cancer J. 2016;6(12):e512. Dingli D, Nowakowski GS, Dispenzieri A, et al. Flow cytometric detection of circulating myeloma cells before transplantation in patients with multiple myeloma: a simple risk stratification system. Blood. 2006; 107(8):3384-3388. Nowakowski GS, Witzig TE, Dingli D, et al. Circulating plasma cells detected by flow cytometry as a predictor of survival in 302 patients with newly diagnosed multiple myeloma. Blood. 2005;106(7):22762279. Burgos L, Alignani D, Garces J-J, et al. NonInvasive Genetic Profiling Is Highly Applicable in Multiple Myeloma (MM) through Characterization of Circulating Tumor Cells (CTCs). Blood. 2016;128(22):801-801[Meeting Abstract]. Flores-Montero J, Sanoja-Flores L, Paiva B, et al. Next Generation Flow for highly sensitive and standardized detection of minimal residual disease in multiple myeloma. Leukemia. 2017 Mar 10 [Epub ahead of print].

haematologica | 2017; 102(8)

10. Kumar S, Kimlinger T, Morice W. Immunophenotyping in multiple myeloma and related plasma cell disorders. Best Pract Res Clin Haematol. 2010;23(3):433-451. 11. Kumar S, Paiva B, Anderson KC, et al. International Myeloma Working Group consensus criteria for response and minimal residual disease assessment in multiple myeloma. Lancet Oncol. 2016;17(8):e32846. 12. Gertz MA, Dingli D. How we manage autologous stem cell transplantation for patients with multiple myeloma. Blood. 2014;124(6):882-890. 13. Gertz MA, Ansell SM, Dingli D, et al. Autologous stem cell transplant in 716 patients with multiple myeloma: low treatment-related mortality, feasibility of outpatient transplant, and effect of a multidisciplinary quality initiative. Mayo Clin Proc. 2008;83(10):1131-1138. 14. Kaplan EL, Meier P. Nonparametric Estimation from Incomplete Observations. J Am Stat Assoc. 1958;53(282):457-481. 15. Vagnoni D, Travaglini F, Pezzoni V, et al. Circulating plasma cells in newly diagnosed symptomatic multiple myeloma as a possible prognostic marker for patients with standard-risk cytogenetics. Br J Haematol. 2015;170(4):523-531. 16. Sanoja-Flores L, Paiva B, Flores-Montero JA, et al. Next Generation Flow (NGF): A

17.

18.

19. 20.

21.

High Sensitive Technique to Detect Circulating Peripheral Blood (PB) Clonal Plasma Cells (cPC) in Patients with Newly Diagnosed of Plasma Cell Neoplasms (PCN). Blood. 2015;126(23):4180-4180 [Meeting Abstract]. Lohr JG, Kim S, Gould J, et al. Comprehensive Genetic Interrogation of Circulating Multiple Myeloma Cells at Single Cell Resolution. Blood. 2016; 128(22):800-800[Meeting Abstract]. Bretones G, Paiva B, Valdes-Mas R, et al. Genomic Profiles of Bone Marrow (BM) Clonal Plasma Cells (PCs) Vs Circulating Tumor Cells (CTCs) and Extramedullary (EM) Plasmacytomas in Multiple Myeloma (MM). Blood. 2016;128(22):44424442[Meeting Abstract]. van de Donk NW, Lokhorst HM, Anderson KC, Richardson PG. How I treat plasma cell leukemia. Blood. 2012;120(12):2376-2389. Paiva B, Paino T, Sayagues JM, et al. Detailed characterization of multiple myeloma circulating tumor cells shows unique phenotypic, cytogenetic, functional, and circadian distribution profile. Blood. 2013;122(22):3591-3598. Muz B, de la Puente P, Azab F, et al. A CD138-independent strategy to detect minimal residual disease and circulating tumour cells in multiple myeloma. Br J Haematol. 2016;173(1):70-81.

1445

ARTICLE EUROPEAN HEMATOLOGY ASSOCIATION

Cell Therapy & Immunotherapy

Ferrata Storti Foundation

Tbet is a critical modulator of FoxP3 expression in autoimmune graft-versus-host disease Shoba Amarnath,1 Arian Laurence,1 Nathaniel Zhu,2 Renato Cunha,2 Michael A. Eckhaus,3 Samuel Taylor,2 Jason E. Foley,2 Monalisa Ghosh,2 Tania C. Felizardo2 and Daniel H. Fowler2

Institute of Cellular Medicine, Newcastle University, Newcastle Upon Tyne, UK, Experimental Transplantation Immunology Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA and 3Division of Veterinary Resources, Office of Research Services, Bethesda, MD, USA 1 2

Haematologica 2017 Volume 102(8):1446-1456

ABSTRACT

Correspondence: shoba.amarnath@ncl.ac.uk

Received: September 3, 2016 Accepted: May 3, 2017. Pre-published: May 4, 2017.

D4+ T-helper subsets drive autoimmune chronic graft-versus-host disease, a major complication after allogeneic bone marrow transplantation. However, it remains unclear how specific T-helper subsets contribute to chronic graft-versus-host disease. T-helper type 1 cells are one of the major disease-mediating T-cell subsets and require interferon-Îł signaling and Tbet expression for their function. Regulatory T cells on the other hand can inhibit T-helper type 1 cell-mediated responses. Using an established murine model that isolates the autoimmune component of graft-versus-host disease, we hypothesized that Thelper type 1 cells would restrict FoxP3-driven regulatory T cells. Upon transfer into immune-deficient syngeneic hosts, alloreactive Tbx21-/CD4+ T cells led to marked increases in FoxP3+ cells and reduced clinical evidence of autoimmunity. To evaluate whether peripheral induction contributed to regulatory T-cell predominance, we adoptively transferred Tbx21-/- T cells that consisted of fate mapping for FoxP3: recipients of flow-purified effector cells that were Foxp3- and Tbx21-/- had enhanced Tregulatory-cell predominance during autoimmune graft-versus-host disease. These data directly demonstrated that peripheral T-regulatory-cell induction was inhibited by Tbet. Finally, Tbx21-/- T-regulatory cells crossregulated autoimmune wild-type T-effector-cell cytokine production in vivo. The Tbet pathway therefore directly impairs T-regulatory-cell reconstitution and is consequently a feasible target in efforts to prevent autoimmune graft-versus-host disease.

doi:10.3324/haematol.2016.155879

Introduction Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1446 ÂŠ2017 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.

1446

T-helper (Th) 1, Th2, and Th17 cells mediate distinct acute graft-versus-host disease (GvHD) syndromes.1 In contrast, T-regulatory (Treg) cells prevent acute GvHD2 and are associated with reduced clinical GvHD. T-helper subsets are primarily driven by cross-regulatory transcription factors, namely: deficiency of Th1- and Th17-driving transcription factors allows FoxP3+ Treg reconstitution and prevents acute GvHD.3,4 However, transcription factor cross-regulation is less clear for chronic GvHD, which is distinct from acute GvHD in part due to an autoimmune mechanism.5 Chronic GvHD is propagated by donor T cells that recognize host peptides presented by donor antigen-presenting cells. The pathological manifestations of chronic GvHD therefore resemble those of autoimmunity.6-9 The mechanism by which autoimmunity arises from alloimmunity remains unresolved.10 In animal models, a decrease in Treg cells occurs along with an expansion of Th1 and Th17 cells,11 which leads to autoimmune pathology. During alloimmunity, donor T cells respond to shared antigens, thereby resulting in repertoire skewing and recognition of nonpolymorphic antigens.12 There are no data indicating whether this process results in prevention of the induction of Treg cells thereby resulting in long-term loss in immune regulation. While animals models of chronic GvHD exist,13 there is a paucihaematologica | 2017; 102(8)

T-bet restricts FoxP3

ty of data reproducing the break in T-cell regulation. Hence, we utilized a chronic GvHD model11,12,14 that captures these two processes and allows understanding of the specific regulatory signals that prevent immune tolerance of alloreactive donor T cells and consequently cause autoimmunity. Chronic GvHD autoimmunity can be experimentally dissected from acute GvHD by sequential allogeneic and syngeneic T-cell transfer. In this secondary transfer autoimmune chronic GvHD, Th1 and Th17 subsets predominate with a relative deficiency of Treg cells;11 a similar immune imbalance was also observed in a sclerodermatous chronic GvHD model.15 However, the underlying molecular mechanism by which such Treg insufficiency occurs in chronic GvHD is still unknown but therapies that enhance Treg numbers in patients with chronic GvHD are promising.16 The existence of a Th1 component in the pathogenesis of chronic GvHD diverges from initial models that emphasized a Th2-dominant disease17 but is not inconsistent with the current understanding of Th1mediated autoimmunity18 and knowledge that type I immunity drives lethality in genetic Treg deficiency.19 Taken together, Th1-driving signaling molecules may restrict peripheral Treg generation. We hypothesized that Th1-driving signals inhibit Treg generation during chronic GvHD. Because interferon (IFN)-γ signaling induces Tbet,20 which propagates Th1mediated inflammation, sequential deletion of IFN-γR followed by Tbet may identify checkpoints that inhibit peripheral Treg generation. We evaluated key Th1 transcription factors, including STAT1 and STAT4, in modulating chronic GvHD.11

hosts, spleens were harvested and single cell suspensions were obtained to test for autoreactivity, as outlined in the Online Supplementary Methods.

Evaluation of in vivo tissue RNA genes after secondary transfer On the indicated days, recipients’ skin tissue was evaluated for chronic GvHD biomarkers as outlined in the Online Supplementary Methods.

Isolation of T cells from lamina propria lymphocytes and skin Intraepithelial lymphocytes were removed and lamina propria lymphocytes were separated.21 Lymphocytes were extracted from the skin as previously described and outlined in the Online Supplementary Methods.22

Chromatin immunoprecipitation Chromatin was immunoprecipitated as previously described and outlined in the Online Supplementary Methods.23

Histopathology Tissue from Rag2−/− recipients (colon, skin) was fixed in 10% (vol/vol) formyl saline and embedded in paraffin blocks. Tissue sections were stained with eosin and hematoxylin and evaluated by a pathologist (MAE). Skin GvHD scores were either 0 (normal) or 1 (acanthosis, hyperkeratosis); skin stages 2, 3, and 4, which include mononuclear infiltrates and epidermal loss, were not observed. Intestinal GvHD was scored (0 to 4) according to degree of mononuclear cell infiltration and crypt destruction (stage 1, focal and mild; stage 2, diffuse and mild; stage 3, diffuse and moderate; stage 4, diffuse and severe).

Statistics Methods Mice Female C57BL/6 (B6, H2Kb) and BALB/c (H2kd) mice 8- to 10weeks old were obtained from Frederick Cancer Research Facility. Rag2−/−, Tbx21−/−, Ifnγ and Ifnγr−/− animals were from Jackson Laboratories. B6.Tbx21−/−Foxp3-GFP mice were generated by crossing B6.Tbx21−/− with B6.Foxp3GFP; B6.FoxP3GFP littermate controls were simultaneously maintained. Mice were maintained in a specific pathogen-free facility at the National Institutes of Health. Drinking water was supplemented with ciprofloxacin from day -1 to day +14 after bone marrow transplantation. Experiments were carried out in accordance with National Institutes of Health animal health and safety guidelines and approved by the Animal Care and Use Committee, National Cancer Institute, National Institutes of Health.

Bone marrow transplant Bone marrow experiments were performed as outlined in the Online Supplementary Methods.

Flow cytometry On day 14 after allogeneic bone marrow transplantation, splenocytes were stained with CD4 PE-Cy5 (H129.19), H2Kb PE (AF6.88.5), CD3 FITC (145-2C11) and FoxP3 APC (FJK.16s; eBioscience). After secondary transfer to Rag2−/− recipients, splenic T cells were isolated and intracellular flow cytometry was performed as outlined in the Online Supplementary Methods.

Cytokine secretion assay after secondary transfer On the indicated day after secondary transfer into syngeneic haematologica | 2017; 102(8)

Kaplan-Maier survival analysis was performed and survival curves were compared using log-rank testing. Statistical significance was determined for normally distributed data using a twotailed Student t test or one-way analysis of variance (ANOVA) followed by Dunn post-hoc tests. For statistical analysis of histology, a Mann-Whitney-U test was performed. P values <0.05 were considered statistically significant.

Results Type I signaling is important for autoimmune chronic graft-versus-host disease Experiments were designed to identify whether Th1 cells contributed to autoimmune chronic GvHD. A brief schema of the chronic GvHD model is shown in Figure 1A. Wild-type (WT), IFN-γ-/-, IFNγR-/-, STAT1-/-, or STAT4-/CD4+ T cells were transferred into Rag2-/- recipients in the post-alloreactive phase. Contradicting prior reports suggesting that chronic GvHD primarily resembles a Th2 process,17,24 IFNγR-/- CD4+ T cells abrogated chronic GvHD pathogenesis (Figure 1B); recipients had increased Tregcell numbers but no difference in IFNγ−producing T-effector cells; [#CD4+FoxP3+ (x103); mean±SEM, WT versus IFNγ-/- versus IFNγR-/-; 3.35±1.5 versus 1.78±4.8 versus 37.8±11.4; #CD4+IFNγ+ (x103); mean±SEM, WT versus IFNγ-/- versus IFNγR-/-; 46±19.7 versus 6±9 versus 83.7±20) (Figure 1C-E). Recipients of STAT1-/- or STAT4-/- T cells had increased Treg cells, decreased cytokine expression [#CD4+FoxP3+ (x105); mean±SEM, WT versus STAT1-/- ver1447

S. Amarnath et al. A

Figure 1. Lack of Th1 signaling in T cells alleviates autoimmune graft-versus-host disease. Schematic representation of the experimental murine model for autoimmune chronic GvHD. (A) BALB/c mice were subjected to lethal total body irradiation (TBI) and then reconstituted with T-depleted (TD) bone marrow (BM) and CD4+ T cells from C57BL/6 (B6) mice. On day 14 after transplant, murine recipients were euthanized and B6.CD4+ T cells were isolated and adoptively transferred into B6.Rag2-/- mice. The occurrence of autoimmune GvHD was monitored over a period of 60 days. Host BALB/c mice were subjected to TBI (950 cGy) and then reconstituted with allogeneic WT B6 TDBM (10M) and CD4+ T cells (2M). In certain cohorts, mice were reconstituted with BM and T cells deficient in IFN-γR, STAT1, STAT4, or IFN-γ. On day 14 after allogeneic bone marrow transplantation, splenic CD4+ T cells from WT or KO recipients were isolated and adoptively transferred into B6 Rag2−/− mice. (B) Survival curves after secondary transfer of either WT or KO T cells (n=8-10 per cohort. (C-H) Representative flow image and absolute numbers of FoxP3+ CD4+T cells and IFNγ+ CD4+T cells in the WT and KO cohorts (n=10 per cohort).

1448

haematologica | 2017; 102(8)

T-bet restricts FoxP3

sus STAT4-/-; 0.64±0.13 versus 1.87±0.3 versus 3±0.8; #CD4+IFNγ+ (x105); mean±SEM, WT versus STAT1-/- versus STAT4-/-; 7.7±2.6 versus 0.2±0.2 versus 0.9±0.3) (Figure 1F,H), and reduced lethality (Figure 1B). These results suggested that deficiency of Th1-cell signaling (IFN-γR) or transcription factors (STAT1, STAT4) directly impaired chronic GvHD. However, such deficiencies may have reduced chronic GvHD indirectly, namely, via reduction in Th1 cytokines. To address this, we evaluated a transplant cohort that received IFN-γ-deficient T cells: such recipients had low IFN-γ and reduced Treg cells (Figure 1CE) and chronic GvHD lethality similar to that of WT controls (Figure 1B). Recipients of IFNγR-/- T cells, even though protected against chronic GvHD lethality, had similar IFN-γ production to that of WT controls (Figure 1E). Recipients of IFNγR-/-, STAT1-/-, and STAT4-/- T cells had similar numbers of CD4+ T cells as the WT cohort [#CD4+ (x104); mean±SEM, WT versus IFNγ-/- versus IFNγR-/-; 55.1±16.5 versus 50.8±14.3 versus 76.59±15.9; WT versus STAT1-/- versus STAT4-/133.8±18.4 versus 180.7±14.1 versus 129.4±14.6] (Online Supplementary Figure S1A,C) and similar interleukin-17 (IL17) secretion [#CD4+IL17+ (x104); mean±SEM, WT versus IFNγ-/- versus IFNγR-/- ; 25±10.6 versus 42.5±10.7 versus 31±7.6; WT versus STAT1-/- versus STAT4-/-; 0.51±0.005 versus 51.5±25 versus 75.9±54.4] (Online Supplementary Figure S1B,D). These data indicate that the reduction in chronic GvHD was primarily attributable to a deficiency in Th1 cell signaling and Th1 cell transcription factors rather than a secondary deficiency in Th1 cytokines.

Autoimmune graft-versus-host disease requires T-cell Tbx21

Increased Treg numbers in IFNγR and STAT cohorts suggested an inhibitory mechanism by which Th1 signaling molecules prevented peripheral Treg generation. Because an increased number of Treg ceslls may be beneficial in decreasing chronic GvHD,25-29 we characterized the mechanism by which Treg cells were inhibited during chronic GvHD. Experiments were performed with CD4+ T-effector cells from Tbx21-/- mice (Tbet). Tbet is a master regulator of Th1 cells; lack of Tbet results in Th1-cell deficiency. To study the role of Tbet in chronic GvHD, we utilized Tbx21/T cells.30 Consistent with published results,3 WT CD4+ T cells caused acute GvHD in the alloreactive phase; in contrast, recipients of Tbx21−/− T cells were partially protected against acute GvHD [WT: n=10/10 succumbed to acute GvHD; knockout (KO): n=5/10 succumbed to acute GvHD]. At the time of splenic T-cell harvest after allogeneic bone marrow transplantation, the frequencies of Treg cells in WT and Tbet-deficient recipients were similar (0.40% versus 0.17%, respectively; P=NS). However, upon secondary transfer to Rag2−/− recipients, Tbx21−/− CD4+ T cells did not mediate lethality (Figure 2A) or cause autoimmune pathology (Figure 2B, representative result; Figure 2C, pooled results). To evaluate potential cellular mechanisms, subsequent cohorts were euthanized prior to lethality (day 60 after secondary transfer). Protection against autoimmune GvHD was not associated with reduced numbers of CD4+ T cells [#CD4+ (x104); mean±SEM, WT versus Tbx21-/-, 17.2±0.5 versus 6.8±4.1] (Online Supplementary Figure S1E) or Th17 effectors [#CD4+IL17+ (x103); mean±SEM, WT versus Tbx21-/-, 2.5±1.3 versus 1.1±0.7] (Online Supplementary Figure S1F) but was associated with reduced CD4+ IFN-γ+ cells [#CD4+IFNγ+ (x104); mean±SEM, WT versus Tbx21-/-, 102.8±18.4 versus 27.1±4] (Figure 2D representative result, -/-

haematologica | 2017; 102(8)

-/-

2E pooled results). Next, we evaluated the number of Treg cells in WT and Tbx21-/- cohorts. We reasoned that any survival advantage might be attributable to defective homing of T cells to target tissues, thereby limiting GvHD-mediated pathology.31 Homing defects can be directly attributed to Tbet deficiency which is activated downstream of IFN-γR signaling.20,32 However, contrary to this reasoning, we found increased FoxP3+Treg cells in spleen [#CD4+FoxP3+ (x103); mean±SEM, WT versus Tbx21-/-, 2.7±0.1 versus 18.8±4.6] (Figure 2F), mesenteric lymph nodes [#CD4+FoxP3+ (x103); mean±SEM, WT versus Tbx21-/-, 16.9±7.1 versus 169.9±52.8] (Figure 2G), lamina propria [#CD4+FoxP3+ (x103); mean±SEM, WT versus Tbx21-/-, 0.78±0.3 versus 1.72±0.2] (Figure 2H), and skin [#CD4+FoxP3+ (x103); mean±SEM, WT versus Tbx21-/-, 1.6±0.5 versus 3.8±0.7] (Figure 2I). STAT1 deficiency has been associated with enhanced Treg proliferation;33 however, a similar biology was not operational in our model, as Tbet deficiency did not increase the Treg pro(x103); liferative phenotype [#CD4+Ki67+FoxP3+ mean±SEM, WT versus Tbx21-/-, 9.5±4.5 versus 17.7±5.6; #CD4+Ki67+FoxP3- (x103); mean±SEM, WT versus Tbx21-/-, 165.7±92.8 versus 106.6±42.7] (Online Supplementary Figure S2). Furthermore, Tbet deficiency did not alter Treg cell Bcl2 expression [#CD4+bcl2+FoxP3+ (x103); mean±SEM, WT versus Tbx21-/-, 36±4.3 versus 228.1±101.5; mean±SEM, #CD4+bcl2+FoxP3- (x105); mean±SEM, WT versus Tbx21-/-, 73.9±42.7 versus 70.9±41.5] (Online Supplementary Figure S3). Therefore, Tbx21−/− Treg cells are similar to WT Treg cells with respect to their proliferation and apoptotic tendency during chronic GvHD.

Tbet restricts CD4+ peripheral T-regulatory-cell generation Tbet cross-regulation of FoxP3 might occur through several mechanisms, one of which involves the ability of Tbet to inhibit FoxP3 expression and subsequent peripheral Treg generation.34 We considered the possibility that bona fide effector CD4+ T cells might be more amenable to FoxP3 expression and acquiring a Treg phenotype in the absence of Tbet. To elucidate the intrinsic mechanistic implications of Tbet deficiency in FoxP3 expression, we determined whether direct Tbet inhibition of FoxP3 occurs. In light of the report by Eckerstorfer et al.,35 we sought to identify whether Tbet regulates FoxP3 expression and has binding sites in the evolutionary conserved region (ECR) upstream of the Foxp3 promoter site. ECR1, 2 and 3 induce Foxp3 promoter activity in human cells by luciferase assays. In particular, ECR3, which is located in close proximity to ppp1r3f, enhanced Foxp3 promoter activity with negligible ppp1r3f activity. Using chromatin immunoprecipitation sequencing analysis in Th1 polarized cells (GSM836124),36 we determined that Tbet has a binding site in the ECR3 region upstream of the Foxp3 promoter (Figure 3A). To validate this site, naïve CD4+ T cells from WT and Tbx21-/- mice were polarized with IL-2, TGF-β1, IL-12 and IL-18 (Figure 3B) prior to chromatin immunoprecipitation analysis. Lack of Tbet preferentially increased Foxp3 expression in induced Treg (iTreg) cells that were exposed to Th1 cytokines [% CD4+FoxP3+; mean±SEM, WT versus Tbx21-/-, 21±2.0 versus 45.7±2.2] (Figure 3C). In contrast, FoxP3 expression in iTreg cells generated from WT and Tbx21-/- cohorts were 61% versus 59%, respectively. WT iTreg cells cultured with IL-12 and IL-18 had significant Tbet co-expression with FoxP3 and enhanced binding to the ECR locus of Foxp3 [%Tbet 1449

S. Amarnath et al. B

Figure 2. Lack of Tbet alleviates autoimmune graft-versus-host disease. Host BALB/c mice were subjected to total body irradiation (950 cGy) and then reconstituted with allogeneic B6 T-cell-depleted (TD) bone marrow (BM) (10M) and CD4+ T cells (2M). In certain cohorts, mice were reconstituted with BM and T cells deficient in Tbet. On day 14 after allogeneic bone marrow transplantation, splenic CD4+ T cells from WT or Tbx21-/- recipients were isolated and adoptively transferred into B6 Rag2−/− mice. (A) Survival curve after secondary transfer (n=10 per cohort). (B) Representative pictures of colon at day 60 after secondary transfer of WT (left panel) or Tbx21−/− (right panel) T cells. (C) Summary of histological scores (n=5 per cohort). Immune cell phenotype was also evaluated at day 60 after secondary transfer. (D) Representative flow plots showing frequency of IFNγ+ and FoxP3+ T cells in the WT and Tbx21−/− cohorts. (E) Pooled results (n=10 per cohort) for the absolute numbers of splenic CD4+ T cells secreting IFN-γ and (F) absolute number of splenic CD4+ FoxP3+ T cells. (G) Absolute numbers of mesenteric lymph node CD4+ FoxP3+ T cells. (H) Absolute numbers of CD4+ FoxP3+ T cells from the lamina propria (n=10 per cohort). (I) Absolute numbers of CD4+ FoxP3+ T cells from the skin (n=5 per cohort).

1450

haematologica | 2017; 102(8)

T-bet restricts FoxP3

B C

Figure 3. Peripheral Treg numbers are increased in the absence of Tbx21. Chromatin immunoprecipitation (CHIP) sequencing analysis of naïve CD4+ T cells stimulated for 3 days under Th1 conditions. (A) Peaks denote regions of DNA associated with Tbet binding. The contralateral DNA sequence upstream of the first exon of ppp1r3f is shown below, with potential Tbet binding sites indicated in blue. (B) Naive CD4+ T cells from WT and Tbx21−/− mice were isolated and polarized under iTreg conditions [αCD3, αCD28, IL-2 (100 IU) and TGF-β1 (5 ng/mL)] in the presence of rmIL-12 and rmIL-18 (10 ng/mL). On day 3, cells were characterized using intracellular flow cytometry for co-expression of T-bet and FoxP3. (C) Percentage of FoxP3+ iTreg cells in WT and Tbx21-/- mice following stimulation with Th1 conditions. (D) Histograms denote percentage of input binding of T-bet to the ECR3 locus upstream of the Foxp3 promoter in WT or Tbx21−/− iTreg polarized in the presence or absence of rmIL-12 and IL-18 measured by CHIP. Data shown are replicates of three experiments. Bone marrow transplantation experiments were carried out with either WT T cells (B6.Foxp3GFP) or T cells deficient in Tbet (B6.Tbx21-/-Foxp3GFP) to allow flow sorting for FoxP3- T cells for secondary transfer. Splenocytes were harvested 60 days after transplant and then flow cytometry was performed in unstimulated cells to detect FoxP3 frequency. (E) Representative flow cytometry plot of peripheral Treg induction in WT and KO cohorts. (F) Pooled analysis of frequency of Treg induction in WT and KO (n=5 per cohort). (G) Absolute numbers of induced Treg cells in WT and KO cohorts during autoimmune GvHD.

haematologica | 2017; 102(8)

1451

S. Amarnath et al.

bound to DNA; WT iTreg versus WT iTreg + Th1 cytokines; 0.015±0.001 versus 0.025±0.002] relative to control Tbx21-/- iTreg cells (Figure 3C). Allogeneic bone marrow transplantation was performed using CD4+ T cells harvested from B6.Tbx21-/-FoxP3GFP mice; then, 14 days after the transplant, effector T cells were purified by flow cytometry (CD4+GFP-) and transferred into Rag2-/- recipients. At day 60 after transfer, recipients of Tbet-deficient T cells were devoid of clinical autoimmune GvHD and had increased Treg cells [%CD4+FoxP3+, mean±SEM, 1.4±0.3 versus 4.8±0.6; # CD4+FoxP3+ (x104), mean±SEM, 2.1±0.5 versus 8±1.7] (Figure 3E representative data; 3F,G pooled data). Tbet is, therefore, a critical checkpoint and prevents peripheral Treg generation during ongoing autoimmune GvHD. These results stand in contrast to those of studies showing the importance of Tbet32 or GATA337,38 expression in FoxP3+ Treg cells. However, there is emerging literature indicating that this may not be the case in autoimmune syndromes in which acquisition of Tbet generates dysfunctional Treg cells.39 As such, these data illustrate that Tbet can bind to the ECR3 locus of the FoxP3 promoter in vitro and demonstrate that lack of Tbet positively regulates peripheral Treg generation during chronic GvHD.

Tbet-deficient T-regulatory cells cross-regulate pathogenic T cells Infectious disease models suggest that Treg cells that do not express Tbet have limited functional capacity.32,40 Such functional Treg defects have been attributed to homing defects that occur in the absence of Tbet rather than to a Treg suppressor defect.41,42 We thus sought to identify whether Treg cells generated in the absence of Tbet were functional. Cell mixing studies incorporating congenic donor cells were performed to evaluate whether induced peripheral Treg cells from Tbet-deficient CD4+ cells might inhibit otherwise pathogenic WT CD4 effectors (see experimental design, Online Supplementary Figure S4A). Recipients of a 1:1 mix of WT and Tbet-deficient T cells had decreased CD4+IFNγ+ cells in the WT compartment as compared to the WT cohort alone [% CD4+IFNγ+, mean±SEM; WT versus Tbx21-/- versus WT (1:1) versus WT (1:10) versus Tbx21-/- (1:1) versus Tbx21-/- (1:10); 41.3±2.9 versus 3.8±0.6 versus 21.4±0.8 versus 28.2±4.1 versus 1.9±0.6 versus 2.5±0.8] (Figure 4A,B). Therefore, Tbet-deficient Treg cells dampened WT cytokine secretion in vivo thereby cross-regulating otherwise pathogenic effectors. Also, although increased FoxP3 frequency was noted in the WT compartment in the presence of Tbet-deficient cells [% CD4+FoxP3+, mean±SEM; WT versus Tbx21-/- versus WT (1:1) versus WT (1:10) versus Tbx21-/- (1:1) versus Tbx21-/(1:10); 1.7±0.3 versus 4.2±0.5 versus 4.5±1.1 versus 1.5±0.5 versus 3.7±0.8 versus 1.9±0.8] (Figure 4C,D), this was not reflected in absolute numbers of FoxP3 cells [#CD4+FoxP3+ (x103), mean±SEM; WT versus Tbx21-/- versus WT (1:1) versus WT (1:10) versus Tbx21-/- (1:1) versus Tbx21-/- (1:10); 14.4±5 versus 51.9±24 versus 35.8±3.6 versus 17.1±8.8 versus 59.6±23.6 versus 24.7±13.2] (Figure 4E); this result is consistent with a model in which there is a cell-intrinsic regulation of FoxP3 by Tbet.

Tbet-deficient T-regulatory cells modulate clinical graft-versus-host disease manifestations in the presence of wild-type pathogenic cells An additional experiment was performed to confirm Tbet-deficient cell cross-regulation of pathogenic WT 1452

effectors and to further characterize the resultant modulation of chronic GvHD. Consistent with the results shown in Figure 4, we found that: (i) recipients of WT T cells had reduced absolute numbers of Treg cells relative to recipients of non-alloreactive T cells during chronic GvHD; (ii) recipients of Tbet-deficient alloreactive T cells had increased Treg cells during the autoimmune phase relative to recipients of alloreactive WT cells; and (iii) Tbet-deficient T cells cross-regulate pathogenic WT cells, as evidenced by increased Treg cells (Figure 5A). Furthermore, recipients of Tbet-deficient T cells and recipients of the mix of T-bet deficient and WT T cells had reduced secretion of IFN-γ in response to syngeneic dendritic cells relative to WT T-cell recipients (Figure 5B). Syngeneic recipients of WT T cells had increased weight loss at day 17 after transfer (Figure 5C) relative to both recipients of Tbet-deficient T cells (P=0.0012) and recipients of the 1:1 mix (P=0.0003); however, weight of WT cell recipients recovered to values similar to that of other cohorts. Nonetheless, recipients of WT T cells later developed extensive hair loss, primarily across the back (see photographs, Online Supplementary Figure S5); the other three cohorts did not have clinical hair loss. To characterize the apparent cutaneous autoimmune chronic GvHD, histology was performed. In syngeneic recipients of WT cells, there was mild acanthosis and mild hyperkeratosis (Figure 5D; stage 1 of 4); in contrast, there was no acanthosis or hyperkeratosis in syngeneic recipients of Tbet-deficient cells or recipients of both WT and Tbet-deficient cells. These results were observed consistently (Figure 5E). The histological evidence of chronic skin GvHD was observed in clinically affected and non-affected skin samples in WT recipients; this latter result indicates that histological evidence of chronic GvHD is a more sensitive parameter than the clinical sign of hair loss. In this experiment, in which the clinical presentation was mild relative to that of previous experiments, we did not observe evidence of intestinal GvHD by histology. In parallel with histological studies, we performed RNA quantification of molecules associated with autoimmunity. As shown in Figure 5F, recipients of Tbet-deficient cells or the mix of WT plus Tbet-deficient cells had similar expression of CD4 RNA relative to WT recipients but had reduced expression of TGF-β1; of note, TGF-β represents an effector molecule for cutaneous chronic GvHD.43 In addition, relative to recipients of WT cells, skin samples from recipients of Tbet-deficient cells or a mix of Tbet-deficient cells plus WT cells had reduced RNA expression of inflammatory mediators, including IL6 (WT versus KO: P=0.013; WT versus Mix: P=0.006), TLR2 (WT versus KO: P=0.004; WT versus Mix: P=0.004), and TLR7 (WT versus KO: P=0.0006; WT versus Mix: P=0.0002). These data indicate that Tbet-deficient cells, even when administered with otherwise pathogenic WT cells, increase Treg cells and decrease the molecular, pathological, and clinical features of autoimmune chronic GvHD.

Discussion We used alloreactive T-cell transfer into a syngeneic murine host model to study the autoimmune manifestations of chronic GvHD. Development of autoimmunity was dependent on intact Th1 signaling in the transferred T cells. The ability of the cells to secrete IFN-γ did not haematologica | 2017; 102(8)

T-bet restricts FoxP3

affect the disease course, and we did not find any differences in the ability of the cells to proliferate or traffic to target organs in the absence of IFN-γ signaling. By contrast, the absence of IFN-γ signaling and deletion of Tbet was associated with the generation of peripheral Treg cells from CD4+ T cells that had been primed in an alloreactive environment. By comparing IFN-γ−deficient with IFN-γ signaling-deficient T cells, we dissected the critical role of Th1 signaling molecules from the less contributory role of the Th1 cytokine IFN-γ. The inflammatory syndrome associated with alloimmunity and autoimmunity has historically

been attributed to the presence of Th1 and Th17 cells.14,44 Moreover, IFN-γ is pivotal to Treg function during alloimmunity,45 while Treg cells expressing IFN-γ are dysfunctional in autoimmunity.39 Here, we show that the complete abrogation of Th1 differentiation factors rather than the absence of the Th1 cytokine IFN-γ reduces autoimmune chronic GvHD. We identified a novel regulatory mechanism by which Tbet modulates peripheral Treg generation, namely, that Tbet binds to the ECR3 locus of the Foxp3 promoter. There is a paucity of data regarding the functionality of the ECR regions upstream of the Foxp3 promoter. There

haematologica | 2017; 102(8)

Figure 4. Functional Treg cells are generated in the absence of Tbx21. In co-adoptive transfer experiments, WT T cells isolated on day 14 after allogeneic bone marrow transplantation were adoptively transferred (0.5M) into B6 Rag2-/- recipients either alone or in combination with KO T cells at a ratio of 1:1 (WT: Tbx21−/−) or 10:1 (WT: Tbx21−/−). On day 60 after secondary transfer, splenocytes were characterized (n=10 per cohort). (A) CD4+IFNγ+ cells in WT (i) and Tbx21−/− (ii), WT in the presence of Tbx21−/− (iii) cells at a 1:1 ratio, WT in the presence of Tbx21−/− cells at a 10:1 ratio (v). The bottom middle and right panels show the percentages of CD4+IFNγ+ cells in the Tbx21−/− fraction after co-existence with WT cells at a ratio of 1:1 (iv) or 1:10 (vi). (B) Pooled data are shown for CD4+IFNγ+ cell frequency, (C) FoxP3+ T cells in WT (i) and Tbx21−/− (ii), WT in the presence of Tbx21−/− (iii) cells at a 1:1 ratio, WT in the presence of Tbx21−/− cells at a 10:1 ratio (v). The bottom middle and right panels show the percentages of FoxP3+ T cells in the Tbx21−/− fraction after co-existence with WT cells at a ratio of 1:1 (iv) or 1:10 (vi). (D) Pooled data are shown for FoxP3+ Treg cell frequency (n=10 per cohort). (E) Pooled data are shown for FoxP3+ Treg cell number (n=10 per cohort).

1453

S. Amarnath et al.

Figure 5. Tbet-deficient cells have reduced chronic graft-versus-host disease potential and modulate otherwise pathogenic WT cell-induced chronic graft-versus-host disease. WT and Tbet-deficient T cells were isolated on day 14 after allogeneic bone marrow transplantation and adoptively transferred into B6 Rag2-/- recipients either alone or combined in a ratio of 1:1 (WT:Tbx21-/-). An additional control cohort (“Control”) first received a syngeneic transplant with WT cells followed by a secondary transfer into the syngeneic B6 Rag2-/- host. (A) On day 50 after secondary transfer, recipients were euthanized, and splenic T cells were evaluated by flow cytometry and the absolute number of CD4+ Foxp3+ cells derived from the WT or Tbx21-/- inoculum was calculated. (B) Post-transfer splenic T cells were also stimulated ex vivo with syngeneic dendritic cells and the 24 h supernatant was evaluated for IFN-γ content. (C) After transfer into B6 Rag2-/- hosts, the four cohorts were monitored for weight loss. (D) On day 50 after syngeneic transfer, recipients were killed and skin samples were evaluated for histological evidence of cutaneous chronic GvHD by hematoxylin and eosin staining; as illustrated in these representative photomicrographs, there was mild acanthosis and hyperkeratosis in recipients of the WT cells (affected areas indicated by arrows). (E) Cumulative data from the cohorts indicate the consistency of the histology findings; samples in the WT cohort were evaluated from both clinically affected and non-affected regions of the skin. (F) RNA expression of inflammatory molecules was evaluated in the skin tissues using QuantiGene Plex assay; results for control CD4 expression and TGF-β1 expression are shown. For these experiments, there were five replicates in the control cohort and seven replicates in the other cohorts; however, for the RNA studies, there were five replicates in each cohort. Results shown are mean values ± SEM.

1454

haematologica | 2017; 102(8)

T-bet restricts FoxP3

are three highly conserved ECR regions (ECR1, 2 and 3) which are located upstream of the transcriptional start site.35 ECR3, which lies proximal to the pppr13f gene, surprisingly, had transcriptional activity in the direction of FoxP3. While no specific role for ECR3 has been defined by our experiments, we show for the first time that the ECR3 region upstream of the Foxp3 gene possesses Tbet binding sites. The importance of Tbet binding to this locus is reflected in the in vivo experiments in which lack of Tbet allowed for the generation of peripheral Treg cells, thereby abrogating the typical Treg deficiency during chronic GvHD. Hence both the chromatin immunoprecipitation and in vivo data are consistent with a model whereby Tbet possesses a vital regulatory function in FoxP3 expression in the context of chronic GvHD. Tbet is vital for the development of acute GvHD,3 while its role during chronic GvHD is less clear. Here, for the first time, Tbet has been shown to be critical not only in the pathogenesis of chronic GvHD but also in limiting Treg cells. Indeed, a cumulative increase in Treg cells was noted in the secondary lymphoid organs and GvHD target tissues in Tbx21-/- recipients, which correlated with decreased pathogenesis. Our observations are consistent with previous reports according to which: (i) Treg cells in chronic GvHD patients with a favorable prognosis failed to express Th1 chemokine factors;46 and (ii) expression of the Th1 phenotype in Treg cells from patients with multiple sclerosis was associated with diminished Treg function.39 Our results elucidate the specific role of Th1 transcription factors in Treg function during chronic GvHD. Although the presence of Th1 transcription factors within Treg cells allows these cells to combat Th1-mediated damage during infectious disease,32,47 our data suggest that in protracted autoimmune disorders such as chronic GvHD, Tbet becomes a negative regulator of FoxP3 expression in Treg cells. Co-transfer experiments using WT and Tbx21-/- alloreactive cells confirmed the generation of functional Treg cells in the absence of Tbet. Such co-adoptive transfer decreased IFNγ+ T cells in the WT compartment, increased Treg cells during autoimmune chronic GvHD, and reduced the molecular, pathological, and clinical evidence of chronic GvHD. Our experiments also show that the GvHD autoimmune model that we utilized can result in a diversity of clinical presentations, the variability of which is not currently known but may involve factors such as host microbial status. Tbet-deficient cells, including when used in combination with otherwise pathogenic WT cells, had the potential to reduce both manifestations such as autoimmune colitis and resultant lethality and more protracted manifestations such as cutaneous chronic GvHD. We focused on cytokine-mediated events in autoimmune chronic GvHD, namely the role of IFN-γ, which we found was relatively non-contributory given the lack of disease modulation in IFN-γ KO recipients. It should be noted that cytolytic pathways, namely perforin/granzyme, fas ligand, and TNF-α contribute to GvHD, although these path-

References 1. Yi T, Chen Y, Wang L, et al. Reciprocal differentiation and tissue-specific pathogenesis of Th1, Th2, and Th17 cells in graft-versushost disease. Blood. 2009;114(14):31013112.

haematologica | 2017; 102(8)

ways are less well characterized in chronic GvHD models. Our experiments did not evaluate the role of these cytolytic pathways, and as such, it is possible that deficiency in these pathways might ameliorate autoimmune GvHD and perhaps likely that Tbet-deficiency would reduce these cytolytic effector mechanisms for amelioration of disease. However, the capacity of the Tbet-deficient cells to downregulate the otherwise pathogenic WT cells in cell mixing studies provides evidence for a cross-regulatory mechanism to the Tbet-deficiency finding rather than a more direct mechanism involving lack of cytotoxic effector molecule expression. We further clarified the immunopathology of autoimmunity associated with GvHD. Early reports suggested that both Th1 and Th2 cells have specific roles in causing acute GvHD and chronic GvHD24 but not until recently has the immunopathology of chronic GvHD been widely investigated. The autoimmune manifestations in GvHD are primarily thought to be of Th1 origin11,14,48,49 with minimal involvement of Th17 cells.14,50 Here, we found that autoimmune chronic GvHD is indeed caused by CD4+ cells of Th1 origin. Our data are also consistent with clinical observations in chronic GvHD patients where poor prognosis is correlated to decreased Tregs.16,29 In summary, our study identifies a novel molecular mechanism that controls the T-bet/FoxP3 axis in the context of chronic GvHD. The data presented here suggest that adoptive Treg cell therapy strategies currently being pursued to treat chronic GvHD post-BMT may not be fully effective until methods to control Th1 signaling can be harnessed. Inhibition of transcriptions factors such as Tbet and STAT4 represent key molecular targets for the treatment of autoimmune GvHD. In addition, our work specifically delineated the IFNγ signaling pathway in preventing autoimmunity. To boost anti-viral responses, an intact type I IFNγ signaling and STAT1 activation is critical. The data presented here suggest that by specifically inhibiting type II IFNγ receptor signaling, one might be able to augment Treg cells during chronic GvHD while maintaining type I mediated anti-viral responses. We envision that use of antagonists against IFNγ type II receptor might be particularly beneficial for therapy of autoimmune chronic GvHD. Therefore, the Tbet pathway, including the IFN-γ receptor and STAT1/STAT4 as upstream pathway members, drives experimental autoimmune GvHD. Interventions that restrict the Tbet pathway might either be used alone or in combination with adoptive Treg-cell therapy for treatment of autoimmune chronic GvHD. Funding This study was funded by the Intramural Research Program, National Cancer Institute, National Institutes of Health, USA and Newcastle University Research Fellowship, Newcastle University, UK. AL is supported by the Crohns and Colitis Foundation of America (CCFA).

2. Taylor PA, Lees CJ, Blazar BR. The infusion of ex vivo activated and expanded CD4(+)CD25(+) immune regulatory cells inhibits graft-versus-host disease lethality. Blood. 2002;99(10):3493-3499. 3. Yu Y, Wang D, Liu C, et al. Prevention of GVHD while sparing GVL effect by target-

ing Th1 and Th17 transcription factor T-bet and RORgammat in mice. Blood. 2011;118 (18):5011-5020. 4. Laurence A, Amarnath S, Mariotti J, et al. STAT3 transcription factor promotes instability of nTreg cells and limits generation of iTreg cells during acute murine graft-versus-

1455

S. Amarnath et al. host disease. Immunity. 2012;37(2):209-222. 5. Blazar BR, Murphy WJ, Abedi M. Advances in graft-versus-host disease biology and therapy. Nat Rev Immunol. 2012;12(6):443458. 6. Sugerman PB, Savage NW, Walsh LJ, et al. The pathogenesis of oral lichen planus. Crit Rev Oral Biol Med. 2002;13(4):350-365. 7. Snyder LD, Palmer SM. Immune mechanisms of lung allograft rejection. Semin Respir Crit Care Med. 2006;27(5):534-543. 8. Bharat A, Fields RC, Steward N, Trulock EP, Patterson GA, Mohanakumar T. CD4+25+ regulatory T cells limit Th1-autoimmunity by inducing IL-10 producing T cells following human lung transplantation. Am J Transplant. 2006;6(8):1799-1808. 9. Regauer S. Immune dysregulation in lichen sclerosus. Eur J Cell Biol. 2005;84(2-3):273277. 10. Shlomchik WD, Lee SJ, Couriel D, Pavletic SZ. Transplantation's greatest challenges: advances in chronic graft-versus-host disease. Biol Blood Marrow Transplant. 2007;13(1 Suppl 1):2-10. 11. Chen X, Vodanovic-Jankovic S, Johnson B, Keller M, Komorowski R, Drobyski WR. Absence of regulatory T-cell control of TH1 and TH17 cells is responsible for the autoimmune-mediated pathology in chronic graftversus-host disease. Blood. 2007;110(10): 3804-3813. 12. Rangarajan H, Yassai M, Subramanian H, et al. Emergence of T cells that recognize nonpolymorphic antigens during graft-versushost disease. Blood. 2012;119(26):63546364. 13. Zeiser R, Blazar BR. Preclinical models of acute and chronic graft-versus-host disease: how predictive are they for a successful clinical translation? Blood. 2016;127(25):31173126. 14. Chen X, Das R, Komorowski R, van Snick J, Uyttenhove C, Drobyski WR. Interleukin 17 is not required for autoimmune-mediated pathologic damage during chronic graft-versus-host disease. Biol Blood Marrow Transplant. 2010;16(1):123-128. 15. Radojcic V, Pletneva MA, Yen HR, et al. STAT3 signaling in CD4+ T cells is critical for the pathogenesis of chronic sclerodermatous graft-versus-host disease in a murine model. J Immunol. 2010;184(2):764-774. 16. Koreth J, Matsuoka K, Kim HT, et al. Interleukin-2 and regulatory T cells in graftversus-host disease. N Engl J Med. 2011;365(22):2055-2066. 17. Zhang C, Todorov I, Zhang Z, et al. Donor CD4+ T and B cells in transplants induce chronic graft-versus-host disease with autoimmune manifestations. Blood. 2006;107(7):2993-3001. 18. Rosenblum MD, Remedios KA, Abbas AK. Mechanisms of human autoimmunity. J Clin Invest. 2015;125(6):2228-2233. 19. Suscovich TJ, Perdue NR, Campbell DJ. Type-1 immunity drives early lethality in scurfy mice. Eur J Immunol. 2012; 42(9):2305-2310. 20. Afkarian M, Sedy JR, Yang J, et al. T-bet is a STAT1-induced regulator of IL-12R expression in naive CD4+ T cells. Nat Immunol. 2002;3(6):549-557. 21. Yang XP, Jiang K, Hirahara K, et al. EZH2 is

1456

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

crucial for both differentiation of regulatory T cells and T effector cell expansion. Sci Rep. 2015;5:10643. Hwang J, Kita R, Kwon HS, et al. Epidermal ablation of Dlx3 is linked to IL-17-associated skin inflammation. Proc Natl Acad Sci USA. 2011;108(28):11566-11571. Yang XP, Ghoreschi K, Steward-Tharp SM, et al. Opposing regulation of the locus encoding IL-17 through direct, reciprocal actions of STAT3 and STAT5. Nat Immunol. 2011;12(3):247-254. Nikolic B, Lee S, Bronson RT, Grusby MJ, Sykes M. Th1 and Th2 mediate acute graftversus-host disease, each with distinct endorgan targets. J Clin Invest. 2000;105(9): 1289-1298. Nguyen VH, Zeiser R, Dasilva DL, et al. In vivo dynamics of regulatory T-cell trafficking and survival predict effective strategies to control graft-versus-host disease following allogeneic transplantation. Blood. 2007;109(6):2649-2656. Di Biaso I, Di Maio L, Bugarin C, et al. Regulatory T cells and extracorporeal photochemotherapy: correlation with clinical response and decreased frequency of proinflammatory T cells. Transplantation. 2009;87(9):1422-1425. Giorgini A, Noble A. Blockade of chronic graft-versus-host disease by alloantigeninduced CD4+CD25+Foxp3+ regulatory T cells in nonlymphopenic hosts. J Leukoc Biol. 2007;82(5):1053-1061. Rieger K, Loddenkemper C, Maul J, et al. Mucosal FOXP3+ regulatory T cells are numerically deficient in acute and chronic GvHD. Blood. 2006;107(4):1717-1723. Zorn E, Kim HT, Lee SJ, et al. Reduced frequency of FOXP3+ CD4+CD25+ regulatory T cells in patients with chronic graft-versus-host disease. Blood. 2005;106(8):29032911. Finotto S, Neurath MF, Glickman JN, et al. Development of spontaneous airway changes consistent with human asthma in mice lacking T-bet. Science. 2002;295(5553): 336-338. Choi J, Ziga ED, Ritchey J, et al. IFNgammaR signaling mediates alloreactive T cell trafficking and GvHD. Blood. 2012;120(19):4093-4103. Koch MA, Tucker-Heard G, Perdue NR, Killebrew JR, Urdahl KB, Campbell DJ. The transcription factor T-bet controls regulatory T cell homeostasis and function during type 1 inflammation. Nat Immunol. 2009;10(6): 595-602. Ma H, Lu C, Ziegler J, et al. Absence of Stat1 in donor CD4(+) T cells promotes the expansion of Tregs and reduces graft-versushost disease in mice. J Clin Invest. 2011;121(7):2554-2569. Wei J, Duramad O, Perng OA, Reiner SL, Liu YJ, Qin FX. Antagonistic nature of T helper 1/2 developmental programs in opposing peripheral induction of Foxp3+ regulatory T cells. Proc Natl Acad Sci USA. 2007;104(46):18169-18174. Eckerstorfer P, Novy M, BurgstallerMuehlbacher S, et al. Proximal human FOXP3 promoter transactivated by NFkappaB and negatively controlled by feedback loop and SP3. Mol Immunol.

2010;47(11-12):2094-2102. 36. Nakayamada S, Kanno Y, Takahashi H, et al. Early Th1 cell differentiation is marked by a Tfh cell-like transition. Immunity. 2011;35(6):919-931. 37. Wohlfert EA, Grainger JR, Bouladoux N, et al. GATA3 controls Foxp3(+) regulatory T cell fate during inflammation in mice. J Clin Invest. 2011;121(11):4503-4515. 38. Wang Y, Su MA, Wan YY. An essential role of the transcription factor GATA-3 for the function of regulatory T cells. Immunity. 2011;35(3):337-348. 39. Dominguez-Villar M, Baecher-Allan CM, Hafler DA. Identification of T helper type 1like, Foxp3+ regulatory T cells in human autoimmune disease. Nat Med. 2011;17(6): 673-675. 40. Koch MA, Thomas KR, Perdue NR, Smigiel KS, Srivastava S, Campbell DJ. T-bet(+) Treg cells undergo abortive Th1 cell differentiation due to impaired expression of IL-12 receptor beta2. Immunity. 2012;37(3):501510. 41. Yu F, Sharma S, Edwards J, Feigenbaum L, Zhu J. Dynamic expression of transcription factors T-bet and GATA-3 by regulatory T cells maintains immunotolerance. Nat Immunol. 2015;16(2):197-206. 42. McPherson RC, Turner DG, Mair I, O'Connor RA, Anderton SM. T-bet expression by Foxp3(+) T regulatory cells is not essential for their suppressive function in CNS autoimmune disease or colitis. Front Immunol. 2015;6:69. 43. Du J, Paz K, Flynn R, et al. Pirfenidone ameliorates murine chronic GVHD through inhibition of macrophage infiltration and TGFbeta production. Blood. 2017;129(18):25702580. 44. Amarnath S, Mangus CW, Wang JC, et al. The PDL1-PD1 axis converts human TH1 cells into regulatory T cells. Sci Transl Med. 2011;3(111):111ra120. 45. Koenecke C, Lee CW, Thamm K, et al. IFNgamma production by allogeneic Foxp3+ regulatory T cells is essential for preventing experimental graft-versus-host disease. J Immunol. 2012;189(6):2890-2896. 46. Croudace JE, Inman CF, Abbotts BE, et al. Chemokine-mediated tissue recruitment of CXCR3+ CD4+ T-cells plays a major role in the pathogenesis of chronic graft versus host disease. Blood. 2012;120(20):4246-4255. 47. Hall AO, Beiting DP, Tato C, et al. The cytokines interleukin 27 and interferongamma promote distinct Treg cell populations required to limit infection-induced pathology. immunity. 2012;37(3): 511-523. 48. Broady R, Yu J, Chow V, et al. Cutaneous GVHD is associated with the expansion of tissue-localized Th1 and not Th17 cells. Blood. 2010;116(25):5748-5751. 49. Imanguli MM, Swaim WD, League SC, Gress RE, Pavletic SZ, Hakim FT. Increased T-bet+ cytotoxic effectors and type I interferon-mediated processes in chronic graftversus-host disease of the oral mucosa. Blood. 2009;113(15):3620-3630. 50. Nishimori H, Maeda Y, Teshima T, et al. Synthetic retinoid Am80 ameliorates chronic graft-versus-host disease by down-regulating Th1 and Th17. Blood. 2012;119(1): 285-295.

haematologica | 2017; 102(8)

ARTICLE

Cell Therapy & Immunotherapy

Shorter leukocyte telomere length is associated with higher risk of infections: a prospective study of 75,309 individuals from the general population

EUROPEAN HEMATOLOGY ASSOCIATION

Ferrata Storti Foundation

Jens Helby,1,2,3 BĂ¸rge G. Nordestgaard,1,2,3,4 Thomas Benfield3,5 and Stig E. Bojesen1,2,3,4

Department of Clinical Biochemistry, Herlev and Gentofte Hospital, Copenhagen University Hospital; 2The Copenhagen General Population Study, Herlev and Gentofte Hospital, Copenhagen University Hospital; 3Faculty of Health and Medical Sciences, University of Copenhagen; 4The Copenhagen City Heart Study, Bispebjerg and Frederiksberg Hospital, Copenhagen University Hospital and 5Department of Infectious Diseases, Hvidovre Hospital, Copenhagen University Hospital, Denmark 1

Haematologica 2017 Volume 102(8):1457-1465

ABSTRACT

n the general population, older age is associated with short leukocyte telomere length and with high risk of infections. In a recent study of allogeneic hematopoietic cell transplantation for severe aplastic anemia, long donor leukocyte telomere length was associated with improved survival in the recipients. These findings suggest that leukocyte telomere length could possibly be a marker of immune competence. Therefore, we tested the hypothesis that shorter leukocyte telomere length is associated with higher risk of infectious disease hospitalization and infection-related death. Relative peripheral blood leukocyte telomere length was measured using quantitative polymerase chain reaction in 75,309 individuals from the general population and the individuals were followed for up to 23 years. During follow up, 9228 individuals were hospitalized with infections and infection-related death occurred in 1508 individuals. Shorter telomere length was associated with higher risk of any infection (hazard ratio 1.05 per standard deviation shorter leukocyte telomere length; 95% confidence interval 1.03-1.07) and pneumonia (1.07; 1.03-1.10) after adjustment for conventional infectious disease risk factors. Corresponding hazard ratios for infection-related death were 1.10 (1.04-1.16) for any infection and 1.11 (1.04-1.19) for pneumonia. Telomere length was not associated with risk of skin infection, urinary tract infection, sepsis, diarrheal disease, endocarditis, meningitis or other infections. In conclusion, our findings indicate that leukocyte telomere length may be a marker of immune competence. Further studies are needed to determine whether risk of infections in allogeneic hematopoietic cell transplantation recipients can be reduced by considering donor leukocyte telomere length when selecting donors.

Correspondence: stig.egil.bojesen@regionh.dk

Received: December 11, 2016. Accepted: May 12, 2017. Pre-published: May 18, 2017. doi:10.3324/haematol.2016.161943 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/102/8/1457 ÂŠ2017 Ferrata Storti Foundation

Introduction Telomeres are located at the chromosome tips and are composed of protein and tandem repeats of the nucleotide sequence TTAGGG.1 In most cell types, the telomeric DNA becomes shorter with each mitotic cell division due to the end replication problem.2,3 If telomeres reach a critically short length, further cell divisions may not be possible and cells become senescent or undergo apoptosis.1,4 In the general population, older age is associated with short telomere length in peripheral blood leukocytes and with high risk of infections,5-7 but it is currently unknown whether short leukocyte telomere length is a cause of impaired immune competence.8 In a recent study of allogeneic hematopoietic cell transplantation (allo-HCT) for severe aplastic anemia, long donor leukocyte telomere length was haematologica | 2017; 102(8)

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.

1457

J. Helby et al. Table 1. Baseline characteristics of 75,309 individuals from the general population according to age-adjusted quartiles of leukocyte telomere length.

Characteristic

1st (longest)

Individuals, n. Relative telomere length, T/S-ratio

18,800 0.79 (0.74-0.88) 57 (47-67) 7875 (42) 11,091 (59) 17 (7-31) 7047 (37) 25.4 (23.0-28.2) 3549 (19) 8138 (43) 1.4 (1.0-2.5)

Age, years Male sex, n. Ever smokers, n. Cumulative smoking#, pack-years Alcohol consumption >168/84 g/week†, n. Body mass index, kg/m2 Any comorbidity§, n. Previously hospitalized‡, n. C-reactive protein, mg/L

Telomere length quartiles 2nd 3rd 18,834 0.65 (0.61-0.68) 57 (47-67) 8295 (44) 11,415 (61) 18 (7-33) 7279 (39) 25.5 (23.2-28.5) 3721 (20) 8092 (43) 1.5 (1.1-2.5)

18,816 0.56 (0.52-0.59) 57 (47-67) 8547 (45) 11,646 (62) 19 (8-34) 7301 (39) 25.7 (23.2-28.5) 3820 (20) 8078 (43) 1.5 (1.1-2.6)

4th (shortest) 18,859 0.45 (0.41-0.50) 57 (47-67) 9059 (48) 11,984 (64) 20 (9-35) 7027 (37) 25.8 (23.3-28.8) 3898 (21) 8306 (44) 1.6 (1.1-2.7)

Number (n.) (%) is shown for categorical variables and median (interquartile range, IQR) is shown for continuous variables. #Ever smokers only. †>168 g/week for men and >84 g/week for women. §As defined by the Charlson comorbidity index. ‡Defined as any inpatient hospitalization within ten years before study enrollment for any cause other than infections.

associated with improved long-term survival in the recipients.9 Due to insufficient statistical power in the analysis of cause-specific deaths, the study could not identify the reason why survival was higher in patients receiving transplants from donors with long leukocyte telomere length. Since infections are among the leading causes of death in aplastic anemia patients treated with alloHCT,10,11 a possible explanation could be that leukocyte telomere length is a marker of overall immune competence in the donor. However, no previous studies have examined the association between telomere length and risk of hospitalization for infectious disease in the general population, and studies on telomere length and risk of infection-related death have produced conflicting results.1215 If leukocyte telomere length is a marker of overall immune competence in the general population, it would indicate that shortening of telomeres may be one of the biological mechanisms underlying age-related decline in adaptive immunity and increase in infectious disease susceptibility. It is currently unknown whether a marker of immune competence can be useful for making decisions on treatment or other health interventions in individuals from the general population, but such a marker may potentially be useful when selecting donors for allo-HCT. Therefore, examination of the possible association between leukocyte telomere length and immune competence could provide important information of clinical as well as biological relevance. Given this, we studied 75,309 individuals from the general population to test the hypothesis that shorter telom1458

ere length in leukocytes is associated with higher risk of hospitalization for infectious disease and higher risk of infection-related death. All individuals had telomere length measured in peripheral blood leukocytes at study enrollment and the participants were prospectively followed for up to 23 years for hospitalization for infectious disease and death.

Methods Participants We studied 75,309 individuals from two studies of the general population: 8681 individuals from the Copenhagen City Heart Study16,17 (enrolled between 1991 and 1994) and 66,628 individuals from the Copenhagen General Population Study18,19 (enrolled between 2003 and 2012). Using the Danish Civil Registration System,20 which includes a unique identification number for all individuals with permanent residence in Denmark, the participants were randomly invited to represent the general population aged 20-100 years. In the Copenhagen City Heart study, 61% of the invited individuals participated, and in the Copenhagen General Population Study, 46% participated. At the day of examination, all participants completed a questionnaire on lifestyle and health, underwent a physical examination, and had blood samples drawn. None of the individuals participated in both studies. Among the participants, more than 99% were white and of Danish descent. No participants were lost to follow up. The studies were approved by Danish ethical committees and all participants provided written, informed consent. haematologica | 2017; 102(8)

Leukocyte telomere length and risk of infections

Figure 1. Risk of first hospitalization for any infection and specific infections in the general population per standard deviation shorter telomere length. The sum of specific infections exceeds the number of any infection since some individuals had more than one type of infection. Multivariable models were adjusted for values at study enrollment of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort. CI: confidence interval

Covariates All included covariates were chosen a priori based on previous studies reporting them to be associated with telomere length and risk of infections.19,21-24 Information on age and sex was obtained from the Danish Civil Registration system, while information on smoking status (current/former/never), cumulative smoking in pack-years (with one pack-year defined as 20 cigarettes or equivalent per day for a year), alcohol consumption (none/moderate/heavy, with heavy defined as >168 g/week for men and >84 g/week for women, as recommended by The Danish Health Authority), and body mass index (measured weight in kilograms divided by measured height in meters squared) was derived from the questionnaire and physical examination. As a marker of inflammation, plasma high-sensitivity Creactive protein level at study enrollment was measured using standard hospital assays.

Telomere length measurements After isolation of DNA from peripheral blood leukocytes using the Qiagen blood kit,25 relative telomere length was measured by a modified monochrome multiplex quantitative polymerase chain reaction (qPCR) method,26 as previously described in detail elsewhere.19,21 The qPCR method was chosen since this is the only available method for high-throughput measurements of telomere length.1,27 In short, the telomere template was amplified simultaneously with the single copy gene for albumin in the same well to adjust for different amounts of DNA in the samples. All samples were run in quadruplicates. For each sample, we calculated the ratio between telomere (T) repeat copy numbers and single gene (S) copy numbers, and the relative telomere length expressed as the T/S ratio was derived through calibration with measurements on K562 cell line DNA, which was included in each plate. Using this method, the T/S ratio of the K562 cell line DNA is by definition set to 1, so samples with shorter telomere length than K562 cell line DNA have T/S ratios below 1 while samples with longer haematologica | 2017; 102(8)

telomere length than K562 cell line DNA have T/S ratios above 1.26 To obtain a functional single-calibrator measurement, telomere length measurements were adjusted across calibrator lots, as described in detail in the Online Supplementary Methods. The laboratory technician performing all measurements was blinded to infectious disease endpoints and deaths.

Genotypes An inherent limitation of observational studies on biomarkers and risk of disease is their inability to determine whether a biomarker are having a causal effect on the risk of disease, or if any found associations could be due to confounding or reverse causation. A way of overcoming this limitation is to study whether single nucleotide polymorphisms (SNPs) that influence the level of a biomarker are also associated with risk of disease. Using this approach, we genotyped participants for the three SNPs, rs1317082, rs7726159, and rs2487999, to examine whether a genetic predisposition to shorter telomere length is associated with risk of infections. The 3 SNPs were chosen based on a genome wide association study including 26,089 individuals, which found they were the SNPs most strongly associated with measured leukocyte telomere length.28 The three genotypes were combined into an overall unweighted allele score, taking values from 0 to 6, which was calculated as the sum of the number of telomere length shortening alleles for each of the three genotypes. Since there were only 68 individuals with 0 telomere length shortening alleles, individuals with an unweighted allele score of 0 and 1 were combined into a single group. Similarly, a weighted allele score was constructed by first assigning each SNP a weight according to the SNPs average per allele effect size on measured T/S ratio, and then calculating the sum of weighted telomere shortening alleles for the three SNPs combined. To increase power, genotyping was also performed on participants from the two general population studies for whom measured telomere length was not available, leading to a total of 107,693 individuals genotyped. 1459

J. Helby et al.

Figure 2. Risk of first hospitalization for any infection and pneumonia in the general population according to age-adjusted quartiles of telomere length. Multivariable models were adjusted for values at study enrollment of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort. CI: confidence interval.

More details on genotyping are described in the Online Supplementary Methods.

Infectious disease endpoints Using the national Danish Patient Registry,29 which covers all Danish hospitals, we obtained information for each individual participant on inpatient hospitalizations with a primary discharge diagnosis of an infectious disease from January 1st 1977 until November 5th 2014. Likewise, we obtained information from the national Danish Patient Registry on emergency room visits with a primary diagnosis of an infectious disease from January 1st 1994 until November 5th 2014. Information on vital status and date of death until November 5th 2014 was retrieved from the Danish Civil Registration system. For participants who died before December 31st 2012, information on infection-related deaths was obtained from the national Danish Register of Causes of Death,30 which covers all deaths in Denmark and registers all diagnoses that a physician has listed on the death certificate as contributing to the cause of death. More details on classification of infectious disease endpoints are described in Online Supplementary Table S1 and in the Online Supplementary Methods.

Comorbidities Non-infectious comorbidities could possibly confound the association between telomere length and risk of infections, since short leukocyte telomere length has been found in individuals with chronic conditions such as heart failure31 and chronic obstructive pulmonary disease,32 and individuals with these conditions are also at high risk of infections.22,24 To reduce such confounding, comorbidities at study enrollment were assessed using the 1460

Charlson comorbidity index,33 which is a severity weighted measure of comorbid conditions that has been validated for its ability to predict mortality.34,35 Individuals at high risk of undiagnosed comorbidities were identified using measurements of white blood cell differential count, platelet count, blood hemoglobin, plasma alanine aminotransferase, plasma creatinine, and non-fasting plasma glucose at study enrollment. These measurements were chosen as a broad screening for hematologic, hepatic, renal, and metabolic diseases. More details on assessment of comorbidities are described in the Online Supplementary Methods.

Statistical analysis Statistical analyses were performed with Stata v. 13.1. All statistical tests were two-sided. Since several studies have documented a strong association between higher age and shorter telomere length,5,6,19 calculation of telomere length quartiles was adjusted for age by computing quartiles of telomere length separately for each one-year age span. To reduce risk of reverse causation (i.e. that subclinical infections already present at date of examination may influence leukocyte telomere length), individuals were only included in the analyses if they were not hospitalized with any infection during the first 180 days after the date of examination. Risk of infectious disease hospitalization and risk of infectionrelated death were modeled separately by Cox proportional hazards regression using left-truncated age as the timescale, as described in detail in the Online Supplementary Methods. For the analyses on measured telomere length and risk of hospitalization for infectious disease, and for risk of infection-related death, follow up began 180 days after the date of examination. Multivariable models were adjusted for values at study enrollment haematologica | 2017; 102(8)

Leukocyte telomere length and risk of infections

Figure 3. Risk of first hospitalization for any infection and pneumonia in the general population per standard deviation shorter telomere length stratified according to follow-up interval. Multivariable models were adjusted for values at study enrollment of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort. P for interaction was calculated using a likelihood ratio test, comparing models with and without an interaction term. CI: confidence interval.

of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort.

Results Among 75,309 individuals from the general population, older age was strongly associated with shorter leukocyte telomere length with a mean lower T/S ratio of 0.0032 [95% confidence interval (CI): 0.0032-0.0033; linear regression P<1*10-300; R-squared=0.073] for each year of older age. Therefore, baseline characteristics are shown after adjustment for age by calculation of age-adjusted quartiles of telomere length (Table 1).

Telomere length and risk of infections During a median follow up of seven years (range 0-23 years), 9228 individuals were hospitalized due to infections. When examining risk of first hospitalization for infection after study enrollment, a one standard deviation shorter leukocyte telomere length was associated with higher risk of any infection [hazard ratio (HR) 1.05; 95%CI: 1.03-1.07] and pneumonia (HR 1.07; 95%CI: 1.031.10) (Figure 1). We found no association between telomere length and risk of skin infection, urinary tract infection, sepsis, diarrheal disease, endocarditis, meningitis, or other infections. Similarly, when examining risk of first hospitalization for infection according to age-adjusted quartiles of telomere length, increasing quartiles of telomere length were associated with higher risk of any infection (P for trend=1*10-7) and pneumonia (P for trend=3* 10-7) (Figure 2). Hazard ratios for the shortest versus the haematologica | 2017; 102(8)

longest quartile of telomere length were 1.16 (95%CI: 1.09-1.23) for any infection and 1.25 (95%CI: 1.15-1.36) for pneumonia. When including both first and recurrent infectious disease hospitalizations (Online Supplementary Figure S1), risk estimates per standard deviation shorter telomere length were similar to those from the analysis on risk of first hospitalization for infectious disease (Figure 1).

Stratified analyses on risk of infections To investigate whether the association between shorter telomere length and higher risk of any infection and pneumonia may be due to reverse causation, we performed stratified analyses according to follow-up interval. We found no indication of reverse causation, as risk estimates for any infection became more pronounced as more time elapsed after study enrollment (P for interaction with follow-up interval=0.006), while risk estimates for pneumonia were largely stable across follow-up intervals (P for interaction with follow-up interval=0.09) (Figure 3). To further examine the robustness of the association between shorter telomere length and higher risk of any infection and pneumonia, we performed stratified analyses according to strata of the covariates that were included in the multivariable models (Figure 4 and Online Supplementary Figure S2). To additionally reduce confounding by comorbidities that may be undiagnosed at study enrollment, we also performed stratified analyses according to whether or not participants had normal blood laboratory tests at study enrollment. In all of the above mentioned strata, risk estimates remained stable for any infection (Figure 4) and pneumonia (Online Supplementary Figure S2). As previous studies have found that shorter telomere 1461

J. Helby et al.

Figure 4. Stratified analyses for risk of first hospitalization for any infection per standard deviation shorter telomere length. Number of individuals at risk and number of infections vary slightly among the stratifications due to varying numbers of individuals with missing data on each of the covariates. Multivariable models were adjusted for values at study enrollment of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort. P for interaction was calculated using a likelihood ratio test, comparing models with and without an interaction term. CI: confidence interval. #Ever smokers only. †>168 g/week for men and >84 g/week for women. §As defined by the Charlson comorbidity index. ‡Defined as any inpatient hospitalization within ten years before study enrollment for any cause other than infections. ¶Includes measurements of white blood cell differential count, platelet count, blood hemoglobin, plasma alanine aminotransferase, plasma creatinine, and non-fasting plasma glucose.

length was associated with higher risk of cardiovascular disease,21,36-39 and since cardiovascular disease is associated with high risk of infections,22,24 we investigated whether the observed association between shorter telomere length and risk of infections may be secondary to cardiovascular disease. Risk estimates for any infection and pneumonia remained stable when the analyses were stratified according to whether or not individuals were diagnosed with any type of cardiovascular disease at study enrollment or during follow up (Online Supplementary Figure S3). Similarly, risk estimates for any infection and pneumonia were stable when including cardiovascular disease diagnosed at study enrollment or during follow up as a timedependent variable in the multivariable model (Online Supplementary Figure S4).

Telomere length and risk of infection-related death One standard deviation shorter telomere length was associated with higher risk of death related to any infection (HR 1.10; 95%CI: 1.04-1.16) and pneumonia (HR 1.11; 95%CI: 1.04-1.19) (Figure 5). We found no association between shorter telomere length and risk of death related to sepsis, urinary tract infection, diarrheal disease, endocarditis, skin infection, or other infections. For meningitis, exact calculations of risk estimates were not 1462

possible since there were only 2 meningitis-related deaths. Since previous studies on telomere length and risk of infection-related death have produced conflicting results which could potentially be explained by limited statistical power,12-15 we performed power calculations based on the risk estimates from our present study and the number of participants and deaths reported in previous studies (Online Supplementary Table S2). All four previous studies each had less than 15% power to detect a hazard ratio for any infection-related death of 1.10 per standard deviation shorter telomere length at two-sided P<0.05.

Genetic predisposition to shorter telomere length and risk of infections Among the 107,693 genotyped individuals, 21,317 individuals were hospitalized due to any infection during follow up (see Online Supplementary Table S3 for baseline characteristics). Mean T/S ratio decreased by 0.012 (95%CI: 0.011-0.014) per telomere length shortening allele among the 75,018 individuals who had both leukocyte telomere length measurements and genotyping performed (linear regression on telomere length as a function of unweighted allele score: standardized β=-0.076; P=2*10-97; Rsquared=0.0058; same regression for weighted allele score: standardized β=-0.077; P=6*10-99; R-squared=0.0059). For haematologica | 2017; 102(8)

Leukocyte telomere length and risk of infections

Figure 5. Risk of death related to any infection and death related to specific infections in the general population per standard deviation shorter telomere length. The sum of deaths related to specific infections exceeds the number of deaths related to any infection since some individuals had more than one type of infection listed on the death certificate. For meningitis, exact calculations of risk estimates were not possible since there were only 2 meningitis-related deaths. Multivariable models were adjusted for values at study enrollment of age, sex, smoking status, cumulative smoking in pack-years, alcohol consumption, body mass index, plasma C-reactive protein level, Charlson comorbidity index, number of non-infectious disease hospitalizations within ten years before study enrollment, and study cohort. CI: confidence interval.

both the unweighted and the weighted allele score, we found no association between a genetic predisposition to shorter telomere length and risk of hospitalization for any infection, pneumonia, skin infection, urinary tract infection, diarrheal disease, sepsis, meningitis, endocarditis or other infections (Online Supplementary Figures S5-S7). Importantly, however, genetic risk estimates did not differ from observational estimates, but as seen from the power calculation in the Online Supplementary Results, the rather modest influence of the SNPs on telomere length leads to very limited power in the analyses on a genetic predisposition to shorter telomeres and risk of infections, despite the large number of individuals genotyped.

Discussion In this prospective study of 75,309 individuals from the general population, we found that shorter leukocyte telomere length was associated with higher risk of hospitalization due to any infection and pneumonia. Likewise, shorter leukocyte telomere length was associated with higher risk of death related to any infection and pneumonia. Our findings on telomere length and risk of hospitalization for infectious disease are novel, while our findings on risk of infection-related death corroborate the previous studies on this subject, which had produced conflicting results.12-15 The higher risk of infections in individuals with shorter leukocyte telomere length could possibly be caused by impaired adaptive immune function as a consequence of short lymphocyte telomere length. This possible mechanism is supported by a study on the immune response after influenza vaccination,40 which found that individuals with long telomere length in B-lymphocytes produced a more robust antibody response when compared to individuals with short B-lymphocyte telomere length. Likewise, when prehaematologica | 2017; 102(8)

sented with a synthetic influenza peptide in vitro, influenzaspecific CD8+ T-lymphocytes with long telomere length showed higher proliferative capacity than those with short telomere length.40 This indicates that telomere length in subtypes of lymphocytes may be directly related to the effectiveness of the adaptive immune response. Furthermore, aging leads to increasing numbers of CD8+ Tlymphocytes undergoing replicative senescence, and these senescent lymphocytes are characterized by short telomere length and no expression of the co-stimulatory molecule CD28 which is necessary for proliferation.41-44 Combined with our results, these findings suggest that shortening of telomeres is likely to be one of the mechanisms underlying age-related decline in adaptive immunity and increased susceptibility to infectious disease.7,8 Our finding that shorter leukocyte telomere length was associated with higher risk of hospitalization due to any infection and pneumonia is novel, as no previous studies have examined the association between leukocyte telomere length and risk of hospitalization for infectious disease in the general population. Nonetheless, our results are supported by a study on 152 individuals,45 which found that shorter telomere length in peripheral blood mononuclear cells was associated with higher risk of rhinovirus infection in healthy volunteers who were quarantined and experimentally exposed to nasal drops containing rhinovirus. When we investigated risk of specific types of infection, shorter telomere length was associated exclusively with higher risk of pneumonia. However, hazard ratios per standard deviation shorter telomere length was consistently above 1 for skin infection, diarrheal disease, endocarditis and meningitis, although these findings were not statistically significant at a level of P<0.05. Hypothetically, our finding that pneumonia was the only specific type of infection with higher risk in individuals with shorter telomeres could 1463

J. Helby et al.

simply be due to the high number of hospitalizations for pneumonia in the general population, leading to more statistical power in the analysis on risk of pneumonia than for other infections. Somewhat similar to our findings, individuals with humoral immunodeficiencies such as X-linked agammaglobulinemia, common variable immunodeficiency, and selective immunoglobulin A deficiency have especially pronounced risk of bacterial respiratory tract infections, often caused by encapsulated bacteria such as Streptococcus pneumoniae and Haemophilus influenzae.46 We did not have information on specific causative pathogens for hospitalizations for pneumonia in the present study, but other studies have found Streptococcus pneumoniae and Haemophilus influenzae to be among the most common causes of bacterial pneumonia in the general population.47,48 Hence, the pattern of risk of infectious disease associated with shorter telomere length may to some degree resemble what has been reported for specific deficiencies of the humoral immune response. Our finding that shorter leukocyte telomere length was associated with higher risk of death related to any infection is supported by two independent prospective cohort studies, which together included a total of 1260 individuals, of whom 83 died from infectious diseases.12,13 However, two other prospective cohort studies including a total of 3319 individuals and 90 infectious disease deaths found no association between telomere length and risk of death due to infectious disease.14,15 When calculating power based on the risk estimates from our present study and the number of participants and deaths reported in previous studies, the four previous studies each had less than 15% power to detect the risk estimate for any infection-related death found in the present study. This suggests that the conflicting results from previous studies could possibly be related to limited statistical power because of the modest number of deaths from infectious disease in the study populations, which is less of a concern in our present study with 75,309 individuals and 1508 infection-related deaths. Importantly, the precision of the telomere length measurements also affects power, which is especially relevant since studies on telomere length and infection-related death have used different methods for measuring telomere length, and since the precision of the measurements may differ substantially between methods.49,50 The results from our study and those from previous studies suggest that leukocyte telomere length may be a marker of overall immune competence in the general population. It is still not known whether such a marker can be useful for making decisions on treatment or other health interventions in individuals from the general population. However, a marker of overall immune competence could potentially be useful when selecting donors for allogeneic hematopoietic cell transplantation (allo-HCT), where infections are among the leading causes of death in the recipients.10,11 Theoretically, it may be possible to reduce the risk of serious infections and infection-related death in allo-HCT recipients by including donor telomere length in the selection criteria when selecting donors. Obviously, this hypothesis cannot be tested in a study of the general population, so further studies on donor telomere length and the risk of serious infections and infection-related death in allo-HCT recipients are needed. Among the strengths of the current study is the prospective general population design, the large number of individuals studied, and the availability of detailed information on 1464

possible confounders such as smoking, alcohol intake, body mass index, hospital diagnosed comorbidities, and measurements of C-reactive protein as a marker of inflammation. Furthermore, as we have performed measurements of white blood cell differential count, platelet count, blood hemoglobin, plasma alanine aminotransferase, plasma creatinine, and non-fasting plasma glucose in a large subgroup of 66,818 individuals, we were also able to minimize confounding by undiagnosed comorbidities by stratifying the analyses on whether individuals had normal or abnormal values of these biomarkers. Due to the observational nature of the study, we are unable to determine with any certainty whether the association between leukocyte telomere length and risk of infections are due to a causal effect. In an attempt to investigate the question of causality, we examined the association between a genetic disposition to shorter telomere length and risk of infections in 107,693 individuals genotyped for the three SNPs that were reported to be most strongly associated with leukocyte telomere length in a genome wide association study of 26,089 individuals.28 We found no association between a genetic predisposition to shorter telomere length and risk of any infection or any type of specific infections, which in principle indicates that the observational association between shorter telomere length and higher risk of infections is not due to a causal effect. However, these results should be interpreted with caution, as the rather modest influence of the SNPs on telomere length leads to low statistical power despite the large number of individuals genotyped, and this limits our ability to confirm or disprove any hypotheses of causality. Another limitation of our study is that we only have information on hospitalizations for infectious disease while information on cases of less serious infections that are typically treated by general practitioners is lacking. In theory, our findings could be biased if individuals with short telomere length appear frailer than other individuals, which could lead general practitioners to have a lower threshold for admitting these patients for hospital treatment. However, it is unlikely that our findings are caused solely by such a differential referral bias, as shorter telomere length was also associated with higher risk of death related to any infection and pneumonia. In conclusion, we prospectively followed 75,309 individuals from the general population for up to 23 years and found that shorter leukocyte telomere length was associated with higher risk of any infection and pneumonia. These findings indicate that leukocyte telomere length may be a marker of immune competence among individuals from the general population. Further studies are needed to determine whether risk of infections in allo-HCT recipients can be reduced by considering donor leukocyte telomere length when selecting donors. Acknowledgments We are indebted to laboratory technician Anja Jochumsen for her valuable assistance with the telomere length measurements. We thank the participants and staff of the Copenhagen City Heart Study and the Copenhagen General Population Study for their important contributions. Funding This work was supported by the Danish Council for Independent Research, the research foundation for health research of the Capital Region of Denmark and by Herlev and Gentofte Hospital, Copenhagen University Hospital. haematologica | 2017; 102(8)

Leukocyte telomere length and risk of infections

References 1. Bojesen SE. Telomeres and human health. J Intern Med. 2013;274(5):399-413. 2. Olovnikov AM. A theory of marginotomy. The incomplete copying of template margin in enzymic synthesis of polynucleotides and biological significance of the phenomenon. J Theor Biol. 1973;41(1):181-190. 3. Hills M, Lucke K, Chavez EA, Eaves CJ, Lansdorp PM. Probing the mitotic history and developmental stage of hematopoietic cells using single telomere length analysis (STELA). Blood. 2009;113(23):5765-5775. 4. Wang J, Lu X, Sakk V, Klein CA, Rudolph KL. Senescence and apoptosis block hematopoietic activation of quiescent hematopoietic stem cells with short telomeres. Blood. 2014;124(22):3237-3240. 5. Hastie ND, Dempster M, Dunlop MG, Thompson AM, Green DK, Allshire RC. Telomere reduction in human colorectal carcinoma and with ageing. Nature. 1990;346(6287):866-868. 6. Frenck RW Jr, Blackburn EH, Shannon KM. The rate of telomere sequence loss in human leukocytes varies with age. Proc Natl Acad Sci USA. 1998;95(10):5607-5610. 7. Weiskopf D, Weinberger B, GrubeckLoebenstein B. The aging of the immune system. Transpl Int. 2009;22(11):10411050. 8. Weng NP. Telomeres and immune competency. Curr Opin Immunol. 2012;24(4):470475. 9. Gadalla SM, Wang T, Haagenson M, et al. Association between donor leukocyte telomere length and survival after unrelated allogeneic hematopoietic cell transplantation for severe aplastic anemia. JAMA. 2015;313(6):594-602. 10. Robin M, Porcher R, De Castro AR, et al. Risk factors for late infections after allogeneic hematopoietic stem cell transplantation from a matched related donor. Biol Blood Marrow Transplant. 2007; 13(11):1304-1312. 11. Wingard JR, Majhail NS, Brazauskas R, et al. Long-term survival and late deaths after allogeneic hematopoietic cell transplantation. J Clin Oncol. 2011;29(16):2230-2239. 12. Cawthon RM, Smith KR, O'Brien E, Sivatchenko A, Kerber RA. Association between telomere length in blood and mortality in people aged 60 years or older. Lancet. 2003;361(9355):393-395. 13. Fitzpatrick AL, Kronmal RA, Kimura M, et al. Leukocyte telomere length and mortality in the Cardiovascular Health Study. J Gerontol A Biol Sci Med Sci. 2011; 66(4):421-429. 14. Martin-Ruiz CM, Gussekloo J, van Heemst D, von Zglinicki T, Westendorp RG. Telomere length in white blood cells is not associated with morbidity or mortality in the oldest old: a population-based study. Aging Cell. 2005; 4(6):287-290. 15. Njajou OT, Hsueh WC, Blackburn EH, et al. Association between telomere length, specific causes of death, and years of healthy life in health, aging, and body composition, a population-based cohort study. J Gerontol A Biol Sci Med Sci. 2009;64(8):860-864. 16. Benfield T, Jensen JS, Nordestgaard BG. Influence of diabetes and hyperglycaemia on infectious disease hospitalisation and outcome. Diabetologia. 2007;50(3):549554. 17. Weischer M, Nordestgaard BG, Cawthon RM, Freiberg JJ, Tybjaerg-Hansen A,

haematologica | 2017; 102(8)

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29. 30. 31.

32.

33.

Bojesen SE. Short telomere length, cancer survival, and cancer risk in 47102 individuals. J Natl Cancer Inst. 2013;105(7):459-468. Zacho J, Benfield T, Tybjaerg-Hansen A, Nordestgaard BG. Increased Baseline CReactive Protein Concentrations Are Associated with Increased Risk of Infections: Results from 2 Large Danish Population Cohorts. Clin Chem. 2016; 62(2):335-342. Rode L, Nordestgaard BG, Bojesen SE. Peripheral blood leukocyte telomere length and mortality among 64,637 individuals from the general population. J Natl Cancer Inst. 2015;107(6):djv074. Pedersen CB, Gotzsche H, Moller JO, Mortensen PB. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull. 2006; 53(4):441-449. Weischer M, Bojesen SE, Cawthon RM, Freiberg JJ, Tybjaerg-Hansen A, Nordestgaard BG. Short telomere length, myocardial infarction, ischemic heart disease, and early death. Arterioscler Thromb Vasc Biol. 2012;32(3):822-829. Almirall J, Bolibar I, Serra-Prat M, et al. New evidence of risk factors for community-acquired pneumonia: a populationbased study. Eur Respir J. 2008;31(6):12741284. Lange P, Vestbo J, Nyboe J. Risk factors for death and hospitalization from pneumonia. A prospective study of a general population. Eur Respir J. 1995;8(10):16941698. Laupland KB, Gregson DB, Church DL, Ross T, Pitout JD. Incidence, risk factors and outcomes of Escherichia coli bloodstream infections in a large Canadian region. Clin Microbiol Infect. 2008; 14(11):1041-1047. Scherczinger CA, Bourke MT, Ladd C, Lee HC. DNA extraction from liquid blood using QIAamp. J Forensic Sci. 1997; 42(5):893-896. Cawthon RM. Telomere length measurement by a novel monochrome multiplex quantitative PCR method. Nucleic Acids Res. 2009;37(3):e21. Savage SA, Bertuch AA. The genetics and clinical manifestations of telomere biology disorders. Genet Med. 2010;12(12):753764. Pooley KA, Bojesen SE, Weischer M, et al. A genome-wide association scan (GWAS) for mean telomere length within the COGS project: identified loci show little association with hormone-related cancer risk. Hum Mol Genet. 2013;22(24):5056-5064. Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health. 2011;39(7 Suppl):30-33. Helweg-Larsen K. The Danish Register of Causes of Death. Scand J Public Health. 2011;39(7 Suppl):26-29. van der Harst P, van der Steege G, de Boer RA, et al. Telomere length of circulating leukocytes is decreased in patients with chronic heart failure. J Am Coll Cardiol. 2007;49(13):1459-1464. Savale L, Chaouat A, Bastuji-Garin S, et al. Shortened telomeres in circulating leukocytes of patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2009;179(7):566-571. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J

Chronic Dis. 1987;40(5):373-383. 34. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):11301139. 35. Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011; 173(6):676-682. 36. Burnett-Hartman AN, Fitzpatrick AL, Kronmal RA, et al. Telomere-associated polymorphisms correlate with cardiovascular disease mortality in Caucasian women: the Cardiovascular Health Study. Mech Ageing Dev. 2012;133(5):275-281. 37. Haycock PC, Burgess S, Nounu A, et al. Association Between Telomere Length and Risk of Cancer and Non-Neoplastic Diseases: A Mendelian Randomization Study. JAMA Oncol. 2017;3(5):636-651. 38. Hunt SC, Kimura M, Hopkins PN, et al. Leukocyte telomere length and coronary artery calcium. Am J Cardiol. 2015; 116(2):214-218. 39. Scheller MA, Rode L, Nordestgaard BG, Bojesen SE. Short Telomere Length and Ischemic Heart Disease: Observational and Genetic Studies in 290 022 Individuals. Clin Chem. 2016;62(8):1140-1149. 40. Najarro K, Nguyen H, Chen G, et al. Telomere Length as an Indicator of the Robustness of B- and T-Cell Response to Influenza in Older Adults. J Infect Dis. 2015;212(8):1261-1269. 41. Boucher N, Dufeu-Duchesne T, Vicaut E, Farge D, Effros RB, Schachter F. CD28 expression in T cell aging and human longevity. Exp Gerontol. 1998;33(3):267282. 42. Chou JP, Effros RB. T cell replicative senescence in human aging. Curr Pharm Des. 2013;19(9):1680-1698. 43. Effros RB. Role of T lymphocyte replicative senescence in vaccine efficacy. Vaccine. 2007;25(4):599-604. 44. Monteiro J, Batliwalla F, Ostrer H, Gregersen PK. Shortened telomeres in clonally expanded CD28-CD8+ T cells imply a replicative history that is distinct from their CD28+CD8+ counterparts. J Immunol. 1996;156(10):3587-3590. 45. Cohen S, Janicki-Deverts D, Turner RB, et al. Association between telomere length and experimentally induced upper respiratory viral infection in healthy adults. JAMA. 2013;309(7):699-705. 46. Furst DE. Serum immunoglobulins and risk of infection: how low can you go? Semin Arthritis Rheum. 2009;39(1):18-29. 47. Woodhead MA, Macfarlane JT, McCracken JS, Rose DH, Finch RG. Prospective study of the aetiology and outcome of pneumonia in the community. Lancet. 1987; 1(8534):671-674. 48. Jokinen C, Heiskanen L, Juvonen H, et al. Microbial etiology of community-acquired pneumonia in the adult population of 4 municipalities in eastern Finland. Clin Infect Dis. 2001;32(8):1141-1154. 49. Verhulst S, Susser E, Factor-Litvak PR, et al. Response to: Reliability and validity of telomere length measurements. Int J Epidemiol. 2016;45(4):1298-1301. 50. Verhulst S, Susser E, Factor-Litvak PR, et al. Commentary: The reliability of telomere length measurements. Int J Epidemiol. 2015;44(5):1683-1686.

1465

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation

The origin of a name that reflects Europe’s cultural roots.

Ancient Greek

aÂma [haima] = blood a·matow [haimatos] = of blood lÒgow [logos]= reasoning

Scientific Latin

haematologicus (adjective) = related to blood

Scientific Latin

haematologica (adjective, plural and neuter, used as a noun) = hematological subjects

Modern English

The oldest hematology journal, publishing the newest research results. 2016 JCR impact factor = 7.702

Haematologica, as the journal of the European Hematology Association (EHA), aims not only to serve the scientific community, but also to promote European cultural identify.