Haematologica, Volume 107, Issue 3 by Haematologica

haematologica Journal of the Ferrata Storti Foundation

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

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

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

Statistical Consultant Catherine Klersy (Pavia)

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

Managing Director Antonio Majocchi (Pavia)

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

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

haematologica Journal of the Ferrata Storti Foundation

Brief information on Haematologica Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org). Haematologica publishes Editorials, Original articles, Review articles, Perspective articles, Editorials, Guideline articles, Letters to the Editor, Case reports & Case series and Comments. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org). Manuscripts should be submitted online at http://www.haematologica.org/. Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail at www.haematologica.org/content/policies. Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties. Subscription. Detailed information about subscriptions is available at www.haematologica.org. Haematologica is an open access journal and access to the online journal is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org). Rates of the printed edition for the year 2022 are as following: Institutional: Euro 700 Personal: Euro 170 Advertisements. Contact the Advertising Manager, Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, e-mail: marketing@haematologica.org). Disclaimer. Whilst every effort is made by the publishers and the editorial board to see that no inaccurate or misleading data, opinion or statement appears in this journal, they wish to make it clear that the data and opinions appearing in the articles or advertisements herein are the responsibility of the contributor or advisor concerned. Accordingly, the publisher, the editorial board and their respective employees, officers and agents accept no liability whatsoever for the consequences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage, and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published literature.

Direttore responsabile: Prof. Carlo Balduini; Autorizzazione del Tribunale di Pavia n. 63 del 5 marzo 1955. Printing: Press Up, zona Via Cassia Km 36, 300 Zona Ind.le Settevene - 01036 Nepi (VT)

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

haematologica Journal of the Ferrata Storti Foundation

Table of Contents Volume 107, Issue 3: March 2022 About the Cover 565

Images from the Haematologica Atlas of Hematologic Cytology: bone marrow metastases from malignant melanoma Rosangela Invernizzi

https://doi.org/10.3324/haematol.2021.280521

Landmark Paper in Hematology 566

Another Philadelphia story* Jerald P. Radich

https://doi.org/10.3324/haematol.2021.280581

Editorials 568

Metchnikoff’s legacy: the dysplastic nature of innate immunity in myelodysplastic syndromes Peter L. Greenberg

https://doi.org/10.3324/haematol.2021.279419

570

Timely diagnosis and treatment of acute promyelocytic leukemia should be available to all Andre C. Schuh

https://doi.org/10.3324/haematol.2021.279052

356

It is time to adapt anti-CD20 administration schedules to allow efficient anti-SARS-CoV-2 vaccination in patients with lymphoid malignancies Caroline Besson

https://doi.org/10.3324/haematol.2021.279457

Articles Platelet Biology & its Disorders 574 Identification of a novel genetic locus associated with immune-mediated thrombotic thrombocytopenic purpura Matthew J. Stubbs et al.

https://doi.org/10.3324/haematol.2020.274639

Acute Myeloid Leukemia 583 Clinical significance of RAS pathway alterations in pediatric acute myeloid leukemia Taeko Kaburagi et al. https://doi.org/10.3324/haematol.2020.269431

Chronic Lymphocytic Leukemia 593 Chromosome banding analysis and genomic microarrays are both useful but not equivalent methods for genomic complexity risk stratification in chronic lymphocytic leukemia patients Silvia Ramos-Campoy et al.

https://doi.org/10.3324/haematol.2020.274456

604

Clonal evolution in chronic lymphocytic leukemia is scant in relapsed but accelerated in refractory cases after chemo(immune) therapy Marc Zapatka et al.

https://doi.org/10.3324/haematol.2020.265777

615

Integrative prognostic models predict long-term survival after immunochemotherapy in chronic lymphocytic leukemia patients Johannes Bloehdorn et al.

https://doi.org/10.3324/haematol.2020.251561

625

Safety and efficacy of the BNT162b mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia Ohad Benjamini et al.

https://doi.org/10.3324/haematol.2021.279196

Haematologica 2022; vol. 107 no. 3 - March 2022 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Complications in Hematology 635 Methotrexate-related central neurotoxicity: clinical characteristics, risk factors and genome-wide association study in children treated for acute lymphoblastic leukemia Marion K. Mateos et al.

https://doi.org/10.3324/haematol.2020.268565

Hematopoiesis 644 Ddx41 inhibition of DNA damage signaling permits erythroid progenitor expansion in zebrafish Joshua T. Weinreb et al.

https://doi.org/10.3324/haematol.2020.257246

655

Reduced frequencies and functional impairment of dendritic cell subsets and non-classical monocytes in myelodysplastic syndromes Nathalie van Leeuwen-Kerkhoff et al.

https://doi.org/10.3324/haematol.2020.268136

Hemostasis 668 Sialylation on O-linked glycans protects von Willebrand factor from macrophage galactose lectin-mediated clearance Soracha E. Ward et al.

https://doi.org/10.3324/haematol.2020.274720

Myelodysplastic Syndromes 680 ZRSR1 co-operates with ZRSR2 in regulating splicing of U12-type introns in murine hematopoietic cells Vikas Madan et al.

https://doi.org/10.3324/haematol.2020.260562

Non-Hodgkin Lymphoma 690 Subtype-specific and co-occurring genetic alterations in B-cell non-Hodgkin lymphoma Man Chun John Ma et al.

https://doi.org/10.3324/haematol.2020.274258

702

Deregulation of JAK2 signaling underlies primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma Armando N. Bastidas Torres et al.

https://doi.org/10.3324/haematol.2020.274506

715

Humoral serological response to the BNT162b2 vaccine is abrogated in lymphoma patients within the first 12 months following treatment with anti-CD2O antibodies Ronit Gurion et al.

https://doi.org/10.3324/haematol.2021.279216

Plasma Cell Disorders 721 The innate sensor ZBP1-IRF3 axis regulates cell proliferation in multiple myeloma Kanagaraju Ponnusamy et al.

https://doi.org/10.3324/haematol.2020.274480

Letters to the Editor 733

Early mortality and survival improvements for adolescents and young adults with acute promyelocytic leukemia in California: an updated analysis Renata Abrahão et al.

https://doi.org/10.3324/haematol.2021.278851

737

Dual pyroptotic biomarkers predict erythroid response in lower-risk non-del(5q) myelodysplastic syndromes treated with lenalidomide and recombinant erythropoietin Chen Wang et al.

https://doi.org/10.3324/haematol.2021.278855

Haematologica 2022; vol. 107 no. 3 - March 2022 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

740

https://doi.org/10.3324/haematol.2021.279037

744

Mechanical unloading aggravates bone destruction and tumor expansion in myeloma Kotaro Tanimoto et al.

https://doi.org/10.3324/haematol.2021.278295

750

GNE-related thrombocytopenia: evidence for a mutational hotspot in the ADP/substrate domain of the GNE bifunctional enzyme Roberta Bottega et al.

https://doi.org/10.3324/haematol.2021.279689

755

https://doi.org/10.3324/haematol.2021.280139

759

Guideline for management of non-Down syndrome neonates with a myeloproliferative disease on behalf of the I-BFM AML Study Group and EWOG-MDS^ Eline J.M. Bertrums et al.

https://doi.org/10.3324/haematol.2021.279507

765

Biallelic CXCR2 loss-of-function mutations define a distinct congenital neutropenia entity Viviana Marin-Esteban et al.

https://doi.org/10.3324/haematol.2021.279254

770

A phase I study of the fully human, fragment crystallizable-engineered, anti-CD-33 monoclonal antibody BI 836858 in patients with previously-treated acute myeloid leukemia Sumithira Vasu et al.

https://doi.org/10.3324/haematol.2020.274118

Errata Corrige 774

Isatuximab plus pomalidomide and dexamethasone in elderly patients with relapsed/refractory multiple myeloma: ICARIA-MM subgroup analysis Fredrik Schjesvold et al.

https://doi.org/10.3324/haematol.2021.279160

Haematologica Reviewers in 2021 776

List of the 1117 reviewers who in 2021 made an essential contribution to the high scientific quality of the journal

Haematologica 2022; vol. 107 no. 3 - March 2022 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

The origin of a name that reflects Europe’s cultural roots.

Ancient Greek

Scientific Latin

Modern English

haematologicus (adjective) = related to blood haematologica (adjective, plural and neuter, used as a noun) = hematological subjects The oldest hematology journal, publishing the newest research results. 2020 JCR impact factor = 9.94

ABOUT THE COVER Images from the Haematologica Atlas of Hematologic Cytology: bone marrow metastases from malignant melanoma Rosangela Invernizzi University of Pavia, Pavia, Italy E-mail: ROSANGELA INVERNIZZI - rosangela.invernizzi@unipv.it doi:10.3324/haematol.2021.280521

n some patients with malignant melanoma, neutrophils and monocytes can phagocytize the melanin deposits released by melanoma cells infiltrating the bone marrow and carry them into circulation, thus suggesting that the bone marrow is affected. In this patient with severe pancytopenia, who, 10 years before, had undergone surgery for cutaneous melanoma, a buffy coat smear shows granules of melanic pigment in the cytoplasm of a neutrophil (Figure A) and a monocyte (Figure B). The bone marrow smear demonstrates the presence of some large cells with round nuclei, prominent nucleoli, and abundant cytoplasm containing many very fine black granules. Melanin is also present in macrophages (Figure C). These findings confirm the diagnosis of a bone marrow relapse. It should be noted, however, that, not infrequently, metastatic malignant melanoma is amelanotic with cells that cannot be distinguished morphologically from other neoplastic cells.1 Disclosures No conflicts of interest to disclose

Reference 1. Invernizzi R. Metastases of solid tumors. Haematologica. 2020;105(Suppl 1):261-269.

haematologica | 2022; 107(3

565

LANDMARK PAPER IN HEMATOLOGY Another Philadelphia story* Jerald P. Radich Fred Hutchinson Cancer Research Center, Seattle, WA; USA E-mail: jradich@fredhutch.org doi:10.3324/haematol.2021.280581

*“The Philadelphia Story” is a classic 1940 American romantic comedy starring a blockbuster lineup of Katharine Hepburn, Cary Grant, and James Stewart. TITLE

A minute chromosome in human chronic granulocytic leukemia.

AUTHORS

Nowell PC, Hungerford DA

JOURNAL

Science 1960;132:1497.

hiladelphia, in the USA, is a city steeped in history. It was in Philadelphia that the Declaration of Independence and the US Constitution were signed. The Philly cheesesteak sandwich was born there, and Benjamin Franklin died there (presumably not from the sandwich). In 1960, Philadelphia was also home of Drs. Peter Nowell, a pathologist at the University of Pennsylvania and David Hungerford, a graduate student at the Institute for Cancer Research. 1960 was an eventful year. In the USA, John Kennedy was elected president, and 3,500 troops were sent to Vietnam, initiating a grueling and divisive 13-year war. The Olympic Games, held in Rome, were televised for the first time. Alfred Hitchcock made the horror classic, Psycho. And tucked in the Journal Science was a modest, two paragraph, 243word abstract that would revolutionize the field of genetics and cancer, entitled “A minute chromosome in human chronic granulocytic leukemia.”1 The piece described a recurrent chromosomal abnormality found in seven cases of chronic granulocytic (now, myelogenous) leukemia. No other chromosome abnormalities were found in the cases, and some cells had a normal karyotype, suggesting that the new chromosome was not constitutional. The abstract con-

cludes with the understated blockbuster, “The findings suggest a causal relationship between the chromosomal abnormality observed and chronic granulocytic leukemia.” Earlier in 1960, Nowell and Hungerford had previously reported on two of these cases in a paper that included cases of acute and chronic leukemia, in which they suggested the minute chromosome might be an altered Y. In the Science abstract, five more cases of chronic myelogenous leukemia (CML) were added and, remarkably, each had the altered chromosome, making this the first demonstration of a consistent chromosomal abnormality found in a specific type of cancer. The rival cytogenetic research group in Edinburgh proposed that the chromosome be dubbed “the Philadelphia chromosome.” In Peter Nowell’s own underplayed words, “the next decade was frustrating.” Further identification of the chromosome awaited the advent of new banding techniques in the 1970s pioneered by Janet Rowley, which demonstrated that the Philadelphia chromosome was actually a reciprocal translocation between chromosomes 22 and 9.2 Jump another decade into the 1980s, and the story is further detailed with the discovery that the translocation fused the BCR region from chromosome 22 with the human homologue of the

Figure 1. The Philadelphia chromosome. The left panel is the first image of the Philadelphia chromosome from Nowell and Hungerford’s first paper.5 This was initially thought to be a Y chromosome. The right panel shows the breakpoints involving the BCR gene from chromosome 22, and the ABL gene from chromosome 9. The bottom right panels show the fusion mRNA breakpoints found in leukemia.

566

haematologica | 2022; 107(3

Landmark Paper in Hematology

recently discovered v-ABL gene.3 A further decade leap into the 1990s and it was demonstrated that BCR-ABL did indeed cause leukemia in a murine model.4 The foundation of “precision medicine” had been laid. CML is perhaps the best example of the old “bench to bedside” adage and is an excellent model of the power of precision medicine since BCR-ABL is a target of treatment, and a biomarker of response. Soon after the invention of the polymerase chain reaction, BCR-ABL mRNA was used to track disease response and predict relapse following allogeneic transplantation. Tyrosine kinase

inhibitors were found to inhibit BCR-ABL in vitro and were soon championed into the clinic. The combination of having a remarkably effective drug and a remarkably robust way of monitoring response completely changed the natural history of CML from a disease associated with a median lifespan of <7 years to one where patients now enjoy essentially a normal lifespan. It is difficult to underestimate the accomplishments of these scientists and patients over a relatively short period of time. In 1776 a revolution officially began in Philadelphia. In 1960, another one started.

References 1. Nowell PC, Hungerford DA. A minute chromosome in human chronic granulocytic leukemia. Science. 1960;142:1497. 2. Rowley JD. A new consistent chromosomal abnormality in chronic myelogenous leukaemia identified by quinacrine fluorescene and Giemsa staining. Nature. 1973;243(5405):290-293. 3. De Klein A, van Kessel AG, Grosveld G, et al. A cellular oncogene is translocated to the Philadelphia chromosome in chronic myelocytic leukaemia. Nature. 1982;300(5894):765-767. 4. Daley GQ, Van Etten RA, Baltimore D. Induction of chronic myelogenous leukemia in mice by the P210 bcr/abl gene of the Philadelphia chromosome. Science. 1990;247(4944):834-830. 5. Nowell PC, Hungerford DA. Chromosome studies on normal and leukemic human leukocytes. J Natl Cancer Inst. 1960;25(1):85-109.

haematologica | 2022; 107(3)

567

EDITORIALS Metchnikoff’s legacy: the dysplastic nature of innate immunity in myelodysplastic syndromes Peter L. Greenberg Stanford Cancer Institute, Stanford, CA, USA E-mail: PETER L. GREENBERG - peterg@stanford.edu doi:10.3324/haematol.2021.279419

ver a century ago, Metchnikoff’s seaside investigations into a eukaryotic organism’s response to induced inflammation led to the beginning of our understanding of the innate immune system.1 The imbalance of the lymphoid/macrophage components of this system in myelodsyplastic syndromes (MDS) generates an adverse immunological mileau for the development of autoimmune disorders in this spectrum of diseases. Given the inherent contribution of myeloid and lymphoid cells to innate immunity, it is not unexpected that the dysregulation of these cells has an impact on chronic myeloid clonal blood disorders, particularly MDS.2 The cytopenias and potential for progression of these disorders are generated predominantly by their immunological abnormalities, inflammatory bone marrow microenvironment, hematopoietic stem cell mutation status and vulnerability to inhibitory cytokines. Multiple epidemiological and clinical studies have demonstrated an increased incidence (10-30%) of autoimmune and inflammatory disorders in association with MDS,3-5 ranging from limited hematologic manifestations, such as autoimmune hemolytic anemia and immune thrombocytopenic purpura, to systemic diseases affecting multiple organs, including vasculitis, connective tissue diseases, inflammatory arthritis and neutrophilic diseases.3-5 Some of these disorders may be associated with adverse outcomes (e.g., vasculitis) or progression of the MDS. Conversely, patients with autoimmune disorders are more likely to develop MDS than are members of the general population.6 Studies evaluating the deranged biological processes involving innate immunity which underlie the meshing of these neoplastic and autoimmune/inflammatory diseases have provided important insights into the pathogenesis of their co-occurence. Regulatory T cells (T ) play a critical role in controlling inflammation and autoimmune disorders7 and are present at a high frequency in the bone marrow. In lowerrisk MDS, the number of T was shown to be decreased, thereby potentially permitting the emergence of autoimmune responses, including those directed against the dysplastic clone.8 In addition, it was separately demonstrated that there are interleukin (IL)-17 producing T cells and elevated serum levels of the pro-inflammatory cytokines IL-7, IL12, RANTES and interferon-γ in lower-risk MDS. In chronic myelomonocytic leukemia (CMML) and some MDS patients, monocytes demonstrate a strikingly abnormal functional imbalance, comprised of >90% classical type monocytes,9 which, upon pathogen stimulation, produce high levels of a broad range of cytokines, including granulocyte colony-stimulating factor, IL-10, CCL2, IL-6 and S100 inflammatory proteins. The latter proteins are generated in response to activation of pyroptosis, an inflammasomemediated process of cell death in myeloid clonal disorders.2 Increased responsiveness of neoplastic CMML hematopoietic precursor cells to microenvironmental inflammatory cytokines, such as granulocyte-macrophage colony-stimulating factor, has also been demonstrated. reg

reg

568

The size of a cell population is orchestrated by apoptosis, an ordered form of programmed cell death, which occurs variably during different stages of a disease trajectory. However, in the clinical setting in which inflammatory disorders occur, pyroptosis generated by activated inflammasomes also contributes substantially to cell death in MDS.2 Inflammasomes are a class of intracellular poly-protein complexes primarily composed of a sensor, an adaptor protein and an effector.10,11 The nucleotide-binding domain-(NOD) like receptor NLRP3, is a redox-sensitive cytosolic sensor that recruits the ASC (apoptosis-associated speck-like protein containing a caspase-recruitment domain) adaptor protein. NEK7, a member of the NIMA-related kinase (NEK) family, is implicated in the control of inflammasome effector function.12 In response to diverse pathogenic stimuli that trigger a cascade of downstream reactions, disordered cellular homeostasis, including mitochondrial dysfunction and toll receptor signaling via reactive oxygen species (ROS) are signals that regulate NEK7-mediated NLRP3 inflammasome activation.1012 This interaction in turn causes polymerization of ASC into large cytoplasmic aggregates referred to as ASC specks, permitting docking and activation of caspase-1, which produces mature IL-1β and IL-18 (interferon-γ inducing factor) proinflammatory cytokines that are secreted into the extracellular space as inflammatory effectors of pyroptosis. In this issue of Haematologica, Wang et al.13 describe the presence in plasma of a marker (ASC specks) of pyroptotic cell death generated by activation of the inflammasome within MDS bone marrow cells from patients treated with recombinant erythropoietin and lenalidomide. The authors used confocal and electron microscopy to visualize and flow cytometry to quantify these specks, which are released upon cytolysis and circulate in peripheral blood for extended periods because of their inherent resistance to degradation. They provide data suggesting the potential utility of such measurements to define inflammasome activation, identified by this pyroptotic biomarker (ASC specks) and suggest that this feature, along with assessment of serum erythropoietin levels, may represent a method to detect lower-risk MDS patients whose anemia could benefit from treatment with lenalidomide and erythropoietin. These findings provide a potentially useful approach to clinical assessment of inflammation. However, they require further confirmation, especially regarding their specificity and sensitivity for MDS patients’ responsiveness to therapy. There is a genetic basis for the inflammatory phenomena contributing to some of the clinical conditions associated with MDS. Both germline and somatic mutations have been associated with myeloid-associated inflammatory diseases, including Schwachman-Diamond syndrome, an autosomal recessive inherited disease with bone marrow failure and inflammatory symptoms, and the myeloid-restricted cryopyrin-associated periodic syndrome (CAPS), an autoinflammatory disease related to mutations in the NLRP3 gene.14 Polymoprhisms in this gene may play a role in the variable

haematologica | 2022; 107(3

Editorials

inflammatory clinical features in MDS patients. TET2 and splice gene mutations, common in MDS, contribute to inflammatory gene expression in macrophages and are associated with cardiovascular inflammatory comorbidities. Acute leukemic transformation is more frequent in MDS patients with autoinflammatory features than in those without. Recently, a clinically severe autoinflammatory disease associated with MDS and other myeloid disorders termed VEXAS syndrome (characterized by Vacuoles in myeloid precursors, E1-ubiquitinating enzyme abnormal function, X-linked, Autoinflammatory disorders, Somatic mutation) has been ascribed to a somatic mutation in the UBA1 gene.15 This disorder has escaped much prior clinical attention since the gene is not captured by most current next-generation sequencing mutation panels. Treatment of MDS patients wih disordered immunological and inflammatory components has been problematic. For certain associated diseases, such as CAPS and Schnitzler syndrome with NLRP3 activation, IL-1 and IL1 receptor antagonists have been beneficial in disease management and are now being considered for MDS.11,12,16 Although some MDS patients with both disease elements may respond to therapy with hypomethylating agents or to antagonists of IL-1 or IL-6 or their respective receptors, these drugs appear to have only temporizing effects in this disease setting; nevertheless, they may be steroidsparing as an aid to symptom management.17 Other molecular targets have been evaluated for the treatment of such patients, including inhibition of the Toll receptor or Bruton tyrosine kinase signaling.2 The more recently discovered NEK7 component of NLRP3 activation may provide a novel target for inhibitors of the inflammasome’s upstream effector arm.12 In addition, given the important role of T in controlling inflammation and of their deficiency in lower-risk MDS patients, consideration of T usage as a feature of cellular therapeutic approaches for such patients may prove valuable in this neoplastic disease with disordered innate immunity.18 Thus, the paper by Wang et al.13 heralds methods to improve understanding of pathogenic mechanisms underlying critical interactions between inflammation and myeloid neoplasia. Such advances should facilitate the development of more effective approaches to the treatment of the dysplastic innate immunity involved in the hemato-inflammatory nature of MDS. reg

reg

haematologica | 2022; 107(3)

Disclosures No conflicts of interest to disclose.

References 1. Gordon S. Elie Metchnikoff: father of natural immunity. Eur J Immunol. 2008;38(12):3257-3264. 2. Sallman D, List A. The central role of inflammatory signaling in the pathogenesis of myelodysplastic syndromes. Blood. 2019;133(10): 1039-1048. 3. Mekinian A, Grignano E, Braun T, et al. Systemic inflammatory and autoimmune manifestations associated with myelodysplastic syndromes and chronic myelomonocytic leukaemia: a French multicentre retrospective study. Rheumatology (Oxford). 2016;55(2):291-300. 4. Enright H, Jacob HS, Vercellotti G. Paraneoplastic autoimmune phenomena in patients with myelodysplastic syndromes: response to immunosuppressive therapy. Br J Haematol. 1995;91(2):403-408. 5. De Hollanda A, Beucher A, Henrion D, et al. Systemic and immune manifestations in myelodysplasia: a multicenter retrospective study. Arthritis Care Res (Hoboken). 2011;63(8):1188-1194. 6. Ertz-Archambault N, Kosiorek H, Taylor GE, et al. Association of therapy for autoimmune disease with myelodysplastic syndromes and acute myeloid leukemia. JAMA Oncol. 2017;3:936-943. 7. Plitas G, Rudensky A. Regulatory T cells in cancer. Ann Rev Cancer Biol. 2020;4:459-477. 8. Kordasti SY, Ingram W, Hayden J, et al. CD4+CD25high Foxp3+ regulatory T cells in myelodysplastic syndrome. Blood. 2007;110(3): 847-850. 9. Selimoglu-Buet D, Wagner-Ballon O, Saada V, et al. Characteristic repartition of monocyte subsets as a diagnostic signature of chronic myelomonocytic leukemia. Blood. 2015;125(23):3618-3626. 10. Basiorka AA, McGraw KL, Eksioglu EA, et al. The NLRP3 inflammasome functions as a driver of the myelodysplastic syndrome phenotype. Blood. 2016;128(25):2960-2975. 11. Shen HH, Yang YX, Meng X, et al. NLRP3: a promising therapeutic target for autoimmune diseases. Autoimmun Rev. 2018;17:694-702. 12. Liu G, Chen X, Wang, Q , Yuan L. NEK7: a potential therapy target for NLRP3-related diseases. BioScience Trends. 2020;14(2):74-82. 13. Wang C, McGraw KL, McLemore AF, et al. Dual pyroptotic biomarkers predict erythroid response in lower risk non-del(5q) myelodysplastic syndromes treated with lenalidomide and recombinant erythropoietin. Haematologica. 2022;107(3):737-739. 14. Nigrovic PA, Lee PY, Hoffman HM. Monogenic autoinflammatory disorders: conceptual overview, phenotype, and clinical approach. J Allergy Clin Immunol. 2020;146(5):925-937. 15. Beck DB, Ferrada MA, Sikora KA, et al. Somatic mutations in UBA1 and severe adult-onset autoinflammatory disease. N Eng J Med. 2020;383(27):2628-2638. 16. Dinarello, CA A, van de Meer J. Treating inflammation by blocking interleukin-1 in a broad spectrum of diseases. Nat Rev Drug Discov. 2012;11(8):633-652. 17. Fraison JB, Mekinian A, Grignano E, et al. Efficacy of azacitidine in autoimmune and inflammatory disorders associated with myelodysplastic syndromes and chronic myelomonocytic leukemia. Leuk Res. 2016;43:13-17. 18. Sharabi A, Tsokos M, Ding Y, et al. Regulatory T cells in the treatment of disease. Nat Rev Drug Discov. 2018;17(11): 823-844.

569

Editorials

Timely diagnosis and treatment of acute promyelocytic leukemia should be available to all Andre C. Schuh University Health Network/Princess Margaret Cancer Centre, and University of Toronto, Toronto, Ontario, Canada E-mail: ANDRE C. SCHUH - andre.schuh@uhn.ca doi:10.3324/haematol.2021.279052

n the first report of acute promyelocytic leukemia (APL) as a discrete entity in 1957, Hillestad1 characterized it as having “a very rapid fatal course of only a few weeks’ duration”, largely due to a severe bleeding tendency. He concluded that the disease “seems to be the most malignant form of acute leukemia”. In the decades since that time, the treatment of APL has improved progressively with the sequential introduction of acute myeloid leukemia-type chemotherapy, all-trans retinoic acid (ATRA), and arsenic trioxide, such that APL is now considered one of the great successes in leukemia treatment. With the use of regimens containing ATRA plus arsenic trioxide, long-term overall survival rates of at least 95% are achievable in both low-/intermediate-2 and highrisk3 APL. In the face of such results, it is sobering to remember, however, that not all patients with APL fare this well. Indeed, APL remains extremely deadly up front, with such outstanding outcomes restricted largely to those patients in whom the diagnosis is actually suspected, treatment with ATRA is started immediately at the first suspicion of APL, coagulation abnormalities are corrected aggressively, and referral/transfer to a leukemia-treating center is initiated promptly, together with timely prevention/management of APL differentiation syndrome. In the absence of such timely diagnosis and treatment, outcomes remain poor. More than 60 years after its initial description, early death remains the major problem in APL management. Over the last 10 years, several groups, both in Europe and in North America, have confirmed that APL early death rates are higher in real life than would be suggested by clinical trial outcomes, have identified factors and management gaps potentially contributing to early death, and have suggested possible interventions to improve outcomes.4-12 Such studies have been heterogeneous in design and execution, and have included populationbased registry analyses, as well as single-centre and multiple-centre retrospective reviews, from multiple jurisdictions, so inter-study comparisons have been difficult. While there are some discrepancies among studies, triggering attempts to explain these differences,4,5 common themes include reported early death rates ranging from 930%, with most early deaths occurring within the first 710 days, and most commonly related to hemorrhage (commonly intracerebral). In addition, some (but not all) studies have suggested that early deaths increased with delay in APL diagnosis, delay in ATRA initiation, delay in hospital admission, older age of the patients, and admission to a non-teaching hospital. Consistent with the last point, when considered together with the rarity of APL, physicians’ awareness and experience have also been suggested as gaps. While delay in starting ATRA treatment was identified by some groups as a key factor in early

570

deaths, and 30-day mortality was shown to decrease significantly from the pre-ATRA to the post-ATRA era,7 two population-based studies did not observe decreasing early death rates over time, contrary to what would have been expected with increased ATRA usage over the same time period.5,6 Early death in APL patients is clearly multifactorial in origin. Differences among study outcomes remain poorly explained, but are presumably related to patient, institutional, and jurisdictional differences, among others, including, speculatively, interrelated differences in healthcare availability and socio-economic status. As a result of these reports, strategies to reduce early deaths have been introduced in many jurisdictions. The overarching goal of such initiatives has been medical provider education and mentorship, with a view towards better disease awareness and diagnosis, the early administration of ATRA at the first suspicion of APL, aggressive blood product support, early consultation and transfer (or co-management) as appropriate, and timely prevention and management of APL differentiation syndrome.13-15 At our institution, in addition to the longstanding 24/7 access to an acute leukemia physician, a policy of recurrent community APL teaching and prompt APL transfer, and an “ATRA Program” (whereby all emergency rooms in our acute leukemia catchment area have ATRA on hand), has been in place for almost 10 years. While the outcomes of such programs are difficult to assess in the short term, early reports are promising.14 In the current issue of Haematologica, Abrahão et al.15 help to clarify some of these issues. The authors previously analyzed early deaths in adolescents and young adults with APL in California,12 and showed that among patients aged ≤39 years diagnosed with APL, 30-day mortality decreased from 26% pre-ATRA (1988-1995) to 14% post-ATRA (2004-2011). In contrast, however, 7-day mortality did not differ between pre- and post-ATRA eras. Notably, a higher risk of 30-day mortality and inferior overall survival were observed among patients without health insurance and those of Black and Hispanic race/ethnicity.12 The authors concluded that efforts to achieve equal outcomes in young patients with APL should focus on improving access to effective treatment, mainly among these underserved groups. The authors’ current report underscores this chilling recommendation. The authors now report an update of early APL outcomes in the California group of adolescents and young adults, dividing patients into subgroups based on health insurance status. Patients could be privately insured, could be enrolled under Medicaid (a program that covers healthcare costs for non-elderly people with low incomes), or could be uninsured. With respect to Medicaid coverage, patients were divided into three timebased groups (pre-, early-, and full-) based on the introduction and adoption of the Affordable Care Act (ACA;

haematologica | 2022; 107(3)

Editorials

also known as ‘Obamacare’), which expanded Medicaid eligibility. The authors report reduced early mortality in post-ACA versus pre-ACA patients, and also that location of care played a key role: Patients diagnosed/treated at National Cancer Institute-Designated Cancer Centers had lower mortality rates and better overall survival than did patients treated elsewhere. This report is important in several areas that help to clarify previous observations (and the discordance among observations). First, it underscores that in APL, healthcare access is at least as important as are other factors influencing early deaths. Indeed, in the absence of timely access, none of the other factors - early diagnosis, immediate ATRA availability, vigorous correction of coagulopathy, etc. - really matters. As a Canadian enjoying universal healthcare access, I am moved by this realization. Second, treaters’ experience and expertise with APL are important. This study confirms previous suggestions regarding better APL outcomes at ‘teaching’ or ‘university’ hospitals. And third, the improved outcomes over time in this report suggest that efforts to eradicate early deaths have been effective, at least in part, and provide hope that with further efforts, we can solve the problem of early deaths in patients with APL. In the right circumstances, APL is a curable disease. Potential cure is an opportunity that should be available to all afflicted individuals. Disclosures No conflicts of interest to disclose.

References 1. Hillestad LK. Acute promyelocytic leukemia. Acta Med Scand. 1957;159:189-194. 2. Platzbecker U, Avvisati G, Cicconi L, et al. Improved outcomes with retinoic acid and arsenic trioxide compared with retinoic acid and chemotherapy in non-high-risk acute promyelocytic leukemia: final

haematologica | 2022; 107(3)

results of the randomized Italian-German APL0406 trial. J Cin Oncol. 2017;35(6):605-612. 3. Iland HJ, Collins M, Bradstock K, et al. Use of arsenic trioxide in remission induction and consolidation therapy for acute promyelocytic leukaemia in the Australasian Leukaemia and Lymphoma Group (ALLG) APML4 study: a non-randomised phase 2 trial. Lancet Hematol. 2015;2(9):e357-366. 4. Lehmann S, Deneberg S, Antunovic P, et al. Early death rates remain high in high-risk APL: update from the Swedish Acute Leukemia Registry 1997-2013. Leukemia. 2017;31(6):1457-1459. 5. Park JH, Qiao B, Panageas KS, et al. Early death rate in acute promyelocytic leukemia remains high despite all-trans retinoic acid. Blood. 2011;118(5):1248-1254. 6. McClellan JS, Kohrt HE, Coutre S, et al. Treatment advances have not improved the early death rate in acute promyelocytic leukemia. Haematologica. 2012;97(1):133-136. 7. Altman JK, Rademaker A, Cull E, et al. Administration of ATRA to newly diagnosed patients with acute promyelocytic leukemia is delayed contributing to early hemorrhagic death. Leuk Res. 2013;37(9):1004-1009. 8. Breccia M, Latagliata R, Cannella L, Minotti C, Meloni G, Lo-Coco F. Early hemorrhagic death before starting therapy in acute promyelocytic leukemia: association with high WBC count, late diagnosis and delayed treatment initiation. Haematologica. 2010;95(5):853-854. 9. Rahmé R, Thomas X, Recher C, et al. Early death in acute promyelocytic leukemia (APL) in French centers: a multicenter study in 399 patients. Leukemia. 2014;28(12):2422-2424. 10. Rashidi A, Goudar RK, Sayedian F, et al. All-trans retinoic acid and early mortality in acute promyelocytic leukemia. Leuk Res. 2013;37(10):1391-1392. 11. Rashidi A, Riley M, Goldin TA, et al. Delay in the administration of all-trans retinoic acid and its effects on early mortality in acute promyelocytic leukemia: final results of a multicentric study in the United States. Leuk Res. 2014;38(9):1036-1040. 12. Abrahão R, Ribeiro RC, Medeiros BC, Keogh RH, Keegan TH. Cancer disparities in early death and survival in children, adolescents, and young adults with acute promyelocytic leukemia in California. Cancer. 2015;121(22):3990-3997. 13. Tallman MS, Manji GA. Don't just stand there, do something: strategies for the prevention of early death in acute promyelocytic leukemia: a commentary. Blood Cells Mol Dis. 2011;46(2):173-174. 14. Jillella AP, Arellano ML, Gaddh M, et al. Comanagement strategy between academic institutions and community practices to reduce induction mortality in acute promyelocytic leukemia. JCO Oncol Pract. 2021;17(4):e497-e505. 15. Abrahão R, Ribeiro RC, Malogolowkin MH, Wun T, Keegan THM. Early mortality and survival improvements for adolescents and young adults with acute promyelocytic leukemia in California: an updated analysis. Haematologica. 2022;107(3):733-736.

571

Editorials

It is time to adapt anti-CD20 administration schedules to allow efficient anti-SARS-CoV-2 vaccination in patients with lymphoid malignancies Caroline Besson Service d’Hematologie Oncologie, Centre Hospitalier de Versailles, Le Chesnay, 78150, France; UVSQ, Inserm, CESP, 94805, Villejuif, France E-mail: CAROLINE BESSON - cbesson@ch-versailles.fr doi:10.3324/haematol.2021.279457

atients with comorbidities are especially sensitive to coronavirus disease 2019 (COVID-19). This is notably true for patients with cancer, including patients with a recent (<5 years) diagnosis of a hematologic malignancy who have a ≥2.5-fold increased risk of death from COVID19.1 Patients with non-Hodgkin lymphoma (NHL) and chronic lymphocytic leukemia (CLL) share immune-system deficiencies due to the biological features of NHL/CLL per se (hypogammaglobulinemia, frequent neutropenia, lymphopenia or lymphocytic dysfunction) and to their treatments (chemotherapy, anti-CD20 monoclonal antibodies [anti-CD20], Bruton tyrosine kinase [BTK] inhibitors or BCL2 inhibitors), leading to an increased incidence and severity of infections. NHL/CLL patients are more likely to develop severe2 and/or prolonged forms of COVID-19.3 The COVID-19-related mortality of NHL patients was shown to increase with age, relapsed/refractory disease and administration of anti-CD20 therapy within 1 year3 while it was shown to be related to age, comorbidities but not with therapy (mainly BTK inhibitors) among CLL patients.4 Therefore, these populations need to be particularly protected against COVID19. Vaccination against severe acquired respiratory syndrome coronavirus 2 (SARS-CoV-2) was shown to prevent COVID-19 in the general population. The efficacy of vaccination in the NHL/CLL population requires further evaluation as immunocompromised patients were excluded from initial studies of SARS-CoV-2 mRNA vaccines. Only limited data on the efficacy of vaccination in these populations have been published. Herishanu et al.5 studied 167 CLL patients from a single center and reported that their antibody response to the BNT162b2 mRNA COVID-19 vaccine was affected by disease activity and by treatments. It decreased from 55.2% in treatmentnaive patients to 16.0% in patients on treatment at the time of vaccination. Remarkably, none of the 22 patients exposed to anti-CD20 within less than 12 months before vaccination had an antibody response.5 This raises particular concerns about these drugs. In this issue of Haematologica, Benjamini et al. report on a larger multicenter series of 373 CLL patients, followed in nine Israeli medical centers, who received two doses of BNT162b2 mRNA COVID-19 vaccine.6 Consistently with the study by Herishani et al.,5 61% of the treatmentnaive patients had a serological response to vaccine, compared to 23% and 24% among patients on BTK inhibitors or BCL2 inhibitors, and only 5% among patients who received anti-CD20 antibodies during the year before vaccination. Deepening the analysis to clinical and biological factors, they demonstrate that age <70 years, normal IgM (≥40 mg/dL), IgA (≥80 mg/dL) and IgG (≥700 mg/dL) levels, normal hemoglobin level (≥13.5 g/dL for

572

males or ≥12 g/dL for females) are associated with an antibody response. This allowed the construction of a specific score that predicted response to vaccination. In the same issue, Gurion et al., analyze the antibody response after vaccination with two doses of BNT162b2 mRNA COVID-19 vaccine of 162 patients with lymphoma enrolled in two medical centers in Israel.7 Positive serological responses were observed in 51% of the patients. In a multivariate analysis, active lymphoma and administration of anti-CD20 treatment within 1 year before the second dose of vaccine were identified as negative predictors for antibody response. Interestingly, the rate of seropositivity increased according to the time between anti-CD20 administration and vaccination, from 3% within 45 days, to 22% between 45 days and 1 year and to 80% if the vaccine was given more than 1 year after anti-CD20. Remarkably, the last percentage was equal to that of patients never exposed to anti-CD20. The lack of robust data from large and multicenter cohorts available so far in these high-risk populations renders the present studies of the upmost importance for physicians taking care of NHL/CLL patients worldwide. Two important messages can be drawn from the results reported by these studies. First, patients with NHL/CLL frequently fail to develop an effective humoral response to BNT162b2 vaccine. The striking observation that recent anti-CD20 therapy strongly impairs the development of antibody response after vaccination should be at the forefront of concerns. The second major information is the identification of other risk factors associated with lack of humoral response in this setting. Besides older age, a risk factor for lack of antibody response, which had already been identified in the general population, some NHL/CLL-specific factors also seem to affect serological response, such as active disease, and, among CLL patients only, lower hemoglobin and/or immunoglobulin levels. The usefulness of the CLL score built with these factors needs to be determined in clinical practice. The two studies suffer from significant limitations, mostly related to the short follow-up after vaccination (26 weeks and 2-3 weeks after the second dose of vaccine in patients with NHL and CLL, respectively). With longer follow-up, it will be especially important to obtain data on the occurrence of COVID-19 after vaccination in these cohorts of patients. As B-cell depletion may also affect the generation of both B- and T-cell memory responses,8 the serological data should be supplemented by the exploration of T-cell immune responses. Indeed, T-cell immunity has a major role in generating durable protective immunity after viral vaccination. Recent studies, performed among non-immunocompromised patients, show that two doses of BNT162b1 can elicit solid CD4+ and CD8+ T-cell responses.9 Although the evaluation of T-cell

haematologica | 2022; 107(3)

Editorials

responses is not as robust and reproducible as serological responses, T-cell responses should be evaluated in treatment-naïve NHL/CLL patients as well as among those receiving BTK inhibitors, BCL2 inhibitors, chemotherapy and/or anti-CD20 to establish whether it could provide additional protection. Overall, the findings raise questions about the management of patients with NHL/CLL during this COVID-19 era, for whom there are currently no consensual guidelines. It is time to consider adapting our therapeutic strategies in these patients. First, in any non-critical clinical situation, SARS-CoV-2 vaccination should be proposed before the onset of treatments with BTK inhibitors, BCL2 inhibitors or anti-CD20. Secondly, to prevent prolonged COVID-19 and lack of vaccination efficacy, avoiding or delaying the administration of anti-CD20 may be considered in patients with indolent NHL/CLL with low tumor burden and mild symptoms or cytopenia, for whom delaying the initiation of the treatment will not place the patient at risk. Furthermore, consideration should be given to not re-administering anti-CD20 in patients with NHL in the relapse/refractory setting when other reasonable options are available. Moreover, as already adopted in many centers, avoidance or suspension of maintenance therapy with anti-CD20 in patients with indolent B-cell lymphoma in complete remission to allow their vaccination should also be recommended. This decision should not preclude a patient from receiving the most efficacious treatment strategy and requires consideration of the disease characteristics and the patient's history. Lastly, systematic vaccination of the patients’ relatives and close associates and hospital workers should also benefit the patients directly. Other vaccination strategies should also be explored in these patients such as the effect of a third vaccine dose in nonresponding patients or in those with a low serological response. This approach is currently recommended in some countries, for example France, although its efficacy has not yet been demonstrated. Additional large studies are required to address the question of vaccination in cancer patients, such as that supported by the “COVID-19 and Cancer Global Taskforce”10 and, more specifically, among vaccinated NHL/CLL patients, to specify the level of cellular protection against infection and to determine the risk of clinical COVID-19 and its severity. Meanwhile, individ-

haematologica | 2022; 107(3)

uals with NHL/CLL should receive the COVID-19 vaccine, be informed that they are unlikely to be protected and continue social distancing and adhere to other proven mitigation strategies such as mask wearing. Finally, these findings should contribute to the production of guidelines for the management of NHL/CLL patients during the COVID-19 pandemic, an essential step towards improving the efficiency of vaccination in this setting. Disclosures CB reports research funding from Roche and non-financial support from Takeda and Roche outside the submitted work. Acknowledgments The author thanks Rémy Duléry and Sylvain Lamure for their collaboration on COVID-19 in lymphoma patients.

References 1. Williamson EJ, Walker AJ, Bhaskaran K, et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature. 2020;584 (7821):430-436. 2. Scarfo L, Chatzikonstantinou T, Rigolin GM, et al. COVID-19 severity and mortality in patients with chronic lymphocytic leukemia: a joint study by ERIC, the European Research Initiative on CLL, and CLL Campus. Leukemia. 2020;34(9):2354-2363. 3. Duléry R, Lamure S, Delord M, et al. Prolonged in-hospital stay and higher mortality after Covid-19 among patients with non-Hodgkin lymphoma treated with B-cell depleting immunotherapy. Am J Hematol. 2021;96(8):934-944. 4. Mato AR, Roeker LE, Lamanna N, et al. Outcomes of COVID-19 in patients with CLL: a multicenter international experience. Blood. 2020;136(10):1134-1143. 5. Herishanu Y, Avivi A, Aharon A, et al. Efficacy of the BNT162b2 mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia. Blood. 2021;137(23):3165-3173. 6. Benjamini O, Rokach L, Itchaki G, et al. Safety and efficacy of BNT162b mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia. Haematologica. 2022;107(3):625-634. 7. Gurion R, Rozovski U, Itchaki G, et al. Humoral serological response to the BNT162b2 vaccine is abrogated in lymphoma patients within the first 12 months following treatment with anti-CD2O antibodies. Haematologica. 2022;107(3):715-720. 8. Durali D, de Goër de Herve M-G, Gasnault J, Taoufik Y. B cells and progressive multifocal leukoencephalopathy: search for the missing link. Front Immunol. 2015;6:241. 9. Sahin U, Muik A, Derhovanessian E, et al. COVID-19 vaccine BNT162b1 elicits human antibody and TH1 T cell responses. Nature. 2020;586(7830):594-599. 10. Yusuf A, Sarfati D, Booth CM, et al. COVID-19 and Cancer Global Taskforce. Cancer and COVID-19 vaccines: a complex global picture. Lancet Oncol. 2021;22(6):749-751.

573

ARTICLE Ferrata Storti Foundation

Haematologica 2022 Volume 107(3):574-582

Platelet Biology & its Disorders

Identification of a novel genetic locus associated with immune-mediated thrombotic thrombocytopenic purpura Matthew J. Stubbs,1,2 Paul Coppo,3 Chris Cheshire,2 Agnès Veyradier,4 Stephanie Dufek,2 Adam P. Levine,2 Mari Thomas,1,5 Vaksha Patel,2 John O. Connolly,2 Michael Hubank,6 Ygal Benhamou,3 Lionel Galicier,3 Pascale Poullin,3 Robert Kleta,2# Daniel P. Gale,2# Horia Stanescu2# and Marie A. Scully1,5# 1

Hemostasis Research Unit, ULC, London, UK; 2Department of Renal Medicine, UCL, London, UK; 3Centre de Référence des Microangiopathies Thrombotiques, Hôpital Saint-Antoine, Paris, France; 4Department d’Hematologie, Centre de Référence des Microangiopathies Thrombotiques, Hôpital Lariboisière, Paris, France; 5National Institute for Health Research Cardiometabolic Programme, UCLH/UCL Cardiovascular BRC, London, UK and 6Clinical Genomics, Royal Marsden Hospital, London, UK. #

RK, DPG, HS and MAS contributed equally as co-senior authors.

ABSTRACT

Correspondence: MATTHEW JAMES STUBBS m.stubbs@doctors.org.uk Received: October 21, 2020. Accepted: February 2, 2021. Pre-published: February 18, 2021. https://doi.org/10.3324/haematol.2020.274639

©2022 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.

574

mmune thrombotic thrombocytopenic purpura (iTTP) is an ultra-rare, life-threatening disorder, mediated through severe ADAMTS13 deficiency causing multi-system micro-thrombi formation, and has specific human leukocyte antigen associations. We undertook a large genome-wide association study to investigate additional genetically distinct associations in iTTP. We compared two iTTP patient cohorts with controls, following standardized genome-wide quality control procedures for single-nucleotide polymorphisms and imputed HLA types. Associations were functionally investigated using expression quantitative trait loci (eQTL), and motif binding prediction software. Independent associations consistent with previous findings in iTTP were detected at the HLA locus and in addition a novel association was detected on chromosome 3 (rs9884090, P=5.22x10-10, odds ratio 0.40) in the UK discovery cohort. Meta-analysis, including the French replication cohort, strengthened the associations. The haploblock containing rs9884090 is associated with reduced protein O-glycosyltransferase 1 (POGLUT1) expression (eQTL P<0.05), and functional annotation suggested a potential causative variant (rs71767581). This work implicates POGLUT1 in iTTP pathophysiology and suggests altered post-translational modification of its targets may influence disease susceptibility.

Introduction Thrombotic thrombocytopenic purpura (TTP) is an ultra-rare, life-threatening illness, with an annual incidence of approximately 6/million, and with an untreated mortality approaching 90% (10-20% with prompt intervention). It can affect patients of any age, but often affects young adults (30-40 years) and is more common in women.1 The initial diagnosis of TTP is based on clinical suspicion, but ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13) activity <10 IU/dL confirms the diagnosis. Severe deficiency of ADAMTS13 results in failure to cleave ultra-large von Willebrand Factor multimers (UL-VWF), crucial for normal hemostatic function and proteolytic regulation of VWF. ADAMTS13 deficiency in immune TTP (iTTP) is mediated through immunoglobulin G (IgG) autoantibodies.2,3 The precipitant of the disease in most cases is unclear.4 As with many autoimmune diseases, human leukocyte antigen (HLA) type is associated with the risk of developing iTTP, with HLA-DRB1*11, HLA-DQB1*03 and HLADRB3* increasing risk, and HLA-DRB1*04 and HLA-DRB4 (HLA-DR53) being protective in Europeans.5,6,7 No genetic risk factors outside the HLA genes have previously been shown to be associated with iTTP.

haematologica | 2022; 107(3)

Genome wide association study in Immune TTP

We performed a genome-wide association study (GWAS) in UK and French iTTP cohorts and identified association of alleles both within and beyond the HLA locus.

loblock was performed using ChipSeq data via the UCSC genome browser (https://genome.ucsc.edu). Binding sites of transcription factors (highlighted through genome annotation) were obtained from FactorBook,20 and position weight matrix (PWM) binding motifs generated. Binding motifs were generated using Mast-Meme.21

Methods Cohorts As part of the UK TTP registry, patients were consented for DNA analysis (MREC: 08/H0810/54) (see the Online Supplementary Appendix). Patients on the UK TTP registry were screened for the clinical diagnosis, and confirmed with an ADAMTS13 level <10 IU/dL at diagnosis (utilizing FRETS methodology)8 and the presence of an anti-ADAMTS13 autoantibody.2,3 The French replication cohort TTP samples were obtained from the French Reference Center for TMA (CNRMAT) and informed consent was obtained from each patient with confirmed iTTP (see above criteria) (Institutional Review Board of Pitié Salpêtrière Hospital; clinicaltrials gov. Identifier: NCT00426686). The European control genotypes were obtained from the Wellcome Trust Case Control Consortium (WTCCC), both the 1958 British Birth Cohort and National Blood Service control samples.9 In addition, controls were used from the Illumina reference panel10 and Oxford controls.11,12

Genotyping, quality control and imputation TTP samples were genotyped on the Illumina Human Omni Express single-nucleotide polymorphisms (SNP) chips and controls were genotyped on different SNP chips (see the Online Supplementary Appendix). Pre-imputation quality control was performed in all data sets separately, and then in a combined cohort (Online Supplementary Figure S1). Quality control (QC) was performed for individuals and SNP. Individuals were selected for further analysis by European ancestry principal component analysis (PCA) (see the Online Supplementary Figure S2). Only SNP present in all data sets were subsequently analyzed. Genome-wide imputation was performed on markers that had passed quality control, and were present in all datasets using Beagle (version 5.0) utilizing the 1.000 Genome Project Phase 3 as a reference panel.13 In addition to standardized QC, only SNP with a dosage R2 (DR2) >0.8 were included.

Genome-wide association study and loci characterization GWAS was performed using SNP & Variation Suite v8, using logistic regression with principal component correction.14,15 The logistic regression P-values, odds ratios (OR) were calculated in addition to l inflation factors, and QQ plots are shown (Online Supplementary Figure S3). A standardized genome wide significance level of 5x10-8 was applied.15 For discovery and replication analysis meta-data please contact the authors. Conditional analyses were undertaken using a full versus reduced regression model. Lead SNP at each locus were used as conditional inputs to determine independence, with results plotted using Locus Zoom software.16 Imputation of HLA types was performed utilizing SNP2HLA with previously genotyped markers. 17 Imputed HLA types were excluded if the R2 (confidence) was <0.80. Conditional analyses were subsequently performed as described above. Expression quantitative trait locus (eQTL) analysis was performed to associate identified SNP with differential gene expression.18 Additional markers in linkage disequilibrium with the lead SNP at the chromosome 3 locus were identified by LD-link (https://ldlink.nci.nih.gov).19 Functional annotation of the hap-

haematologica | 2022; 107(3)

Results Discovery cohort Following quality control as outlined in the methods (Online Supplementary Figure S1) there were 241 TTP cases and 3,200 controls in the UK discovery cohort. Following imputation and quality control 3,649,347 SNP were available for analysis. Association testing was performed using a logistic regression model with PCA correction, and the genomic inflation factor (l) was 1.0239 (Online Supplementary Figure S3). In the UK discovery cohort two peaks were identified (Figure 1) (Online Supplementary Figure S4) (lead SNP are summarized in Table 1). The peak with the strongest association corresponded to the class II HLA region on chromosome 6, with 1,017 SNP reaching genome wide significance. The lead SNP rs28383233 located in the intergenic region between HLA-DRB1 and HLA-DQA1 (P=2.20x10-23, OR 3.12, 95% Confidence Interval [CI]: 2.49-3.93) (Table 1; Figure 2). Conditional analysis was performed on rs28383233 and the lead SNP following this was rs1064994 (within HLADQA1), with a P-value of 1.13x10-10 (OR 2.20, 95% CI: 2.06-3.37). Following conditioning on both rs28383233 and rs1064994 no further markers reached significance within the class II HLA region, indicating that there are two detectable independent genetic associations with iTTP within the HLA region. HLA imputation was performed on the UK discovery cohort, and following quality control, 95 imputed HLA alleles remained. HLA-DRB1*11:01 was the allele most strongly associated with iTTP, with a P-value of 3.25x1017 (OR 2.79, 95% CI: 2.23-3.50). Following conditional analysis of HLA-DRB1*1101, no other HLA types reached genome wide significance, but HLA-DQA1*03:01 remained significant (with a HLA-only Bonferroni correction, P<5.26x10-4) at 1.49x10-6 (OR 0.47, 95% CI: 0.330.65) suggesting that the protective effect of this allele is independent of HLA-DRB1*11:01. In addition to the class II HLA peak on chromosome 6, a novel association was observed on chromosome 3. Sixteen markers reached genome wide significance, with the lead SNP, rs9884090(A), having a P-value of 5.22x10-10 (OR 0.40, 95% CI: 0.29-0.56) (Table 1; Figure 3). Upon conditional analysis of the lead SNP no markers reached genome wide significance indicating one detectable signal at this locus. No statistical epistasis was seen between the chromosome 3 and chromosome 6 associations, with each association being independent. Five genes are annotated within this chromosome 3 haploblock: ARHGAP31, TMEM39A, POGLUT1, TIMMDC1, and CD80.

Replication cohort Within the French replication cohort there were 112 cases and 2,603 controls following quality control as outlined in the methods (Online Supplementary Figures S1 and S2). 3,649,546 SNP were available for analysis, and asso575

M.J. Stubbs et al.

Table 1. Lead single nucleotide polymorphisms identified in the UK discovery cohort.

rsID (position)

Minor allele / MAF cases / Major allele MAF controls

rs9884090 (ch3:119116150) rs28383233 (ch6:32584153) rs1064994 (ch6:32611195)

A/G

0.08/0.19

G/A

0.64/0.40

C/T

0.25/0.11

Logistic Odds regression Ratio P-value (95% CI) P = 5.22x10-10

0.40 (0.29-0.56) P = 2.20x10-23 3.12 (2.49-3.93) P = 1.13x10-10 2.20 (2.06-3.37)

Displayed are Minor/Major Alleles, Minor Allele Frequencies (MAF), logistic regression P-value (corrected for principal component analysis stratification), and odds ratio (with 95% Confidence Intervals [CI]). Genomic positions refer to Human Assembly GRCh37/hg19.

ciation testing was performed using a logistic regression model with PCA correction, and l was 1.0830 (Online Supplementary Figure S5). The association with the lead SNP in the chromosome 3 haploblock, rs9884090(A) was replicated with a P-value of 0.001 (OR 0.52), and the two independent lead SNP with the class II HLA peak on chromosome 6 were also replicated (Table 2). The locus zoom plots are shown (Online Supplementary Figures S6 to S8). Imputed HLA type analysis was also consistent with the UK discovery cohort with HLA-DRB1*11:01 and HLA-DQA1*03:01 representing two independent HLA signals. In addition, a meta-analysis was performed combining the UK and French cohorts (cases 241/112, controls 3,200/2,603 respectively), which demonstrated strengthening of the previously observed signal (rs9884090 P=1.60x10-10, OR 0.47, rs28383233 P=1.22x10-42, OR 3.70, rs1064994 P=5.03x10-25, OR 2.89) (Table 3; Online Supplementary Figure S9).

Expression quantitative trait loci and functional DNA analysis eQTL data from the Genotype Tissue Expression Project and Blood eQTL Browser for the lead SNP at the chromosome 3 locus (rs9884090) demonstrated significant reduction in expression of POGLUT1 with the protective allele in the majority of tissues tested, including blood cells (P<0.001).18,22 LD-link identified 20 markers found to be in tight linkage disequilibrium (R2 and D’ >0.80) with rs9884090 contained within the chromosoTable 2. French cohort replication of lead single nucleotide polymorphisms identified in the UK discovery cohort.

rsID (position)

Minor Allele / MAF Cases / Logistic Major Allele MAF Controls Regression P-value

rs9884090 (ch3:119116150) rs28383233 (ch6:32584153) rs1064994 (ch6:32611195)

A/G

0.10/0.18

P = 0.001 -9

G/A

0.68/0.40

P = 3.87x10

C/T

0.42/0.11

P = 5.015x10-9

Odds Ratio (95% CI) 0.52 (0.34-0.81) 2.57 (1.87-3.53) 2.86 (2.06-3.99)

576

mal region (see the Online Supplementary Table S1).19 All markers were functionally annotated with information from the UCSC Genome Browser (Human Assembly GRCh37/hg19)23,24) (see the Online Supplementary Table S1). One variant was particularly noted, rs71767581 (Ch3, 119187422 AC/-del), which is a 2-basepair deletion in the promoter of POGLUT1. This may be functionally important as the haploblock identified is associated with reduced expression in POGLUT1. Upon analysis of ChipSeq data in UCSC Genome Browser 14 transcription factors were predicted to bind at this site (see the Online Supplementary Table S2), adding further evidence that rs71767581 may be functionally important for POGLUT1 expression.

Discussion This GWAS, involving two European populations, is the first to be performed in iTTP and shows consistent evidence of association at loci on chromosome 6 and chromosome 3. The associated alleles on chromosome 6 lie within the HLA region and imputation of HLA types and conditional analyses indicated independent association between HLA-DRB1*11:01 (OR 2.79; P=3.25x10-17) and HLA-DQA1*03:01 (OR 0.47; P=1.49x10-6, post conditional analysis), which are consistent, and in linkage with previously published risk and protective associations with iTTP at this locus.5–7 A recent case-control study comparing frequency of alleles only at immune loci in 190 Italian TTP patients and 1,255 controls identified the HLA variant rs6903608, (in addition to HLA-DQB1*05:03) as conferring a 2.5-fold increase of developing TTP.25 Here we also identified a novel association of iTTP with alleles on chromosome 3 tagged by the lead SNP rs9884090. Five genes are located within the associated haploblock: ARHGAP31, TMEM39A, POGLUT1, TIMMDC1, and CD80. ARHGAP31 (rho GTPase activating protein 31) is associated with the autosomal dominant condition Adams-Oliver Syndrome (OMIM 100300).26 Mutations within ARHGAP31 have been implicated with abnormal vascular development and VEGF (vascular endothelial growth factor) angiogenesis.27 Little is understood regarding the function of TMEM39A (transmembrane protein 39A). While variants have been implicated in autoimmune disease such as systemic lupus erythematosus28,29 and multiple sclerosis,30,31 understanding of Table 3. Meta-analysis combining UK and French Cohorts, showing lead single nucleotide polymorphisms identified in the UK discover cohort.

rsID (position)

Logistic Odds Ratio Minor Allele / MAF Cases / Major Allele MAF Controls Regression (95% CI) P-value

rs9884090 (ch3:119116150) rs28383233 (ch6:32584153) rs1064994 (ch6:32611195)

A/G

0.08/0.19

G/A

0.64/0.41

C/T

0.22/0.11

0.47 (0.36-0.60) P = 1.22x10-42 3.70 (2.81-4.03) P = 5.03x10-25 2.89 (2.39-3.49) P =1.60x10-10

haematologica | 2022; 107(3)

Genome wide association study in Immune TTP

its function is lacking. TIMMDC1 is a membrane embedded mitochondrial complex factor, and is associated with mitochondrial disorders.32 The protein encoded by the CD80 gene functions as a membrane receptor being activated by CTLA-4 or CD28, both of which are T-cell receptors. The downstream mechanisms are T-cell proliferation and cytokine production. CD80 and its receptors have been associated with focal segmental glomerulosclerosis33 and systemic lupus erythematosus.34,35 POGLUT1 (protein O-glucosyltransferase 1) is mutated in DowlingDegos disease-4 (an autosomal dominant genodermatosis with progressive and disfiguring reticulate hyperpigmentation and muscular dystrophy, OMIM 615696) and POGLUT1 has been shown to catalyse O-glycosylation of epidermal growth factor (EGF)-like repeats.36,37 EGFlike repeats are well conserved structures, and highly represented with proteins involved in coagulation.38,39 In vitro work has demonstrated POGLUT1 binds and glycosylates specific coagulation factors including factor VII and factor IX.37,40 The haploblock identified in this analysis of iTTP (which is tagged by rs9884090(A)) is associated with significantly decreased POGLUT1 expression by eQTL.41 Several other genetic variants contained within this haploblock have been associated with other autoimmune diseases, and the majority of these variants have been shown to be in linkage with our lead variant rs9884090 (see the Online Supplementary Appendix), supporting the findings described here.28,29,31,42,43,44 eQTL analysis is a robust tool, that can associate gene expression with spe-

cific genetic variants. Our analysis found rs9884090(A) to have a reduced frequency in iTTP, and rs9884090(A) was shown to be associated with significantly decreased POGLUT1 expression in different eQTL resources.18,22 In order to locate the underlying genetic variant implicated in this reduced POGLUT1 expression we used LD-link to identify additional variants, and located a 2-basepair deletion with the POGLUT1 upstream promoter region that is in tight linkage disequilibrium with the lead associated variant (R2/D’>0.80). As rs9884090(A) confers reduced risk of developing iTTP, we hypothesize that reduced expression of POGLUT1 leads to altered posttranslational modification (O-glycosylation) of key POGLUT1 targets to reduce the risk of iTTP. The evidence we present supports POGLUT1 as the gene of interest, but we cannot exclude other genes within the associated haploblock. The pathway through which POGLUT1’s effects could be mediated remains to be determined. Given there are several reported variants with this haploblock associated with different autoimmune disease, it is likely the downstream functional consequences medicated through POGLUT1 influence immune-regulatory pathways which may generally increase the risk of other autoimmune disease, in addition to iTTP, and may provide insights into potential therapies.45–56 In summary, we have identified a novel genetic variant, rs9884090(A), in two independent populations, which is associated with reduced risk of iTTP. Utilizing linkage disequilibrium we have identified a functional variant in tight LD with the lead SNP in the POGLUT1 promoter

Figure 1. Manhattan plot of genome wide association analysis comparing UK immune thrombotic thrombocytopenic purpura discovery cohort compared with controls. The x-axis shows chromosome location, and the y-axis shows negative logarithmic P-values. Standardized genome wide significant 5x10-8 is depicted by the red line. The human leukocyte antigen peak is visualized on chromosome 6 (black), in addition to the novel chromosome 3 association (orange).

haematologica | 2022; 107(3)

577

M.J. Stubbs et al.

Figure 2. Locus zoom plots of the chromosome 6 peak in the UK discovery cohort. The upper plot (A) shows the unconditioned analysis with the lead singlenucleotide polymorphisms rs28383233, and the middle plot (B) shows analysis conditioned on the lead SNP rs28383233, revealing independent association with rs1064994. The lower plot (C) shows analysis conditioned on both rs28323233 and rs1064994. Genomic positions refer to Human Assembly GRCh37/hg19. chr6: chromosome 6.

578

haematologica | 2022; 107(3)

Genome wide association study in Immune TTP

Figure 3. Locus zoom plots of the chromosome 3 peak in the UK discovery cohort. The upper plot (A) shows the unconditioned analysis and the lower plot (B) shows associations of the same markers when conditioned on the lead single-nucleotide polymorphisms (SNP), rs9884090. Genomic positions refer to Human Assembly GRCh37/hg19. chr3: chromosome 3.

site and eQTL demonstrates reduced POGLUT1 expression associated with this variant. We therefore hypothesize this leads to altered O-glycosylation on POGLUT1 targets. Whilst the exact role of POGLUT1 in the pathophysiology of iTTP requires further downstream functional analysis, this work represents an important step forward in our understanding of iTTP. Disclosures MJS received research funding from Shire/Takeda; PC sits on the advisory board and received symposia fees from Sanofi, Alexion and Roche, received fees from Octapharma; AV sits on the advisory board of Ablynx/Sanofi, Roche-Chugai, and haematologica | 2022; 107(3)

Shire/Takeda, received course fees and awards from LFB Biomédicaments, Octapharma and CSL-Behring; MT sits on the advisory board of Sanofi; YB sits on the advisory board of Sanofi and Octapharma; PP sits on the advisory board of Sanofi; DPG received honoraria and sits on the advisory board of Alexion; MS consults, received honoraria, sits on the advisory board, received speakers fees from Novarits, received honoraria, sits on the advisory board, received research funding and speakers fees from Shire/Takeda, consults for, received honoraria, sits on the advisory board, received speakers fees from Ablynx/Sanofi and Shire/Takeda, received honoraria, sits on the advisory board and speakers bureau of Alexion, received research funding from Baxalta: All other authors have no conflicts of interest to disclose. 579

M.J. Stubbs et al.

Contributions MJS designed research, recruited patients, performed research, collected data, analyzed and interpreted data, wrote the manuscript; PC designed research, recruited patients, analyzed and interpreted data, wrote the manuscript; CC, SD, VP and APL performed research, collected data, analyzed and interpreted data, wrote the manuscript; AV designed research, recruited patients, analyzed and interpreted data, wrote the manuscript; MT designed research, recruited patients, analysed and interpreted data, wrote the manuscript; JOC designed research, wrote the manuscript; MH designed research, wrote the manuscript; YB, LG and PP designed research, recruited patients, wrote the manuscript; RK, DPG, HS and MAS designed research, performed research, analyzed and interpreted data, wrote the manuscript. Acknowledgments We would like to thank Drs Fairfax, Makino, and Knight from the University of Oxford for sharing ethnically matched control data. We would also like to acknowledge the significant contribution of our collaborators within in UK TTP Registry and the French Reference Center for Thrombotic Microangiopathies, who are listed below: The members of the UK TTP Registry are: Aberdeen Royal Infirmary (H. Watson, M. Greiss, J. Dixon, S. Rodwell, M. Fletcher), Addenbrookes Hospital (W. Thomas, S. MacDonald), Bath Royal United Hospitals (S. Moore, C. Cox), Birmingham University Hospitals (W. Lester, G. Lowe, C. Percy, E. Dwenger, M. Pope), Birmingham Children’s Hospital (J. Motwani), Bournemouth and Christchurch Hospitals, (S. Killick, M. Serrano), Bristol University Hospitals (A. Clark, A. Mumford, L. Humphrey, S. Mulligan), Cardiff & Vale University Hospitals (R. Rayment, P. Collins, M. Norton, A. Guerrero, S. Cunningham), Cornwall Hospitals NHS Trust (D. Beech, S. Hunter, B. Mills), Coventry & Warwickshire University Hospitals (O. Chapman, B. Bailiff, A. Pearson, D. Morris), Edinburgh Royal Infirmary (L. Manson, N. Priddee), Glasgow Royal Infirmary (K. Douglas, C. Tait, C. Bagot), Glasgow Royal Hospital for Children (E. Chalmers), Great Ormand Street Hospital (R. Liesner, K. Sibson, A. Taylor, A. Griffien), Guy's & St Thomas' NHS Trust (B. Hunt, J. Young), Imperial College NHS Trust (N. Cooper, A. Luqmani, C. Vladescu, D. Paul), Leeds Teaching Hospital NHS Trust (Q. Hill), Leicester Hospitals (R. Gooding, K. Siguake), Royal Liverpool Hospital (T. Dutt, C. Powell), Manchester Children’s Hospital (J. Thachil, J. Granger, S. Boydell), Newcastle Hospitals (T. Biss, J. Wallis, J. Hanley, K. Talks, A. Charlton), Norfolk and Norwich University Hospitals NHS Foundation Trust (H. Lyall, E. Malone, M. Sheridan), Nottingham University Hospitals NHS Trust (G. Swallow, J. Hermans), Oxford University Hospital NHS Trust (S. Benjamin, C. Deane, A. Eordogh, K. Santos), Oxford Children’s Hospital (N. Bhatnager, S. Pavord, G. Hall, P. Baker), Plymouth University Hospitals NHS Trust (T. Nokes), Poole Hospital NHS Foundation Trust (F. Jack, N. Beamish, A. Wandowski), Portsmouth Hospitals NHS Trust (T. Cranfield, C. James, S. Liu), Royal Devon & Exeter NHS Foundation Trust (J. Coppell, L. Ngu), Sheffield Teaching Hospital NHS Foundation Trust (J. Vanveen, M. Makris, R. MacLean, K. Harrington, S. Megson, R. Fretwell), South Tees Hospitals NHS Foundation Trust (R. Dang, M. David, J. Maddox, D. Winterburn), South Warwickshire NHS Foundation Trust (P. Rose), St George’s University Hosptials NHS Foundation Trust (S. Austin, J. Uprichard, J. Chackathayil, A. Lee), University College London Hospitals (M. Scully, J.P. Westwood, M. Thomas, R. Newton, 580

S. McGuckin, I. Obu, C. Vendramin, L. Keogh, J. Shin, M. Stubbs), NHSBT (K. Pendry, K. Slevin) and Southampton University Hospital NHS Foundation Trust (S. Boyce). The members of the French Reference Center for Thrombotic Microangiopathies (CNR-MAT) are: Augusto Jean-François (Service de Néphrologie, Dialyse et Transplantation; CHU Larrey, Angers); Azoulay Elie (Service de Réanimation Médicale, Hôpital Saint-Louis, Paris); Barbay Virginie (Laboratoire d’Hématologie, CHU Charles Nicolle, Rouen); Benhamou Ygal (Service de Médecine Interne, CHU Charles Nicolle, Rouen); Bordessoule Dominique (Service d’Hématologie, Hôpital Dupuytren, Limoges); Charasse Christophe (Service de Néphrologie, Centre Hospitalier de Saint-Brieuc); Charvet-Rumpler Anne (Service d’Hématologie, CHU de Dijon); Chauveau Dominique (Service de Néphrologie et Immunologie Clinique, CHU Rangueil, Toulouse); Choukroun Gabriel (Service de Néphrologie, Hôpital Sud, Amiens); Coindre Jean-Philippe (Service de Néphrologie, CH Le Mans); Coppo Paul (Service d’Hématologie, Hôpital SaintAntoine, Paris); Corre Elise (Service d’Hématologie, Hôpital Saint-Antoine, Paris); Delmas Yahsou (Service de Néphrologie, CHU de Bordeaux, Bordeaux); Deschenes Georges (Service de Néphrologie Pédiatrique, Hôpital Robert Debré, Paris); Devidas Alain (Service d’Hématologie, Hôpital Sud-Francilien, CorbeilEssonnes); Dossier Antoine (Service de Néphrologie, Hôpital Bichat, Paris); Fain Olivier (Service de Médecine Interne, Hôpital Saint-Antoine, Paris); Fakhouri Fadi (Service de Néphrologie, CHU Hôtel-Dieu, Nantes); Frémeaux-Bacchi Véronique (Laboratoire d’Immunologie, Hôpital Européen Georges Pompidou, Paris); Galicier Lionel (Service d’Immunopathologie, Hôpital Saint-Louis, Paris); Grangé Steven (Service de Réanimation Médicale, CHU Charles Nicolle, Rouen); Guidet Bertrand (Service de Réanimation Médicale, Hôpital Saint-Antoine, Paris); Halimi Jean-Michel (Service de Néphrologie Pédiatrique, Hôpital Bretonneau, Tours); Hamidou Mohamed (Service de Médecine Interne, Hôtel-Dieu, Nantes); Hié Miguel (Service de Médecine Interne, Groupe Hospitalier Pitié-Salpétrière, Paris); Jacobs Frédéric (Service de Réanimation Médicale, Hôpital Antoine Béclère, Clamart); Joly Bérangère (Service d’Hématologie Biologique, Hôpital Lariboisière, Paris); Kanouni Tarik (Unité d’Hémaphrèse, Service d’Hématologie, CHU de Montpellier); Kaplanski Gilles (Service de Médecine Interne, Hôpital la Conception, Marseille); Lautrette Alexandre (Hôpital Gabriel Montpied, Service de Réanimation Médicale, ClermontFerrand); Le Guern Véronique (Unité d’Hémaphérèse, Service de Médecine Interne, Hôpital Cochin, Paris); Mariotte Eric (Service de Réanimation, Hôpital Saint-Louis, Paris); Moulin Bruno (Service de Néphrologie, Hôpital Civil, Strasbourg); Mousson Christiane (Service de Néphrologie, CHU de Dijon); Ojeda Uribe Mario (Service d’Hématologie, Hôpital Emile Muller, Mulhouse); Ouchenir Abdelkader (Service de Réanimation, Hôpital Louis Pasteur, Le Coudray); Parquet Nathalie (Unité de Clinique Transfusionnelle, Hôpital Cochin, Paris); Peltier Julie (Urgences Néphrologiques et Transplantation Rénale, Hôpital Tenon, Paris); Pène Frédéric (Service de Réanimation Médicale, Hôpital Cochin, Paris); Perez Pierre (Service de Réanimation Polyvalente, CHU de Nancy); Poullin Pascale (Service d’Hémaphérèse et d’Autotransfusion, Hôpital la Conception, Marseille); PouteilNoble Claire (Service de Néphrologie, CHU Lyon-Sud, Lyon); Presne Claire (Service de Néphrologie, Hôpital Nord, Amiens); Provôt François (Service de Néphrologie, Hôpital Albert Calmette, Lille); Rondeau Eric (Urgences Néphrologiques et Transplantation Rénale, Hôpital Tenon, Paris); Saheb Samir haematologica | 2022; 107(3)

Genome wide association study in Immune TTP

(Unité d’Hémaphérèse, Hôpital la Pitié-Salpétrière, Paris); Seguin Amélie (Service de Réanimation Médicale, Centre Hospitalier de Vendée); Servais Aude (Service de Néphrologie, CHU Necker-Enfants Malades); Stépanian Alain (Laboratoire d’Hématologie, Hôpital Lariboisière, Paris); Vernant Jean-Paul (Service d’Hématologie, Hôpital la Pitié-Salpétrière, Paris); Veyradier Agnès (Service d’Hématologie Biologique, Hôpital Lariboisière, Paris); Vigneau Cécile (Service de Néphrologie,

References 1. Scully, M, Yarranton H, Liesner R, et al. Regional UK TTP registry: correlation with laboratory ADAMTS 13 analysis and clinical features. Br J Haematol. 2008; 142(5):819-826. 2 Tsai H-M, Raoufi M, Zhou W, et al. ADAMTS13-binding IgG are present in patients with thrombotic thrombocytopenic purpura. Thromb Haemost. 2006;95(5):886-892. 3. Rieger M, Mannucci PM, Kremer Hovinga JA, et al. ADAMTS13 autoantibodies in patients with thrombotic microangiopathies and other immunomediated diseases. Blood. 2005;106(4):1262-1267. 4. Scully M, Hunt BJ, Benjamin S, et al. Guidelines on the diagnosis and management of thrombotic thrombocytopenic purpura and other thrombotic microangiopathies. Br J Haematol. 2012;158(3):323335. 5 Scully M, Brown J, Patel R, McDonald V, Brown CJ, Machin S. Human leukocyte antigen association in idiopathic thrombotic thrombocytopenic purpura: evidence for an immunogenetic link. J Thromb Haemost. 2010;8(2):257-262. 6. Coppo P, Busson M, Veyradier A, et al. HLA-DRB1*11: a strong risk factor for acquired severe ADAMTS13 deficiencyrelated idiopathic thrombotic thrombocytopenic purpura in Caucasians. J Thromb Haemost. 2010;8(4):856-859. 7. John M-L, Hitzler W, Scharrer I. The role of human leukocyte antigens as predisposing and/or protective factors in patients with idiopathic thrombotic thrombocytopenic purpura. Ann Hematol. 2012;91(4):507-510. 8. Kokame K, Nobe Y, Kokubo Y, Okayama A, Miyata T. FRETS-VWF73, a first fluorogenic substrate for ADAMTS13 assay. Br J Haematol. 2005;129(1):93-100. 9. Wellcome Trust Case Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature. 2007; 447(7145):661-678. 10. Illumina. “TOP/BOT” Strand and “A/B” Allele - Technical Note. Illumina SNP Genotyping. 2006. 11. Fairfax BP, Makino S, Radhakrishnan J, et al. Genetics of gene expression in primary immune cells identifies cell type-specific master regulators and roles of HLA alleles. Nat Genet. 2012;44(5):502-510. 12. Fairfax BP, Humburg P, Makino S, et al. Innate immune activity conditions the effect of regulatory variants upon monocyte gene expression. Science. 2014; 343(6175):1246949. 13. Browning BL, Zhou Y, Browning SR. A one-penny imputed genome from nextgeneration reference panels. Am J Hum Genet. 2018;103(3):338-348. 14. Golden Helix, Inc., Bozeman, MT www.goldenhelix.com. SNP & Variation

haematologica | 2022; 107(3)

Hôpital Pontchaillou, Rennes); Wynckel Alain (Service de Néphrologie, Hôpital Maison Blanche, Reims) and Zunic Patricia (Service d’Hématologie, Groupe Hospitalier SudRéunion, la Réunion). Funding This Study was supported by research funding from Shire/Takeda to MAS.

Suite TM (Version 8.8.1). Golden Helix, Inc. 15. Fadista J, Manning AK, Florez JC, Groop L. The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants. Eur J Hum Genet. 2016;24(8): 1202-1205. 16. Pruim RJ, Welch RP, Sanna S, et al. LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics. 2010;26(18):2336-2337. 17. Jia X, Han B, Onengut-Gumuscu S, et al. Imputing amino acid polymorphisms in human leukocyte antigens. PLoS One. 2013;8(6):e64683. 18. Lonsdale, J, Thomas, J, Salvatore, M, et al. The genotype-tissue expression (GTEx) project. Nat Genet. 2013;45(6):580-585. 19. Machiela MJ, Chanock SJ. LDlink: a webbased application for exploring populationspecific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics. 2015;31(21):35553557. 20. Wang J, Zhuang J, Iyer S, et al. Sequence features and chromatin structure around the genomic regions bound by 119 human transcription factors. Genome Res. 2012;22(9):1798-1812. 21. Bailey TL, Johnson J, Grant CE, Noble WS. The MEME suite. Nucleic Acids Res. 2015;43(W1):W39-49. 22. Westra H-J, Peters MJ, Esko T, et al. Systematic identification of trans eQTLs as putative drivers of known disease associations. Nat Genet. 2013;45(10):1238-1243. 23. Rosenbloom KR, Sloan CA, Malladi VS, et al. ENCODE data in the UCSC Genome Browser: year 5 update. Nucleic Acids Res. 2013;41(database issue):D56-63. 24. Kent WJ, Sugnet CW, Furey TS, et al. The human genome browser at UCSC. Genome Res. 2002;12(6):996-1006. 25. Mancini I, Ricaño-Ponce I, Pappalardo E, et al. Immunochip analysis identifies novel susceptibility loci in the human leukocyte antigen region for acquired thrombotic thrombocytopenic purpura. J Thromb Haemost. 2016;14(12):2356-2367. 26. Hassed S, Li S, Mulvihill J, Aston C, Palmer S. Adams-Oliver syndrome review of the literature: Refining the diagnostic phenotype. Am J Med Genet A. 2017;173(3):790800. 27. Caron C, DeGeer J, Fournier P, et al. CdGAP/ARHGAP31, a Cdc42/Rac1 GTPase regulator, is critical for vascular development and VEGF-mediated angiogenesis. Sci Rep. 2016;6:27485. 28. Cai X, Huang W, Liu X, Wang L, Jiang Y. Association of novel polymorphisms in TMEM39A gene with systemic lupus erythematosus in a Chinese Han population. BMC Med Genet. 2017;18(1):43. 29. Lessard CJ, Adrianto I, Ice JA, et al. Identification of IRF8, TMEM39A, and IKZF3-ZPBP2 as susceptibility loci for systemic lupus erythematosus in a large-scale multiracial replication study. Am J Hum

Genet. 2012;90(4):648-660. 30. Wagner M, Sobczyński M, Bilińska M, et al. Preliminary study on the role of TMEM39A gene in multiple sclerosis. J Mol Neurosci. 2017;62(2):181-187. 31. D’Cunha MA, Pandit L, Malli C. CD6 gene polymorphism rs17824933 is associated with multiple sclerosis in Indian population. Ann Indian Acad Neurol. 2016; 19(4):491-494. 32. Guarani V, Paulo J, Zhai B, Huttlin EL, Gygi SP, Harper JW. TIMMDC1/C3orf1 functions as a membrane-embedded mitochondrial complex I assembly factor through association with the MCIA complex. Mol Cell Biol. 2014;34(5):847-861. 33. Ishimoto T, Shimada M, Araya CE, Huskey J, Garin EH, Johnson RJ. Minimal change disease: a CD80 podocytopathy? Semin Nephrol. 2011;31(4):320-325. 34. Kow NY, Mak A. Costimulatory pathways: physiology and potential therapeutic manipulation in systemic lupus erythematosus. Clin Dev Immunol. 2013;2013: 245928. 35. Saverino D, Simone R, Bagnasco M, Pesce G. The soluble CTLA-4 receptor and its role in autoimmune diseases: an update. Auto Immun Highlights. 2010;1(2):73-81. 36. Fernandez-Valdivia R, Takeuchi H, Samarghandi A, et al. Regulation of mammalian Notch signaling and embryonic development by the protein O-glucosyltransferase Rumi. Development. 2011; 138(10):1925-1934. 37. Servián-Morilla E, Takeuchi H, Lee TV, et al. A POGLUT1 mutation causes a muscular dystrophy with reduced Notch signaling and satellite cell loss. EMBO Mol Med. 2016;8(11):1289-1309. 38. Stenflo J, Stenberg Y, Muranyi A. Calciumbinding EGF-like modules in coagulation proteinases: function of the calcium ion in module interactions. Biochim Biophys Acta. 2000;1477(1-2):51-63. 39. Wouters MA, Rigoutsos I, Chu CK, Feng LL, Sparrow DB, Dunwoodie SL. Evolution of distinct EGF domains with specific functions. Protein Sci. 2005;14(4):1091-1103. 40. Li Z, Fischer M, Satkunarajah M, Zhou D, Withers SG, Rini JM. Structural basis of Notch O-glucosylation and O-xylosylation by mammalian protein-O-glucosyltransferase 1 (POGLUT1). Nat Commun. 2017;8(1):185. 41. Basmanav FB, Oprisoreanu A-M, Pasternack SM, et al. Mutations in POGLUT1, encoding protein O-glucosyltransferase 1, cause autosomal-dominant Dowling-Degos disease. Am J Hum Genet. 2014;94(1):135-143. 42. International Multiple Sclerosis Genetics Consortium (IMSGC). Comprehensive follow-up of the first genome-wide association study of multiple sclerosis identifies KIF21B and TMEM39A as susceptibility loci. Hum Mol Genet. 2010;19(5):953-962. 43. Sheng Y, Xu J, Wu Y, et al. Association

581

M.J. Stubbs et al. analyses confirm five susceptibility loci for systemic lupus erythematosus in the Han Chinese population. Arthritis Res Ther. 2015;17(1):85. 44. Yao Q, Wang B, Qin Q, Jia X, Li L, Zhang JA. Genetic variants in TMEM39A gene are associated with autoimmune thyroid diseases. DNA Cell Biol. 2019;38(11):12491256. 45. Lenting PJ, Christophe OD, Denis CV. von Willebrand factor biosynthesis, secretion, and clearance: connecting the far ends. Blood. 2015;125(13):2019-2028. 46. Verbij FC, Stokhuijzen E, Kaijen PHP, van Alphen F, Meijer AB, Voorberg J. Identification of glycans on plasma-derived ADAMTS13. Blood. 2016;128(21):e51-58. 47. Sorvillo N, Kaijen PH, Matsumoto M, et al. Identification of N-linked glycosylation and putative O-fucosylation, C-mannosylation sites in plasma derived ADAMTS13. J

582

Thromb Haemost. 2014;12(5):670-679. 48. Nowak AA, O’Brien HER, Henne P, et al. ADAMTS-13 glycans and conformationdependent activity. J Thromb Haemost. 2017;15(6):1155-1166. 49. Janghorban M, Xin L, Rosen JM, Zhang XH-F. Notch signaling as a regulator of the tumor immune response: to target or not to target? Front Immunol. 2018;9:1649. 50. Rong H, Shen H, Xu Y, Yang H. Notch signalling suppresses regulatory T-cell function in murine experimental autoimmune uveitis. Immunology. 2016;149(4):447-459. 51. Yang H, Zheng S, Mao Y, et al. Modulating of ocular inflammation with macrophage migration inhibitory factor is associated with notch signalling in experimental autoimmune uveitis. Clin Exp Immunol. 2016;183(2):280-293. 52. Kuksin CA, Minter LM. The link between autoimmunity and lymphoma: does

NOTCH signaling play a contributing role? Front Oncol. 2015;5:51. 53. Bassil R, Orent W, Elyaman W. Notch signaling and T-helper cells in EAE/MS. Clin Dev Immunol. 2013;2013:570731. 54. Sandy AR, Stoolman J, Malott K, Pongtornpipat P, Segal BM, Maillard I. Notch signaling regulates T cell accumulation and function in the central nervous system during experimental autoimmune encephalomyelitis. J Immunol. 2013;191(4): 1606-1613. 55. Radtke F, MacDonald HR, Tacchini-Cottier F. Regulation of innate and adaptive immunity by Notch. Nat Rev Immunol. 2013; 13(6):427-437. 56. Hao X, Li Y, Wang J, et al. Deficient OGlcNAc glycosylation impairs regulatory T cell differentiation and Notch signaling in autoimmune hepatitis. Front Immunol. 2018;9:2089.

haematologica | 2022; 107(3)

ARTICLE

Acute Myeloid Leukemia

Clinical significance of RAS pathway alterations in pediatric acute myeloid leukemia

Ferrata Storti Foundation

Taeko Kaburagi,1,2 Genki Yamato,1,2 Norio Shiba,3 Kenichi Yoshida,4 Yusuke Hara,2 Ken Tabuchi,5 Yuichi Shiraishi,6 Kentaro Ohki,7 Manabu Sotomatsu,1 Hirokazu Arakawa,2 Hidemasa Matsuo,8 Akira Shimada,9 Tomohiko Taki,10 Nobutaka Kiyokawa,7 Daisuke Tomizawa,11 Keizo Horibe,12 Satoru Miyano,13 Takashi Taga,14 Souichi Adachi,8 Seishi Ogawa4 and Yasuhide Hayashi1,15 Department of Hematology/Oncology, Gunma Children's Medical Center, Gunma; Department of Pediatrics, Gunma University Graduate School of Medicine, Gunma; 3 Department of Pediatrics, Yokohama City University Hospital, Kanagawa; 4Department of Pathology and Tumor Biology, Graduate School of Medicine, Kyoto University, Kyoto; 5 Department of Pediatrics, Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Tokyo; 6Division of Cellular Signaling, National Cancer Center Research Institute, Tokyo; 7Department of Pediatric Hematology and Oncology Research, National Research institute for Child Health and Development, Tokyo; 8Human Human Health Sciences, Graduate School of Medicine, Kyoto University, Kyoto; 9Department of Pediatrics, Okayama University, Okayama; 10Department or Medical Technology, Kyorin University Faculty of Health Sciences, Tokyo; 11Division of Leukemia and Lymphoma, Children’s Cancer Center, National Center for Child Health and Development, Tokyo; 12 Clinical Research Center, National Hospital Organization Nagoya Medical Center, Aichi; 13 Laboratory of Sequence Analysis, Human Genome Center, Institute of Medical Science, Tokyo University, Tokyo; 14Department of Pediatrics, Shiga University of Medical Science, Shiga and 15Institute of Physiology and Medicine, Jobu University, Gunma, Japan. 1 2

Haematologica 2022 Volume 107(3):583-592

ABSTRACT

AS pathway alterations have been implicated in the pathogenesis of various hematological malignancies. However, their clinical relevance in pediatric acute myeloid leukemia (AML) is not well characterized. We analyzed the frequency, clinical significance, and prognostic relevance of RAS pathway alterations in 328 pediatric patients with de novo AML. RAS pathway alterations were detected in 80 (24.4%) of 328 patients: NF1 (n=7, 2.1%), PTPN11 (n=15, 4.6%), CBL (n=6, 1.8%), NRAS (n=44, 13.4%), KRAS (n=12, 3.7%). Most of these alterations in the RAS pathway were mutually exclusive also together with other aberrations of signal transduction pathways such as FLT3-ITD (P=0.001) and KIT mutation (P=0.004). NF1 alterations were frequently detected in patients with complex karyotype (P=0.031) and were found to be independent predictors of poor overall survival (OS) in multivariate analysis (P=0.007). At least four of seven patients with NF1 alterations had biallelic inactivation. NRAS mutations were frequently observed in patients with CBFB-MYH11 and were independent predictors of favorable outcomes in multivariate analysis (OS, P=0.023; event-free survival [EFS], P=0.037). Patients with PTPN11 mutations more frequently received stem cell transplantation (P=0.035) and showed poor EFS than patients without PTPN11 mutations (P=0.013). Detailed analysis of RAS pathway alterations may enable a more accurate prognostic stratification of pediatric AML and may provide novel therapeutic molecular targets related to this signal transduction pathway.

Introduction Acute myeloid leukemia (AML) is characterized by considerable genetic heterogeneity. Several chromosomal aberrations and gene alterations have been identified in these patients; some of these have been found useful for risk stratification.1 Aberrations of signal transduction pathways (such as RAS family members, KIT, and FLT3) are considered as one of the most important pathogenetic factors in AML.2

haematologica | 2022; 107(3)

Correspondence: YASUHIDE HAYASHI hayashiy@jobu.ac.jp hayashiy-tky@umin.ac.jp Received: August 28, 2020. Accepted: March 12, 2021. Pre-published: March 18, 2021. https://doi.org/10.3324/haematol.2020.269431

583

T. Kaburagi et al.

Recently, aberrations of NF1 and PTPN11 were reported to be associated with a poor prognosis in adult patients with AML.3,4 NF1 and PTPN11 are the family of RAS pathway genes and constitute the granulocyte-macrophage colony stimulating factor signaling pathway. Among the broad family of RAS pathway genes, mutations of CBL, NRAS and KRAS were also commonly detected in AML.2 These RAS pathway alterations have also been implicated in the causation of juvenile myelomonocytic leukemia (JMML).5 Mutations of PTPN11, NRAS, and KRAS have been reported in 3–4%,6,7 7–13%, 6–11%8,9 of pediatric patients with AML, respectively. However, there is no clear consensus on the clinical significance of RAS pathway gene mutations especially NF1 and CBL mutations.10,11 The reported frequency of detection of CBL mutations and NF1 mutations or deletions in adult patients with AML is 0.6–0.7%12,13 and 3.5–10.5%,14-16 respectively. However, the prognostic relevance of these mutations is not well characterized, particularly in pediatric AML patients. In this study, we analyzed NF1, PTPN11, CBL, NRAS, and KRAS alterations in 328 pediatric patients with AML to determine the clinical significance of these alterations. We also examined the correlation of RAS pathway alterations with other genetic aberrations, cytogenetic alterations, and clinical characteristics.

Copy number analysis Copy number (CN) analysis was performed as previously reported34 using an in-house pipeline CNACS (https://github.com/papaemmelab/toil_cnacs); the total number of reads covering each bait region and the allele frequency of heterozygous single-nucleotide polymorphisms (SNP) (n=1,216) detected by targeted sequencing were used as input data. Based on the previous reports,15 we set the total CN <1.5 as the definition of NF1 deletion.

Statistical methods All statistical analyses were performed using the EZR software (version 1.35; Saitama Medical Center, Jichi Medical University, Saitama, Japan).35 Between-group differences with respect to clinical characteristics were assessed using the Fisher’s exact and Mann-Whitney U tests. Survival rates were estimated using the Kaplan–Meier method and compared using the log-rank test. Overall survival (OS) was defined as the time from diagnosis to death or last follow-up. Event-free survival (EFS) was defined as the time from diagnosis to the date of failure (induction failure, relapse, second malignancy, or death) for patients who experienced treatment failure or to the date of last contact for all other patients. Cox proportional hazards model was used to estimate hazard ratios and 95% Confidence Intervals (CI). For all analyses, two tailed P-values <0.05 were considered indicative of statistical significance.

Methods

Results

Patients

Frequencies of RAS pathway alterations in 328 pediatric acute myeloid leukemia patients

Between November 2006 and December 2010, 443 pediatric patients with de novo AML (age <18 years) participated in the Japanese AML-05 trial conducted by the Japanese Pediatric Leukemia/Lymphoma Study Group (JPLSG). Treatment, data collection, and other details of the AML-05 study are presented in the Online Supplementary Appendix and the Online Supplementary Figure S1. This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the Gunma Children’s Medical Center and the Ethical Review Board of the JPLSG.

Mutation analysis of RAS pathway alterations We analyzed PTPN11 (exons 2–4, and 13), CBL (exons 8–9), NRAS (exons 1–2), and KRAS (exons 1–2) mutations using Sanger sequencing as previously described.9,17,18 All coding exons of the NF1 were captured using the SureSelect custom kit (Agilent Technologies, Santa Clara, CA, USA), and sequenced using Hiseq 2500. Somatic mutations in NF1 were identified as described elsewhere.19

Molecular characterization We analyzed KIT (exons 8, 10, and 17),20 NPM1 (exon 12),21 CEBPA (exons 1–4),22 CSF3R (exons 14 and 17),23 WT1 (exons 7– 10),24 ASXL1 (exon 12), ASXL2 (exons 11 and 12),25 all exons of BCOR, BCORL126, RAD21, SMC3, STAG2,27 RUNX1,28 FLT3-ITD,29 and gene rearrangement of NUP98-NSD130 and FUS-ERG31 using Sanger sequencing. KMT2A-partial tandem duplication (PTD) was analyzed using the multiplex ligation-dependent probe amplification (MLPA) method.32 Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of the PRDM16 and MECOM genes was performed using the 7900HT Fast Real Time PCR System, TaqMan Gene Expression Master Mix, and TaqMan Gene Expression Assay (Applied Biosystems, Foster City, CA, USA), as described elsewhere.33

584

Out of the 443 patients, 115 patients were excluded from this study because of unavailability of genomic DNA samples. Therefore, 328 samples were analyzed in this study. We did not analyze germline alterations because of the lack of non-hematological or remission samples. The clinical characteristics of patients with available samples (n=328) and those with no available samples (n=115) are summarized in the Online Supplementary Table S1. White blood cell (WBC) count at diagnosis was significantly higher in the “sample available group” than in the “sample unavailable group” (P<0.001). There were more patients who were at a low risk and there were less patients who were at an intermediate risk in the “sample available group” as compared with the “sample unavailable group” (low risk, P=0.046; intermediate risk, P=0.003). Cytogenetic features and prognosis were not significantly different between the available and unavailable samples (Online Supplementary Table S1). RAS pathway alterations were detected in 80 (24.4%) of the 328 patients; most of these alterations were mutually exclusive (Figure 1). The mutation sites and clinical characteristics of patients with RAS pathway alterations are summarized in Figure 2, Tables 1 and 2 and the Online Supplementary Tables S2 and S3, respectively. We detected six NF1 mutations in four patients; all of these were frameshift or nonsense mutations (Figures 1 and 2). Two patients concomitantly had two types of mutations, respectively (Table 1). In addition, we also detected four patients with a microdeletion within chromosome 17q containing NF1 (Table 1; Online Supplementary Figure S2). One patient had both an NF1 mutation and CN alteration and NF1 alterations were detected in seven (2.1%) patients (Figure 1; Table 1). Two haematologica | 2022; 107(3)

RAS pathway alterations in pediatric AML

patients: unique patient number (UPN) 57 and UPN 415 also had a heterozygous deletion. Additionally patient UPN 57 with variant allele frequency (VAF) 0.83 had nonsense mutations in the remaining allele, while UPN 50 with VAF 0.94 had 17q uniparental disomy (UPD) (Online Supplementary Figure S2). UPN 105 and UPN 333 had two or three different CN regions in NF1, respectively with partially homozygous deletions. Other two patients (UPN 262 and 367) had two types of mutations each. However, it was not clear whether these alterations were monoallelic or bi-allelic (UPN 262, VAF 0.28 and 0.26; UPN 367, VAF 0.28 and 0.08). Thus, we concluded that at least four patients (UPN 50, UPN 57, UPN 105, and UPN 333) had bi-allelic NF1 inactivation. Next, on the basis of the VAF of each mutation, we estimated whether NF1 mutations were somatic or germline. If a mutation is a heterozygous germline mutation, then the VAF is around 0.5.36 We considered mutations in UPN 262 (VAF 0.28 and 0.26) and UPN 367 (VAF 0.28 and 0.08) as somatic mutations. Regarding UPN 50 (VAF 0.94) and UPN 57 (VAF 0.83), it was impossible to predict whether these mutations were somatic or germline because their VAF were high owing to their co-existence with heterozygous deletion or UPD. On the contrary, these two mutations were determined as somatic in the COSMIC v90 (URL: https://cancer.sanger.ac.uk/cosmic). R1241X detected in UPN 57 was previously observed in adult AML and E1561X detected in UPN 50 was previously detected in non-hematological malignancies.37,38 PTPN11 mutations were detected in 15 (4.6%) patients (Figure 1). Of these, 14 were located in exon 3 or exon 13,

which are known mutation hotspots in AML and JMML (Figure 2).39 As previously observed,39 codon 76 represented a mutational hot spot (four of 15, 27%) with three different amino acid substitutions (Figure 2), and 13 of the 15 mutations have been reported as somatic mutations.39-41 Although the remaining two mutations (V45L and T493I) have not been confirmed as somatic mutations, V45L was earlier detected in lung carcinoma and showed an association with activation of protein-tyrosine phosphatase.41 However, T493I has not been reported in any hematological or other disease. These two variants have not been reported as SNP on any database such as COSMIC v90, ClinVar, mutations taster, Ensembl GRCh37, or db SNP (URL: https://www.ncbi.nlm.nih.gov/snp/, https://www.ncbi.nlm.nih.gov/snp/, http://grch37.ensembl.org/index.html, https://www.ncbi.nlm.nih.gov/clinvar/, and http://mutationtaster.org/); therefore, we identified these as novel disease-causing mutations. CBL mutations were found in six (1.8%) patients (Figure 1). Among these, four were deletions or insertions and deletions in exon 8 and two were missense mutations in exon 9. Five of these mutations were in the linker region or the RING finger domain which were previously reported as the affected regions in myeloid malignancies with CBL mutations (Figure 2).12,13,18 None of the six mutations have been reported as SNP or germline mutations in any online databases or previous reports.42 CBL mutations especially missense mutations were shown to exhibit a strong association with 11q-acquired UPD.18 11q UPD was detected in only one patient with a missense mutation (UPN 97) by CN analysis (Online Supplementary Figure S2).

Figure 1. Molecular and cytogenetic aberrations in 80 pediatric acute myeloid leukemia patients with RAS pathway alterations. Each column displays the cytogenetic aberration pattern and clinical status of an individual sample. Orange indicates RAS pathway and other genetic alterations. Blue indicates chromosomal aberrations. Purple indicates gene expression. Gray indicates clinical outcome. Blanks indicate the absence of the chromosomal aberration, genetic alteration, or prognostic event. CR: complete remission.

haematologica | 2022; 107(3)

585

T. Kaburagi et al.

Figure 2. Gene diagrams depicting RAS pathway mutations in pediatric patients with acute myeloid leukemia. (A) NF1 mutations (NCBI reference sequence; NM_000267); (B) PTPN11 mutations (NCBI reference sequence; NM_002834); (C) CBL mutations (NCBI reference sequence; NM_005188); (D) NRAS mutations (NCBI reference sequence; NM_002524); (E) KRAS mutations (NCBI reference sequence; NM_004985).

NRAS and KRAS mutations were detected in 44 (13.4%) and 12 (3.7%) patients, respectively (Figure 1). All NRAS and KRAS mutations were missense mutations in codon 12, 13, or 61, which are well known hotspots (Figure 2).43 Six patients concomitantly had two missense mutations in NRAS.

Clinical and cytogenetic characteristics of patients with RAS pathway alterations The clinical characteristics of patients with RAS pathway alterations are summarized in the Online Supplementary Table S4. Patients with RAS pathway alterations showed a significantly higher frequency of detection of monosomy 7 as compared to those without RAS pathway alterations (P<0.001). FLT3-ITD and KIT mutations were significantly less frequent in patients with RAS pathway alterations (FLT3-ITD, P=0.001; KIT mutations, P=0.004). Age, sex, or relapse rate were not significantly different between patients with or without each specific RAS pathway alteration. Patients with CBL mutations had significantly higher WBC count at diagnosis (P=0.026; Online Supplementary Table S4; Online Supplementary Figure S3). The frequency of stem cell transplantation (SCT) was significantly higher in patients with PTPN11 mutations (P=0.035), and significantly lower in patients with NRAS mutations (P=0.022; Figure 1; Online Supplementary Table S4). PTPN11 mutations were significantly fewer (P=0.024) in patients with low risk, i.e., core binding factor (CBF)-AML, and NRAS mutations were significantly higher (P=0.017) in these patients (Figure 1; Online Supplementary Table S4). The frequency of detection of NF1 alterations was significantly higher in patients with complex karyotype (P=0.031) and MECOM high expression (P=0.013, Figure 1; Online 586

Supplementary Table S4). PTPN11 mutations were significantly more frequently detected in patients with monosomy 7 (P=0.047), RUNX1 mutations (P=0.004), PRDM16 high expression (P=0.002), and MECOM high expression (P=0.004) (Figure 1; Online Supplementary Table S4). NRAS mutations were frequently detected in inv(16)(p13q22)/CBFB-MYH11 (P=0.001) and monosomy 7 (P=0.013). NRAS mutations were also mutually exclusive with FLT3-ITD (P=0.005) and KIT mutations (P=0.040) (Figure 1; Online Supplementary Table S4). Although there was no significant difference, three of six patients with CBL mutations were identified in CBF-AML (P=0.411) (Figure 1; Online Supplementary Table S4).

Prognosis of patients with RAS pathway alterations We analyzed the prognosis of patients with or without RAS pathway alterations using the Kaplan–Meier method (Figure 3; Online Supplementary Figure S4). Despite the small sample size, alterations of NF1 and PTPN11 showed a significant association with poor prognosis. Although there was no significant difference in EFS between patients with or without NF1 alterations, the OS of patients with NF1 alterations was significantly worse than that of patients without NF1 alterations (2year OS, 42.9% vs. 82.3%, P=0.003) (Figure 3A and B). Although no significant differences were observed in OS, PTPN11 mutations were significantly associated with poor EFS (2-year EFS, 30.0% vs. 59.8%, P=0.013) (Figure 3C and D). The OS and EFS of patients with NRAS mutations were significantly better than those of patients without NRAS mutations (2-year OS, 97.7% vs. 79.0%, P=0.014; 2-year EFS, 74.9% vs. 55.9%, P=0.021) (Figure 3E and F). The presence of CBL or KRAS mutations showed no significant impact on prognosis (Online haematologica | 2022; 107(3)

RAS pathway alterations in pediatric AML

Table 1. Summary of characteristics of pediatric acute myeloid leukemia patients with NF1 alteration.

UPN

Nucleotide change*

Amino acid change*

VAF

Copy number

Start to end

Sex

Age, WBC, y ×109/L

c.G4681T p.E1561X c.2027dupC p.I679Dfs21X c.6862_6863insCG p.P2289Rfs10X 367 c.966_967insGA p.A323Efs54X c.2027dupC p.I679Dfs21X

0.94 0.28 0.26 0.28 0.08

M M

13.7 12.3

c.C3721T

R1241X

0.83

15.2

10.8

415

1225849-29422297 29485961-30325657 27009658-29588669 29626467-29679186 29683418-30325657 1225849-29422297 29485961-30325657

105

1.16 0.99 1.02 0.29 0.95 1.26 1.04

12.3

333

0.22 29485961-29588669 0.94 29626467-30325657

9.8

50 262

Cytogenetics

Additional genetic aberrations

19.9 45,XY,-7[13]/46,XY[7] KIT 159.3 46,XY,inv(16) CBL, (p13q22)[20] NRAS 9.9 47,XY,+11[18]/54, PTPN11 idem,+X,+10,+11,+13,+14, +20,+21[1]/46,XY[1] 69.0 #1 RUNX1, BCORL1 15.5 #2 ASXL1

1.9

45,XX,ins(1;?) PTPN11 (q21;?), add(4)(q12), add(7)(q36), der(17;18) (q10;q10)[20] 4.1 46,XX,t(8;12) BCORL1 (q11.2;p11.2)[20]

CR Relapse Event SCT Prognosis

+ -

Death Alive

Alive

Death

Alive

UPN: unique patient number; VAF: variant allele frequency; WBC: white blood cell count; CR: complete remission; SCT: stem cell transplantation; M: male, F: female; y: years; SCT: stem cell transplantation. *NCBI reference sequence; NM_00267. #1 47,X,-Y,add(3)(q11.2),+6, add(6)(p21)x2,+7,del(8)(q24)der(8)t(1;8)(q11;q24),del(11)(q?),add(17)(p11.2)[7]/48,sl,+22[6]/47,sl,-14,+mar1[2] #2 46,XY,+Y,add(1)(p11),del(2)(q?),del(5)(q?),add(8)(p11.2),-9,-9,-11,-17,add(18)(q21),-19,add(22)(q11.2),+del(?)t(?;11)(?;q13),+mar1,+mar2,+mar3[2]/88,sl,×2,-3,-del(5)×2,-6,+9,-20,-20,-21, -mar1,-mar3×2,+5mar[1]/47,XY,+Y[9]

Supplementary Figure S4). With respect to prognosis, patients with CBL mutations were divided into two distinct groups based on the presence of CBF. All CBF-AML patients with CBL mutations achieved complete remission and were alive. However, all non-CBF-AML patients relapsed and died (Table 2). Next, we performed multivariate analysis using the Cox regression analysis to determine the prognostic impacts of RAS pathway alterations (Table 3). Besides RAS pathway mutations, we used t(8;21)(q22;q22)/RUNX1-RUNX1T1, CBFB-MYH11, monosomy 7, complex karyotype, FLT3ITD, 5q-, FUS-ERG, NUP98-NSD1, and PRDM16 high expression as explanatory variables in the multivariate analysis; these cytogenetic aberrations were used for risk classification in the AML-05 trials (Online Supplementary Figure S1) or were recently shown to affect the prognosis.33,44 Remarkably, NF1 alterations were associated with inferior OS in multivariate analysis (hazard ratio [HR] 4.109; 95% CI:, 1.471–11.48; P=0.007] (Table 3). In univariate analysis, PTPN11 mutation was associated with inferior EFS (HR 2.142; 95% CI: 1.157-3.965; P=0.015) (Table 3). However, PTPN11 mutation was not associated with inferior EFS (HR 1.239; 95% CI: 0.616–2.494; P=0.548) in multivariate analysis; this indicated that cooccurring aberrations contributed to worse outcomes (Table 3). In multivariate analysis, NRAS mutation was a favorable prognostic factor for both OS and EFS (OS: HR 0.309; 95% CI: 0.112–0.849; P=0.023; EFS: HR, 0.530; 95% CI: 0.293–0.961; P=0.037) (Table 3). These results suggested that alterations of NF1 and NRAS were independent predictors of prognosis in pediatric patients with AML. CBFB-MYH11 could not be evaluated accurately for OS in the Cox regression analysis because 27 patients with CBFB-MYH11 enrolled in this study were all alive. The OS of patients with CBFB-MYH11 was significantly better than that of patients without CBFB-MYH11 in the Kaplan–Meier method (P=0.005). (Online Supplementary Figure S5) haematologica | 2022; 107(3)

Discussion In this study, we detected RAS pathway alterations in 80 (24.4%) of the 328 patients with AML (NF1 [n=7, 2.1%], PTPN11 [n=15, 4.6%], CBL (n=6, 1.8%], NRAS [n=44, 13.4%], KRAS [n=12, 3.7%]). Most of these were mutually exclusive and were also mutually exclusive with aberrations involving other signal transduction pathways such as FLT3-ITD and KIT mutation (Figure 1). Loss of the wild-type allele of NF1, either through deletions or mutations, has been implicated in the pathogenesis of hematological malignancies.11 We have summarized previous reports on NF1 alterations in adult and pediatric AML in the Online Supplementary Table S5. NF1 deletions have been reported in 3.5–10.5% of adult patients with AML; in addition, 20-50% of patients with NF1 deletions had concomitant NF1 mutations in the remaining allele.14-16 In this study, the frequency of NF1 alterations was less than that in previous reports pertaining to adult patients. In addition, at least four of the seven (57%) patients with NF1 alterations had bi-allelic NF1 inactivation (Table 1). NF1 alterations have been frequently reported in complex karyotype AML; in addition, NF1 alterations were shown to be associated with poor prognosis in adult AML.3 In the contemporary literature, there are few reports about NF1 alteration in pediatric AML. Balgobind et al. detected NF1 deletion in two of the 71 AML patients with KMT2A rearrangement, one of whom experienced relapse.11 Consistent with previous reports, NF1 alterations were frequently detected in complex karyotype, and were associated with poor OS in this study (Figure 3; Table 3). None of the four patients with relapse or induction failure were rescued by SCT (Figure 1; Table 1). Our findings suggest that more intensive primary chemotherapy may be an option to rescue AML patients with NF1 alterations including use of novel molecular targeted therapy such as mTOR inhibitors. In a study by Parkin et al., NF1 null blasts showed sensitivity to rapamycin-induced apoptosis.3,14 587

T. Kaburagi et al.

Figure 3. Prognostic significance of NF1, PTPN11, and NRAS alterations in pediatric patients with acute myeloid leukemia. (A), (C), and (E) show Kaplan–Meier curves of overall survival of patients with and without NF1, PTPN11, and NRAS alterations. (B), (D), and (F) show Kaplan–Meier curves of event-free survival of patients with and without NF1, PTPN11, and NRAS alterations.

We also detected 15 PTPN11 mutations including two novel mutations (Table 2). In several previous studies, PTPN11 mutations have been reported to be associated with acute monoblastic leukemia (FAB-M5),4,7 however, no such tendency was observed in this study (data was not shown). PTPN11 mutations in our cohort were frequently detected in AML, minimally differentiated (FAB-M0) (P=0.026) and erythroleukemia (FAB-M6) (P=0.047). Goemans et al. also reported that the prevalence of PTPN11 was not increased in acute monoblastic leukemia (FAB-M5) suggesting that differences could exist in the ethnic background of the patients studied.45 In a study by Alfayez et al., PTPN11 mutation in adult AML patients was associated with adverse prognosis.4 However, the prognostic relevance of PTPN11 has not been reported in pediatric AML.6,7 In this study, patients with PTPN11 mutations had a high frequen588

cy of RUNX1 mutations, MECOM high expression, and PRDM16 high expression which are strongly associated with poor prognosis (Figure 1; Online Supplementary Table S4).30,33,46-48 In our study, PTPN11 mutations were associated with poor EFS in univariate analysis; however, multivariate analysis revealed no significant impact of PTPN11 mutations on EFS or OS (Figure 3; Table 3). A significantly greater proportion of patients with PTPN11 mutations received SCT (Online Supplementary Table S4); in addition, five of 11 patients with events were rescued by SCT (Figure 1). We consider that AML patients with PTPN11 mutations tended to have a high frequency of relapse or induction failure, and some of these patients were successfully rescued by SCT. Consistent with a previous report,49 NRAS mutations were significantly more frequently detected in CBFBMYH11 (Figure 1; Online Supplementary Table S4). Previous haematologica | 2022; 107(3)

RAS pathway alterations in pediatric AML

Table 2. Summary of characteristics of pediatric acute myleoid leukemia patients with PTPN11 and CBL mutations.

Gene

PTPN11

CBL

UPN

Nucleotide change*

A227T

E76V

52 113

G133C C215T

V45L A72V

F F

127 142 156 177

C218T G1508C C215A A227G

T73I G503A A72D E76G

F M F M

249

G179T

G60V

300

C1478T

T493I

367

G226A

E76K

375 415

G181T G1508C

D61Y G503A

M F

417

G1508C

G503A

425 438

G205A A227T

E69K E76V

M F

2 c.1174_1181delins TTATCATCCTTATCAT TATCACAGGT 67 c.A1405G 97 c.T1248G 167 c.1096-75_1218 delinsAAAGGCT 184 c.1183_1227+27del 262 c.1096-40_1227+35del

Amino acid Sex change*

Age, WBC, y ×109/L 4.8

Cytogenetics

Additional genetic aberrations

33.9

45,XY,-7[1]/45,sl,t(3;12) (q26;p13)[18]/46,XY[1] 14.1 16.5 46,XX[20] WT1, KMT2A-PTD 10.3 17.8 46,XX,add(12)(p11)[12]/46,XX[8} CBL, KRAS, KMT2A-ELL, WT1, STAG2 0.4 17.1 47,XX,t(7;12)(q36;p13),+19[20] RAD21 6.9 190.5 N/A KMT2A-MLLT3 11.5 4.5 46,XX[20] FLT3-ITD, NPM1 2.9 25.2 45,XY,-7[1]/45,sl,t(11;21) (q13;q22)[19] 11.8 60.1 46,XX[20] NRAS, KMT2A-PTD, RUNX1 4.2 4.6 46,XY,t(8;21)(q22;q22)[2]/46,sl, del(9)(q?)[7]/46,XY[11] 7 9.9 47,XY,+11[18]/54,idem,+X, NF1 +10,+11,+13,+14,+20,+21[1]/46,XY[1] 1.9 16.1 46,XY,-7,+mar[17]/46,idem,del(6)(q?)[3] RUNX1 12.3 1.9 45,XX,ins(1;?)(q21;?),add(4)(q12) NF1 add(7)(q36),der(17;18)(q10;q10)[20] 5.6 51.7 46,XY,t(11;19)(q23;p13.1)[17]/ KMT2A-ELL, 47,idem,+8[1]/46,XY[2] STAG2 9.8 73.2 46,XY[20] NPM1 13.6 161.0 49,XX,+8,+10,+12[20] FUS-ERG 2.3 172.0

p.392-394 M delins LSSLSLSQV p.M469V M p.C416W F p.366_406del M

7.4 168.1 11.6 38.2 9.9 20.5

p.395_409del M p.366_409del M

15.1 54.2 12.3 159.3

Death

+ +

Death Alive

+ + + -

+ + -

+ + +

+ + + +

Alive Death Alive Death

Alive

+ +

Alive Death

Alive

+ +

Alive Death

Death

NPM1 KIT

+ +

+ + -

+ -

Death Death Alive

NRAS, NF1

+ +

Alive Alive

46,XY,t(9;11)(p22;q23)[16]/46,XY[4] KMT2A-MLLT3 47,XY,+8[20] 47,XX,+18[1]/46,XX[19] 46,XY,t(8;21)(q22;q22)[17]/ 45,X,-Y,t(8;21)(q22;q22)[3] 47,XY,inv(16)(p13.1q22),+22[20] 46,XY,inv(16)(p13q22)[20]

CR Relapse Event SCT Prognosis

UPN: unique patient number; WBC: white blood cell count; CR: complete remission; SCT: stem cell transplantation; N/A: not applicable; M: male; F: female.; y: years. *NCBI reference sequence; PTPN11, NM_002834; CBL, NM_005188.

studies have found inconsistent evidence of the clinical significance of NRAS mutations.8,9 In the present study, NRAS mutations were associated with favorable prognosis. This seemed attributable to the characteristics of patients with NRAS mutations, i.e., high frequency of CBFB-MYH11 with no other poor prognostic factors. 11q-UPD was detected in only one patient with a CBL missense mutation, which might be consistent with a previous study reporting that somatically acquired CBL deletions are frequently heterozygous, whereas most missense mutations are homozygous as a consequence of 11q-UPD.50 We summarized previous reports on CBL mutations in AML in the Online Supplementary Table S6. CBL mutations were previously shown to be associated with CBF-AML.13 In the present study, three of six patients with CBL mutations had CBF-AML; however, there was no significant association in this respect (Figure 1, Table 2). Owing to the low incidence of CBL mutation, its prognostic significance is not well characterized.10,12,13 Although we did not observe any significant prognostic haematologica | 2022; 107(3)

impact of CBL mutations in our cohort, all three patients without CBF experienced relapse and died (Table 2). These results might suggest that non-CBF patients with CBL mutation show poor prognosis. RAS pathway alterations are also a major cause of JMML; in addition, each of these alterations are of prognostic relevance in patients with JMML.51,52 In previous studies, JMML patients with PTPN11 and NF1 mutations showed significantly poor prognosis.51,52 On the other hand, JMML patients with NRAS mutations exhibited favorable outcomes.51,52 In our study, the prognostic impact of NF1, PTPN11, and NRAS was similar to that observed in JMML. However, we are unable to explain this similarity because the transformation of JMML to AML is rare.53 There may be some possible limitations in this study. First, we analyzed PTPN11, CBL, NRAS and KRAS mutations by Sanger sequencing because the mutation hotspots of these genes were well known. Although the frequency of these mutations was similar to the previous reports by 589

T. Kaburagi et al.

Table 3. Univariate and multivariate Cox regression analyses of overall survival and event-free survival.

Univariate analysis 95%CI Inferior Superior

P-value

Multivariate analysis 95%CI Inferior Superior

P-value

Overall survival NF1 PTPN11 CBL NRAS KRAS RUNX1-RUNX1T1 CBFB-MYH11 Monosomy 7 Complex karyotype FLT3-ITD 5q– FUS-ERG NUP98-NSD1 PRDM16 high expression

4.104 2.027 2.145 0.305 1.201 0.173 0.000 1.655 2.230 3.051 2.442 10.19 5.232 3.427

1.492 0.880 0.676 0.111 0.379 0.075 0.000 0.522 1.270 1.833 0.339 3.671 2.605 2.203

11.29 4.670 6.800 0.833 3.808 0.398 Inf 5.250 3.916 5.076 17.60 28.26 10.510 5.331

0.006 0.097 0.195 0.021 0.756 <0.001 0.995 0.392 0.005 <0.001 0.376 <0.001 <0.001 <0.001

4.109 0.694 2.617 0.309 2.064 0.250 0.000 2.617 1.812 1.853 1.627 6.007 2.941 1.921

1.471 0.260 0.794 0.112 0.618 0.106 0.000 0.781 0.991 0.985 0.207 2.096 1.366 1.165

11.48 1.851 8.630 0.849 6.892 0.590 Inf 8.775 3.312 3.486 12.82 17.22 6.331 3.168

0.007 0.466 0.114 0.023 0.239 0.002 0.995 0.119 0.054 0.056 0.644 0.001 0.006 0.010

Event-free survival NF1 PTPN11 CBL NRAS KRAS RUNX1-RUNX1T1 CBFB-MYH11 Monosomy 7 Complex karyotype FLT3-ITD 5q– FUS-ERG NUP98-NSD1 PRDM16 high expression

1.794 2.142 1.215 0.506 0.966 0.466 0.422 1.539 1.926 2.250 5.587 4.179 8.056 2.797

0.664 1.157 0.387 0.280 0.396 0.306 0.186 0.569 1.222 1.460 1.366 1.533 4.180 1.990

4.852 3.965 3.813 0.914 2.359 0.708 0.956 4.161 3.037 3.469 22.86 11.39 15.53 3.931

0.249 0.015 0.739 0.024 0.940 <0.001 0.039 0.395 0.005 <0.001 0.017 0.005 <0.001 <0.001

1.621 1.239 1.527 0.530 1.084 0.659 0.603 2.275 1.810 1.236 4.441 3.191 5.017 2.172

0.588 0.616 0.471 0.293 0.432 0.420 0.255 0.786 1.111 0.716 1.009 1.144 2.463 1.489

4.469 2.494 4.948 0.961 2.725 1.035 1.427 6.589 2.948 2.133 19.54 8.903 10.220 3.167

0.351 0.548 0.480 0.037 0.863 0.070 0.250 0.130 0.017 0.447 0.049 0.027 <0.001 <0.001

HR: hazard ratio; CI: confidence interval.

Sanger sequencing,6-9 it appears to be lower than that of the recent pediatric report by targeted deep sequencing.1 Next, there were a small number of patients harboring NF1 alterations. Further investigation is needed to determine the clinical significance of NF1 alterations in pediatric AML. Since there have been few reports on NF1 alterations, especially in pediatric AML (Online Supplementary Table S5), our results might be valuable for future analysis. Lastly, we could not analyze germline alterations because of the lack of non-hematopoietic cells. Congenital alterations of RAS pathway genes are known as RASopthies predisposing to hematological malignancies.54 Especially for NF1 and CBL, it is difficult to distinguish between somatic and germline mutations because the mutation hotspots overlap. While it is sometimes difficult to diagnose RASopathy because of minor clinical symptoms, patients with distinct clinical features of AML predisposing diseases, such as neurofibromatosis, Noonan syndrome, or CBL syndrome were excluded from the AML-05 trial according to its eligibility criteria. Also, we estimated that most of NF1 and CBL mutations might be somatic from online databases and previous reports. Since there have been few reports of detailed 590

analysis on NF1 and CBL alterations in pediatric AML (Online Supplementary Tables S5 and S6), further analyses are needed. In conclusion, NF1 alteration is possibly a poor prognostic factor and NRAS mutation is a favorable prognostic factor in pediatric patients with AML. Pediatric AML patients with PTPN11 mutations may show a greater tendency for relapse and induction failure. Detailed analysis of RAS pathway alterations may enable a more accurate prognostic stratification of pediatric AML and may provide novel therapeutic molecular targets related to this signal transduction pathway. Disclosures No conflicts of interest to disclose. Contributions TK, GY, NS, and YH designed and performed the research, analyzed the data, and wrote the paper; YHayashi designed the research, led the project, and wrote the paper; KY, YS, SM, and SO performed the research; KT performed the research and bioinformatics analysis; KO, MS, HA, HM, AS, TTabi, NK, DT, KH, haematologica | 2022; 107(3)

RAS pathway alterations in pediatric AML

TTaga and SA provided patient samples and data. All authors critically reviewed and revised the manuscript. Acknowledgements The authors thank Yuki Hoshino for her valuable assistance in performing the experiments. The authors would like to thank Enago (www. enago.jp) for the English language review. Funding This work was supported by a Grant-in-Aid for Scientific

References 1. Bolouri H, Farrar JE, Triche T Jr, et al. The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions. Nat Med. 2018;24(1):103112. 2. Fröhling S, Scholl C, Gilliland DG, Levine RL. Genetics of myeloid malignancies: Pathogenetic and clinical implications. J Clin Oncol. 2005;23(26):6285-6295. 3. Eisfeld AK, Kohlschmidt J, Mrózek K, et al. NF1 mutations are recurrent in adult acute myeloid leukemia and confer poor outcome. Leukemia. 2018;32(12):2536-2545. 4. Alfayez M, Issa GC, Patel KP, et al. The clinical impact of PTPN11 mutations in adults with acute myeloid leukemia. Leukemia. 2021;35(3):691-700. 5. Murakami N, Okuno Y, Yoshida K, et al. Integrated molecular profiling of juvenile myelomonocytic leukemia. Blood. 2018; 131(14):1576-1586. 6. Tartaglia M, Martinelli S, Iavarone I, et al. Somatic PTPN11 mutations in childhood acute myeloid leukaemia. Br J Haematol. 2005;129(3):333-339. 7. Loh ML, Reynolds MG, Vattikuti S, et al. PTPN11 mutations in pediatric patients with acute myeloid leukemia: results from the Children's Cancer Group. Leukemia. 2004;18(11):1831-1834. 8. Meshinchi S, Stirewalt DL, Alonzo TA, et al. Activating mutations of RTK/ras signal transduction pathway in pediatric acute myeloid leukemia. Blood. 2003;102(4): 1474-1479. 9. Sano H, Shimada A, Taki T, et al. RAS mutations are frequent in FAB type M4 and M5 of acute myeloid leukemia, and related to late relapse: a study of the Japanese Childhood AML Cooperative Study Group. Int J Hematol. 2012;95(5):509-515. 10. Coenen EA, Driessen EM, Zwaan CM, et al. CBL mutations do not frequently occur in paediatric acute myeloid leukaemia. Br J Haematol. 2012;159(5):577-584. 11. Balgobind BV, Van Vlierberghe P, van den Ouweland AM, et al. Leukemia-associated NF1 inactivation in patients with pediatric T-ALL and AML lacking evidence for neurofibromatosis. Blood. 2008;111(8):43224328. 12. Sargin B, Choudhary C, Crosetto N, et al. Flt3-dependent transformation by inactivating c-Cbl mutations in AML. Blood. 2007;110(3):1004-1012. 13. Abbas S, Rotmans G, Löwenberg B, Valk PJ. Exon 8 splice site mutations in the gene encoding the E3-ligase CBL are associated with core binding factor acute myeloid leukemias. Haematologica. 2008;93(10): 1595-1597 14. Parkin B, Ouillette P, Wang Y, et al. NF1

haematologica | 2022; 107(3)

Research on Innovative Areas from the Ministry of Health, Labor and Welfare of Japan (15H05909), a grant for project for development of innovative research on cancer therapeutics (PDIRECT) from the Japanese Agency for Medical Research and Development (AMED; JP16ck0106064, JP 19ck0106329), the Japanese Society for the Promotion of Science (KAKENHI grants 17K10130, 18H06234, 19K21333, 20K08744), a research grant from the Japanese Society of Hematology, and the Kawano Masanori Memorial Public Interest Incorporated Foundation for Promotion of Pediatrics.

inactivation in adult acute myelogenous leukemia. Clin Cancer Res. 2010;16(16): 4135-4137. 15. Boudry-Labis E, Roche-Lestienne C, Nibourel O, et al. Neurofibromatosis-1 gene deletions and mutations in de novo adult acute myeloid leukemia. Am J Hematol. 2013;88(4):306-311. 16. Haferlach C, Grossmann V, Kohlmann A, et al. Deletion of the tumor-suppressor gene NF1 occurs in 5% of myeloid malignancies and is accompanied by a mutation in the remaining allele in half of the cases. Leukemia. 2012;26(4):834-839. 17. Tartaglia M, Kalidas K, Shaw A, et al. PTPN11 mutations in Noonan syndrome: molecular spectrum, genotype-phenotype correlation, and phenotypic heterogeneity. Am J Hum Genet. 2002;70(6):1555-1563. 18. Sanada M, Suzuki T, Shih LY, et al. Gain-offunction of mutated C-CBL tumour suppressor in myeloid neoplasms. Nature. 2009;460(7257):904-908. 19. Matsuo H, Yoshida K, Fukumura K, et al. Recurrent CCND3 mutations in MLLrearranged acute myeloid leukemia. Blood Adv. 2018;2(21):2879-2889. 20. Shimada A, Taki T, Tabuchi K, et al. KIT mutations, and not FLT3 internal tandem duplication, are strongly associated with a poor prognosis in pediatric acute myeloid leukemia with t(8;21): a study of the Japanese Childhood AML Cooperative Study Group. Blood. 2006;107(5):1806-1809. 21. Döhner K, Schlenk RF, Habdank M, et al H. Mutant nucleophosmin (NPM1) predicts favorable prognosis in younger adults with acute myeloid leukemia and normal cytogenetics: interaction with other gene mutations. Blood. 2005;106(12):3740-3746. 22. Mizushima Y, Taki T, Shimada A, et al. Prognostic significance of the BAALC isoform pattern and CEBPA mutations in pediatric acute myeloid leukemia with normal karyotype: a study by the Japanese Childhood AML Cooperative Study Group. Int J Hematol. 2010;91(5):831-837 23. Sano H, Ohki K, Park MJ, et al. CSF3R and CALR mutations in paediatric myeloid disorders and the association of CSF3R mutations with translocations, including t(8; 21). Br J Haematol. 2015;170(3):391-397. 24. Sano H, Shimada A, Tabuchi K, et al. WT1 mutation in pediatric patients with acute myeloid leukemia: a report from the Japanese Childhood AML Cooperative Study Group. Int J Hematol. 2013;98(4):437-445 25. Yamato G, Shiba N, Yoshida K, et al. ASXL2 mutations are frequently found in pediatric AML patients with t(8;21)/ RUNX1-RUNX1T1 and associated with a better prognosis. Genes Chromosomes Cancer. 2017;56(5):382-393. 26. Damm F, Chesnais V, Nagata Y, et al.

BCOR and BCORL1 mutations in myelodysplastic syndromes and related disorders. Blood. 2013;122(18):3169-3177. 27. Thol F, Bollin R, Gehlhaar M, et al. Mutations in the cohesin complex in acute myeloid leukemia: clinical and prognostic implications. Blood. 2014;123(6):914-920. 28. Shiba N, Hasegawa D, Park MJ, et al. CBL mutation in chronic myelomonocytic leukemia secondary to familial platelet disorder with propensity to develop acute myeloid leukemia (FPD/AML). Blood. 2012;119(11):2612-2614. 29. Xu F, Taki T, Yang HW, et al. Tandem duplication of the FLT3 gene is found in acute lymphoblastic leukaemia as well as acute myeloid leukaemia but not in myelodysplastic syndrome or juvenile chronic myelogenous leukaemia in children. Br J Haematol. 1999;105(1):155-162. 30. Shiba N, Ichikawa H, Taki T, et al. NUP98NSD1 gene fusion and its related gene expression signature are strongly associated with a poor prognosis in pediatric acute myeloid leukemia. Genes Chromosomes Cancer. 2013;52(7):683-693. 31. Kong XT, Ida K, Ichikawa H, et al. Consistent detection of TLS/FUS-ERG chimeric transcripts in acute myeloid leukemia with t(16;21)(p11;q22) and identification of a novel transcript. Blood. 1997;90(3):1192-1199. 32. Balgobind BV, Hollink IH, Reinhardt D, et al. Low frequency of MLL-partial tandem duplications in paediatric acute myeloid leukaemia using MLPA as a novel DNA screenings technique. Eur J Cancer. 2010;46(10):1892-1899. 33. Jo A, Mitani S, Shiba N, et al. High expression of EVI1 and MEL1 is a compelling poor prognostic marker of pediatric AML. Leukemia. 2015;29(5):1076-1083. 34. Yoshizato T, Nannya Y, Atsuta Y, et al. Genetic abnormalities in myelodysplasia and secondary acute myeloid leukemia: impact on outcome of stem cell transplantation. Blood. 2017;129(17):2347-2358. 35. Kanda, Y. Investigation of the freely available easy-to-use software 'EZR' for medical statistics. Bone Marrow Transplant. 2013; 48(3):452-458. 36. Smith KS, Yadav VK, Pei S, et al. SomVarIUS: somatic variant identification from unpaired tissue samples. Bioinformatics. 2016;32(6):808-813. 37. 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. 38. Giannakis M, Mu XJ, Shukla SA, Genomic correlates of immune-cell Infiltrates in colorectal carcinoma. Cell Rep. 2016;15(4): 857865. 39. Tartaglia M, Niemeyer CM, Fragale A, et al. Somatic mutations in PTPN11 in juvenile

591

T. Kaburagi et al. myelomonocytic leukemia, myelodysplastic syndromes and acute myeloid leukemia. Nat Genet. 2003;34(2):148-150. 40. Tartaglia M, Martinelli S, Cazzaniga G, et al. Genetic evidence for lineage-related and differentiation stage-related contribution of somatic PTPN11 mutations to leukemogenesis in childhood acute leukemia. Blood. 2004;104(2):307-13. 41. Bentires-Alj M, Paez JG, David FS, et al. Activating mutations of the noonan syndrome-associated SHP2/PTPN11 gene in human solid tumors and adult acute myelogenous leukemia. Cancer Res. 2004;64(24): 8816-8820. 42. Martinelli S, Stellacci E, Pannone L, et al. Molecular diversity and associated phenotypic spectrum of germline CBL mutations. Hum Mutat. 2015;36(8):787-796. 43. Bos JL. ras oncogenes in human cancer: a review. Cancer Res. 1989;49(17):4682-4689. 44. Shiba N, Ohki K, Kobayashi T, et al. High PRDM16 expression identifies a prognostic subgroup of pediatric acute myeloid

592

leukaemia correlated to FLT3-ITD, KMT2APTD, and NUP98-NSD1: the results of the Japanese Paediatric Leukaemia/Lymphoma Study Group AML-05 trial. Br J Haematol. 2016;172(4):581-591. 45. Goemans BF, Zwaan CM, Martinelli S, et al. Differences in the prevalence of PTPN11 mutations in FAB M5 paediatric acute myeloid leukaemia. Br J Haematol. 2005;130(5):801-803 46. 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. 47. Yamato G, Shiba N, Yoshida K, et al. RUNX1 mutations in pediatric acute myeloid leukemia are associated with distinct genetic features and an inferior prognosis. Blood. 2018;131(20):2266-2270. 48. Barjesteh van Waalwijk van DoornKhosrovani S, Erpelinck C, van Putten WL, et al. High EVI1 expression predicts poor survival in acute myeloid leukemia: a study

of 319 de novo AML patients. Blood. 2003;101(3):837-845. 49. Goemans BF, Zwaan CM, Miller M, et al. Mutations in KIT and RAS are frequent events in pediatric core-binding factor acute myeloid leukemia. Leukemia. 2005;19(9): 1536-1542. 50. Kales SC, Ryan PE, Nau MM, Lipkowitz S. Cbl and human myeloid neoplasms: the Cbl oncogene comes of age. Cancer Res. 2010;70(12):4789-4794. 51. Locatelli F, Niemeyer CM. How I treat juvenile myelomonocytic leukemia. Blood. 2015;125(7):1083-1090. 52. Niemeyer CM, Flotho C. Juvenile myelomonocytic leukemia: who's the driver at the wheel? Blood. 2019;133(10):10601070. 53. Luna-Fineman S, Shannon KM, Atwater SK, et al. Myelodysplastic and myeloproliferative disorders of childhood: a study of 167 patients. Blood. 1999;93(2):459-466. 54. Niemeyer CM. RAS diseases in children. Haematologica. 2014;99(11):1653-1662.

haematologica | 2022; 107(3)

ARTICLE

Chronic Lymphocytic Leukemia

Chromosome banding analysis and genomic microarrays are both useful but not equivalent methods for genomic complexity risk stratification in chronic lymphocytic leukemia patients Silvia Ramos-Campoy,1,2* Anna Puiggros,1,2* Sílvia Beà,3 Sandrine Bougeon,4 María José Larráyoz,5 Dolors Costa,3 Helen Parker,6 Gian Matteo Rigolin,7 Margarita Ortega,8 María Laura Blanco,9 Rosa Collado,10 Rocío Salgado,11 Tycho Baumann,3 Eva Gimeno,1,12 Carolina Moreno,9 Francesc Bosch,8 Xavier Calvo,1,2 María José Calasanz,5 Antonio Cuneo,7 Jonathan C. Strefford,6 Florence Nguyen-Khac,13 David Oscier,14 Claudia Haferlach,15 Jacqueline Schoumans4 and Blanca Espinet1,2

Ferrata Storti Foundation

Haematologica 2022 Volume 107(3):593-603

Molecular Cytogenetics Laboratory, Pathology Department, Hospital del Mar, Barcelona, Spain; 2Translational Research on Hematological Neoplasms Group, Cancer Research Program, Institut Hospital del Mar d’Investigacions Mèdiques (IMIM), Barcelona, Spain; 3Hematopathology Unit, Hospital Clínic, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), CIBERONC, Barcelona, Spain; 4 Oncogenomic Laboratory, Hematology Service, Lausanne University Hospital, Lausanne, Switzerland; 5Cytogenetics and Hematological Genetics Services, Department of Genetics, University of Navarra, Pamplona, Spain; 6Cancer Sciences, Faculty of Medicine, University of Southampton, Southampton, UK; 7Hematology Section, St. Anna University Hospital, Ferrara, Italy; 8Department of Hematology, University Hospital Vall d'Hebron, Barcelona, Spain; 9Department of Hematology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain; 10Department of Hematology, Consorcio Hospital General Universitario, Valencia, Spain; 11Cytogenetics Laboratory, Hematology Department, Fundación Jiménez Díaz, Madrid, Spain; 12Applied Clinical Research in Hematological Malignances, Cancer Research Program, IMIM-Hospital del Mar, Barcelona, Spain; 13Hematology Department and Sorbonne Université, Hôpital Pitié-Salpêtrière, APHP, INSERM U1138, Paris, France; 14Department of Molecular Pathology, Royal Bournemouth Hospital, Bournemouth, UK and 15MLL Munich Leukemia Laboratory, Munich, Germany 1

*SR-C and AP contributed equally as co-first authors.

ABSTRACT

enome complexity has been associated with poor outcome in patients with chronic lymphocytic leukemia (CLL). Previous cooperative studies established five abnormalities as the cut-off that best predicts an adverse evolution by chromosome banding analysis (CBA) and genomic microarrays (GM). However, data comparing risk stratification by both methods are scarce. Herein, we assessed a cohort of 340 untreated CLL patients highly enriched in cases with complex karyotype (CK) (46.5%) with parallel CBA and GM studies. Abnormalities found by both techniques were compared. Prognostic stratification in three risk groups based on genomic complexity (0-2, 34 and ≥5 abnormalities) was also analyzed. No significant differences in the percentage of patients in each group were detected, but only a moderate agreement was observed between methods when focusing on individual cases (κ=0.507; P<0.001). Discordant classification was obtained in 100 patients (29.4%), including 3% classified in opposite risk groups. Most discrepancies were technique-dependent and no greater correlation in the number of abnormalities was achieved when different filtering strategies were applied for GM. Nonetheless, both methods showed a similar concordance index for prediction of time to first treatment (TTFT) (CBA: 0.67 vs. GM: 0.65) and overall survival (CBA: 0.55 vs. GM: 0.57). High complexity maintained its significance in the multivariate analysis for TTFT including TP53 and IGHV status when defined by CBA (hazard ratio [HR] 3.23; P<0.001) and GM (HR haematologica | 2022; 107(3)

Correspondence: BLANCA ESPINET bespinet@parcdesalutmar.cat ANNA PUIGGROS apuiggros@imim.es Received: November 27, 2020. Accepted: February 26, 2021. Pre-published: March 11, 2021. https://doi.org/10.3324/haematol.2020.274456

593

S. Ramos-Campoy et al.

2.74; P<0.001). Our findings suggest that both methods are useful but not equivalent for risk stratification of CLL patients. Validation studies are needed to establish the prognostic value of genome complexity based on GM data in future prospective studies.

Introduction Deletions of 17p13 region and/or mutations in TP53 as well as the mutational status of the variable region of the immunoglobulin heavy chain (IGHV) gene constitute the most important prognostic and predictive factors in chronic lymphocytic leukemia (CLL) in the era of chemoimmunotherapy.1 However, several studies have highlighted the independent clinical significance of genomic complexity, mainly defined by the detection of complex karyotypes (CK) by chromosome banding analysis (CBA), due to its association with an unfavorable clinical outcome. This has been demonstrated in patients treated not only with standard chemoimmunotherapy regimes2-5 but also in the initial clinical trials with the novel mechanism-based therapeutic agents such as ibrutinib or venetoclax.6-8 Early studies in CLL defined CK as the presence of at least three numerical and/or structural chromosomal abnormalities in the same cell clone detected by CBA.9,10 Of note, the increasing number of chromosomal abnormalities in the karyotype has been correlated with the worsening of clinical evolution of CLL patients.11,12 In this context, a large retrospective study from the European Research Initiative on CLL (ERIC) has reported that patients with five or more abnormalities (the so-called high-CK) display an adverse outcome independently of other known biomarkers such as TP53 abnormalities or the IGHV mutational status.5 On the other hand, it has been demonstrated that CK might have a different clinical impact in CLL patients according to not only the number, but also the type of aberrations detected in the karyotype. In this regard, it has been described that patients with CK carrying +12, +19 display a particularly favorable outcome while the presence of unbalanced rearrangements define a subset with very aggressive disease.13-15 Even though CBA has been the gold standard method to identify CK, in the last decade genomic microarrays (GM) have emerged as a valuable tool for genome-wide screening in CLL.16-20 Indeed, some studies have correlated the genomic complexity detected by GM to progressive disease and poorer response rates to treatment.21-23 Nonetheless, although some European countries have replaced conventional techniques by GM, standard criteria to analyze and define genomic complexity by GM are still needed. According to the published guidelines for GM analysis in acquired hematologic neoplasms, recurrent aberrations with known clinical relevance in the disease irrespective of their size as well as other copy number abnormalities (CNA) ≥5 Mb should be considered in order to reduce the detection of benign constitutional variants and avoid the reporting of anomalies with uncertain clinical significance.24 However, it remains unclear whether this threshold is the optimal to analyze CLL patients or whether potentially relevant chromosomal imbalances are disregarded by applying this highly conservative size cut-off. Besides, another multicenter study conducted by ERIC suggested that CLL patients could be divided into three distinct prognostic subgroups based on the number of CNA.25 According to Leeksma et al., the so594

called high genomic complexity (high-GC) subgroup, which is defined by carrying ≥5 CNA, emerged as prognostically adverse, independently of other biomarkers. Nevertheless, to the best of our knowledge, the comparison of genomic complexity for risk stratification using CBA and GM in parallel has not been performed in a large CLL cohort. In the present multicenter retrospective study we aimed to compare the usefulness of CBA and GM techniques in a series of 340 CLL patients with and without CK to determine both their concordance and their equivalence in the prognostic stratification of CLL patients with CK. Moreover, we have analyzed the detected aberrations using different counting strategies to ascertain whether other parameters, such as the type of the aberrations or their size, might have an influence on the risk stratification of CLL patients.

Methods Patient cohort A total of 340 previously untreated CLL (n=327; 96.2%) and monoclonal B-cell lymphocytosis (n=13; 3.8%) patients from 18 European institutions were included. All had CBA results at diagnosis or before treatment. GM data were already available or obtained from DNA extracted within 1 year. Analyses were performed on peripheral blood (PB) (n=286) or bone marrow (BM) (n=54). Due to the purpose of the study, this cohort was enriched in patients with CK (n=158; 46.5%). Demographic, clinical and biological characteristics are summarized in Table 1. The study was carried out in accordance with national and international guidelines (Professional Code of Conduct, Declaration of Helsinki) and approved by the Hospital del Mar Ethics Committee (2017/7565/I).

Chromosome banding analyses CBA was performed on G- or R-banded chromosomes. Karyotypes were described according to the International System for Human Cytogenetic Nomenclature (ISCN 2016).26 A complex karyotype was defined as the presence of three or more numerical and/or structural chromosomal abnormalities (abn.) detected in the same cell clone. Patients were stratified in three categories: non-CK (0-2 abn.), low/intermediate-CK (3-4 abn.) and high-CK (≥5 abn.)5

Genomic microarray analyses Microarray platforms used are summarized in the Online Supplementary Table S1. All aberrations found irrespective of size were collected, although non-classical CLL abnormalities (other than gain of chromosome 12 and losses of 11q, 13q and 17p) were filtered in the CNA count for prognostic stratification following the 5 Mb cut-off size recommended.24 Three subgroups were defined according to genomic complexity (GC): low-GC (0-2 CNA), intermediate-GC (3-4 CNA) and high-GC (≥5 CNA).25 This strategy was compared with other CNA counting methodologies, such as the inclusion of smaller abnormalities (no size filter or 1 Mb as cut-off) or counting as a unique CNA small contiguous abnormalities (with a distance ≤1 Mb between them) or those included in a chromothripsis event.

haematologica | 2022; 107(3)

Genome complexity in CLL: karyotype versus microarrays

Statistical analyses Descriptive statistics were used to provide frequency distributions of discrete variables while statistical measures were used to provide median values and ranges for quantitative variables. Groups were compared using Chi-square or Fisher exact tests for discrete variables and Mann-Whitney U test for continuous variables. The Kappa coefficient was used for assessing the agreement in patients categorization among techniques. Survival analysis was restricted to 259 patients. A total of 81 non-CK cases from three institutions were excluded as CBA was performed at recruitment for clinical trials, introducing a bias in the results. Time to first treatment (TTFT), the end point of the study, was calculated from the date of cytogenetic study to the date of first treatment or last follow-up while overall survival (OS) was defined from date of cytogenetic study until last follow-up or death. Kaplan–Meier method was used to estimate the distribution of TTFT and OS. Comparisons among patient subgroups were performed with the log-rank test. The concordance statistic (C-index) was calculated to assess the accuracy of CBA and GM for predicting TTFT and OS. Multivariate analysis using Cox proportional hazards regression model was used to assess the maintenance of the independent prognostic impact on TTFT and OS. Statistical analyses were performed using SPSS v.23 software (SPSS Inc, Chicago, IL, USA) and R v3.5.2. P-values <0.05 were considered statistically significant. Additional information regarding the methodology of the study is detailed in the Online Supplementary Appendix.

Results Number and type of abnormalities detected by chromosome banding analysis and genomic microarrays Regarding CBA, 270 of 340 (79.4%) patients showed an abnormal karyotype. Overall, 182 were considered non-CK (0-2 abn.) while 158 displayed a CK (≥3 abn.). The vast majority of non-CK aberrant cases carried only one aberration (75 of 112; 66.9%), while the median number of abnormalities among CK patients was four (range, 3-19). Abnormal karyotypes from the non-CK group mainly included known recurrent CLL aberrations, the most frequent being trisomy of chromosome 12 (15.4% patients). In contrast, the CK group showed a wide variety of abnormalities affecting all chromosomes and included unbalanced structural aberrations (552 of 823; 67.1%), complex abnormalities affecting material of unknown origin (179 of 823; 21.7%) and monosomies (155 of 823; 18.8%). In seven of these, a co-existence of +12 and +19, associated with more indolent course, was found (4.4%). Balanced translocations, potentially missed when studied by GM, were present in only 57 patients (11.5% non-CK; 22.8% CK). Even though they were detected in a minority of cases, 13q14 and 14q32 loci were recurrently involved (in 13 and seven patients, respectively). GM had the highest detection rate of abnormalities, with 309 of 340 (90.9%) cases carrying at least one CNA when all the abnormalities, irrespective of their size, were considered. No significant differences were observed among the GM platforms used. Expectedly, the non-CK group showed a significantly lower median number of CNA compared with CK patients (2 CNA; range, 0-10 vs. 6 CNA; range, 051; P<0.001). Nearly half of the patients carried at least one small (<5 Mb) non-classical CLL CNA (median size 5.38 Mb; range, 0.019-198 Mb) that would not be taken into haematologica | 2022; 107(3)

consideration following the current microarray recommendations (35.7% non-CK vs. 64.6% CK patients; P<0.001). Although they were less frequent, similar results were observed regarding the presence of non-classical CLL CNA below 1 Mb (26.9% vs. 44.9%; P<0.001). Patterns suggestive of chromothripsis or chromothripsis-like were identified in 30 patients (20 and ten cases, respectively). Fluorescence in situ hybridization (FISH), the gold standard method for the detection of the four genetic abnormalities included in the Döhner et al. prognostic hierarchical model,27 confirmed the higher incidence of high-risk aberrations in the CK group. Specifically, del(11q) was found in 12.4% (22 of 177) of non-CK patients and 32.2% (49 of 152) of CK patients (P<0.001) while del(17p) was present in 4.5% (eight of 177) and 40.1% (61 of 152), respectively (P<0.001). Detection of del(13q), del(11q) and del(17p) was lower by CBA compared to FISH although these loci were involved in different type of abnormalities (Online Supplementary Table S2). GM showed a high concordance with FISH results (Online Supplementary Table S3). With regard to commonly detected non-classical CLL abnormalities, similar results were observed by both CBA and GM among non-CK and CK patients. The only recurrent aberrations detected by CBA within the non-CK group were deletions in the long arm of chromosome 14 (6.6%) and unbalanced translocations affecting 2p arm (5.5%), which were detected as losses at 14q and gains of 2p region by GM, respectively. Likewise, despite being distributed Table 1. Baseline characteristics of patients at diagnosis and last follow-up.

NON-CK GROUP n = 182; n (%)

CK GROUP P-value n = 158; n (%)

Sex Men 118 (64.8%) 113 (71.5%) Median age at diagnosis 66 years (29-89) 68 years (33-96) Stage at diagnosis MBL 11 (6.0%) 2 (1.3%) CLL 171 (94.0%) 156 (98.7%) Binet A 117/159 (73.6%) 80/136 (58.8%) Binet B/C 42/159 (26.4%) 56/136 (41.2%) Common CLL genomic aberrations* del(13)(q14) 103/182 (56.6%) 96/158 (60.8%) isolated del(13)(q14) 70/103 (67.9%) 25/96 (26.0%) Trisomy 12 29/182 (15.9%) 27/158 (17.1%) del(11)(q22q23) 25/182 (13.7%) 49/158 (31.0%) Aberrations in TP53** 21/164 (12.8%) 70/156 (44.9%) del(17)(p13) 8/182 (4.4%) 65/158 (41.1%) TP53 mutation 15/161 (9.3%) 45/147 (30.6%) Unmutated IGHV 80/169 (47.3%) 92/138 (66.7%) Median follow-up (range) 68 months (0-261) 29 months (0-160) Time from diagnosis to 3.5 months (0-242) 0 months (0-298) cytogenetic study Treatment Treated patients 32/101 (31.7%) 103/151 (68.2%) Median time to first NR 13 months treatment (95% CI) (8-18) Survival Median overall survival 102 months 81 months (95% CI) (82-121) (58-103)

0.115 0.056 0.024 0.009

0.437 <0.001 0.775 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001

<0.001 <0.001

0.367

*Deletions and trisomy detected by fluorescence in situ hybridization (FISH) and/or genomic microarrays. **Cases in which TP53 mutation screening was not performed and FISH and/or genomic microarrays were negative for deletion were not considered. CLL: chronic lymphocytic leukemia; CK: complex karyotype; MBL: monoclonal B-cell lymphocytosis; CI: confidence interval, NR: not reached.

595

S. Ramos-Campoy et al.

Table 2. Recurrent copy number abnormalities found by genomic microarrays within the non-complex karyotype and complex karyotype subgroups and minimal common altered regions.

Non-CK CNA Gain 2p Loss 14q

n (%)

Cytogenetic bands

Minimal deleted/amplified region (GRCh37/hg19)

16 (8.8) 13 (7.1)

p25.3-p21 q24.1-q32.11

chr2: 29,477 - 45,859,076 chr14: 69,272,718 - 91,882,259

n (%)

Cytogenetic bands

Minimal deleted/amplified region (GRCh37/hg19)

p24.3-p23.1 p22.3-p15 p21.31-p21.31 q26.1-q29 p16.2-p15.2 q16.3-q21 p23.1-p22 q24.21-q24-21 q24.2-q24.3 q15.1-q15.1 q22.31-q26.3 q22-q25.1 p11.31-p11.23 q13.41-q13.42

chr2: 15,664,402 - 30,125,169 chr2: 32,877,675 - 62,206,329 chr3: 47,084,224 - 48,321,854 chr3: 165,375,394 - 196,284,424 chr4: 5,481,786 - 25,640,042 chr6: 103,468,966 - 112,256,460 chr8: 12,617,155 - 15,933,687 chr8: 128,286,744 - 130,380,043 chr14: 70,711,555 - 77,202,084 chr15: 41,755,587 - 42,090,500 chr15: 66,265,674 - 99,711,975 chr17: 56,560,919 - 71,135,799 chr18: 4,853,926 - 7,717,988 chr19: 51,943,080 - 54,499,334

CK CNA Gain 2p Gain 2p Loss 3p Gain 3q Loss 4p Loss 6q Loss 8p Gain 8q Loss 14q Loss 15q Gain 15q Gain 17q Loss 18p Gain 19q

39 (24.7) 13 (8.2) 11 (6.9) 15 (9.5) 15 (9.5) 16 (10.1) 17 (10.8) 13 (8.2) 16 (10.1) 11 (6.9) 12 (7.6) 24 (15.2) 12 (7.6)

In non-complex karyotype (non-CK) group, aberrations were considered recurrent if present in at least ten patients while in CK group, recurrence was set at 10 and 15 patients for gains and losses, respectively. CNA: copy number abnormality.

Table 3. Classification of patients in the previously suggested risk categories according to the number of aberrations detected by chromosome banding analysis and genomic microarrays.

Non-CK (0-2 abn.) GENOMIC MICROARRAYS

Low-GC (0-2 CNA) Intermediate-GC (3-4 CNA) High-GC (≥5 CNA) Total

CHROMOSOME BANDING ANALYSIS Low / Intermediate-CK High-CK (3-4 abn.) (≥5 abn.)

Total

157

192 (56.5%)

83 (24.4%)

65 (19.1%)

182 (53.5%)

81 (23.8%)

77 (22.6%)

340

A moderate agreement was observed between methods (κ=0.507;P<0.001). CK: complex karyotype; GC: genomic complexity; CNA: copy number abnormality; abn.: abnormalities.

along the genome, gains of chromosome arms 2p, 3q, 8q, 15q, 17q and 19q and losses at 3p, 4p, 6q, 8p, 14q, 15q and 18p, usually involved in unbalanced translocations or simple deletions in the karyotype, were the most recurrent CNA detected in CK patients (Online Supplementary Figure S1). Detailed information regarding recurrent CNA found by GM is shown in Table 2.

Risk stratification of the genomic complexity observed by chromosome banding analysis and genomic microarrays In order to compare the concordance among risk stratification based on CBA and GM data, patients were classified into those categories suggested by previous large-scale studies.5,25 Notably, both techniques did not significantly differ in the percentage of patients classified into intermediate-risk categories (3-4 abnormalities; 23.8% by CBA vs. 596

24.4% by GM; P=0.923) or those showing the highest risk (≥5 abnormalities; 22.6% and 19.1%, respectively; P=0.299). However, when focusing in individual cases, only a moderate agreement was observed between methods (κ=0.507; P<0.001). Discordant classification was obtained in 100 patients (29.4%), including eight cases with ≥5 abnormalities in the karyotype which would not be considered complex by GM and two patients with high-GC who did not have CK (2.9%) (Table 3). Next, we evaluated if the CNA filtering strategy used for GM results could underlie the differences observed in the assessment of the complexity. Regardless of the CNA filtering strategy, the proportion of patients with CNA <5 Mb was similarly represented among those patients with increased complexity scored by CBA (n=58) or by GM (n=42) (55.2% vs. 64.3%, respectively; P=0.413). When less strict filtering strategies were applied for GM, no greater haematologica | 2022; 107(3)

Genome complexity in CLL: karyotype versus microarrays

B Figure 1. Distribution of the number of copy number abnormalities detected by genomic microarrays among the groups identified by chromosome banding analysis. Patients were divided according to the risk groups defined by chromosome banding analysis (CBA) in noncomplex karyotype (non-CK) (0-2 abnormalities [abn.]), low/intermediate-CK (3-4 abn.) or highCK (≥5 abn.). Each plot represents copy number abnormality (CNA) counts found when non-classical chronic lymphocytic leukemia (CLL) abnormalities were filtered by different strategies: (A) Current recommended criteria for genomic microarrays (GM) analysis (cut-off size: ≥5 Mb); (B) considering all the CNA irrespectively of their size; (C) considering only those CNA ≥1 Mb; (D) filtering by 1 Mb cut-off and grouping small contiguous abnormalities or considering those CNA included in a chromothripsis event as a unique CNA. Spearman correlation coefficient between CBA and GM counts is shown for each GM analysis.

correlation in the number of abnormalities counted by CBA and GM was achieved. Similar results were observed when including all the abnormalities irrespective of their size, using 1 Mb as a cut-off for non-classical CLL CNA, or if CNA separated by <1 Mb or chromothripsis patterns were counted as one event to evaluate the effect of joining consecutive aberrations (Figure 1). Parallel analyses of the abnormalities detected by CBA and GM were also performed to identify other potential causes of discrepancy. Among those abnormalities recorded only by CBA, differences were mainly due to the presence of balanced translocations (n=28 patients) or abnormalities haematologica | 2022; 107(3)

represented in a minor proportion of tumor cells probably expanded during the cytogenetics culture which were missed by GM (n=40 patients). Moreover, FISH with chromosome painting probes performed in two high-CK cases by CBA, who showed low-GC, revealed that some apparently unbalanced abnormalities were complex balanced rearrangements that ultimately did not lead to loss of material (Online Supplementary Figure S2). On the other hand, when assessed by GM, most of the more complex cases showed aberrations <10 Mb, which is the resolution threshold of CBA, or multiple CNA that corresponded to complex rearrangements recorded as single abnormalities in 597

S. Ramos-Campoy et al.

Figure 2. Kaplan-Meier plots for time to first treatment and overall survival based on genomic complexity stratification assessed by chromosome banding analysis and genomic microarrays. Kaplan-Meier estimation for time to first treatment (TTFT) (A) and overall survival (OS) (B) in patients classified in each category based on total number of aberrations found by chromosome banding analysis (CBA): non-complex karyotype (non-CK) (0-2 abnormalities [abn.]), low/intermediate-CK (3-4 abn.) or high-CK (≥5 abn.) (plots on the left) and based on total number of copy number aberrations (CNA) detected by genomic microarrays (GM): low-genomic complexity (GC) (0-2 CNA), intermediate-GC (3-4 CNA) or high-GC (≥5 CNA)] (plots on the right).

the karyotype (73 and 19 cases, respectively). No division of the tumor clone during the cytogenetics culture is the most feasible explanation for 50 patients who carried CNA ≥10 Mb that should have been identified by CBA, of which 17 presented a normal karyotype. Detailed comparison for the ten patients who only displayed high complexity by one method is shown in the Online Supplementary Table S4. The genetic analysis using both methods allowed the correction of the karyotype in six patients after GM interpretation (Online Supplementary Table S5). Although it resulted in a change of the number of abnormalities recorded by CBA for three of them, initial counts were considered for the present analysis.

Prognostic impact of complex karyotype stratification by chromosome banding analysis and genomic microarrays As previously stated in the ERIC studies, significant differences in terms of TTFT were observed within the three risk groups according to the number of aberrations found by CBA and GM.5,25 Whereas the highest risk group defined by both techniques displayed a similar short median TTFT 598

(5 and 3 months by CBA and GM, respectively), TTFT was shorter for the intermediate risk group when defined by CBA (18 months vs. 35 months) (Figure 2A). Indeed, both methods showed a similar accuracy for predicting TTFT (Cindex: 0.67 by CBA vs. 0.65 by GM). With regard to OS, only the highest risk groups defined by each technique displayed a poorer outcome (68 months in both cases) (Figure 2B) although differences were only statistically significant in GM defined groups. Equivalent C-indexes were obtained for OS (0.55 by CBA vs. 0.57 by GM). In order to compare the discriminatory power for outcome prediction of both techniques, patients were first classified according to CBA to assess TTFT of GM-defined groups within each category. Of note, those non-CK and low/intermediate-CK patients by CBA who carried ≥5 CNA (high-GC) showed a poor outcome equivalent to that observed in the high-CK by CBA (median TTFT: 2 and 1 months, respectively). However, within the high-CK group, low-GC patients did not show a better evolution (TTFT: 2 months) while cases with intermediate-GC displayed an unexpected median TTFT of 22 months (Figure 3A). When these analyses were performed in the reverse order, CBA haematologica | 2022; 107(3)

Genome complexity in CLL: karyotype versus microarrays

Figure 3. Kaplan-Meier plots for time to first treatment of the genomic risk stratification within each category defined by the alternative technique. (A) Patients classified in each category based on total number of aberrations found by chromosome banding analysis (CBA): non-complex Karyotype (CK) (0-2 abnormalities [abn.]), low/intermediate-CK (3-4 abn.) or high-CK (≥5 abn.) are represented in different plots. Time to first treatment (TTFT) of genomic microarrays (GM) defined groups was assessed. Within non-CK and low/intermediate-CK, cases classified as high-GC (≥5 copy number abnormalities [CNA] by GM) showed a poor outcome. In the high-CK group, those low-GC patients did not display a better evolution while intermediate-GC cases showed an unexpected median TTFT of 22 months. (B) Each plot represents patients classified in each category based on total number of CNA detected by GM: low-GC (0-2 CNA), intermediate-GC (3-4 CNA) or high-GC (≥5 CNA). Within each subgroup, TTFT of CBA defined groups was assessed. Low-GC patients could be stratified in three risk categories when reclassified by CBA, while no significant differences were observed when intermediate-GC and high-GC subsets were reclassified.

reclassification within the low-GC patients allowed the distinction of three risk categories showing similar outcomes to those observed when applied to the global cohort (P<0.001). No significant differences were observed when the intermediate-GC and high-GC categories were reclassified (Figure 3B). It is noteworthy that the ten cases categorized in opposite risk groups displayed the poor prognosis haematologica | 2022; 107(3)

predicted by the technique that classified them in the higher risk category. Expectedly, the frequency of TP53 abnormalities (deletions and/or mutations) increased together with the complexity by both methods. In contrast, intermediate and high risk categories showed a similar increased proportion of unmutated IGHV (U-IGHV) and del(11q) compared 599

S. Ramos-Campoy et al.

Table 4. Univariate and multivariate analysis for time to first treatment.

Variable

CBA low/intermediate-CK vs. non-CK high-CK vs. non-CK GM intermediate-GC vs. low-GC high-GC vs. low-GC Del/mut TP53 U-IGHV del(11)(q22q23)

Univariate analysis Median TTFT P-value in months (95% CI)

Multivariate analysis for CBA Hazard ratio P-value (95% CI)

Multivariate analysis for GM Hazard ratio P-value (95% CI)

18 (11-25) vs. NR 5 (1-9) vs. NR

<0.001 <0.001

2.54 (1.47-4.41) 3.23 (1.81-5.76)

<0.001 <0.001

35 (0-74) vs. NR 3 (0-6) vs. NR 4 (0-9) 12 (4-20) 17 (9-25)

0.022 <0.001 <0.001 <0.001 0.111

1.72 (1.14-2.60) 1.71 (1.12-2.61) NA

0.010 0.012 NA

1.24 (0.76-2.04) 2.74 (1.61-4.67) 1.44 (0.92-2.26) 2.12 (1.39-3.22) NA

0.395 <0.001 0.109 <0.001 NA

CBA: chromosome banding analysis; CK: complex karyotype; non-CK: 0-2 abnormalities detected by CBA; low/intermediate-CK: 3-4 abnormalities; high-CK: ≥5 abnormalities; GM: genomic microarrays; GC: genomic complexity; low-GC: 0-2 copy number abnormalities (CNA) detected by genomic microarrays; intermediate-GC: 3-4 CNA; high-GC: ≥5 CNA; del/mut TP53: abnormalities in TP53 include deletion in 17p13 and/or mutation in TP53 gene; U-IGHV: CLL with unmutated IGHV; NR: not reached; NA: not assessed.

Table 5. Univariate and multivariate analysis for overall survival.

Variable

CBA low/intermediate-CK vs. non-CK high-CK vs. non-CK GM intermediate-GC vs. low-GC high-GC vs. low-GC Del/mut TP53 U-IGHV del(11)(q22q23)

Univariate analysis Median OS in months P-value (95% CI)

Multivariate analysis for GM* Hazard ratio P-value (95% CI)

114 (65-163) vs. 102 (83-121) 68 (25-111) vs. 102 (83-121)

0.729 0.133

114 (64-164) vs. 103 (55-151) 68 (32-104) vs. 103 (55-151) 50 (29-71) 79 (58-100) 79 (53-105)

0.741 0.003 <0.001 0.008 0.255

0.69 (0.36-1.34) 1.51 (0.76-3.01) 1.89 (1.05-3.42) 1.97 (1.15-3.36) NA

0.275 0.244 0.034 0.013 NA

*Multivariate analysis for CBA-defined risk categories was not performed due to the lack of statistical significance in the univariate analysis. OS: overall survival; CBA: chromosome banding analysis; CK: complex karyotype; non-CK: 0-2 abnormalities detected by CBA; low/intermediate-CK: 3-4 abnormalities; high-CK: ≥5 abnormalities; GM: genomic microarrays; GC: genomic complexity; low-GC: 0-2 copy number abnormalities (CNA) detected by genomic microarrays; intermediate-GC: 3-4 CNA; high-GC: ≥5 CNA; del/mut TP53: abnormalities in TP53 include deletion in 17p13 and/or mutation in TP53 gene; U-IGHV: CLL with unmutated IGHV; NR: not reached; NA: not assessed.

with non-CK/low-GC patients (Online Supplementary Table S6). Despite being highly associated with these known prognostic factors, three groups with significant differences on TTFT could be established by CBA and GM when patients were categorized depending on their TP53 status (Online Supplementary Figure S3). Regarding IGHV status, similar results were obtained within the mutated IGHV (M-IGHV) group while no clear discrimination was observed in the U-IGHV subset (Online Supplementary Figure S4). No prognostic impact was observed for del(11q) in the entire cohort (Table 4). High complexity defined by both CBA and GM maintained its significance when a multivariate analysis for TTFT including TP53 and IGHV status was performed. Conversely, genomic complexity by GM lost its significance in the multivariate analysis for OS (Table 5). Finally, the impact of other genetic findings was also analyzed. In this regard, the presence of unbalanced rearrangements was associated with shorter TTFT in the entire cohort (11 months vs. NR; P<0.001) and within the non-CK subgroup (10 months vs. NR; P=0.001) (Online Supplementary Figure S5). A negative impact was also observed for chromothripsis (2 months vs. 37 months; P<0.001), which was mainly found among CK patients (29 of 30). Indeed, tendency to this worse evolution was maintained within this subset (5 months vs. 15 months; P=0.062) (Online Supplementary Figure S6). As expected, these cases showed a high frequency of abnormalities in 600

TP53 (22 of 30; 73.3%) and U-IGHV (21 of 29; 72.4%). In the multivariate analysis including these features and genomic complexity, only the latter defined by both CBA and GM retained its negative impact (Online Supplementary Table S7).

Discussion In recent years, there has been a rising interest in identifying CLL patients with CK since they may pursue a more aggressive clinical course and respond less well to treatment.2,4,28 Large co-operative studies within ERIC have demonstrated that five abnormalities is the optimal cut-off which better predicts an impaired outcome by both CBA and GM.5,25 However, data comparing the risk stratification based on genomic complexity by both methods in the same patients are scarce. Indeed, a small cohort of 122 patients from the Leeksma et al. study was analyzed by GM and CBA, but the proportion of CK cases was very low, as expected in an unselected CLL population. To the best of our knowledge, the present study is the largest report conducted to date in which a cohort of CLL patients enriched in CK cases has been simultaneously analyzed by CBA and GM, comparing the usefulness of both methods in their prognostic stratification. By clustering patients according to criteria previously defined by ERIC, we confirmed that both techniques did haematologica | 2022; 107(3)

Genome complexity in CLL: karyotype versus microarrays

not differ in the proportion of patients classified into each risk category. Notwithstanding, it should be pointed out that only moderate agreement was observed between them. Discordances in the risk assigned to nearly one third of the patients were found, including around 3% of patients classified in either high or low risk groups depending on the methodology employed for their study. We have demonstrated that most of these discordances are the consequence of known limitations intrinsic to each technique. In this regard, the clonal architecture in the sample could mask some alterations by GM, if present in low percentages below their limited sensitivity (~20%), while the CBA result would rely on the in vitro proliferative capacity of the altered clones.29,30 In addition, balanced abnormalities are only detectable by CBA and, on the contrary to expectations, our FISH painting studies confirmed that not all the complex rearrangements described in the karyotype ultimately imply gain or loss of genomic material. On the contrary, some abnormalities recorded as a single monosomy or unbalanced translocation in the karyotype turned out to be multiple CNA or even chromothripsis events when assessed by GM. Thus, our results suggested that discrepancies were not predictable by the type of abnormalities detected by any of the methods. Conversely, we discarded a global underestimation of the genomic complexity associated with the application of the recommended filtering criteria for non-classical CLL CNA by GM.24 Small abnormalities (<5 Mb) were equally found by GM among concordant and discordant patients, and greater agreement in the number of abnormalities could not be achieved when smaller CNA were also considered. Thus, we have confirmed that the present recommendations for GM analysis are robust for complexity assessment.24 The observed differences did not represent a poorer performance for CBA or GM in patients risk stratification. In both cases, the established risk groups showed significant differences in terms of TTFT, which were independent of TP53 and IGHV mutational status. Concerning OS, only high complex groups displayed a dismal evolution. In addition, the heterogeneity regarding the GM platforms employed could be a limitation of this study. However, all GM results were reviewed and uniformly interpreted using the same criteria to filter CNA and similar findings were obtained among different GM companies. Regarding CBA data, previous publications have investigated whether specific cytogenetic patterns not identifiable by GM (presence of balanced or unbalanced rearrangements) may correlate with dismal outcome. Initial studies suggested that carrying chromosomal translocations was associated with poorer clinical course.31 More recently, this negative impact has been attributed to the presence of unbalanced rearrangements and its association with CK.2,32 Indeed, Rigolin et al. proved that CK carrying unbalanced rearrangements constituted a very poor risk subset with particular features such as an increased proportion of TP53 aberrations and a lower frequency of 11q deletions. The presence of these aberrations has also been associated with a deregulated expression of genes involved in cell cycle control and DNA damage response.14 Visentin et al. showed that the combination of the presence of CK and/or unbalanced rearrangements by CBA and IGHV mutational status improved their risk stratification.15 In our cohort, we observed a shorter TTFT for those patients with unbalanced rearrangements but the poor outcome was not confirmed within CK group. haematologica | 2022; 107(3)

Unexpectedly, GM were unable to detect CNA related to all the apparently unbalanced rearrangements. Indeed, the eight patients with high-CK classified as low-GC by GM carried this type of abnormality and showed a dismal evolution. On the other hand, our GM analyses identified patients with patterns of chromothripsis who showed a short TTFT. As previously reported, there was a high association between chromothripsis and CK or TP53 aberrations.19,33,34 Our study is based in a retrospective cohort highly enriched in patients carrying CK, which was necessary to extensively compare both genetic methodologies in the detection of these prognostically relevant highly complex cases. Therefore, as it is not representative of a real-life CLL cohort, it hinders the development of more accurate genetic prognostic scores. Additionally, potential confounding effects of different therapeutic agents could be attributed to the retrospective and multicenter nature of the cohort enriched in treated patients. These could underlie the lack of statistical significance of genome complexity in the analyses for OS. To date, most of the survival analyses of genomic complexity included in clinical trials have been reported using CBA data. Even though the prognostic/predictive value of CK for TTFT and progression-free survival in patients treated with chemoimmunotherapy has been extensively demonstrated,2-5,10 its clinical relevance in patients receiving the new treatment modalities has not been fully established. Initial data from clinical trials with novel agents, mainly developed in older relapsed/refractory and/or in high risk patients (TP53 del/mut, U-IGHV) suggested an adverse significance of CK.6-8,35 In contrast, a number of recent studies including extended follow-ups of older trials, pooled analyses or new drug combinations have not confirmed its adverse significance.36-43 However, most of these studies have analyzed CK impact taking into account patients with ≥3 aberrations but not those with high complexity (≥5 aberrations), compared a low number of patients and showed relatively short follow-ups.44 Thus, additional analyses are needed to clarify the prognostic/predictive impact of genomic complexity. One particular finding of this study is that, even though overall concordance between FISH and GM is high (90%), GM do not detect around 20% of cases with TP53 deletion due to their low sensitivity.23,29 The presence of 17p13 deletions and/or mutations in TP53 predicts the poorest outcome and its assessment is currently mandatory in CLL study. Our results confirm that FISH should be maintained for the study of CLL patients complemented with one genome-wide technique to assess genome complexity for risk stratification. The choice between CBA and GM will depend on each laboratory, which should take into account the methods and equipment availabilities, personnel expertise and the economic costs, among others. In conclusion, we have confirmed that both CBA and GM are valuable tools to assess the prognosis of CLL patients based on genomic complexity. Nevertheless, a considerable proportion of cases are discordantly classified depending on the technique employed. Consequently, previous findings generated from CBA, currently the gold standard for cytogenetic assessment, are not directly applicable to GM or other promising cytogenomic methodologies such as optical genome mapping. Additional validation studies are needed to establish the prognostic value of genomic complexity by GM in future prospective studies and clinical trials. 601

S. Ramos-Campoy et al.

Disclosures No conflicts of interest to disclose.

MARGenomics Platform from Institut Hospital del Mar d’Investigacions Mèdiques (Barcelona) for performing part of the genomic microarrays.

Contribution AP and BE designed the research study; SR, AP and BE provided patients data, analyzed the data and wrote the manuscript; SBo and JS performed a great proportion of genomic microarrays from cases which lacked this information; SBo, JS and SBe were involved in analysis, interpretation and critical discussion of the results; SBe, MJL, DC, HP, GMR, MO, MLB, RC, RS, TB, EG, CM, FB, XC, MJC, AC, JCS, FNK, DO and CH provided patient data and samples. All authors read the last version of the manuscript.

Funding This work was partly supported by grants from Generalitat de Catalunya (17SGR437), Gilead Sciences Fellowship (GLD17/00282), Ministerio de Educación, Cultura y Deporte of Spain (FPU17/00361), Fundación Española de Hematología y Hemoterapia (FEHH-Janssen), Instituto de Salud Carlos III/FEDER (PT17/0015/0011) and the “Xarxa de Bancs de tumors“ sponsored by Pla Director d’Oncologia de Catalunya (XBTC).

Acknowledgements The authors would like to thank Idoya Ancín, Andrea Campeny, María Dolores García-Malo, Alberto Valiente, Marco Moro, Gonzalo Blanco, Ferran Nadeu and Julio Delgado for their contribution to the study providing patients samples and data and the

Data sharing statement Detailed chromosome banding analyses and genomic microarrays profiles for selected cases are provided in the Online Supplementary Tables. Please contact either bespinet@parcdesalutmar.cat or apuiggros@imim.es for additional data.

References 1. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760. 2. Baliakas P, Iskas M, Gardiner A, et al. Chromosomal translocations and karyotype complexity in chronic lymphocytic leukemia: a systematic reappraisal of classic cytogenetic data. Am J Hematol. 2014;89(3): 249-255. 3. Herling CD, Klaumünzer M, Rocha CK, et al. Complex karyotypes and KRAS and POT1 mutations impact outcome in CLL after chlorambucil-based chemotherapy or chemoimmunotherapy. Blood. 2016;128(3): 395-404. 4. Puiggros A, Collado R, Calasanz MJ, et al. Patients with chronic lymphocytic leukemia and complex karyotype show an adverse outcome even in absence of TP53/ATM FISH deletions. Oncotarget. 2017;8(33): 54297-54303. 5. Baliakas P, Jeromin S, Iskas M, et al. Cytogenetic complexity in chronic lymphocytic leukemia: definitions, associations, and clinical impact. Blood. 2019;133(11):12051216. 6. Thompson PA, O'Brien SM, Wierda WG, et al. Complex karyotype is a stronger predictor than del(17p) for an inferior outcome in relapsed or refractory chronic lymphocytic leukemia patients treated with ibrutinibbased regimens. Cancer. 2015;121(20):36123621. 7. Chanan-Khan A, Cramer P, Demirkan F, et al. Ibrutinib combined with bendamustine and rituximab compared with placebo, bendamustine, and rituximab for previously treated chronic lymphocytic leukaemia or small lymphocytic lymphoma (HELIOS): a randomised, double-blind, phase 3 study. Lancet Oncol. 2016;17(2):200-211. 8. Anderson MA, Tam C, Lew TE, et al. Clinicopathological features and outcomes of progression of CLL on the BCL2 inhibitor venetoclax. Blood. 2017;129(25):3362-3370. 9. Haferlach C, Dicker F, Schnittger S, Kern W, Haferlach T. Comprehensive genetic characterization of CLL: a study on 506 cases analysed with chromosome banding analysis, interphase FISH, IgV(H) status and

602

immunophenotyping. Leukemia. 2007;21 (12):2442-2451. 10. Badoux XC, Keating MJ, Wang X, et al. Cyclophosphamide, fludarabine, alemtuzumab, and rituximab as salvage therapy for heavily pretreated patients with chronic lymphocytic leukemia. Blood. 2011;118(8): 2085-2093. 11. Van Den Neste E, Robin V, Francart J, et al. Chromosomal translocations independently predict treatment failure, treatment-free survival and overall survival in B-cell chronic lymphocytic leukemia patients treated with cladribine. Leukemia. 2007;21(8):1715-1722. 12. Jaglowski SM, Ruppert AS, Heerema NA, et al. Complex karyotype predicts for inferior outcomes following reduced-intensity conditioning allogeneic transplant for chronic lymphocytic leukaemia. Br J Haematol. 2012;159(1):82-87. 13. Baliakas P, Puiggros A, Xochelli A, et al. Additional trisomies amongst patients with chronic lymphocytic leukemia carrying trisomy 12: the accompanying chromosome makes a difference. Haematologica. 2016;101 (7):e299-302. 14. Rigolin GM, Saccenti E, Guardalben E, et al. In chronic lymphocytic leukaemia with complex karyotype, major structural abnormalities identify a subset of patients with inferior outcome and distinct biological characteristics. Br J Haematol. 2018;181(2): 229-233. 15. Visentin A, Bonaldi L, Rigolin GM, et al. The combination of complex karyotype subtypes and IGHV mutational status identifies new prognostic and predictive groups in chronic lymphocytic leukaemia. Br J Cancer. 2019;121(2):150-156. 16. Ouillette P, Erba H, Kujawski L, Kaminski M, Shedden K, Malek SN. Integrated genomic profiling of chronic lymphocytic leukemia identifies subtypes of deletion 13q14. Cancer Res. 2008;68(4):1012-1021. 17. Gunn SR, Bolla AR, Barron LL, et al. Array CGH analysis of chronic lymphocytic leukemia reveals frequent cryptic monoallelic and biallelic deletions of chromosome 22q11 that include the PRAME gene. Leuk Res. 2009;33(9):1276-1281. 18. Kolquist KA, Schultz RA, Slovak ML, et al. Evaluation of chronic lymphocytic leukemia by oligonucleotide-based microarray analysis uncovers novel aberrations not detected

by FISH or cytogenetic analysis. Mol Cytogenet. 2011;4:25. 19. Edelmann J, Holzmann K, Miller F, et al. High-resolution genomic profiling of chronic lymphocytic leukemia reveals new recurrent genomic alterations. Blood. 2012;120 (24):4783-4794. 20. Chun K, Wenger GD, Chaubey A, et al. Assessing copy number aberrations and copy-neutral loss-of-heterozygosity across the genome as best practice: an evidencebased review from the Cancer Genomics Consortium (CGC) working group for chronic lymphocytic leukemia. Cancer Genet. 2018;228-229:236-250. 21. Kujawski L, Ouillette P, Erba H, et al. Genomic complexity identifies patients with aggressive chronic lymphocytic leukemia. Blood. 2008;112(5):1993-2003. 22. Gunnarsson R, Mansouri L, Isaksson A, et al. Array-based genomic screening at diagnosis and during follow-up in chronic lymphocytic leukemia. Haematologica. 2011;96(8): 1161-1169. 23. Ouillette P, Collins R, Shakhan S, et al. Acquired genomic copy number aberrations and survival in chronic lymphocytic leukemia. Blood. 2011;118(11):3051-3061. 24. Schoumans J, Suela J, Hastings R, et al. Guidelines for genomic array analysis in acquired haematological neoplastic disorders. Genes Chromosomes Cancer. 2016;55(5):480-491. 25. Leeksma AC, Baliakas P, Moysiadis T, et al. Genomic arrays identify high-risk chronic lymphocytic leukemia with genomic complexity: a multicenter study. Haematologica. 2020;106(1):87-97. 26. McGowan-Jordan J, Simons A, Schmid M, International Standing Committee on Human Cytogenetic Nomenclature. ISCN: An International System for Human Cytogenomic Nomenclature. Basel; New York: Karger; 2016. 27. Döhner H, Stilgenbauer S, Benner A, et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med. 2000;343(26):1910-1916. 28. Rigolin GM, Cavallari M, Quaglia FM, et al. In CLL, comorbidities and the complex karyotype are associated with an inferior outcome independently of CLL-IPI. Blood. 2017;129(26):3495-3498. 29. Puiggros A, Puigdecanet E, Salido M, et al.

haematologica | 2022; 107(3)

Genome complexity in CLL: karyotype versus microarrays

Genomic arrays in chronic lymphocytic leukemia routine clinical practice: are we ready to substitute conventional cytogenetics and fluorescence in situ hybridization techniques? Leuk Lymphoma. 2013;54(5): 986-995. 30. Urbankova H, Papajik T, Plachy R, et al. Array-based karyotyping in chronic lymphocytic leukemia (CLL) detects new unbalanced abnormalities that escape conventional cytogenetics and CLL FISH panel. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub. 2014;158(1):56-64. 31. Mayr C, Speicher MR, Kofler DM, et al. Chromosomal translocations are associated with poor prognosis in chronic lymphocytic leukemia. Blood. 2006;107(2):742-751. 32. Heerema NA, Muthusamy N, Zhao Q, et al. Prognostic significance of translocations in the presence of mutated IGHV and of cytogenetic complexity at diagnosis of chronic lymphocytic leukemia. Haematologica. 2021;106(6):1608-1615. 33. Puente XS, Beà S, Valdés-Mas R, et al. Noncoding recurrent mutations in chronic lymphocytic leukaemia. Nature. 2015;526 (7574):519-524. 34. Salaverria I, Martín-Garcia D, López C, et al. Detection of chromothripsis-like patterns

haematologica | 2022; 107(3)

with a custom array platform for chronic lymphocytic leukemia. Genes Chromosomes Cancer. 2015;54(11):668-680. 35. O’Brien S, Furman RR, Coutre S, et al. Single agent ibrutinib in treatment-naıve and relapsed/refractory chronic lymphocytic leukemia: a 5-year experience. Blood. 2018;131(17):1910-1919. 36. Kreuzer KA, Furman RR, Stilgenbauer S, et al. Outcome of patients with complex karyotype in a phase 3 randomized study of idelalisib plus rituximab for relapsed chronic Lymphocytic Leukemia. Blood. 2016;128 (22):192. 37. Brown JR, Hillmen P, O'Brien S, et al. Extended follow-up and impact of high-risk prognostic factors from the phase 3 RESONATE study in patients with previously treated CLL/SLL. Leukemia. 2018;32(1):8391. 38. Mato AR, Thompson M, Allan JN, et al. Real-world outcomes and management strategies for venetoclax-treated chronic lymphocytic leukemia patients in the United States. Haematologica. 2018;103(9):15111517. 39. Woyach JA, Ruppert AS, Heerema NA, et al. Ibrutinib regimens versus chemoimmunotherapy in older patients with untreat-

ed CLL. N Engl J Med. 2018;379(26):25172528. 40. Kipps TJ, Fraser G, Coutre SE, et al. Longterm studies assessing outcomes of ibrutinib therapy in patients with del(11q) chronic lymphocytic leukemia. Clin Lymphoma Myeloma Leuk. 2019;19(11): 715-722. 41. Munir T, Brown JR, O'Brien S, et al. Final analysis from RESONATE: up to six years of follow-up on ibrutinib in patients with previously treated chronic lymphocytic leukemia or small lymphocytic lymphoma. Am J Hematol. 2019;94(12):1353-1363. 42. Al-Sawaf O, Lilienweiss E, Bahlo J, et al. High efficacy of venetoclax plus obinutuzumab in patients with complex karyotype and chronic lymphocytic leukemia. Blood. 2020;135(11):866-870. 43. Kreuzer KA, Furman RR, Stilgenbauer S, et al. The impact of complex karyotype on the overall survival of patients with relapsed chronic lymphocytic leukemia treated with idelalisib plus rituximab. Leukemia. 2020;34(1):296-300. 44. Jondreville L, Krzisch D, Chapiro E, NguyenKhac F. The complex karyotype and chronic lymphocytic leukemia: prognostic value and diagnostic recommendations. Am J Hematol. 2020;95(11):1361-1367.

603

ARTICLE Ferrata Storti Foundation

Chronic Lymphocytic Leukemia

Clonal evolution in chronic lymphocytic leukemia is scant in relapsed but accelerated in refractory cases after chemo(immune) therapy Marc Zapatka,1* Eugen Tausch,2* Selcen Öztürk,1 Deyan Yordanov Yosifov,2,3 Martina Seiffert,1 Thorsten Zenz,4 Christof Schneider,2 Johannes Bloehdorn,2 Hartmut Döhner,2 Daniel Mertens,2,3 Peter Lichter1 and Stephan Stilgenbauer2

Haematologica 2022 Volume 107(3):604-614

Division of Molecular Genetics, German Cancer Research Center, Heidelberg, Germany; Department of Internal Medicine III, Ulm University Hospital, Ulm, Germany; 3 Mechanisms of Leukemogenesis, German Cancer Research Center (DKFZ), Heidelberg, Germany and 4University Hospital, University of Zürich, Zürich, Switzerland 1 2

*MZ and ET contributed equally as co-first authors.

ABSTRACT

lonal evolution is involved in the progression of chronic lymphocytic leukemia (CLL). In order to link evolutionary patterns to different disease courses, we performed a long-term longitudinal mutation profiling study of CLL patients. Tracking somatic mutations and their changes in allele frequency over time and assessing the underlying cancer cell fraction revealed highly distinct evolutionary patterns. Surprisingly, in long-term stable disease and in relapse after long-lasting clinical response to treatment, clonal shifts are minor. In contrast, in refractory disease major clonal shifts occur although there is little impact on leukemia cell counts. As this striking pattern in refractory cases is not linked to a strong contribution of known CLL driver genes, the evolution is mostly driven by treatment-induced selection of sub-clones, underlining the need for novel, non-genotoxic treatment regimens.

Correspondence: PETER LICHTER Peter.Lichter@dkfz-heidelberg.de STEPHAN STILGENBAUER Stephan.Stilgenbauer@uniklinik-ulm.de Received: July 7, 2020. Accepted: February 26, 2021. Pre-published: March 11, 2021. https://doi.org/10.3324/haematol.2020.265777

604

Introduction Cancer can be conceptualized as an evolutionary process within a given organism.1,2 By increasing the fitness of cancer cells, mutations enable sub-clones to outcompete non-malignant cells and less adapted cancer cell clones. Furthermore, clonal evolution allows the selection of cell populations that are resistant to therapy or responsible for disease recurrence. For some tumor entities like acute myeloid leukemia (AML), the concept of cancer initiating cells seems to account for tumor relapses without further genetic evolution. For other malignancies however, it is more likely that additional mutations play a crucial role in tumor recurrence. This is also true for chronic lymphocytic leukemia (CLL), where progression and clonal evolution have been analyzed in the context of treatment induced genetic changes.3,4 Clinically, CLL is characterized by a highly variable course. The survival time of patients varies between months and decades. Often patients remain untreated for many years until clinical symptoms require therapeutic intervention.5 Despite high rates of initial treatment response, a major clinical challenge is the occurrence of refractory disease that does not respond to treatment. Refractory cases are often characterized by a deletion and/or a mutation in the tumor suppressor gene TP53 located on the short arm of chromosome 17 (del17p/TP53mut). Although a number of recurrently mutated genes were identified in CLL that are of prognostic relevance4,6-9 del17p/TP53mut remains the strongest adverse prognostic factor for progression-free and overall survival in CLL.4,10,11 The incidence of mutated or deleted TP53 is below 3% in Binet A stage CLL representing cases with good prognosis or in the pre-malignant monoclonal B-cell lymphocytosis (MBL) state, but increases to 12% at time of first treatment initiation, and to more than 37% in chemotherapy refractory cohorts.8,12,13 Despite this increase in cases with mutated or deleted TP53 at later disease stages, clonal evolution has been considered rare in CLL.

haematologica | 2022; 107(3)

Clonal evolution in relapsed/refractory CLL cases

Early cytogenetic and molecular cytogenetic studies reported unequivocal evidence for occurrence of clonal evolution in CLL, albeit rare.14-17 More recently, high-resolution microarray and next-generation sequencing (NGS) based approaches were applied to track subclonal heterogeneity and clonal evolution in CLL. Based on a single nucleotide polymorphism (SNP) micro-array analysis of pretreatment and relapsed samples from 42 patients, DNA copy number variations (CNV) were reported that expand or newly occur at relapse.18 The respective genomic regions contain candidate driver genes of relapse and/or chemotherapy resistance. Somatic mutation profiling of CLL by NGS revealed recurrent gene alterations19 and confirmed molecular heterogeneity.20 The comprehensive analysis of 149 CLL cases allowed to distinguish clonal (MYD88, trisomy 12, and del(13q)) and subclonal (SF3B1 and TP53) driver mutations20 and this order was validated by the same group in a huge clinical study.4 While a considerable number of driver genes and recurrent genomic alterations were identified via whole-exome sequencing (WES) analysis of a cumulative number of more than 1,000 CLL patients, there are only few studies that decipher changes of drivers over the course of disease. Mutation profiling of three CLL patients over time indicated heterogeneous clonal evolution patterns.21 By a similar approach, ten of 12 CLL cases treated with chemotherapy were shown to undergo evolution of sub-clones with respective driver mutations (SF3B1 and TP53), while this was detected in only one of six cases that were not treated.22 While one study reported that clonal composition remained stable at disease progression and relapse23 another study referred that 13 of 28 sequentially sampled cases underwent genetic change of >20% with nine of them (but none of the nonevolving cases) also displaying epigenetic evolution.24 A number of deep sequencing studies focused on a targeted panel for candidate genes in CLL and provided evidence of clonal outgrowth over time i.e., of TP53 after treatment.25-27 Despite that, their major focus was on untreated patient samples and the response to therapy was not considered as a predictor of evolution. In addition, targeted analysis of a restricted number of drivers can give an idea of clonal rigidity, but fail to show emergence and outgrowth of new subclones characterized by variants not covered with the panel. A similar approach considering aberrations in addition to known driver mutations deciphered the history of these alterations by integrating longitudinal and cross-sectional data in 70 patients.28 While the distinction of evolutionary early and late events showed a similar pattern to Landau et al., again the association with patient outcome was not addressed. The biggest WES cohort with sequential sampling in CLL included 59 patients from CLL8 with samples before and after relapse to FC/FCR (fludarabine, cyclophosphamide, rituximab) showing changes of cell fractions characterized by specific drivers as well as linear versus branched evolution patterns in 57 of 59 cases.4 However, this group consisted only of relapsed cases with a missing control of refractory and long-term untreated patients. Due to the fact, that again type and duration of response were not considered as parameters, a link between treatment, outcome and dynamic genomic changes in CLL is barely explored. Although a connection of response to therapy and dynamic genomic changes is plausible, it remains unclear how clonal evolution is linked to long-term stable, to relapsed or to refractory disease. haematologica | 2022; 107(3)

In order to elucidate the clonal evolution of CLL cell populations in the presence or absence of therapy, we performed a long-term longitudinal mutation profiling study of a multifarious cohort of CLL patients with a well annotated patient history. Aberrant TP53 dictates the clinical course of the disease, it is a key driver of acquired resistance and potentially supersedes other parameters. Therefore, we excluded patients with del17p or mutated TP53 status at baseline as we presumed that these patients had acquired the most relevant evolution marker already. Samples were obtained at different time points before and after treatment in three different clinical groups: i) long-term untreated cases with stable disease and no need for treatment over at least 4 years, ii) relapsed cases with durable response to therapy of at least 2 years, and iii) refractory cases without response to treatment (stable disease [SD], progressive disease [PD]) or cases that progressed with requirement of a subsequent therapy within 1 year. WES was performed and data were subsequently partially validated by targeted resequencing of identified mutations.

Methods Sample collection We compiled an inventory of CLL patient samples before and after treatment and sequenced tumor and non-tumor control DNA (25 patients and 54 tumor samples including 21 patients with baseline samples prior to any therapy). Our inclusion criteria were: (i) no del17p or mutated TP53 status at baseline, (ii) patients fitting to any of the three groups (a) long-term untreated cases with stable disease and no need for treatment over at least 4 years, (b) relapsed cases with durable response to therapy of at least 2 years, and (c) refractory cases without response to treatment (SD, PD) or cases that progressed with requirement of a subsequent therapy within 1 year. All patients gave informed consent according to the Helsinki Declaration. Sample acquisition for sequencing purposes was approved by a local Ethics Review Committee (Ethikkommision Ulm University, ethik-kommission@uni-ulm.de, 17.06.2008, 96/08-UBB/se). Peripheral blood mononuclear cell (PBMC) samples were enriched for tumor (CD19+) and normal CD19-cells using MACS microbead cell separation (Miltenyi Biotec, Bergisch Gladbach, Germany). Genomic DNA was isolated from unsorted and sorted CLL cells using All Prep Kit (Qiagen, Hilden, Germany). Quality and quantity of the purified DNA were assessed with the Qubit dsDNA BR Assay Kit (Lifetech technologies, Carlsbad, CA).

Sequencing WES was performed on Illumina HiSeq 2000 machines. Exome libraries were created using the TruSeq Exome Library Prep Kit or Agilent SureSelect enrichment Human Exome V4 Kit according to the manufacturer's protocols. Alignment and variant calling were performed as previously described in29.

Allele frequency changes in patient groups Per patient single nucleotide variants (SNV) with genotype change were identified and differences in alternative allele frequency (aAF) calculated between consecutive time points. aAF were clustered per patient into six clusters to give each time point equal weight regardless of the number of SNV detected. Each change in aAF was grouped according to the status (untreated,

605

M. Zapatka et al.

relapse, refractory) at the second time point. Differences in the distribution of the allele frequency changes between the three patient groups were identified using a bootstrapped Kolmogorov-Smirnov test with n=10,000.

Copy number variations calling and calculation of absolute copy numbers Estimation of the copy number state based on the exome sequencing data was achieved using Varscan 2 on the target regions.30 Absolute copy numbers were calculated as previously described.31

Calculation of cancer cell fraction Cancer cell fractions (CCF) integrating sample purity (estimated by fluorescence-activated cell sorting [FACS]), ploidy inferred from fluorescence in situ hybridization (FISH), copy number states calculated from WES and allelic fraction and coverage of somatic variants were calculated for the patients with available germline samples following the approach previously outlined in20.

Estimation of clonal composition by TrAP Changes in clonal tumor composition were calculated integrating the CCF at the respective time points using TrAP (tree approach to clonality).32

Quantification of DNA methylation and estimation of correlation between time points DNA methylation from the first and second time point of ten patient phases (three long-term untreated, two relapsed and five refractory) was assessed by Illumina Infinium HumanMethylation450 BeadChips according to the manufacturer’s protocol. Details on the individual approaches are further described in the Online Supplementary Appendix.

Results The clinical course of patients grouped into distinct phases The clonal evolution in malignant B-cell populations of CLL patients was studied by longitudinal analyses in a total of 25 patients and 54 tumor samples. For 21 patients, the baseline sample was obtained prior to any CLL therapy, whereas four additional patients were pretreated before enrollment in our study. A common case history in CLL can consist of different phases including an untreated phase with a watch and wait strategy in the beginning followed by one or more therapies with either durable or very short responses or even refractoriness to the ongoing treatment. We observed such clinical phases in our patients throughout their individual medical history. For example, some of the long-term untreated patients required therapy at a later stage (e.g., HU-1-06) and some patients with initially long-lasting response became refractory after a subsequent treatment (e.g., HU-1-11). Therefore, we divided the individual patient histories into different clinical phases rather than using a rigid division of patients into categories. Individuals can go through several of these phases with sampling at the beginning and at the end of each phase. Three clinical disease patterns were distinguished and in total we identified 29 phases: six phases were evaluated as long-term untreated, five as relapsed after initially durable response to therapy, and 18 as treatment refractory. Details of the clinical course of 606

patients and patient phases including treatment, treatment response and sampling, as well as cytogenetic grouping and the immunoglobulin heavy-chain variable region gene (IGHV) mutation status are presented in the Online Supplementary Tables S1 to S3 and in the Online Supplementary Figure S1.

Increased mutation rate is associated with refractory disease Identification of mutations was performed by comparative WES of CD19+ enriched PBMC and, as non-malignant control, the sorted CD19-negative fraction of PBMC from the same patient. Over the course of this longitudinal study, no IGHV status switch was identified. In IGHVmutated cases, the major IGHV clone did not change, and IGHV mutations and SNP fingerprinting were used to confirm sample identity. Based on limited material for sorting of non-neoplastic cells, for 19 of 25 patients a non-tumor control was available for mutation detection. Applying established algorithms33 for the calling of SNV and small insertions and deletion (Indels), we observed an average of 15.1 mutations per sample (range, 2-36) (Online Supplementary Tables S4 to S6). A prediction of the response to therapy was not possible based on mutation numbers, as samples taken before long lasting response to therapy and before refractory disease had similar numbers of mutations (11.3 [range, 1-30] and 15.8 [range, 2-34] respectively P-value Mann-Whitney test P=0.36; Figure 1A and B). Samples obtained before any therapy as well as post-therapeutic samples from relapsed patients had the lowest number with 13.5 (range, 2-30) and 13.0 (range, 6-25) mutations in contrast to refractory patients with 17.9 (range, 4-36) mutations, respectively (Figure 1B) (Kruskal Wallis test P=0.30). We identified 1.5 known driver events per sample with the largest variation and highest number of SNV/Indels in refractory CLL samples. All cases except HU-1-08, HU-1-11, and HU-1-21 harbored SNV/Indels in known or candidate CLL driver genes.4,34 Indeed, candidates previously associated with adverse outcome like BIRC3, EGR2 and SAMHD were identified predominantly in refractory cases, but preceded good response to (chemo)therapy and therefore did not determine outcome (e.g., patients HU-1-19 or HU-1-15). In addition, this study revealed genes that had so far not been associated with CLL but were mutated in more than one of the analyzed patients: MC5R, MYH2, RFX7, ROBO2 and SLITRK5.

Clonal evolution of leukemic cells is dominant in patients with refractory disease Clonal evolution was modeled on the basis of single nucleotide variants that were assessed in longitudinal sample collections. FISH analysis with a panel of diagnostic probes10 in a subset of samples revealed near diploidy of the neoplastic cells. Interestingly, no changes in cytogenetic aberrations in long-term untreated phases could be identified based on FISH data (Online Supplementary Table S1). Most patients retained their karyotype after treatment, but HU-1-19 acquired a deletion in chromosome 17p. Since neoplastic B-cell content was generally higher than 80%, AF were used as basis for modeling evolution over time. To this aim, SNV were identified that displayed variable AF between the time points of molecular analysis. During long-term untreated phases, AF remained stable, which is in accordance with an unchanged clonal compohaematologica | 2022; 107(3)

Clonal evolution in relapsed/refractory CLL cases

Figure 1. Single nucleotide variants as well as insertions and deletion (Indels) in patients across phases. (A) Somatic single nucleotide variants (SNV) and insertions and deletion (Indels) identified in chronic lymphocytic leukemia (CLL) samples (SNV in blue, Indels in green). Genes with recurrent (patients n>2) somatic SNV and Indels identified in our study or CLL drivers from previously published CLL cohorts highlighted by black boxes (4,26, COSMIC 19/03/14). Only patients with matched control sample were considered. (B) Each symbol shows the number of variants identified in one patient sample. Samples are grouped and colored in dependence of treatment and outcome of the subsequent phase. Black bars represent group means.

haematologica | 2022; 107(3)

607

M. Zapatka et al.

sition. Although one might expect the occurrence of clonal evolution with the acquisition of new variants in tumors with relapse after therapy, we observed the opposite: such clinical phases show the same mutational landscapes both at baseline and at relapse, and major shifts in AF occurred only exceptionally (Figure 2). In sharp contrast, during phases of therapy refractoriness, we found dramatic alterations in clonal composition (Figures 2 and 3). Notably, in refractory phases these increases and reductions in AF occurred within relatively short time intervals (median phase length: refractory 707 days, relapse 2,395 days, untreated 2,088 days, time span refractory phases vs. time span untreated, treated and relapsed Mann-Whitney test P=0.00014; Online Supplementary Table S1), that were particularly much shorter than the phases in stable or relapsed cases. The high degree of AF changes and the short time window over which these changes occurred indicate marked dynamics in the clonal shift, often notable in tumors that appeared clinically unaffected by therapy (i.e., without remission and subsequent

regrowth). These clonal shifts clearly indicate a change in the clonal composition, and strikingly they occur mostly during refractory phases, i.e., during treatment that does not successfully affect the clinical outcome. We further quantified overall AF changes in the three different groups of clinical phases independent of the further course of disease (Figure 3, changes in AF over time provided in the Online Supplementary Figure S2). Clearly, a substantially higher variation of AF is seen in the samples that reach the therapy refractory phase. Comparison of 28 subsequent time points in 25 patients identified significant differences in the AF changes over time between different types of clinical phases, which indicates that the degree of change in the clonal composition is different in the three clinical groups (Kruskal-Wallis test P=0.00262, corrected based on permutation of phase labels). Furthermore, the untreated phases showed a significantly lower AF change over time independent of the extent of time between sampling (P<0.01 for untreated vs. relapsed and untreated vs. relapsed/refractory; Online Supplementary Table S7).

Figure 2. Allele frequency changes during the clinical course of chronic lymphocytic leukemia patients. Overall clinical phase is depicted by colored bars on top (green=untreated, yellow=relapsed after treatment, red=refractory to treatment). The individual treatments and disease progressions are depicted in the first row of each graph with gray bars representing treatment, while green, yellow, orange and red bars indicate the type of treatment response and disease progression. Richter transformation in HU-1-32 (blue) is highlighted in the second row, above the alternative allele frequencies (aAF). aAF changes are colored based on hierarchical clustering of the trajectories following by identifying the six major clusters (R function cutree). y-axis: indicates allele frequencies, x-axis: indicates time course in years.

608

haematologica | 2022; 107(3)

Clonal evolution in relapsed/refractory CLL cases

Clonal expansion or reduction of individual sub-clones over time and with treatment Integrating tumor purity, copy number state (Online Supplementary Figure S3) and AF, we inferred the CCF (Online Supplementary Table S8) affected by individual mutations as described before.4 In some patients and phases, known cancer drivers listed in the COSMIC mutation database could be linked to the changes in CCF (Online Supplementary Figure S4; gene symbols from COSMIC highlighted in purple). For example, HU-1-13 showed a mutation in the cancer driver EGFR only in the untreated and refractory sample (CCF 7% coverage 33 and CCF 66%, coverage 57) (Online Supplementary Figure S4). The mutation was undetected (coverage 48) at relapse indicating a reduction of this clone below the detection limit at relapse. Furthermore, the fraction of cells carrying an ANO1 mutation steadily increased from 0% over 2.3% to 22.0% in the refractory sample (coverage 48, 42 and 41). Interestingly, the major clone present at the relapse and characterized by an NLRP13 mutation (CCF 28.5%, coverage 21) was not detected any more in the refractory sample (coverage 25), indicating that this clone was lost during treatment or during progress. HU-1-19 displayed similar shifts albeit with a different clonal composition, but also with elimination of a clone

after treatment. The EGR2 variant changed from 2.1% to 27% mutant allele frequency (CCF 4%, coverage 47 and CCF 55%, coverage 37) during the treatment-free interval, but dropped below the detection limit after first treatment (coverage 56). In addition, the major clone at the first time point characterized by MARK2 (CCF 37%, coverage 74) without treatment (“untreated”) was slightly less prominent at the second time point without treatment (CCF 34%, coverage 54) and undetectable after treatment (coverage 60). These observations indicate a gradual change in clonal composition during an untreated phase of 6 years and a significant clonal replacement after treatment.

Refractory chronic lymphocytic leukemia is associated with a branched evolution of leukemic cells In order to group observed clonal changes into different patterns of evolution, we analyzed overall AF changes between all possible pairs of consecutive samples that we grouped into disease phases for different types of evolution. On the basis of time-dependent changes of CCF, when significant AF changes were unidirectional, these evolution patterns were classified as co-evolution (also termed “linear evolution”).22 In contrast, evolution was classified as “branched” when different significant changes concomitantly increased and

Figure 3. Allele frequency changes across phases. Between two consecutive samplings in a patient, changes in allele frequency of clustered somatic single nucleotide polymorphism (SNP) shown as circles. Circles are grouped according to the clinical phase at the second time point. In order to weigh each patient identically, regardless of the number of mutations, the changes in allele frequency of all single nucleotide variants (SNV) were clustered into six groups per patient and the average of these groups is depicted resulting in six circles per patient phase. Black line represents phase mean. Statistical significance of allele frequency change differences between clinical phases was tested using Kruskal-Wallis test and P-value corrected using 100,000 permutations of the phase labels for the six mutation clusters representing each patient (P=0.00262).

haematologica | 2022; 107(3)

609

M. Zapatka et al.

Figure 4. Changes in cancer cell fractions and evolution types. (A) Evolution patterns in 12 patients. Probability distribution of the cancer cell fraction (CCF) for each somatic single nucleotide variant (SNV) revealed clonal (red) or subclonal (blue) SNV (left side). The changes in CCF are depicted in the last column. Changes (a mutation with a change in CCF of greater than 0.2 (DCCF>0.2) with probability >0.5) are highlighted in green (increased CCF) or red (reduced CCF). On the basis of time dependent changes of CCF (right side), evolution patterns were considered as (unbranched) co-evolution (C) when significant changes were unidirectional (up or down), or branched (B) when significant changes were in both directions (up and down) indicating that a dominant clone is replaced by its siblings. Time between samplings is indicated in years at the top. (B) Difference in occurrence of evolution types across clinical phases (coevolution = blue, branched evolution = brown).

610

haematologica | 2022; 107(3)

Clonal evolution in relapsed/refractory CLL cases

decreased in CCF between consecutive samplings and were thus “bidirectional”. As depicted in Figure 4A, we observed in the group of long-term untreated patients phases in the clinical course where substantial shifts in the clonal composition occurred only after treatment. These shifts resulted either in co-evolution of few sub-clones while other sub-clones were lost (e.g., HU-1-15), or a simultaneous decrease and increase of different subclones indicative of branched evolution. Longterm untreated phases display mainly co-evolution patterns (four of five), while only one patient (HU-1-19), who required treatment after 6 years, followed a branched evolution pattern and became refractory to a subsequent treatment. In phases preceding relapse, co-evolution appeared less frequent (three of five). Instead, relapsed and refractory CLL showed a tendency towards more frequent branched evolution than untreated CLL (six of ten vs. one of five, Fisher exact test P=0.28), a pattern that was e.g., observed in the relapsed and treatment refractory phases of a single patient (HU-1-13). Interestingly, this patient was treated with FCR (fludarabine, cyclophosphamide, rituximab) in 2005 and in 2009 again achieving a complete response (CR) each time, but at the second time with shorter duration.

The differences in evolution types shows a trend towards more branched evolution in relapsed and even more in refractory phases (Figure 4B). In order to assess the dynamics on a cellular level we inferred the clonal composition based on the cancer cell fractions using TrAP (Figure 5). Interestingly the major subclone at refractory time point in HU-1-19 (clone 3, 41.5% clone fraction) is already present to a minor extend at the second untreated time point (clone fraction 2.0%) whereas the major clone at the first time point further evolved gaining an additional set of mutations subdividing into clones 4, 5 and 6 (clone fraction of 13.7%, 21.3% and 10.5%). In contrast a clone (19.0% clone fraction) present at the second untreated time point defined by an EGR2 mutation was undetected at the other last time point. In order to confirm the evolutionary changes with an additional method, we performed epigenetic analysis of 20 samples corresponding to ten phases. In line with the genetic data, large-scale evolution of methylation patterns was not present in any of the evaluated long-term untreated (n=3) and relapsed (n=2) phases displaying clonal changes of linear type while three of the five examined refractory phases featured profound changes in DNA methylation (Online Supplementary Figure

Figure 5. Clonal composition as inferred by TrAP. Exemplary changes in clone fraction for individual patients across samplings. Clonal composition was estimated by applying TrAP on the cancer cell fraction calculated for individual single nucleotide variants (SNV). Different clones are represented by the respective colors and individual time points are indicated on the x-axis. Highlighted with the respective gene symbols are inferred clones linked to a known cancer driver gene. y-axis represents clonal fraction of the individual clones identified for the best fit TrAP solution.

haematologica | 2022; 107(3)

611

M. Zapatka et al.

S5). In all patients, even in the ones with few methylation changes, hypermethylation was concentrated in poised promoters and polycomb-repressed regions, whereas hypomethylation occurred mostly in heterochromatin (data not shown; assignment of chromatin states was according to the published reference epigenome of CLL35). In spite of this common pattern, we could not identify any specific CpGs that consistently changed the methylation status throughout different patients or phases. In summary, based on the patients analyzed here, the clonal evolution pattern seems to be linked to the disease phases, and increased changes in AF and a branched evolution are significantly more frequent after treatment compared to untreated patient phases (Figure 6).

Discussion Medical history and disease course of patients with CLL is very individual. In this study, we examined WES data of CLL patients acquired at several time points during their disease and treatment course. Comparing consecutive samples from individual patients, we identified somatic mutations that were present in the leukemia cells and tracked over time the changes in AF of these mutations and the underlying fraction of cancer cells that carried the respective mutations. By modeling the clonal composition using the software TrAP,32 we discovered different clonal evolution patterns and disease progression courses that were linked to the treatment and response history of the patients (Figure 5). From the mutations and clonal changes that occur during CLL disease progression, we draw the following conclusions with respect to groups of genes, but also more conceptually with respect to clonal composition and evolution over time. Recurrent mutations in genes were linked to CLL relapse in three different time- and treatment-dependent patterns.36 First, one subset of genes initially displays subclonal mutations that are enriched after therapy. In contrast, mutations in a second set of genes remained clonally stable upon relapse. Finally, mutations in a third set of genes that are stable in most patients show clonal enrichment only in rare cases. However, these groups of mutations were not linked to a clinical phenotype. Furthermore, exponential-like growth patterns were recently associated with a larger number of CLL drivers and short time to first treatment.37 Of note, these data were derived from untreated CLL patients followed over time. In our patient cohort under the selective pressure of treatment, neither common genetic risk factors like IGHV or recurrent aberrations, nor variants or typically affected pathways are characteristic for a specific clinical course. And although the number of mutations increased slightly after treatment, this did not reflect or even predict outcome after therapy, nor did the number of (sub-)clones. Furthermore, clonal evolution was associated with treatment and indeed branched evolution was found more often in refractory cases, but not exclusively. These results reflect published data for relapsed cases after FCR therapy, which could also not link progression-free survival to an evolution pattern after FC(R) therapy.4,22 Dividing our patient groups in long term responder and refractory cases allowed us in contrast to prior attempts to match the duration of response to the extent of the clonal shift. Counterintuitively, clonal evolution that was mostly 612

dynamic and occurred primarily in patients who displayed refractory disease, i.e., where major changes in clonal evolution happened under the guise of a clinically stable or progressing disease. Therefore, what correlated most with the duration of response to treatment was a highly dynamic evolutionary change among sub-clones, and this change was directly associated with refractory disease. In contrast and unexpectedly, relapse after initially durable response occurred mostly with the same sub-clones. We identified three distinctly different courses of clonal evolution that occurred under distinctly different treatment and response patterns. In refractory cases, clonal composition changed dramatically upon treatment failure and in patients 2, 4, and 18 this happened within only 3 months of therapy. Furthermore, in refractory phases, change in clonal composition was often accompanied by a profound shift in the bulk DNA methylation profile of the tumor, most probably reflecting different methylation profiles of the competing clones rather than de novo methylation changes, as it was previously shown that established CLL clones are epigenetically stable and changes in DNA methylation are unlikely to occur without genetic evolution.24 As an example, patient HU-1-23 did not gain any new mutations between the two time points of his refractory phase but underwent selection of particular pre-existent CLL clones according to the branched genetic evolution model and this was also manifested by a shift in DNA methylation of the bulk tumor. For the clinician managing the patient, “hidden“ selection of a resistant clone is masked by a tumor with a seemingly stable clinical phenotype, i.e., with a persistent lymphadenopathy and leukocytosis. This dynamic clonal change suggests either

Figure 6. Model of the clonal composition changes. Model of the clonal composition changes in the three different treatment phases (long-term untreated, relapsed and treatment refractory). Black lines indicate lymphocyte counts as surrogate marker for tumor load. Arrows indicate times of treatment. The stacked bar plot indicates clonal tumor composition where different colors indicate a different clone defined by a set of mutations.

haematologica | 2022; 107(3)

Clonal evolution in relapsed/refractory CLL cases

an increased evolutionary capacity in these patients or simply the presence of at least one resilient clone. Importantly a selective pressure of therapy is necessary to induce or catalyze this clonal change as provided by the clinical course of patients 15 and 19. Both have a long term untreated phase without marked evolution but a strong shift after becoming refractory. Therefore, neither the underlying risk factors, the natural disease biology nor the type of evolution, which is branched in untreated and refractory phase in both patients, reflects the clinical course while the extent of evolution does. Importantly, we could show in addition that this process occurs also independently of TP53 mutations, i.e., in a cohort of CLL patients without TP53 aberrations before treatment. Thus, in these patients the clonal evolution is driven mostly by treatment that seems to select for resistant CLL clones. Thus, our key finding is the striking observation of a clonal turnover during therapy in those patients, who were considered treatment refractory and therefore are assumed to have a stable tumor load. In contrast, long term untreated cases and late relapses are genomically stable, although they are observed over a much longer period of time. In the latter we found a remarkably stable genomic landscape considering that these patients received a therapy with a subsequent regrowing after a prolonged treatment-free interval. This stability is completely different to a tumor that is refractory and apparently unaffected by therapy, but in contrast displays a dramatic change in clonal composition. This opposing clinical and genomic phenotype at first appears counterintuitive. However, these different courses of clonal dynamics in relapsing and refractory patient phases could be explained by the preexistence of a resistant clone that after removal of the bulk tumor by a treatment intervention will quickly grow out and fill the empty niche. If such resistant clones are absent, competition and outgrowth over time is still possible, so that the tumor regrows with an almost identical clonal composition. This finding mechanistically explains and underlines the relevance of the widely used clinical paradigm of repeating the previous treatment regimen when a good and long-lasting response is achieved: based on the same clonal composition at relapse, the clinician can expect another good response of the tumor to the treatment as the tumor has the same clonal composition as before the treatment. Interestingly, new treatment modalities like venetoclax may behave similarly due to the strong reduction of the tumor load, comparable to chemotherapy: while patients treated with short and effective venetoclax containing combination therapies lack BCL2 mutations at relapse, refractoriness to a long lasting venetoclax treatment associates with the outgrowth of a BCL2 mutated clone displaying the same clonal shift towards drug resistance, that we observe here in chemotherapy refractory cases.38,39 On the other hand, ibrutinib may cause a more decelerated clonal shift due to its slow debulking treatment effect and also only slowly emerging resistant clones (i.e., point mutations in BTK/PLCG240).

haematologica | 2022; 107(3)

In summary despite the small patient cohort (n=25 with 54 time points), we identified a link between changes in the variant AF and changes in clonal architecture, both of which are linked with shortened time to further treatment, i.e., treatment resistance. We found clonal evolution to occur without strong contribution of known CLL driver genes. However, there is a dramatic difference in clonal evolution patterns between relapsed and refractory samples, which highlights the importance of the treatment-induced clonal changes in relation to treatment response. This intrinsic characteristic of CLL evolution underlines the relevance of comparing the benefits of treatment compared to the watch-and-wait strategy that has a very low clonal evolution rate. Furthermore, the substantial clonal evolution in refractory disease highlights the need for novel, non-genotoxic treatment regimens with targeted therapy that are less likely to induce clinical disease resistance by selecting out preexistent refractory sub-clones. Disclosures and Funding DM was supported by DFG (SFB1074 subproject B1 and B2) and ERA-NET “FIRE-CLL”; MZ and MS were supported by ERA-NET “FIRE-CLL” and BMBF “PRECISE”; SS received support from DFG (SFB1074 subproject B1 and B2, BMBF “PRECISE” and ERA-NET “FIRE CLL”, he also received honoraria and research support from AbbVie, AstraZeneca, Celgene, Gilead, GSK, Hoffmann La-Roche, Janssen, Novartis; ET received honoraria from the Speakers Bureau, Advisory board and travel support of Hoffmann LaRoche and AbbVie. Contributions PL and SS developed concepts and ideas; MZ developed software; ET applied methodology; ET validated data; MZ, ET and DYY performed formal analysis; MZ, ET and SÖ performed investigations; TZ, CS, JB, HD, ET, PL and SS provided resources; MZ, ET, and SS cured data; MZ, ET, SÖ, MS, DM, PL and SS wrote the original draft; MZ, ET, SÖ, DYY, MS, TZ, CS, JB, DM, PL and SS wrote, reviewed and edited the manuscript; MZ, ET, SÖ, DYY and DM visualized concepts and data; PL and SS supervised the project; PL and SS were in charge of project administration; MZ, ET, TZ, PL and SS acquired funding. Acknowledgements We thank all patients and physicians, especially Andrea Schnaiter, for donating samples and participating in this study. We thank Michael Hain and Rolf Kabbe for computational support. We thank Stephan Wolf and the High Throughput Sequencing unit of the Genomics & Proteomics Core Facility, German Cancer Research Center (DKFZ), for providing excellent sequencing services. Data availability Sequencing data have been deposited at the European Genome-Phenome Archive hosted at the EBI under accession EGAS00001003652. Data for the methylation arrays are accessible at GSE143411.

613

M. Zapatka et al.

References 1. McGranahan N, Rosenthal R, Hiley CT, et al. Allele-specific HLA loss and immune escape in lung cancer evolution. Cell. 2017;171(6):1259-1271. 2. Nowell PC. The clonal evolution of tumor cell populations. Science. 1976;194(4260): 23-28. 3. Landau DA, Sun C, Rosebrock D, et al. The evolutionary landscape of chronic lymphocytic leukemia treated with ibrutinib targeted therapy. Nat Commun. 2017;8(1):2185. 4. Landau DA, Tausch E, Taylor-Weiner AN, et al. Mutations driving CLL and their evolution in progression and relapse. Nature. 2015;526(7574):525-530. 5. Hallek M, Shanafelt TD, Eichhorst B. Chronic lymphocytic leukaemia. Lancet. 2018;391(10129):1524-1537. 6. Rossi D, Rasi S, Fabbri G, et al. Mutations of NOTCH1 are an independent predictor of survival in chronic lymphocytic leukemia. Blood. 2012;119(2):521-29. 7. Rose-Zerilli MJJ, Forster J, Parker H, et al. The correlation between deletion architecture, ATM mutational status and BIRC3 disruption in 11q-deleted CLL. Blood. 2012;120(21):658-658. 8. Stilgenbauer S, Schnaiter A, Paschka P, et al. Gene mutations and treatment outcome in chronic lymphocytic leukemia: results from the CLL8 trial. Blood. 2014;123(21):32473254. 9. Puente XS, Bea S, Valdes-Mas R, et al. Noncoding recurrent mutations in chronic lymphocytic leukaemia. Nature. 2015;526 (7574):519-524. 10. Dohner H, Stilgenbauer S, Benner A, et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med. 2000;343(26):1910-1916. 11. Zenz T, Eichhorst B, Busch R, et al. TP53 mutation and survival in chronic lymphocytic leukemia. J Clin Oncol. 2010;28(29): 4473-4479. 12. Schnaiter A, Paschka P, Rossi M, et al. NOTCH1, SF3B1, and TP53 mutations in fludarabine-refractory CLL patients treated with alemtuzumab: results from the CLL2H trial of the GCLLSG. Blood. 2013;122(7): 1266-1270. 13. Winkelmann N, Rose-Zerilli M, Forster J, et al. Low frequency mutations independently predict poor treatment-free survival in early stage chronic lymphocytic leukemia and monoclonal B-cell lymphocytosis. Haematologica. 2015;100(6):e237-239. 14. Han T, Ohtaki K, Sadamori N, et al. Cytogenetic evidence for clonal evolution in

614

B-cell chronic lymphocytic leukemia. Cancer Genet Cytogenet. 1986;23(4):321-328. 15. Oscier D, Fitchett M, Herbert T, Lambert R. Karyotypic evolution in B-cell chronic lymphocytic leukaemia. Genes Chromosomes Cancer. 1991;3(1):16-20. 16. Shanafelt TD, Witzig TE, Fink SR, et al. Prospective evaluation of clonal evolution during long-term follow-up of patients with untreated early-stage chronic lymphocytic leukemia. J Clin Oncol. 2006;24(28):46344641. 17. Stilgenbauer S, Sander S, Bullinger L, et al. Clonal evolution in chronic lymphocytic leukemia: acquisition of high-risk genomic aberrations associated with unmutated VH, resistance to therapy, and short survival. Haematologica. 2007;92(9):1242-1245. 18. Knight SJ, Yau C, Clifford R, et al. Quantification of subclonal distributions of recurrent genomic aberrations in paired pretreatment and relapse samples from patients with B-cell chronic lymphocytic leukemia. Leukemia. 2012;26(7):1564-1575. 19. Puente XS, Pinyol M, Quesada V, et al. Whole-genome sequencing identifies recurrent mutations in chronic lymphocytic leukaemia. Nature. 2011;475(7354):101-105. 20. Landau DA, Wu CJ. Chronic lymphocytic leukemia: molecular heterogeneity revealed by high-throughput genomics. Genome Med. 2013;5(5):47. 21. Schuh A, Becq J, Humphray S, et al. Monitoring chronic lymphocytic leukemia progression by whole genome sequencing reveals heterogeneous clonal evolution patterns. Blood. 2012;120(20):4191-4196. 22. Landau DA, Carter SL, Stojanov P, et al. Evolution and impact of subclonal mutations in chronic lymphocytic leukemia. Cell. 2013;152(4):714-726. 23. Ojha J, Ayres J, Secreto C, et al. Deep sequencing identifies genetic heterogeneity and recurrent convergent evolution in chronic lymphocytic leukemia. Blood. 2015;125(3):492-498. 24. Oakes CC, Claus R, Gu L, et al. Evolution of DNA methylation is linked to genetic aberrations in chronic lymphocytic leukemia. Cancer Discov. 2014;4(3):348-361. 25. Rasi S, Khiabanian H, Ciardullo C, et al. Clinical impact of small subclones harboring NOTCH1, SF3B1 or BIRC3 mutations in chronic lymphocytic leukemia. Haematologica, 2016;101(4):135-138. 26. Nadeu F, Delgado J, Royo C, et al. Clinical impact of clonal and subclonal TP53, SF3B1, BIRC3, NOTCH1, and ATM mutations in chronic lymphocytic leukemia. Blood. 2016;127(17):2122-2130. 27. Nadeu F, Clot G, Delgado J, et al. Clinical

impact of the subclonal architecture and mutational complexity in chronic lymphocytic leukemia. Leukemia. 2018;32(3):645653. 28. Wang J, Khiabanian H, Rossi D, et al. Tumor evolutionary directed graphs and the history of chronic lymphocytic leukemia. Elife. 2014;3:e02869. 29. Jones DT, Hutter B, Jager N, et al. Recurrent somatic alterations of FGFR1 and NTRK2 in pilocytic astrocytoma. Nat Genet. 2013;45 (8):927-932. 30. Koboldt DC, Zhang Q, Larson DE, et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res. 2012;22(3):568576. 31. Carter SL, Cibulskis K, Helman E, et al. Absolute quantification of somatic DNA alterations in human cancer. Nat Biotechnol. 2012;30(5):413-421. 32. Strino F, Parisi F, Micsinai M, Kluger Y. TrAp: a tree approach for fingerprinting subclonal tumor composition. Nucleic Acids Res. 2013;41(17):e165. 33. Jones DT, Jager N, Kool M, et al. Dissecting the genomic complexity underlying medulloblastoma. Nature. 2012;488(7409):100105. 34. Bailey MH, Tokheim C, Porta-Pardo E, et al. Comprehensive characterization of cancer driver genes and mutations. Cell. 2018;173 (2):371-385. 35. Beekman R, Chapaprieta V, Russinol N, et al. The reference epigenome and regulatory chromatin landscape of chronic lymphocytic leukemia. Nat Med. 2018;24(6):868-880. 36. Amin NA, Seymour E, Saiya-Cork K, Parkin B, Shedden K, Malek SN. A Quantitative analysis of subclonal and clonal gene mutations before and after therapy in chronic lymphocytic leukemia. Clin Cancer Res. 2016;22(17):4525-4535. 37. Gruber M, Bozic I, Leshchiner I, et al. Growth dynamics in naturally progressing chronic lymphocytic leukaemia. Nature. 2019;570(7762):474-479. 38. Tausch E, Close W, Dolnik A, et al. Venetoclax resistance and acquired BCL2 mutations in chronic lymphocytic leukemia. Haematologica. 2019;104(9):e434-e437. 39. Blombery P, Anderson MA, Gong JN, et al. Acquisition of the recurrent Gly101Val mutation in BCL2 confers resistance to venetoclax in patients with progressive chronic lymphocytic leukemia. Cancer Discov. 2019;9(3):342-353. 40. 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.

haematologica | 2022; 107(3)

ARTICLE

Chronic Lymphocytic Leukemia

Integrative prognostic models predict long-term survival after immunochemotherapy in chronic lymphocytic leukemia patients

Ferrata Storti Foundation

Johannes Bloehdorn,1 Julia Krzykalla,2 Karlheinz Holzmann,3 Andreas Gerhardinger,3 Billy Michael Chelliah Jebaraj,1 Jasmin Bahlo,4 Kathryn Humphrey,5 Eugen Tausch,1 Sandra Robrecht,4 Daniel Mertens,1,6 Christof Schneider,1 Kirsten Fischer,4 Michael Hallek,4 Hartmut Döhner,1 Axel Benner2# and Stephan Stilgenbauer1# Department of Internal Medicine III, University of Ulm, Ulm, Germany; 2Division of Biostatistics, German Cancer Research Center, Heidelberg, Germany; 3Genomics Core Facility, University of Ulm, Ulm, Germany; 4Department I for Internal Medicine and Center for Integrated Oncology, University of Cologne, Cologne, Germany, 5Clinical Development Oncology, Roche Products Ltd, Welwyn Garden City, UK and 6German Cancer Research Center (DKFZ), Heidelberg, Germany 1

Haematologica 2022 Volume 107(3):615-624

AB and SS contributed equally as co-senior authors.

ABSTRACT

hemoimmunotherapy with fludarabine, cyclophosphamide and rituximab (FCR) can induce long-term remissions in patients with chronic lymphocytic leukemia. Treatment efficacy with Bruton's tyrosine kinase inhibitors was found similar to FCR in untreated chronic lymphocytic leukemia patients with a mutated immunoglobulin heavy chain variable (IGHV) gene. In order to identify patients who specifically benefit from FCR, we developed integrative models including established prognostic parameters and gene expression profiling (GEP). GEP was conducted on n=337 CLL8 trial samples, “core” probe sets were summarized on gene levels and RMA normalized. Prognostic models were built using penalized Cox proportional hazards models with the smoothly clipped absolute deviation penalty. We identified a prognostic signature of less than a dozen genes, which substituted for established prognostic factors, including TP53 and IGHV gene mutation status. Independent prognostic impact was confirmed for treatment, β2-microglobulin and del(17p) regarding overall survival and for treatment, del(11q), del(17p) and SF3B1 mutation for progression-free survival. The combination of independent prognostic and GEP variables performed equal to models including only established non-GEP variables. GEP variables showed higher prognostic accuracy for patients with long progression-free survival compared to categorical variables like the IGHV gene mutation status and reliably predicted overall survival in CLL8 and an independent cohort. GEP-based prognostic models can help to identify patients who specifically benefit from FCR treatment. The CLL8 trial is registered under EUDRACT-2004004938-14 and clinicaltrials gov. Identifier: NCT00281918.

Introduction Chemoimmunotherapy with fludarabine, cyclophosphamide and rituximab (FCR) was defined as the standard first-line therapy for patients with chronic lymphocytic leukemia (CLL) who are eligible for intensive treatment.1,2 There is prognostic impact of recurrent genetic alterations and NOTCH1 mutations were identified as a predictive marker for reduced benefit of FCR over FC.3,4,5,6,7 While substantial treatment benefit has been established for FCR in distinct patient populations,1 high efficacy of novel targeted compounds such as the Bruton's tyrosine kinase (BTK) inhibitor ibrutinib was recently reported in previously untreated patients,8,9 and for cohorts with genetic high-risk subgroups or refractory populations.10,11,12,13,14 However, progression-free survival (PFS) in previously untreated patients ≤70 years old with a mutated immunoglobulin heavy chain variable (IGHV) gene was similar

haematologica | 2022; 107(3)

Correspondence: STEPHAN STILGENBAUER stephan.stilgenbauer@uniklinik-ulm.de Received: March 1, 2020. Accepted: March 11, 2021. Pre-published: March 18, 2021. https://doi.org/10.3324/haematol.2020.251561

615

J. Bloehdorn et al.

for the treatment with BTK inhibition or FCR.14 Therefore, identification of young and fit patients who specifically benefit from the treatment with FCR is needed to optimize long-term outcomes, in particular in the light of toxicity and cost associated with lifelong ibrutinib treatment. Additional biological characterization, such as gene expression profiling (GEP), may be helpful for further refinement of prognostic models leading to an increased prognostic accuracy and precise segregation of patients with a high treatment efficacy of FCR. Established markers mostly constitute categorical variables or consensus cut-offs, in the case of IGHV mutation status, and therefore may not fully reflect the underlying biology. In addition, established prognostic markers may loose some of their impact with novel treatments. Since such large-scale studies on randomized trials are scarce, we performed GEP on 337 baseline patient samples from the CLL8 trial and modeled different scenarios for the combined use with established prognostic factors. We identified less than a dozen genes substituting for the prognostic impact of distinct recurrent alterations for PFS and overall survival (OS). Our results provide the basis for refined prognostic models and rational treatment selection.

Methods Patients and samples The study was conducted on peripheral blood samples from 337 previously untreated CLL patients (Table1) collected at enrolment on the CLL8 trial, a prospective, international, multi-center trial comparing first-line treatment with FC or FCR in a 1:1 randomized fashion. Further details for the study are provided online at the ClinicalTrials.gov (CTG) homepage (www.clinicaltrials.gov #NCT00281918).1 Ficoll density gradient centrifugation for isolation of mononuclear cells followed by an immunomagnetic tumor cell enrichment via CD19 (Midi MACS, Miltenyi Biotec®, Bergisch Gladbach, Germany) was performed on all samples. Data on genomic aberrations del(13q), trisomy 12, del(11q), del(17p) and mutation status for IGHV, TP53, SF3B1 and NOTCH1 was assessed as previously described.5 Informed consent and ethics committee approval was obtained in accordance with the Declaration of Helsinki for all patients.

RNA isolation, quality assessment and gene expression profiling on Exon ST 1.0 arrays Total RNA was extracted from whole cell lysate according to the Allprep DNA/RNA mini kit (Qiagen). Quality control was performed using the Agilent 2100 Bioanalyzer with the RNA 6000 Nano LabChip (Agilent Technologies). In order to ensure accuracy and reproducibility, samples with an RNA integrity number (RIN) less than 7.0 were excluded from further analysis. Samples were analyzed for mRNA expression using the Affymetrix GeneChip® Human Exon 1.0 ST Array (Affymetrix, Santa Clara, CA, USA). Further details are provided in the Online Supplementary Appendix.

Normalization of expression data Raw Affymetrix data files were preprocessed by the robust multichip average (RMA) algorithm using the aroma.affymetrix R package (2008).15 Within RMA normalization, background correction and quantile normalization was conducted. Aroma.affymetrix was applied to generate GEP values summarized on the exon/probe set level and on the transcript level using the ‘core’ probe set definition according to Affymetrix. ‘Core’

616

refers to probe sets that are supported by the most reliable evidence from RefSeq and full-length mRNA GenBank records containing complete coding sequences information. We further assessed and excluded the presence of potential batch effects induced by external factors such as time point and location of sampling as well as time point of labeling and hybridization. Quality control was further conducted with "Relative Log Expression” (RLE) and "Normalized Unscaled Standard Errors” (NUSE), where we also did not find any abnormalities indicating potential batch effects.

Statistical analyses Data was analyzed to evaluate improvement of prognostication for PFS and OS by using GEP in addition to prognostic factors del(17p), del(11q), trisomy 12, del(13q), IGHV mutation status, SF3B1, NOTCH1, TP53 mutations, β2-microglobulin (β2-m), thymidine kinase (TK), white blood cell count (WBC), Eastern Cooperative Oncology Group (ECOG) performance status, study medication (FC or FCR), sex and age. For the following analyses,

Table 1. Patient characteristics of the CLL8 gene expression profiling cohort. Baseline characteristics FC FCR Total Target analysis population, N 169 168 337 Age (years), median (range) 62 (36-81) 60 (35-77) 61 (35-81) Female sex, N (%) 41 (24.3) 40 (23.8) 81 (24.0) ECOG performance status, 1 (0-1) 0 (0-2) 0 (0-2) median (range) Total CIRS score, median (range) 2 (0-7) 1.5 (0-7) 2 (0-7) Stage Binet stage, N (%) 169 168 337 A 10 (5.9) 11 (6.5) 21 (6.2) B 107 (63.3) 99 (58.9) 206 (61.1) C 52 (30.8) 58 (34.5) 110 (32.6) Genetic variable Type according to hierarchical 168 167 335 model, N (%) 17p deletion 15 (8.9) 13 (7.8) 28 (8.4) 11q deletion 39 (23.2) 51 (30.5) 90 (26.9) Trisomy 12 21 (12.5) 9 (5.4) 30 (9.0) No abnormalities 30 (17.9) 31 (18.6) 61 (18.2) 13q deletion (single) 63 (37.5) 63 (37.7) 126 (37.6) IGHV mutational status, N (%) 163 164 327 IGHV unmutated 106 (65.0) 109 (66.5) 215 (65.7) IGHV mutated 57 (35.0) 55 (33.5) 112 (34.3) TP53 mutational status, N (%) 167 164 331 TP53 unmutated 140 (83.8) 148 (90.2) 288 (87.0) TP53 mutated 27 (16.2) 16 (9.8) 43 (13.0) TP53 mutation and/or deletion 29 (17.4) 17 (10.4) 46 (13.9) NOTCH1 mutational status, N (%) 163 166 329 NOTCH1 unmutated 152 (93.3) 149 (89.8) 301 (91.5) NOTCH1 mutated 11 (6.7) 17 (10.2) 28 (8.5) SF3B1 mutational status, N (%) 163 165 328 SF3B1 unmutated 126 (77.3) 130 (78.8) 256 (78.0) SF3B1 mutated 37 (22.7) 35 (21.2) 72 (22.0) Biologic variable Telomere length (kb), median 4.2 4.1 4.2 (range) (2.6-11.5) (2.6-15.3) (2.6-15.3) Serum thymidine kinase (U/L), median 23.4 17.1 20.1 (range) (3.5-855.0) (2.7-970.0) (2.7-970.0) Serum β2-microglobulin (mg/L), median 2.9 2.7 2.8 (range) (1.1-9.2) (0.9-8.0) (0.9-9.2) Leukocyte count (G/L), median 94.0 95.6 94.9 (range) (6.7-867.0) (12.6-363.0) (6.7-867.0) ECOG: Eastern Cooperative Oncology Group; CIRS: cumulative illness rating scale; FC: fludarabine and cyclophosphamide FCR: fludarabine, cyclophosphamide and rituximab.

haematologica | 2022; 107(3)

GEP for high dimensional prognostic models in CLL

missing values in the clinical data were imputed using chained equations.16 The algorithm imputes the missing values using a model with all other clinical variables as predictors, thus generating ’plausible’ synthetic values. As the percentage of missingness for each variable was low (maximum of 16 missing values in 337 patients), a single imputation method was adequate. Furthermore, a non-specific filtering was performed selecting the 500 genes with highest variability over all samples. The final model was built by sparsed Cox proportional hazards model using the smoothly clipped absolute deviation (SCAD) penalty.17 The “reference model” for our analysis is a Cox proportional hazards model including variables with confirmed prognostic impact: age (continuous), sex (male or female), study medication (FC or FCR), ECOG performance status (1 or 2 vs. 0), WBC, TK and β2-m (all continuous), IGHV/ NOTCH1/ SF3B1 mutation status (all unmutated vs. mutated), del(11q), del(13q), del(17p), trisomy 12 and TP53 mutation (all present or absent). The analysis is based on updated results from the CLL8 trial.1 Models investigated for possible improvement of prognostication using GEP included, first: the combination of all above-mentioned confirmed prognostic variables without penalization and a subset of the GEP data selected by SCAD penalization (referred to as “fixed model”), and secondly: the combination of confirmed prognostic variables and GEP data in which all variables were equally penalized (“equally penalized model”) allowing for substitution of the confirmed prognostic variables with equally strong prognostic GEP variables. For internal validation bootstrap subsampling with 1.000 subsamples equal to 63.2% of the original sample size was used.18 The prognostic value of the final model was evaluated on the basis of the time-dependent Brier score (as implemented in the R-package pec).19 The Brier score was used to estimate the prediction error at a given time point. Resulting prediction error curves show the time-dependent Brier score over 60 months of follow-up and the integrated Brier score (IBS) was used to summarize prediction accuracy. For external validation the apparent error was calculated. For visualization purposes, survival curves were calculated by means of the Stone-Beran estimator20 using symmetrical nearest neighborhoods around the lowest, the median, and the highest observed values of the prognostic variable combinations using the R-package prodlim,21 both for OS and PFS. Statistical analysis was performed with the R environment for statistical computing, version 3.3.1, using the R packages survival, version 2.39-5, prodlim, version 1.5.7, mice, version 2.25, ncvreg, version 3.6-0, pec, version 2.4.9 and bootstrap, version 2015.2. For validation, the prognostic gene signature established on the CLL8 cohort was tested in an array-based GEP training set of an independent cohort (n=149 unsorted CLL samples from treatmentnaive [83%] and pretreated [17%] patients).22 Unmutated IGHV was reported in 49.3% and del(17p) in 8.6% of tested samples. Further details on cohort characteristics are provided in a previous publication.22

Results Gene expression profiling variables substitute established prognostic markers in multivariate models We first established multivariate models for variables for which the prognostic impact was confirmed in previous studies and is herein referred to as the “reference model”. Results are shown in the Online Supplementary Table S1A for OS and in Online Supplementary Table S1B for PFS, respectively. In order to evaluate the impact for OS including a signature consisting of GEP variables selected in the penalized haematologica | 2022; 107(3)

Cox model (Online Supplementary Table S2A), we tested various combinations of confirmed prognostic variables and GEP. Only model combinations including genetic markers with prognostic impact achieved prediction error estimates similar to the confirmed prognostic variables used in the “reference model” (Figure 1A). Using the “fixed model”, penalization of GEP resulted in selection of only one GEP variable (PITPNC1, phosphatidylinositol transfer protein cytoplasmic 1) and no further improvement as compared to the reference model (IBS: reference model 0.092; fixed model 0.092) (Figure 1A). In contrast, using the “equally penalized model” on all variables from the reference model and GEP data resulted in selection of only three confirmed prognostic markers (FCR, β2-m, del(17p)) along with ten GEP variables comprising the genes CLEC2B, RGS1, LDOC1, L3MBTL4, PRKCA, FHL1, SGCE, DCLK2, VSIG1, CD72 (Online Supplementary Table S3A). When assessing the prediction accuracy, this model performed similarly as the reference model (IBS: reference model 0.092; equally penalized model 0.096) (Figure 1A). When analyzing PFS by prediction models including a signature of selected GEP variables for PFS (Online Supplementary Table S2B) with the same approach, the “fixed model” did not lead to selection of GEP variables besides the confirmed prognostic variables. Conversely, only four confirmed prognostic markers (FCR, del(11q), del(17p), SF3B1 mutation) were selected in the “equally penalized model”, together with 11 GEP variables including the genes RGS1, EIF1AY, LDOC1, L3MBTL4, DCAF12, PLD5, GTSF1L, NIPAL2, CYBRD1, ANXA1 (Online Supplementary Table S3B). Again, variables selected in the “equally penalized model” performed similar to the “reference model” as demonstrated by prediction error estimates (IBS: reference model 0.160; equally penalized model 0.166; fixed model 0.160) (Figure 1B). Of note, strong prognostic markers like TP53 and IGHV mutation status (Online Supplementary Table S1) were substituted in both models by prognostic GEP variables (Online Supplementary Table S3). For the prognostication of PFS, inclusion of GEP data alone or in addition to non-genetic variables (β2-m, TK, WBC, ECOG, study medication, sex and age) compensated for missing genetic information in patients with late disease progression (Figure 1B). In such models, GEP reliably increased prediction accuracy for patients over time as prediction error curves converged with those of the reference model. Prediction accuracy was comparable with the reference model at 60 months. The overall number of prognostic variables remained similar for either model (“reference model”: OS/PFS 15 variables vs. “equally penalized”: OS 13 and PFS 15 variables) and although chromosomal gains or losses covered multiple genes, these variables were substituted by the expression of a few genes only. Furthermore, expression variables selected along with clinical variables in the penalized models for OS and PFS were not derived from genes localized in the recurrently deleted or amplified chromosomal regions (Online Supplementary Table S3A and B).

Gene expression profiling signatures refine prognostic estimation and retain strong prognostic value in an independent cohort of unselected patients In order to illustrate the distribution for OS and PFS within the different prediction models, conditional KaplanMeier estimates were generated and survival curve estimates are shown for lowest, median, and highest values of 617

J. Bloehdorn et al.

the prognostic variable combinations (Figure 2A to F). GEP variables are especially suitable to predict cases with late progression, while established prognostic factors compensate in the remaining cases with early progression (Figure 1A and B). Specifically, patients with long-term PFS were more accurately identified with models using prognostic GEP signatures (Figure 2D and F) when compared with models using established prognostic variables only (Figure 2B) or single genetic characteristics. This aspect was further exemplified in a subgroup analysis for patients <60 years and those receiving FCR (Online Supplementary Figure S1A and B). In order to validate the results we tested our prognostic gene signature in an independent cohort.22 This cohort was selected to be most heterogeneous from CLL8 to confirm the strength and independence of our prognostic score for OS (Online Supplementary Table S2A; Figure 2E and F). While the CLL8 cohort consisted of treatment-naive patients

618

receiving FC/FCR and GEP was derived from CD19+ purified tumor cells, the validation cohort contained samples with heterogeneous tumor cell purity from both treatmentnaive and pretreated patients. The CLL8-based signature was estimated on the validation cohort and evaluated for individual performance. For comparison, we used the gene signature established for the validation cohort with respective weights as provided.22 Notably, the CLL8-derived gene signature performed highly similar to the gene signature originally established for this dataset (Online Supplementary Figure S2).22

Gene expression profiling variables balance prognostic inaccuracy of established markers GEP variables selected both for OS and PFS contained the genes RGS1 (regulator of G protein signaling 1), LDOC1 (LDOC1 regulator of NF-κB signaling) and L3MBTL4 (L3MBTL histone methyl-lysine binding protein 4). While

Figure 1. Prediction error estimates for prognostic model combinations. Prediction error curves for combinations of prognostic variables in models are shown for overall survival (OS) (A) and progression-free survival (PFS) (B). Combinations of prognostic variables contain the confirmed prognostic variables, as used in the reference model (age, sex, study medication, Eastern Cooperative Oncology Group [ECOG], log white blood cells [WBC], β2-microglobulin [β2m], log thymidine kinase [TK], IGHV mutation status, del(11q), del(13q), del(17p), trisomy 12, TP53 mutation, NOTCH1 mutation, SF3B1 mutation) and gene expression profiling (GEP) variables. Prognostic GEP variables were selected in addition to (fixed model) or instead of (equally penalized model) the confirmed prognostic variables. In a separate approach prognostic GEP variables were selected in addition to (fixed model) or instead of (equally penalized model) non-genetic prognostic variables (only age, sex, study medication, ECOG, log WBC, log TK, β2-m). GEP variables selected in the fixed or equally penalized model largely overlap with the full prognostic gene signature (Online Supplementary Table S2), which is separately used in the “GEP data only” prediction error curve. Combination of prognostic variables selected in the equally penalized model performed highly similar to the model containing only confirmed prognostic variables. Strong overlap was found for prediction error curves represented by the red and blue solid lines.

haematologica | 2022; 107(3)

GEP for high dimensional prognostic models in CLL

Figure 2. Conditional Kaplan-Meier survival estimates illustrate the distribution for overall survival and progression-free survival within the different prediction models. Kaplan-Meier estimates were generated for the lowest, the median, and the highest observed values of the prognostic variable combinations. Kaplan-Meier estimates illustrate overall survival (OS) (A, C and E) and progression-free survival (PFS) (B, D and F) with regard to the “reference model” (confirmed prognostic variables only, A and B), the “equally penalized model” (confirmed prognostic variables and GEP equally penalized, C and D) and prognostic GEP signatures only (as represented in the Online Supplementary Table S2A and B) (E and F).

haematologica | 2022; 107(3)

619

J. Bloehdorn et al.

RGS1 was homogeneously distributed across the expression range, LDOC1 and L3MBTL4 expression showed a bimodal distribution (Online Supplementary Figure S3). When evaluating expression level distributions of RGS1, LDOC1 and L3MBTL4 in relation to genetic variables, we could not identify an exclusive association with known prognostic factors (Figure 3; Online Supplementary Table S4A to D). In order to elucidate the biologic context from which the prognostic impact of these three genes may derive, we dichotomized patient samples regarding the upper and lower quartile of RGS1, LDOC1 and L3MBTL4 expression and assessed the differential expression of associated genes. Differentially expressed genes with a false discovery rate (FDR) of <0.01 and a fold-change (FC) of >1.5 were assessed for overlaps of the respective expression signatures (Figure 4A). Only 12 genes were overlapping between all three gene-specific comparisons (Figure 4A). Expression signatures associated with RGS1 were highly distinct from the other profiles and showed only nine of 341 genes exclusively overlapping with the LDOC1 specific signature. Conversely, 51 of 69 genes contained in the L3MBTL4 signature exclusively overlapped with the LDOC1 signature and therefore support a similar biologic context. Genes contained in different signatures showed highly correlated expression profiles (Figure 4B). LDOC123 and other genes overlapping for the L3MBTL4 and LDOC1 signature, such as LPL or CRY1, were previously reported as surrogate markers for the IGHV mutation status.24,25 We specifically investigated ZAP70 in this context, since it has also been identified as a surrogate marker for the IGHV mutation status.25,26,27 While ZAP70 had a foldchange lower than the previously set cut-off (FC>1.5), we found a highly significant (q<1x10-7) association with LDOC1 and L3MBTL4 (Figure 4C). Provided that LDOC1 and L3MBTL4 expression levels did not show an exclusive association with the IGHV mutation status (Figure 3; Online Supplementary Table S4A to D), we wondered if the combined status of these two genes may explain the observed similarities. Notably, expression of LDOC1 and L3MBTL4 was highly correlated with each other and the combination of both variables reliably identified the majority of cases with IGHV homology <98% (Figure 5). However, we observed several “discordant” cases with mutated IGHV and high expression levels of LDOC1 and L3MBTL4 or IGHV unmutated cases with low expression levels (Figure 3; Figure 5). Provided the fact that these continuous variables were selected due to the higher prognostic accuracy instead of the categorical IGHV mutation status, these markers therefore better mirror prognostic effects and the related biology of a variable sequence homology, especially in “discordant” cases.

Discussion In the presented study, we evaluated the significance of GEP as a means for prognostic modeling in CLL. The CLL8 study cohort provides a valid basis for this as it was designed as a large international, multi-center phase III study defining current standard treatment, with full genetic characterization and long follow-up. Importantly, CD19+ purified tumor cells were procured at enrollment allowing valid GEP analysis. While GEP was unable to improve prediction when used in addition to confirmed prognostic variables, GEP substi620

tuted for many of these variables when tested in direct comparison in the equally penalized model and reliably predicted OS and PFS, similar to models integrating only confirmed prognostic variables. Furthermore, for the prognostication of PFS, GEP was able to compensate for missing genetic information in the subgroup with late progression events. High prediction accuracy for late progression and confirmation of the independent prognostic value for previously reported high-risk markers,4,5,28 which were selected in the equally penalized model, implies that GEP-based prognostication can primarily substitute for intermediate and lowrisk prognostic variables. However, GEP-based prognostic modeling was also able to substitute for “unmutated IGHV”, one of the most important variables with negative prognostic impact on OS and PFS.1,6,7,28 GEP variables selected for PFS and OS in the equally penalized models were largely heterogeneous, a finding that may reflect both methodological and biological differences when modeling these endpoints. Conversely, we identified RGS1, LDOC1 and L3MBTL4 to have prognostic value both for PFS and OS. While the combined expression of LDOC1 and L3MBTL4 was highly associated with IGHV homology and therefore may be viewed as surrogate marker of the IGHV mutation status at first, one has to consider that both genes were selected in the prognostic model instead of the IGHV mutation status. This indicates that these genes and the associated biology have a considerable impact on the prognosis and not merely substitute for the IGHV mutation status. This study further demonstrates the potential of GEP to reduce biologic dimensionality. As such, chromosomal aberrations affecting a multitude of genes, also if minimally deleted regions only are considered, can be replaced by less than a dozen genes. The fact that the genes contained in the prognostic GEP scores were not located on recurrently affected chromosomal regions indicates that the deregulated expression does not derive from a mere gene dosage effect but represents a convergence of various biologic traits. Genes of the identified signatures likely constitute important elements in overactive signaling cascades impacting on the clinical course. In addition, GEP variables represent continuous variables and therefore may hold more potential to fine-tune prognostic modeling in contrast to categorical variables such as aberrations and mutations. The efficacy resulting from the addition of rituximab to FC treatment and substantial benefit for patients with distinct genetic features leading to long-term disease control and OS has been confirmed recently in a long-term followup analysis.1 Notably, prognostic variables selected in the equally penalized model or the GEP signature estimated the clinical course of long-term PFS within this cohort better compared to the model using only genetic factors or parameters previously identified to characterize such patients.1 Future studies will provide insight, if prognostic models including GEP also hold advantage over recently reported prognostic models using epigenetic subgrouping.29,30,31 Patients with DNA methylation profiles reflecting memory B-cell-like CLL were reported to strongly benefit from treatment with chemoimmunotherapy on two phase II trials.31 A major strength of our study was the possibility to exclusively use CD19+ sorted patient samples from a randomized phase III trial and extensive characterization for established prognostic variables, including availability of the TP53, SF3B1 and NOTCH1 mutation status in >95% of haematologica | 2022; 107(3)

GEP for high dimensional prognostic models in CLL

Figure 3. Association of RGS1, LDOC1 and L3MBTL4 with genetic variables. Boxplots showing distribution for log expression of genes selected for both overall survival (OS) and progressionfree survival (PFS), namely RGS1, LDOC1 and L3MBTL4. LDOC1 and L3MBTL4 show a bimodal distribution. Distribution of the three genes was not exclusively associated with distinct genetic variables. 2

haematologica | 2022; 107(3)

621

J. Bloehdorn et al.

Figure 4. Assessment of genes showing concordant or discordant expression with RGS1, LDOC1 and L3MBTL4. (A) Venn diagram illustrating overlaps for differentially expressed genes (fold-change [FC] >1.5; false discovery rate [FDR] <0.01) between patient samples with either high or low expression (upper vs. lower quartile) for RGS1, LDOC1 and L3MBTL4. (B) Heatmap showing clustered expression pattern (Pearson correlation and average linkage) of 12 genes found in all three gene specific signatures and heatmap showing expression pattern of 51 genes found in gene specific signatures of LDOC1 and L3MBTL4. (C) Scatter plots for ZAP70 expression with regard to groups showing high and low LDOC1 and L3MBTL4 expression (upper vs. lower quartile).

622

haematologica | 2022; 107(3)

GEP for high dimensional prognostic models in CLL

Figure 5. Combined status of LDOC1 and L3MBTL4 is correlated with IGHV sequence homology and identifies cases with “discordant” clinical course. The figure highlights the correlation between expression levels of LDOC1 (x-axis), L3MBTL4 (y-axis) and the immunoglobulin heavy chain variable (IGHV) gene sequence homology (color coded). Cases with IGHV sequence homology <98% are indicated in blue, cases with IGHV sequence homology ≥98% are indicated in red. LDOC1 and L3MBTL4 expression identifies “discordant” cases with mutated IGHV but poor clinical course (high expression of LDOC1 and/or L3MBTL4) and vice versa.

cases. Future comparative studies assessing the prognostic impact of methylation markers need to include a comprehensive genetic characterization since SF3B1 and NOTCH1 mutations were found to have independent prognostic and predictive impact for chemoimmunotherapy5 and show a heterogeneous distribution within epigenetic subgroups.29,31 In addition, the CLL8 trial design provided an ideal basis to differentiate between the prognostic and predictive value of markers and therefore to specifically assess for the prognostic strength of established and GEP variables. Notably, GEP variables selected in our model also reliably substituted for IGHV mutation status and showed strong prognostic impact irrespective of treatment for both PFS and OS in contrast to the epigenetic subgrouping.31 While storage and workup conditions were found to change expression levels of multiple transcripts in an RNA sequencing-based study on healthy donor samples, prognostic GEP variables selected in our study largely represented transcripts with low reported variability.32 Stable expression of our prognostic GEP variables selected for the respective clinical endpoints is further supported since prognostic markers unaffected by surrounding conditions (e.g., chromosomal aberrations, gene mutation status) were reliably substituted in the multivariate analysis. Validation of the prognostic impact of selected GEP variables was achieved in an independent data set differing with regard to storage conditions, workup and sorting of samples from a patient cohort with heterogeneous treatment,22 further demonstrating the prognostic robustness of selected GEP variables. While novel compounds have revolutionized the landscape of CLL treatment in particular for high-risk patients,10,11,12,13 the long-term benefit and treatment related toxicities still remain to be evaluated. Further, the significant economic burden may limit the access in some healthcare systems.33 In this study, we were able to confirm that GEP variables can achieve a higher prognostic accuracy, better reflect IGHV sequence homology and reliably identify “discordant” patients with mutated IGHV but poor clinical

haematologica | 2022; 107(3)

course and vice versa. This is especially promising since treatment with BTK inhibitors and FCR was reported with similar PFS in patients with mutated IGHV.14 Although the depth of biological characterization has reached a new dimension with the use of RNA sequencing, both array and RNA sequencing-based prognostic modeling were found to perform equally well for the prediction of major clinical endpoints.34 Studies evaluating FCR and BTK inhibitor treatment in a randomized fashion14 would provide an ideal basis for marker validation using RNA sequencing and easy to apply quantitative real-time polymerase chain reaction based approaches in parallel. Prognostic models used here may therefore hold promise for future selection, substitution and harmonization of prognostic markers, which show variable prognostic value within the respective treatment context. Disclosures The authors declare that there are no conflicts of interest to disclose that interfered with the experiments and presentation of data. Contributions JB, AB and SS conceptualized study; JB performed expression profiling; JB, JK and AB analyzed data. Data were gathered by all authors. JB wrote the paper with input from JK, AB and SS and all authors reviewed the manuscript. Acknowledgements The authors thank all patients and their physicians for trial participation and donation of samples; the DCLLSG; Sabrina Schrell and Christina Galler for their excellent technical assistance; and Myriam Mendila, Nancy Valente, Stephan Zurfluh, and Jamie Wingate for their support in conception and conduct of the trial. Funding This work was supported by grants from BMBF (PRECISE), European Commission / BMBF (“FIRE CLL”, 01KT160), Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 1074 project B1 and B2), DJCLS R 11/01, and F. Hoffmann-La Roche.

623

J. Bloehdorn et al.

References 1. Fischer K, Bahlo J, Fink AM, et al. Long-term remissions after FCR chemoimmunotherapy in previously untreated patients with CLL: updated results of the CLL8 trial. Blood. 2016;127(2):208-215. 2. Keating MJ, O`Brien S, Albitar M, et al. Early results of a chemoimmunotherapy regimen of fludarabine, cyclophosphamide, and rituximab as initial therapy for chronic lymphocytic leukemia. J Clin Oncol. 2005;23(18):4079-4088. 3. Rossi D, Rasi S, Fabbri G, et al. Mutations of NOTCH1 are an independent predictor of survival in chronic lymphocytic leukemia. Blood. 2012;119(2):521-529. 4. Döhner H, Stilgenbauer S, Benner A, et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med. 2000;343(26):1910-1916. 5. Stilgenbauer S, Schnaiter A, Paschka P, et al. Gene mutations and treatment outcome in chronic lymphocytic leukemia: results from the CLL8 trial. Blood. 2014;123(21):32473254. 6. Damle RN, Wasil T, Fais F, et al. Ig V gene mutation status and CD38 expression as novel prognostic indicators in chronic lymphocytic leukemia. Blood. 1999;94(6):18401847. 7. Hamblin TJ, Davis Z, Gardiner A, Oscier DG, Stevenson FK. Unmutated Ig V(H) genes are associated with a more aggressive form of chronic lymphocytic leukemia. Blood. 1999;94(6):1848-1854. 8. Woyach JA, Ruppert AS, Heerema NA, et al. Ibrutinib regimens versus chemoimmunotherapy in older patients with untreated CLL. N Engl J Med. 2018;379(26):25172528. 9. Moreno C, Greil R, Demirkan F, et al. Ibrutinib plus obinutuzumab versus chlorambucil plus obinutuzumab in first-line treatment of chronic lymphocytic leukaemia (iLLUMINATE): a multicentre, randomised, open-label, phase 3 trial. Lancet Oncol. 2019;20(1):43-56. 10. 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. 11. Stilgenbauer S, Eichhorst B, Schetelig J, et al. Venetoclax in relapsed or refractory chronic lymphocytic leukaemia with 17p deletion: a

624

multicentre, open-label, phase 2 study. Lancet Oncol. 2016;17(6):768-778. 12. Byrd JC, Furman RR, Coutre SE, et al. Targeting BTK with ibrutinib in relapsed chronic lymphocytic leukemia. N Engl J Med. 2013;369(1):32-42. 13. 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. 14. Shanafelt TD, Wang V, Kay NE, et al. A randomized phase III study of ibrutinib (PCI32765)-based therapy vs. standard fludarabine, cyclophosphamide, and rituximab (FCR) chemoimmunotherapy in untreated younger patients with chronic lymphocytic leukemia (CLL): a trial of the ECOG-ACRIN cancer research group (E1912). Blood. 2018;132(Supplement 1):LBA-4. 15. Bengtsson H, Simpson K, Bullard J, Hansen K. aroma.affymetrix: A generic framework in R for analyzing small to very large Affymetrix data sets in bounded memory. Tech Reports. 2008;745:1-9. 16. van Buuren S, Groothuis-Oudshoorn K. MICE: multivariate imputation by chained equations in R. J Stat Software. 2011;45(3). 17. Fan J, Li R. Variable selection via nonconcave penalized likelihood and its oracle properties. J Am Stat Assoc. 2001;96:1348-1360. 18. Willi Sauerbrei, Buchholz A, Boulesteix AL, Binder H. On stability issues in deriving multivariable regression models. Biom J. 2015;57(4):531-555. 19. Mogensen UB, Ishwaran H, Gerds TA. Evaluating random forests for survival analysis using prediction error curves. J Stat Softw. 2012;50(11):1-23. 20. Beran R. Nonparametric regression with randomly censored survival data. Tech Report. 1981 University of California, Berkeley. 21. Gerds TA. Prodlim: Product-limit estimation for censored event history analysis 2014. URL https//CRAN. R-project. org/package= prodlim. R Packag. version 1, 460 (2016). 22. Herold T, Jurinovic V, Metzeler KH, et al. An eight-gene expression signature for the prediction of survival and time to treatment in chronic lymphocytic leukemia. Leukemia. 2011;25(10):1639-1645. 23. Duzkale H, Schweighofer CD, Coombes KR, et al. LDOC1 mRNA is differentially expressed in chronic lymphocytic leukemia and predicts overall survival in untreated patients. Blood. 2011;117(15):4076-4084.

24. Morabito F, Cutrona G, Mosca L, et al. Surrogate molecular markers for IGHV mutational status in chronic lymphocytic leukemia for predicting time to first treatment. Leuk Res. 2015;39(8):840-845. 25. Rosenwald A, Alizadeh AA, Widhopf G, et al. Relation of gene expression phenotype to immunoglobulin mutation genotype in B cell chronic lymphocytic leukemia. J Exp Med. 2001;194(11):1639-1647. 26. Rassenti LZ, Huynh L, Toy TL, et al. ZAP-70 compared with immunoglobulin heavychain gene mutation status as a predictor of disease progression in chronic lymphocytic leukemia. N Engl J Med. 2004;351(9):893901. 27. Klein U, Tu Y, Stolovitzky GA, et al. Gene expression profiling of B cell chronic lymphocytic leukemia reveals a homogeneous phenotype related to memory B cells. J Exp Med. 2001;194(11):1625-1638. 28. International CLL-IPI working group. An international prognostic index for patients with chronic lymphocytic leukaemia (CLLIPI): a meta-analysis of individual patient data. Lancet Oncol. 2016;17(6):779-790. 29. Kulis M, Heath S, Bibikova M, et al. Epigenomic analysis detects widespread gene-body DNA hypomethylation in chronic lymphocytic leukemia. Nat Genet. 2012;44(11):1236-1242. 30. Oakes CC, Seifert M, Assenov Y, et al. DNA methylation dynamics during B cell maturation underlie a continuum of disease phenotypes in chronic lymphocytic leukemia. Nat Genet. 2016;48(3):253-264. 31. Wojdacz TK, Amarasinghe HE, Kadalayil L, et al. Clinical significance of DNA methylation in chronic lymphocytic leukemia patients: results from 3 UK clinical trials. Blood Adv. 2019;3(16):2474-2481. 32. Dvinge H, Ries RE, Ilagan JO, Stirewalt DL, Meshinchi S, Bradley RK. Sample processing obscures cancer-specific alterations in leukemic transcriptomes. Proc Natl Acad Sc. U S A. 2014;111(47):16802-16807. 33. Chen Q, Jain N, Ayer T, et al. Economic burden of chronic lymphocytic leukemia in the era of oral targeted therapies in the United States. J Clin Oncol. 2017;35(2):166-174. 34. Zhang W, Yu Y, Hertwig F, et al. Comparison of RNA-seq and microarray-based models for clinical endpoint prediction. Genome Biol. 2015;16(1):133.

haematologica | 2022; 107(3)

ARTICLE

Chronic Lymphocytic Leukemia

Safety and efficacy of the BNT162b mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia

Ferrata Storti Foundation

Ohad Benjamini,1,2 Lior Rokach,3 Gilad Itchaki,4 Andrei Braester,5 Lev Shvidel,6 Neta Goldschmidt,7 Shirley Shapira,8 Najib Dally,9 Abraham Avigdor,1,2 Galia Rahav,10,2 Yaniv Lustig,11 Shirley Shapiro Ben David,8 Riva Fineman,12 Alona Paz,13,14 Osnat Bairey,4 Aaron Polliack,7 Ilana Levy15 and Tamar Tadmor14,15 on behalf of the Israeli CLL study group (ICLLSG) Hematology Division, Chaim Sheba Medical Center, Tel-Hashomer, 2Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv; 3Department of Software and Information Systems Engineering, Ben-Gurion University of the Negev, Beer-Sheva; 4Institute of Hematology, Davidoff Cancer Center, Rabin Medical Center, Petah Tikva; 5Department of Hematology, Galilee Medical Center, Nahariya; 6Hematology Institute, Kaplan Medical Center, Rehovot; 7Hematology, Hadassah Medical Center, Jerusalem; 8Health Division, Maccabi Healthcare Services, Tel Aviv; 9Division of Hematology, Ziv Medical Center, Safed; 10The Infectious Disease Unit, Sheba Medical Center, Tel-Hashomer; 11Central Virology Laboratory, Ministry of Health and Sheba Medical Center, Tel-Hashomer; 12 Department of Hematology and BMT, Rambam Health Care Campus, Haifa; 13 Infectious Disease Unit, Bnai Zion Medical Center, Haifa; 14The Ruth and Bruce Rappaport Faculty of Medicine, Technion, Haifa and 15Hematology Unit, Bnai Zion Medical Center, Haifa, Israel. 1

Haematologica 2022 Volume 107(3):625-634

ABSTRACT

atients with chronic lymphocytic leukemia (CLL) have a suboptimal humoral response to vaccination. Recently, BNT162b2, an mRNA COVID-19 vaccine with a high efficacy of 95% in immunocompetent individuals, was introduced. We investigated the safety and efficacy of the BNT162b2 mRNA COVID-19 vaccine in patients with CLL from nine medical centers in Israel, Overall 400 patients were included, of whom 373 were found to be eligible for the analysis of antibody response. The vaccine appeared to be safe and only grade 1-2 adverse events were seen in 50% of the patients. Following the second dose, an antibody response was detected in 43% of the cohort. Among these CLL patients, 61% of the treatment-naïve patients responded to the vaccine, while responses developed in only 18% of those with ongoing disease, 37% of those previously treated with a BTK inhibitor and 5% of those recently given an anti-CD20 antibody. Among patients treated with BCL2 as monotherapy or in combination with anti-CD20, 62% and 14%, respectively, developed an immune response. There was a high concordance between neutralizing antibodies and positive serological response to spike protein. Based on our findings we developed a simple seven-factor score including timing of any treatment with anti-CD20, age, treatment status, and IgG, IgA, IgM and hemoglobin levels. The sum of all the above parameters can serve as a possible estimate to predict whether a given CLL patient will develop sufficient antibodies. In conclusion, the BNT162b2 mRNA COVID-19 vaccine was found to be safe in patients with CLL, but its efficacy is limited, particularly in treated patients.

Introduction The coronavirus disease 2019 (COVID-19) pandemic has become the main healthcare issue worldwide since its appearance at the end of 2019, with the disease affecting millions of people globally.1 International efforts generated a vaccine against the causative virus, severe acute respiratory syndrome coronavirus-2 (SARS-

haematologica | 2022; 107(3)

Correspondence: TAMAR TADMOR Tamar.tadmor@b-zion.org.il Received: May 11, 2021. Accepted: July 8, 2021. Pre-published: July 29, 2021. https://doi.org/10.3324/haematol.2021.279196

625

O. Benjamini et al.

CoV-2), which was both safe and highly effective. In December 2020, results of a clinical trial using the BNT162b2 mRNA COVID-19 vaccine in a large cohort of people (≥16 years old) demonstrated a 95% efficacy in preventing symptomatic infection.2 This study prompted an emergency use authorization from the Food and Drug Administration.3 In a real-world setting, nationwide vaccination data from Israel documented high efficacy of the vaccine even in patients with multiple comorbidities.4 However, the trial excluded immunosuppressed patients, as their immune response to vaccination is usually blunted. Chronic lymphocytic leukemia (CLL) is associated with varying degrees of immune deficiency due to the primary disease or to the therapy administered. These include functional defects of B- and T-lymphocytes, natural killer cells, neutrophils and macrophages,5 as well as hypogammaglobulinemia, which is evident in treatment-naïve patients and develops even more frequently following therapy for CLL.5,6 Large, retrospective studies from both Europe7,8 and the USA9 have already shown that patients with CLL have an increased rate of COVID-19 infection, a higher hospitalization rate and a greater risk of dying from the virus irrespective of disease stage or prior treatment status. The role of vaccination in these patients is therefore of major importance. However, several studies have already shown that patients with CLL mount limited responses to other more frequently used vaccines, such as those for influenza,10 pneumococcal infection11 and varicella zoster.12 Furthermore, only limited data are available on the response to vaccines in the era of novel biological agents now used to treat CLL, such as BTK and BCL2 inhibitors in combination with anti-CD20 antibodies.13-15 A previous publication described a reduced serological response rate to the BNT162b2 vaccine in patients with CLL, compared to that in healthy controls, particularly during therapy.16 The aim of the study we report here was to investigate the safety and efficacy of the BNT162b2 mRNA COVID19 vaccine in patients with CLL and the effect of therapy on the serological response to the vaccine, given in nine medical centers in different parts of Israel.

Methods Patients This was a prospective, interventional, multicenter study that was performed on behalf of the Israeli CLL study group. The data retrieved from nine centers in Israel provided information on a total of 400 CLL patients who had been vaccinated with the BNT162b2 mRNA COVID-19 vaccine. The diagnosis of CLL was established according to International Workshop CLL criteria.5 Data were collected from medical records after approval from all of the individual institutes’ ethics committees and all patients who participated gave written informed consent. The study was registered at ClinicalTrials.gov NCT04862806 The referring physicians collected demographic and clinical data from the patients’ medical records, including patients’ characteristics (gender, age, Binet stage, medical history including Cumulative Illness Rating Scale [CIRS] score), previous treatments (number and type), and biological features of the CLL (serum IgG, IgA, IgM levels, IGHV mutation status, fluorescence in situ hybridization [FISH] cytogenetic profile, and TP53 mutation) whenever available. 626

Vaccination and immune response assessment All patients received two 30 mg doses of BNT162b2 vaccine (Pfizer), administered intramuscularly 3 weeks apart. Blood samples for immune response evaluation to the vaccine were assessed 2-3 weeks (median, 19 days) after the second dose using anti-spike (S) antibody tests. Anti-spike antibody tests were performed in each hospital using one of three available commercial kits: The Liaison SARS-CoV-2 S1/S2 IgG (Diasorin, Saluggia, Italy), with a positive cutoff of >15 U/mL; the Architect AdviseDx SARS-CoV-2 IgG II (Abbot, Lake Forest, IL, USA), with a positive cutoff of >50 U/mL. and an enzyme-linked immunosorbent assay (ELISA) that detects IgG antibodies against the receptorbinding domain (RBD) of SARS-CoV-2 (positive value >1.1; range 1.1-10).17,18 A surrogate viral assay was used to test antiviral humoral response based on a highly infectious recombinant vesicular stomatitis virus (VSV) bearing the SARS-CoV-2 spike glycoprotein S. This recombinant virus, rVSV-SARS-CoV-2 or SARS-CoV-2 pseudo-virus (psSARS-2), closely resembles SARS-CoV-2 in its entry-related properties. The psSARS-2 neutralization assay was performed using a propagationcompetent VSV-spike similar to the one previously published, which was kindly provided by Gert Zimmer (University of Bern, Switzerland).19

Safety On the day of the serological test, patients were asked to report any adverse events and filled in a questionnaire related to the development of local and systemic adverse events. Patients reported in free text if they had had any adverse events after either vaccination and answered a multiplechoice questionnaire with a scale from zero to five, where zero indicated the lack of any adverse events.

Statistical analysis The characteristics of IgG responders and IgG nonresponders were compared using the Mann-Whitney test for continuous variables, while the Wald c2 test was used for the comparison of categorical variables. Some continuous variables were also tested as categorical variables using the thresholds indicated in the tables and text. A P value <0.05 was considered statistically significant. For the predictive model, multivariate logistic regression was used to predict the response to the vaccine and determine which variables were independently associated with the response. Least absolute shrinkage and selection operator (LASSO) regularization was used to avoid over-fitting and obtain a simpler model which consists of only the informative variables while disregarding the remaining variables. We compared the predictive performance of the LASSO logistic regression model with the simple sevenfactor score. We applied ten repeats of stratified 10-fold cross-validation to estimate the various predictive performance metrics (area under the curve, accuracy, specificity and sensitivity) and their variance. This procedure helps to avoid over-estimation of the predictive performance of LASSO logistic regression. LASSO estimates regression coefficients by maximizing the log-likelihood function, like any other logistic regression, but by adding a constraint that the sum of the absolute values of the regression coefficients is less than or equal to a positive constant.20,21 Thus, LASSO prefers a parsimonious model, penalizing models with too many variables. In particular, if there is a subset of highly correlated haematologica | 2022; 107(3)

COVID-19 vaccine in patients with CLL

variables (e.g., white blood cell count and absolute lymphocyte count), then LASSO tends to select one variable from this set and ignore the others. This helps to avoid selection bias and poor predictive performance in relatively small datasets17 and is therefore very useful in medical applications.20,21

Results A total of 400 patients with CLL were recruited into the trial from nine medical institutes in Israel between December 2020 and February 2021. Our vaccine response analysis is based on the data obtained from 373 patients after excluding the following patients: 14 patients whose serology tests were taken too early (<12 days after the second dose), one patient who had received the Moderna vaccine, nine patients whose antibody tests were not collected and three patients who were infected by SARS-CoV-2 after vaccination. Blood samples were analyzed 2-3 weeks (median 19 days; range, 12-53) after the patients had received the second dose of the vaccine. The median age of the entire cohort was 70 years old (range, 40-89), and 222 (58.9%) were male. The median time since diagnosis of CLL was 83 months for the whole

cohort, and 66 months and 97 months in patients with or without a serological response, respectively. (Table 1)

Side effects of the vaccine Figure 1 and Online Supplementary Table S1 provide details of adverse events following administration of the BNT162b2 mRNA COVID-19 vaccine to patients with CLL. Patients were asked about the development of fever, rash, pain at the site of injection or generalized muscle pain. Of the 331 patients who answered the questionnaire, 180 (54.4%) reported no side effects following the two doses of the vaccine. All side effects that were reported were either grade 1 (41.7%) and/or grade 2 (4%). The most frequent was local pain at the site of injection, which was reported by 32.3% of the cohort. The most frequent grade 2 side effects were pain and fever, reported by 4.3% and 3.6% of the patients, respectively. Other side effects of interest noted by the investigators included one case of facial numbness lasting for 12 h, which resolved without sequelae. In the open question of the questionnaire 12 patients reported fatigue, and eight complained of headache. Two patients developed autoimmune hemolytic anemia, which was detected on the day of the serology test at 18 and 35 days after the second vaccine with hemoglobin levels of 5.6 mg/dL and 4.71 mg/dL, respectively. Both patients were in

Table 1. Clinical and demographic parameters and efficacy of the BNT162b2 mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia.

Variable

Median age in years (range) ≤ 70 years, n (%) > 70 years, n (%) Sex, n (%) Female Male Median time since CLL diagnosis in months (range) Binet Stage,* n (%) A B C Diabetes mellitus, n (%) No Yes Ischemic heart disease, n (%) No Yes Hypertension, n (%) No Yes R-CIRS, median (range) <6, n (%) ≥6, n (%) Lymphadenopathy, n (%) No Yes Splenomegaly, n (%) No Yes

Serological response

Total

Odds ratio (95% CI)

P-value

Positive n=160 (43%)

Negative n=213 (57%)

n=373

69 (40-88) 99 (48%) 61 (37%)

71 (44-89) 109 (52%) 104 (63%)

208 165

1 (ref) 0.65 (0.43-0.98)

65 (43%) 95 (43%)

86 (57%) 127 (57%)

151 222

1 (ref) 0.99 (0.65-1.5)

66 (1-362)

97 (3-341)

75 (66%) 13 (45%) 4 (44%)

38 (34%) 16 (55%) 5 (56%)

113 29 9

1 (ref) 0.41 (0.18-1.1) 0.41 (0.1-1.6)

0.06 0.19

131 (44%) 29 (38%)

165 (56%) 48 (62%)

296 77

1 (ref) 0.761 (0.45-1.27)

0.3

140 (44%) 20 (37%)

179 (56%) 34 (63%)

319 54

1 (ref) 0.7521 (0.41-1.36)

0.35

105 (44%) 55 (41%) 4 (0-11) 84 (48%) 54 (36%)

134 (56%) 79 (59%) 5 (0-19) 91 (52%) 98 (64%)

239 134

1 (ref) 0.8885 (0.58-1.36)

175 152

1 (ref) 0.5969 (0.38-0.93)

41 (41%) 32 (38%)

60 (59%) 53 (62%)

101 85

1 (ref) 0.8836 (0.49-1.6)

0.68

47 (39%) 11 (31%)

73 (61%) 24 (69%)

120 35

1 (ref) 0.7119 (0.32-1.59)

0.41

<0.001 0.04

0.96 <10-4

0.59 0.004 0.02

*Binet stage is at time of vaccination only for treatment-naïve patients. 95% CI: 95% confidence interval; CLL: chronic lymphocytic leukemia; CIRS: Cumulative Illness Rating Scale.

haematologica | 2022; 107(3)

627

O. Benjamini et al.

Figure 1. Side effects of the BNT162b2 mRNA COVID-19 vaccine in patients with chronic lymphocytic leukemia. Side effects were scored on a 0-5 scale with 0=no side effect.

progressive disease years after previous CLL-directed therapy and had no history of autoimmune hemolytic anemia. Both patients tested negative for SARS-CoV-2 antibodies. Their hemoglobin levels improved with corticosteroid therapy and recovered to normal after they were given rituximab and venetoclax. There was no correlation between the presence or severity of side effects and positive serological response (Online Supplementary Table S1; Figure 1). Analyses were performed separately for therapy-naïve and treated patients, but still no correlation was found between the presence or severity of side effects and a positive serological response. However, the adverse event rate was found to be statistically higher among treatment-naïve patients than among previously treated and currently treated patients (Online Supplementary Table S2).

Vaccine efficacy A positive antibody response to the vaccine was evident in only 160 (43%) of all the patients with CLL. In univariate analysis, the following variables were found to be highly statistically significantly (P<0.001) associated with the lack of development of an immune response to the vaccine: low IgG (<700 mg/dL), low IgM (<40 mg/dL), low IgA (<80 mg/dL), platelet count <150x109/L, hemoglobin below normal value, number of prior therapies for CLL, recent anti-CD20 antibody treatment, and currently being treated with BTK inhibitors or BCL2 inhibitors. A few other variables were found to be statistically significantly (P<0.05) associated, including CIRS score >6, age >70 and trisomy 12. There was no significant difference with regard to gender, comorbidities, FISH results except trisomy 12 or IGHV mutational status. Additional information is available in Table 2. Multivariate logistic regression was used to predict response to vaccination and determine which variables were independently associated with the response (Table 4). LASSO regularization was used to avoid over-fitting and obtain a simpler model, which retained only the informative variables while disregarding the remaining variables. 628

The following independent variables were found to be statistically significant: age >70 years, recent treatment with anti-CD20 antibody, ongoing treatment with ibrutinib, IgG <700 mg/dL and IgM <40 mg/dL.

Neutralizing antibodies Samples from 45 patients at Sheba Medical Center were also tested for the production of neutralizing antibodies. A pseudo typed virus system based on VSV was developed for the detection of neutralizing antibodies, instead of using infectious and viable viruses, due to safety concerns. Neutralizing antibodies prevent the pseudovirus from entering the host cells. As shown in Figure 2, the amount of neutralizing antibodies (log transformed) is correlated linearly with anti-COVID-19 RBDIgG titer (r=0.83 and P<0.001). Moreover, as demonstrated in the correlation matrix, 25 of 26 patients with positive IgG were also positive for neutralizing antibodies (the 26th patient was not tested for neutralizing antibodies). Similarly, 14 of 17 patients who were negative for anti-COVID-19 IgG were also negative for neutralizing antibodies (the neutralizing antibodies of the remaining 3 patients were not determined). The Choen κ agreement between IgG and neutralizing antibodies was κ=0.75±0.08 (P<0.001) which is indicative of high concordance between the two tests.

Vaccine efficacy according to treatment status and type of anti-leukemia therapy given One hundred fifty eight (42.3%) patients were treatmentnaïve and of these 97 (61%) developed an IgG response to the vaccine. The immune response was better in treatmentnaïve patients than in previously treated patients and was graded according to CLL disease status (vaccine response better in therapy-naïve patients > complete response > partial response > progressive disease) (Figure 3A) In the treated cohort: an inverse correlation was found between number of lines of prior anti-CLL therapy and the development of a serological response. (Table 2). haematologica | 2022; 107(3)

COVID-19 vaccine in patients with CLL

Figure 2. Correlation between neutralizing antibodies and COVID-19 IgG titer. NEUT Ab: neutralizing antibodies.

Of the 143 (38%) patients previously treated with antiCD20 antibodies, only 38 (27%) responded to the vaccine and there was a significantly lower antibody response rate of only 5% in patients treated with anti-CD20 antibodies within the year of vaccination compared to 35% when the time from anti-CD20 therapy was more than 1 year (Table 3) Analysis of the serological response in 106 patients treated with BTK inhibitors revealed positive serological responses in 23% of the patients .There was a statistical difference between the response rate in patients receiving ongoing BTK inhibitor therapy (18%) and that in previously treated patients (37%). However, there was no statistical difference depending on whether the BTK inhibitor was given within 2 years or more than 2 years from the time of vaccination or depending on whether it was given as firstline therapy or in relapsed disease (Table 2, Figure 3B). Sixty-two patients were treated with BCL2 inhibitors and of them 24% developed a positive serological response. Among those in whom BCL2 inhibition was combined with anti-CD20 antibodies only 14% developed a positive serological response. There was no statistical difference depending on whether patients received BCL2 inhibitor therapy within or more than a year before vaccination, or depending on whether the BCL2 inhibitor was given as first-line therapy or at relapse; however when the BCL3 inhibitor was combined with antiCD20 antibodies, vaccine response rates were lower. Therapy with prophylactic intravenous immunoglobulins also correlated with vaccine response. Additional information is available in Table 2 and Figure 3B, C.

The effect of the vaccine on IgG levels We compared IgG levels before and after vaccination. More specifically we compared IgG levels that were meashaematologica | 2022; 107(3)

ured up to 150 days before the first dose of vaccine (when applicable) with the IgG levels of the corresponding patients at the serology test of this study and found that they were similar (mean levels: 768.89 mg/dL vs. 755.74 mg/dL, respectively).

A simple score to predict response to vaccine in individual patients with chronic lymphocytic leukemia In addition to the multivariate logistic regression model, we generated a simple score based on seven factors (Figure 4): (i) anti-CD20 treatment in the 12 months preceding vaccination (no: +30; yes: 0); (ii) treatment status (treatmentnaïve: +10; previous or ongoing treatment: 0), (iii) age (<70 years. +10; ≥ 70 years: 0); (iv) IgM level (≥40 mg/dL: +10; <40 mg/dL: 0); (v) IgA level (≥80 mg/dL: +10; <80 mg/dL: 0), (vi) IgG level (≥700 mg/dL: +10; <700 mg/dL: 0), and (vii) hemoglobin concentration (normal [i.e., ≥13.5 g/dL for males and ≥12 g/dL for females]: +10; low: 0). All this information is readily available from the clinical history and a routine and affordable blood test. The sum of all the above parameters in the scoring model can be used to estimate the probability of a given CLL patient developing sufficient antibodies after vaccination. For example, a 65-year-old (+10), pretreated (+0) patient but not with anti-CD20 in the preceding 12 months (+30) with normal IgA (+10), IgG (+10) and hemoglobin (+10) levels but abnormal IgM (+0) has a score of 70. i.e., a 70% probability of developing antibodies above the cutoff. Note that according to the model, the maximum score that can be obtained is 90 and not 100, highlighting the fact that even patients with the most favorable indicators are still at risk of not developing a response to vaccination. In the case that the value of a certain factor is missing, we redistribute its score among the other known factors according to their weights. 629

O. Benjamini et al.

Table 2. Response to the BNT162b2 mRNA COVID-19 vaccine based on type of prior therapy given for chronic lymphocytic leukemia.

Variable

Prior therapies, median (range) Treatment- naive, n (%) One, n (%) Two or more, n (%) Ongoing therapy None, n (%) Previous antiCD20, n (%) No Yes Time since anti-CD20, months, median (range) ≥12 months, n (%) < 12 months, n (%)

BTK inhibitor Ongoing BTK inhibitors, n (%) BTK inhibitors in the past, n (%) Treatment duration, n (%) < 2 years ≥ 2 years Line of treatment, n (%) First line Second line or beyond

BCL2 inhibitor Ongoing, n (%) In the past, n (%) Treatment duration, n (%) ≤ 1 year > 1 year Line of treatment, n (%) First line Second line or beyond +/- anti CD20, n (%) Without antiCD20 With anti-CD20 IVIG*, n (%) No Yes

Serological response Positive Negative n=160 (43%) n=213 (57%)

Total

Odds ratio (95% CI)

P-value

n=373

0 (0-4) 97 (61%) 35 (29%) 28 (29%)

1 (0-6) 61 (39%) 85 (71%) 67 (71%)

1 158 120 95

1 (ref) 0.26 (0.16-0.43) 0.26 (0.15-0.45)

<10-5

143 (57%)

110 (43%)

253

1 (ref)

98 (50%) 38 (27%)

99 (50%) 105 (73%)

197 143

1 (ref) 0.37 (0.23-0.58)

48 (10-102) 32 (35%) 2 (5%)

27 (1-132) 60 (65%) 37 (95%)

36 92 39

1 (ref) 0.1 (0.02-0.45)

0.0004

n=24(23%)

n=82(77%)

n=106

14 (18%) 10 (37%)

65 (82%) 17 (63%)

79 27

1 (ref) 2.73 (1.03-7.21)

0.0384

14 (26%) 10 (19%)

40 (74%) 42 (81%)

54 52

1 (ref) 0.68 (0.27-1.71)

0.4103

15 (28%) 9 (17%)

38 (72%) 44 (83%)

53 53

1 (ref) 0.5182 (0.2-1.32)

0.1638

n=15 (24%)

n=47(76%)

n=62

2 (6%) 13 (46%)

32 (94%) 15 (54%)

34 28

1 (ref) 13.87 (2.77-69.38)

0.0002

8 (19%) 7 (35%)

34 (81%) 13 (65%)

42 20

1 (ref) 2.2885 (0.69-7.59)

0.1703

8 (24%) 7 (25%)

26 (76%) 21 (75%)

34 28

1 (ref) 1.0833 (0.34-3.48)

0.893

8 (62%) 7 (14%)

5 (38%) 42 (86%)

13 49

1 (ref) 0.1042 (0.03-0.41)

0.0004

145 (46%) 15 (28%)

172 (54%) 39 (72%)

317 54

1 (ref) 0.46 (0.24-0.86)

0.0137

<10-3 <10-5

0.0001 0.142

*Given monthly. 95% CI: 95% confidence interval; IVIG: intravenous immunoglobulins.

Based on the proposed score, patients could be divided into three groups: low vaccine responsive (<20), intermediate responsive (20-70) and high responsive (>70) with significantly different response rates: 6%, 35% and 75% respectively. The proposed scoring model presented in Online Supplementary Figure S1 was formed by adding two new constraints to the logistic regression model: (A) the sum of absolute logistic transformed coefficients is less than or equal to 100; and (B) each coefficient is multiples of 10. The solution of this constrained scoring model was obtained using IBM CPLEX Optimization Studio. We compared the predictive performance of the LASSO logistic regression model with that of the simple seven-factor models and applied ten repeats of stratified 10-fold crossvalidation to estimate predictive performance and its variance (Online Supplementary Table S3). It can be seen that the proposed risk scoring model has almost the same predictive performance as the LASSO logistic regression model. 630

Discussion This study investigated a large series of patients with CLL following vaccination with the BNT162b2 mRNA COVID19 vaccine in an attempt to better define the safety of the vaccine and the extent of the immune response to it in these cases. It was found that the adverse events in CLL patients were similar to those encountered in immunocompetent populations and were mainly of grade 1-2 severity. In terms of efficacy, the proportion of patients with CLL with an adequate response was lower (43%) than that in the healthy population (97.4%).2 Our results are in keeping with those of previous studies on other vaccines which had already shown the limited efficacy of vaccination in patients with CLL.10-15 In our study patients who were more likely to develop an adequate immune response were younger than 70 years old, had normal hemoglobin and immunoglobulin levels and had not previously received CLL-directed treatment. haematologica | 2022; 107(3)

COVID-19 vaccine in patients with CLL

Table 3. Response to the BNT162b2 mRNA COVID-19 vaccine based on laboratory and genetic parameters.

Variable

del17p, n (%) No Yes del11q, n (%) No Yes del13q, n (%) No Yes trisomy 12, n (%) No Yes TP53 mutation, n (%) No Yes IGHV, n (%) Mutated Unmutated WBC (x109/L) ≤100x109/L, n (%) >100x109/L, n (%) Hemoglobin, mg/dL Normal, n (%) Low, n (%) Platelets, x109/L Normal, n (%) Low, n (%) ANC, x109/L ≥1,500 <1,500 IgG, mg/dL ≥700 mg/dL, n (%) <700 mg/dL, n (%) IgM, mg/dL ≥40 mg/dL, n (%) <40 mg/dL, n (%) IgA, mg/dL ≥80 mg/dL, n (%) <80 mg/dL, n (%) Monoclonal protein, n (%) No Yes

Serologic response Positive Negative n=160 (43%) n=213 (57%)

Total

Odds ratio (95% CI)

P-value

n=373

75 (33%) 9 (30%)

151 (67%) 21 (70%)

226 30

1 (ref) 0.86 (0.38-1.98)

0.727

70 (33%) 14 (32%)

141 (67%) 30 (68%)

211 44

1 (ref) 0.94 (0.47-1.89)

0.8617

49 (33%) 28 (29%)

99 (67%) 67 (71%)

148 95

1 (ref) 0.84 (0.48-1.48)

0.5524

65 (35%) 10 (19%)

123 (65%) 42 (81%)

188 52

1 (ref) 0.45 (0.21-0.96)

0.0346

28 (29%) 3 (43%)

69 (71%) 4 (57%)

97 7

1 (ref) 1.85 (0.39-8.8)

0.4345

19 (40%) 27 (32%) 13.5 154 (44%) 6 (24%) 13.6 116 (51%) 44 (30%) 179 104 (51%) 53 (33%) 3.8 119 (43%) 32 (41%) 844 88 (47%) 37 (28%) 46.8 76 (57%) 43 (26%) 109 81 (51%) 42 (29%)

29 (60%) 57 (68%) 10.8 194 (56%) 19 (76%) 13.055 110 (49%) 103 (70%) 145 101 (49%) 110 (67%) 3.355 156 (57%) 47 (59%) 709 99 (53%) 93 (72%) 24.7 58 (43%) 123 (74%) 65 78 (49%) 105 (71%)

48 84

1 (ref) 0.72 (0.35-1.51)

348 25

1 (ref) 0.4 (0.16-1.02)

226 147

1 (ref) 0.41 (0.26-0.63)

205 163

1 (ref) 0.47 (0.31-0.72)

275 79

1 (ref) 0.89 (0.54-1.48)

187 130

1 (ref) 0.45 (0.28-0.72)

134 166

1 (ref) 0.27 (0.16-0.43)

159 147

1 (ref) 0.39 (0.24-0.62)

67 (45%) 10 (45%)

83 (55%) 12 (55%)

150 22

1 (ref) 1.03 (0.42-2.54)

0.3881 0.086 0.0481 0.002 <0.001 <0.001 0.0004 0.754 0.6613 0.001 <0.001 < 0.001 <10-5 < 0.001 <0.001

0.9447

95% CI: 95% confidence interval, del: deletion; WBC: white blood cell count; ANC: absolute neutrophil count..

Contrariwise, an ineffective response was more frequently seen in older patients (not as reported in the healthy population2) who had received several lines of prior therapies. We also report the negative effect of ongoing therapy with novel anti-CLL agents on the immune response to the vaccine. In our cohort there were 79 and 34 patients receiving ongoing therapy with BTK inhibitors and BCL2 inhibitors, respectively, and less than 20% of them had a response to the vaccine. In addition, our findings also support the observations recorded by others in earlier studies regarding the development of B-cell depletion and late Bcell reconstitution following anti-CD20 antibody treatment.15 In our cohort we noted in particular that patients who had been treated with anti-CD20 antibodies in the 12 months preceding vaccination had a clearly much lower haematologica | 2022; 107(3)

response and only 5% responded effectively to the vaccine. Recently, Herishanu et al.16 reported that the humoral immune response to BNT162b2 mRNA COVID-19 vaccine in 167 patients with CLL from a single center was 39.5%. Similar to our results, response to the vaccine was markedly impaired and was affected by prior treatment status and the type of therapy given. Our study documents real-world experience in a large cohort of patients and, for the first time, also examined neutralizing antibodies following vaccination in patients with CLL. This is important because, as recently reported by Garcia-Beltran et al., SARS-CoV-2 neutralizing antibodies predict the severity of COVID-19 and survival.22 While the results reported here are mostly in accordance with previously published observations,21 in our 631

O. Benjamini et al.

cohort neither gender nor IGHV mutation status was a statistically significant factor for positive response. In addition, by examining the correlation of complete blood count results with response rate, we found that both abnormal hemoglobin and platelet levels were associated with a low response rate. During the course of the study and in the 3-month observation period after the second dose of vaccine only three out of 400 vaccinated patients developed COVID19 infection. One patient acquired the infection between the first dose and the second dose (3 weeks) and two patients 14 days and 24 days after the second dose. All three of them recovered uneventfully. Several studies have indicated that not only does the tumor response play a role in the immune response after vaccination with BNT162b2 mRNA COVID-19 vaccine, but that other factors are also involved. Recent research showed that two doses of 1 mg or 50 mg of BNT162b1 can

Table 4. Multivariate analysis of serological response in patients with chronic lymphocytic leukemia.

Variable Age >70 years Male CIRS score ≥6 Prior therapy 1st line Prior therapy ≥ 2nd line Time since last antiCD20 antibodies ≤12 months IgG <700 mg/dL IgM <40 mg/dL IgA <80 mg/dL Ongoing BTK inhibitor Ongoing BCL2 inhibitor

Odds ratio

95% CI

P-value

0.6543 0.9315 1.0697 0.3013 0.1246

(0.4323-0.9867) (0.6138-1.4131) (0.3281-3.533) (0.0208-3.0227) (0.0082-1.2373)

0.0444 0.7385 0.9104 0.3243 0.0891

0.0874 0.7358 0.3944 0.6052 0.0577 0.1516

(0.0046-0.5103) (0.4199-1.2906) (0.2379-0.6493) (0.3555-1.0278) (0.0069-0.3195) (0.005-2.22)

0.0256 0.0012 <.001 0.0631 0.0029 0.1989

95% CI: 95% confidence interval; CIRS: Cumulative Illness Rating Scale.

Figure 3. Vaccine efficacy according to treatment. Vaccine response rate by treatment status, by treatment type and treatment timing and response to vaccine in patients who were treated or are currently being treated with BKT inhibitors or BCL2 inhibitors. mo: months; BKTi: BKT inhibitor; BCL2i: BCL2 inhibitor.

632

haematologica | 2022; 107(3)

COVID-19 vaccine in patients with CLL

Figure 4. A simple scoring model to predict response to the BNT162b2 mRNA COVID-19 vaccine in individual patients with chronic lymphocytic leukemia. IgM, IgA and IgM levels in mg/dL; Hgb: hemoglobin.

elicit robust CD4+ and CD8+ T-cell responses.23 More data regarding the role of cellular responses in patients with CLL are still awaited in order to establish whether this system provides additional protection or whether the Tcell anergy known to occur in these patients also affects this particular arm of the immune response. However, based on our understanding of immunity to virus vaccinations, T-cell immunity has a major role in generating durable immunity. In addition, as detailed in the varicella-zoster vaccine study referenced above,12 CLL patients can generate potentially effective antigen-specific CD4+ T-cell responses to vaccines even when on treatment with BTK inhibitors. In principle, it seems to be important to be able to predict the response to vaccination in patients with CLL and because of this we have formulated an original, simple, seven-parameter score which can be readily applied worldwide. It should, however, be taken into consideration that we based our model on in vitro markers of humoral immunity that do not necessarily predict clinical benefit and it should, therefore, be used with caution. Our study has several limitations: Firstly, we used three different assays to measure immune response in our cohort of patients and differences between these commercial kits and their reference ranges must be taken into consideration. On the other hand, the results obtained appear to reflect the true "real-world" situation accurately, in which several different kits are being used worldwide with all achieving similar results. Indeed, a study comparing the sensitivity of the various serological assays has already been published indicating a sensitivity of 84.7%, 82.4% and 89.4% for the Abbott, DiaSorin and ELISA kits, respectively.17,18 Other research has also shown strong agreement between the results of different kits.16 A second limitation of our study is that it lacks data regarding possible past exposure or asymptomatic illness to SARS-CoV-2 itself. because the “local policy“ was to

haematologica | 2022; 107(3)

vaccinate only the "non-infected/recovering from COVID-19 infection" population. We feel that this decision could possibly have affected our results but only in a very limited manner. In conclusion, the results of this study showed that the humoral immune response to the BNT162b2 mRNA COVID-19 vaccine is impaired in patients with CLL. We were able to generate a simple seven-parameter score which helps to predict individual immune responses. Further studies are still required to define the exact role of the cellular immune response and the possible effect of a third dose of the vaccine in these patients. In the long run it is our responsibility as a society to ensure that a high percentage of the healthy population is vaccinated so that we can protect more vulnerable individuals with underlying disorders such as CLL who are only partially capable of mounting an effective immune response following vaccination. Disclosures No conflicts of interest to disclose. Contributions TT and OB designed, organized and wrote the manuscript. LR performed the statistical and machine learning analysis including the seven-parameter models, and was involved in writing the manuscript. APol helped to write the manuscript. GI, AB, LS, NG, SS, ND, AA, GRYL, SSBD, RF, APaz, and IL contributed patients’ data. Acknowledgments We thank the study coordinators, with special thanks to Mrs. Andrea Shoukair. Funding This study was supported by a grant from Janssen Pharmaceutical, number EV00261620.

633

O. Benjamini et al.

References 1. Zhu N, Zhang D, Wang W, et al. A novel coronavirus from patients with pneumonia in China, 2019. N Engl J Med. 2020;382(8): 727-733. 2. Polack FP, Thomas SJ, Kitchin N, et al. Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine. N Engl J Med. 2020;383(27):2 603-2615. 3. https://www.fda.gov/emergency-preparedness-and-response/coronavirus-disease-2019-covid-19/pfizer-biontech-covid19-vaccine [last accessed May 19, 2021] 4. Dagan N, Barda N, Kepten E, et al. BNT162b2 mRNA Covid-19 vaccine in a nationwide mass vaccination setting. N Engl J Med. 2021;384(15):1412-1423. 5. Tadmor T, Welslau M, Hus I. A review of the infection pathogenesis and prophylaxis recommendations in patients with chronic lymphocytic leukemia. Expert Rev Hematol. 2018;11(1):57-70. 6. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760. 7. Scarfo L, Chatzikonstantinou T, Rigolin GM, et al. COVID-19 severity and mortality in patients with chronic lymphocytic leukemia: a joint study by ERIC, the European Research Initiative on CLL, and CLL Campus. Leukemia. 2020;34(9):23542363. 8. Furstenau M, Langerbeins P, De Silva N, et al. COVID-19 among fit patients with CLL treated with venetoclax-based combina-

634

tions. Leukemia. 2020;34(8):2225-2229. 9. Mato AR, Roeker LE, Lamanna N, et al. Outcomes of COVID-19 in patients with CLL: a multicenter international experience. Blood. 2020;136(10):1134-1143. 10. Whitaker JA, Parikh SA, Shanafelt TD, et al. The humoral immune response to highdose influenza vaccine in persons with monoclonal B-cell lymphocytosis (MBL) and chronic lymphocytic leukemia (CLL). Vaccine. 2021;39(7):1122-1130. 11. Svensson T, Kattstrom M, Hammarlund Y, et al. Pneumococcal conjugate vaccine triggers a better immune response than pneumococcal polysaccharide vaccine in patients with chronic lymphocytic leukemia. A randomized study by the Swedish CLL group. Vaccine. 2018;36(25): 3701-3707. 12. Zent CS, Brady MT, Delage C, et al. Short term results of vaccination with adjuvanted recombinant varicella zoster glycoprotein E during initial BTK inhibitor therapy for CLL or lymphoplasmacytic lymphoma. Leukemia. 2021;35(6):1788-1791. 13. Pleyer C, Ali MA, Cohen JI, et al. Effect of Bruton tyrosine kinase inhibitor on efficacy of adjuvanted recombinant hepatitis B and zoster vaccines. Blood. 2021;137(2):185189. 14. Shadman M, Ujjani C. Vaccinations in CLL: implications for COVID-19. Blood. 2021;137(2):144-146. 15. Eisenberg RA, Jawad AF, Boyer J, et al. Rituximab-treated patients have a poor response to influenza vaccination. J Clin Immunol. 2013;(33):388-396. 16. Herishanu Y, Avivi A, Aharon A, et al.

Efficacy of the BNT162b2 mRNA COVID19 vaccine in patients with chronic lymphocytic leukemia. Blood. 2021;137(23): 3165-3173. 17. Oved K, Olmer L, Shemer-Avni Y, et al. Multi-center nationwide comparison of seven serology assays reveals a SARS-CoV2 non-responding seronegative subpopulation. EClinicalMedicine. 2020;29;100651. 18. Perkmann T, Perkmann-Nagele N, Breyer MK, et al. Side-by-side comparison of three fully automated SARS-CoV-2 antibody assays with a focus on specificity. Clin Chem. 2020;66(11):1405-1413. 19. Dieterle ME, Haslwanter D, Bortz RH 3rd, et al. A replication-competent vesicular stomatitis virus for studies of SARS-CoV-2 spike-mediated cell entry and its inhibition. Cell Host Microbe. 2020;28(3):486-496. 20. Ribbing J, Nyberg J, Caster O, Jonsson EN. The lasso - a novel method for predictive covariate model building in nonlinear mixed effects models. J Pharmacokinet Pharmacodyn. 2007;34(4):485-517. 21. Kim SM, Kim Y, Jeong K, Jeong H, Kim J. Logistic LASSO regression for the diagnosis of breast cancer using clinical demographic data and the BI-RADS lexicon for ultrasonography. Ultrasonoography. 2018;37 (1):36-42. 22. Garcia-Beltran WF, Lam EC, Astudillo MG, et al. COVID-19-neutralizing antibodies predict disease severity and survival. Cell. 2021;184(2):476-488. 23. Sahin U, Muik A, Derhovanessian E, et al. COVID-19 vaccine BNT162b1 elicits human antibody and TH1 T cell responses. Nature. 2020;586(7830):594-599.

haematologica | 2022; 107(3)

ARTICLE

Complications in Hematology

Methotrexate-related central neurotoxicity: clinical characteristics, risk factors and genome-wide association study in children treated for acute lymphoblastic leukemia Marion K. Mateos,1-4 Glenn M Marshall,1-3 Pasquale M. Barbaro,5,6 Michael C.J. Quinn,7 Carly George,8,9 Chelsea Mayoh,2,3 Rosemary Sutton,2,3 Tamas Revesz,10 Jodie E Giles,3 Draga Barbaric,1 Frank Alvaro,11,12 Françoise Mechinaud,13,14 Daniel Catchpoole,15 John A. Lawson,2,16 Georgia Chenevix-Trench,7 Stuart MacGregor,7 Rishi S.Kotecha,8,17,18 Luciano Dalla-Pozza5,19,20 and Toby N. Trahair1-3

Ferrata Storti Foundation

Haematologica 2022 Volume 107(3):635-643

1 Kids Cancer Center, Sydney Children’s Hospital Randwick, Sydney, New South Wales, Australia; 2School of Women and Children’s Health, University of New South Wales (UNSW), Sydney, New South Wales, Australia; 3Children’s Cancer Institute, Lowy Cancer Research Center, UNSW, Sydney, New South Wales, Australia; 4Northern Institute for Cancer Research, Wolfson Childhood Cancer Research Center, Newcastle-Upon-Tyne, UK; 5Children’s Medical Research Institute, University of Sydney, Sydney, New South Wales, Australia; 6Department of Hematology, Queensland Children’s Hospital, Brisbane, Queensland, Australia; 7QIMR Berghofer Medical Research Institute, Brisbane, Queensland, Australia; 8Perth Children’s Hospital, Perth, Western Australia, Australia; 9Division of Pediatrics, School of Medicine, University of Western Australia, Perth, Western Australia, Australia; 10Women’s and Children’s Hospital, Adelaide, New South Wales, Australia; 11John Hunter Children’s Hospital, Newcastle, New South Wales, Australia; 12University of Newcastle, Newcastle, New South Wales, Australia; 13The Royal Children’s Hospital, Melbourne, Victoria, Australia; 14Service d’Immunohématologie Pédiatrique, Hôpital Robert-Debré, Paris, France; 15The Tumor Bank, Children’s Cancer Research Unit, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia; 16Department of Neurology, Sydney Children’s Hospital Randwick, Sydney, New South Wales, Australia; 17Telethon Kids Cancer Center, Telethon Kids Institute, University of Western Australia, Perth, Western Australia, Australia; 18School of Pharmacy and Biomedical Sciences, Curtin University, Perth, Western Australia, Australia; 19Cancer Center for Children, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia and 20Children’s Cancer Research Unit, The Children’s Hospital at Westmead, Sydney, New South Wales, Australia

ABSTRACT ymptomatic methotrexate-related central neurotoxicity (MTX neurotoxicity) is a severe toxicity experienced during acute lymphoblastic leukemia (ALL) therapy with potential long-term neurologic complications. Risk factors and long-term outcomes require further study. We conducted a systematic, retrospective review of 1,251 consecutive Australian children enrolled on Berlin-Frankfurt-Münster or Children's Oncology Group-based protocols between 1998-2013. Clinical risk predictors for MTX neurotoxicity were analyzed using regression. A genome-wide association study (GWAS) was performed on 48 cases and 537 controls. The incidence of MTX neurotoxicity was 7.6% (n=95 of 1,251), at a median of 4 months from ALL diagnosis and 8 days after intravenous or intrathecal MTX. Grade 3 elevation of serum aspartate aminotransferase (P=0.005, odds ratio 2.31 [range, 1.28–4.16]) in induction/consolidation was associated with MTX neurotoxicity, after accounting for the only established risk factor, age ≥10 years. Cumulative incidence of CNS relapse was increased in children where intrathecal MTX was omitted following symptomatic MTX neurotoxicity (n=48) compared to where intrathecal MTX was continued throughout therapy (n=1,174) (P=0.047). Five-year central nervous system relapse-free survival was 89.2±4.6% when intrathecal MTX was ceased compared to 95.4±0.6% when intrathecal MTX was continued. Recurrence of MTX neurotoxicity was low (12.9%) for patients whose intrathecal MTX was continued after their first episode. The GWAS identified single-nucletide polymorphism associated with MTX neurotoxicity near genes regulating neuronal growth, neuronal differentiation and cytoskeletal organization (P<1x10-6). In conclusion, increased serum aspartate aminotransferase and age ≥10 years at diagnosis were independent risk factors for MTX neurotoxicity. Our data do not support cessation of intrathecal MTX after a first MTX neurotoxicity event.

haematologica | 2022; 107(3)

Correspondence: MARION K. MATEOS m.mateos@unsw.edu.au Received: August 18, 2020. Accepted: February 2, 2021. Pre-published: February 11, 2021. https://doi.org/10.3324/haematol.2020.268565

635

M.K. Mateos et al.

Introduction

Table 1. Baseline demographics.

Diagnostic information Methotrexate-related central neurotoxicity (MTX neurotoxicity) occurs in 3-7% of children treated for childhood acute lymphoblastic leukemia (ALL)1-3 and is characterized by seizures, stroke-like symptoms, speech disturbances and encephalopathy.4 Although commonly encountered during clinical practice, questions remain relating to risk factors, choice of further intrathecal (IT) chemotherapy following MTX neurotoxicity and longterm neurological outcomes, which are not well defined. Previously described risk factors for symptomatic MTX neurotoxicity include older age1,3,5 and an interaction with cytarabine and cyclophosphamide treatment blocks.5 Upon re-exposure to MTX in children with ALL, MTX neurotoxicity recurs in 7-24% of cases.1,3,5 However, there is a paucity of data on central nervous system (CNS) relapse rates if CNS-directed therapy is modified following MTX neurotoxicity.6 Genomic risk predictors may relate to neuronal development, MTX clearance or folate metabolism by-products such as homocysteine.1,7-11 Here we sought to analyze clinical and germline DNA factors that may predict susceptibility to MTX neurotoxicity during treatment for childhood ALL. Additionally, we assessed the incidence of neurological sequelae following an episode of MTX neurotoxicity, and, the impact that changes made to therapy following MTX neurotoxicity had on subsequent relapse risk.

Methods We conducted a national Australian retrospective review of children diagnosed with ALL or lymphoblastic lymphoma (LBL). Consecutive patients treated on frontline Berlin-FrankfurtMunster (BFM) or Children’s Oncology Group (COG) ALL protocols between 1998 and 2013, were eligible for inclusion (see the Online Supplementary Appendix). The cohorts were assembled as part of the ERASE (Evaluation of Risk of ALL treatment-related Side-Effects) study.12 Complete clinical data were collected for 1,251 children (Online Supplementary Figure S1). The study was approved by Hunter New England Human Research Ethics Committee (HNEHREC reference number: 12/11/21/4.01). We defined MTX neurotoxicity as symptomatic neurotoxicity, with or without leukoencephalopathy, temporally related to intravenous (IV) or IT MTX, where MTX was deemed clinically or through record review as the most likely cause, and where other causes had been reasonably excluded. Symptomatic neurotoxicity included motor deficits, speech deficits, visual disturbances, seizures and altered level of consciousness. Toxicity was graded according to the Common Terminology Criteria for Adverse Events (CTCAE) v4.03.13 Long-term neurological outcomes occurring after ALL/LBL diagnosis, focusing on epilepsy, were collected on all patients where available (n=522). Statistical analysis was performed using IBM SPSS for Macintosh, Versions 23.0/24.0 (see the Online Supplementary Appendix). Overall survival (OS) was computed from date of diagnosis through to date of last contact or death from any cause. Event-free survival (EFS) was assessed as time from diagnosis to first event or last contact date in first complete remission (CR1). An event was defined as relapse, death from any cause or secondary malignancy. Leukemia-free survival (LFS) was determined from date of remission to first relapse or last contact date in CR1. CNS relapse was defined as leukemic relapse involving the CNS,

636

Male DIAGNOSIS Pre-B-ALL B-lymphoblastic lymphoma (B-LBL) T-ALL T-lymphoblastic lymphoma (L-LBL) Other (ALL/LBL, not specified) TREATMENT PROTOCOL BFM-based protocols ANZCHOG Study 7 (1998-2002)36 BFM-95 (1998-2002)37 ANZCHOG Study 8 (2002-2012)38 COG A5971 (2003-2009)39 iBFM-Study 9 (2012-2016)40 COG-based protocols CCG1882 1991-1995)41 CCG1952 (1996-2000)42 CCG1961 (1996-2002)43 CCG1991 (2000-2005)44,45 AALL0031 (2002-2006)46 AALL0232 (2004-2011)47 AALL0331 (2005-2010)48 AALL0434 (2007-2014)45,48 AALL08P1 (2009-2011)49 AALL0932 (2010 - 2018)45 AALL1131 (2012 –[part-closure])45

Number (n=1,251)

% of cohort

696

55.64

1,068 14 110 39 20

85.37 1.12 8.79 3.12 1.6

(n=1,033) 239 125 608 21 40 (n=218) 1 16 36 54 2 49 25 12 2 17 4

19.10 9.99 48.60 1.68 3.20 0.08 1.28 2.88 4.32 0.16 3.92 2.00 0.96 0.16 1.36 0.32

B-ALL: B-cell acute lymphoblastic leukemia; T-ALL: T-cell acute lymphoblastic leukemia; B-LBL: B-cell lymphoblastic lymphoma; BFM: Berlin-Franfurt-Münster; COG: or Children's Oncology Group.

either as an isolated CNS or combined CNS relapse. Thirty-seven variables were assessed in univariate logistic regression. Factors from univariate logistic regression analysis with a P-value <0.0014 were considered significant, using Bonferroni correction for multiple testing. Variables associated with MTX neurotoxicity (unadjusted P<0.10) were further assessed in multivariable regression analysis, adjusting for gender (see the Online Supplementary Appendix).

Genome-wide association study methods Genotyping was conducted on the Illumina Infinium OncoArray-530K Beadchip (533,000 single-nucleotide polymorphisms [SNP]), using stored bone marrow or peripheral blood DNA, collected in remission14 from children treated on BFM-based protocols. Of the 1,021 patients treated using BFM-based therapy, 932 had an available banked DNA sample. After quality control and filtering (see the Online Supplementary Appendix), the resultant cohort included 707 individuals of European ancestry, as previously described.15 Children who experienced neurotoxicity, either central or peripheral, other than MTX neurotoxicity were excluded (n=122), but those who experienced MTX neurotoxicity in addition to another type of neurotoxicity were included. The resultant genome-wide association study (GWAS) cohort of 585 patients included 48 cases and 537 controls (Online Supplementary Figure S2). After filtering for information score >0.4, the total number of

haematologica | 2022; 107(3)

Methotrexate neurotoxicity: risk factors & outcome

Table 2. Clinical risk factors significantly associated with methotrexate neurotoxicity in regression analysis.

Clinical Risk Factor Age at diagnosis (continuous) Age≥10 years HR/VHR treatmenta Bilirubin at diagnosis Peak bilirubinb Peak bilirubin > Grade 3b Peak ALTb Peak ALT > Grade 3b Peak ASTb Peak AST > Grade 3b Positive blood cultureb Peak creatinine > 2 x baseline and abnormalb Insulin requirementb Hyperglycemiab

Univariate OR

95% CI

<0.001 <0.001 0.003 0.026 0.04 0.03 0.029 0.043 0.018 0.004 0.011 0.005 <0.001 0.001

1.01 2.65 2.05 1.03 1.01 2.6 1.001 1.62 1.001 2.38 1.81 3.27 4.39 2.42

1.006-1.014 1.69-4.14 1.29-3.28 1.003-1.05 1.000-1.014 1.10-6.13 1.000-1.002 1.02-2.57 1.000-1.002 1.33-4.26 1.15-2.85 1.44-7.45 2.36-8.19 1.43-4.10

Multivariate OR

95% CI

0.003

2.43

1.35-4.38

0.005

2.31

1.28-4.16

Factors that were associated with methotrexate (MTX) neurotoxicity in logistic regression analysis at a significance level P<0.05 (2-tailed) are listed. aComparison was made between a combined high-risk cohort (high-risk "HR" and very high-risk "VHR" ALL) versus a combined non high-risk cohort (standard and medium risk ALL). bIn induction/consolidation. ALT: alanine aminotransferase; AST: aspartate aminotransferase; OR: odds ratio; CI: Confidence Interval. ALT and AST elevation > grade 3 refers to levels >5x upper limit of normal (ULN). Bilirubin >grade 3 is >3x ULN based on NCI CTCAE v4.03 criteria. Significance levels were set at P<0.0014 and P<0.05 for univariate and multivariate analyses respectively.

SNP for evaluation in the GWAS for MTX neurotoxicity were 10,838,245. Imputation was conducted on 388,439 SNP (including SNP on the X chromosome) using IMPUTE2.16 GWAS was conducted correlating potential SNP with MTX neurotoxicity, using age, sex and principal components as covariates.

Results Incidence of methotrexate neurotoxicity The clinical cohort consisted of 1,251 children with ALL/LBL (Online Supplementary Figure S1) from six tertiary Australian pediatric oncology centers. There were 1,033 children treated on BFM-based protocols and 218 on COG-based protocols (Table 1). The median age at diagnosis was 59 months (range, 9-218 months) with a median duration of follow-up of 78 months (range, 3-186 months). Five-year OS, LFS and EFS were 92±0.8%, 85.6±1.1% and 83.8±1.1% respectively. Comparative IT, IV MTX and leucovorin (folinic acid) doses according to each protocol are listed in the Online Supplementary Appendix and the Online Supplementary Table S1. IT MTX was given during induction, consolidation, CNS-directed phases, reinduction, reconsolidation and in some protocols IT MTX continued in maintenance. There were a varying number of IT MTX doses in induction. In consolidation, all protocols used IT MTX, while most used cyclophosphamide and cytarabine. Age-directed doses varied between protocols for children aged 9 years and older. For protocols that used IV MTX, doses varied considerably and could be divided into protocols that administered low (0-2.5 g/m2 total IV MTX doses) and high (20-35 g/m2total IV MTX dose) doses. Ninety-five children (7.6% of the cohort) fulfilled the criteria for MTX neurotoxicity. Ninety-one patients experienced MTX neurotoxicity within 21 days following IV or IT MTX (Online Supplementary Table S2). Of four patients with MTX neurotoxicity and leukoencephalopathy >21 days following MTX, three patients had typical symptoms and leukoencephalopathy diagnosed at 22, 30 haematologica | 2022; 107(3)

and 47 days post MTX, while one patient had seizures and leukoencephalopathy 56 days after MTX. Further information is contained in the Online Supplementary Table S2. All ninety-five patients were included in subsequent analyses. The median time from ALL diagnosis to onset of MTX neurotoxicity was 4 months (range, 0-19 months, interquartile range [IQR], 2-6 months), occurring at a median of 8 days following IV or IT MTX administration (range, 0-56 days, IQR, 5-11 days). The first episode of MTX neurotoxicity most frequently occurred after induction/consolidation in 55.8% (n=53 of 95), while 44.2% occurred during induction/consolidation. Specific timing of these events is shown in the Online Supplementary Figure S3. MTX neurotoxicity first events were associated with co-administration of IT MTX, cyclophosphamide and cytarabine in 53.2% patients (n=50 of 94; one patient unknown). In comparison, 23.2% (n=22 of 95) of first events were associated with concurrent IV MTX and IT MTX administration. Grade 1 neurotoxicity occurred in 4.2% of the cohort (n=4); grade 2 in 62.1% (n=59); grade 3 in 31.6% (n=30) and grade 4 in 2.1% (n=2).

Risk factors for methotrexate neurotoxicity - clinical risk factors Clinical variables were assessed for association with MTX neurotoxicity in univariate logistic regression (Table 2 for those associated P<0.05, Online Supplementary Appendix). Variables with a P-value of association <0.0014 were considered significant, taking into account 37 variables as per the Bonferroni method of adjustment for multiple testing (P<0.05/37). Factors listed in Table 2 were analyzed further in a multivariable regression analysis. This demonstrated that age ≥10 years and peak serum AST during induction/consolidation were independent risk associations for MTX neurotoxicity (Table 2). Overall, 56.2% (n=703 of 1,251) of the cohort were available for multivariable analysis, due to availability of AST values during induction/consolidation. These two risk variables 637

M.K. Mateos et al.

were associated with 48.4% of MTX neurotoxicity events. There was no association between body mass index (BMI) Z scores and neurotoxicity, using BMI from diagnosis (P=0.133), end of induction (P=0.788), or end of consolidation (P=0.375).

ing the IV MTX phases (P=0.792), use of vincristine during consolidation versus no vincristine (P=0.50) or higher ageadjusted IT MTX doses for children aged ≥9 years, were not significantly associated with MTX neurotoxicity (P=0.778).

Central nervous system disease at diagnosis

Outcome following first episode of methotrexate neurotoxicity

Leukemic involvement of the CNS at diagnosis was not significantly higher for children who experienced MTX neurotoxicity (P=0.414, Pearson c2) (Online Supplementary Table S3).

Dosing of methotrexate MTX levels at 24, 48 or 54 hours following a first course of HD MTX did not correlate with incidence of MTX neurotoxicity (P=0.964, P=0.81 and P=0.866 respectively). Incidence of delayed MTX excretion is outlined in the Online Supplementary Appendix. There was a significantly higher incidence of MTX neurotoxicity on protocols with a lower cumulative IV MTX dose (total dose 0-2.5 g/m2, n=20 of 148, 13.5%) compared to protocols with a higher cumulative IV MTX dose (2035 g/m2, n=69 of 863, 8.0%, P=0.031, odds ratio [OR] 1.8, 95% Confidence Interval [CI]: 1.06-3.06). This difference remained when stratified for high-risk protocols, with cumulative incidence of MTX neurotoxicity of 33% (n=11 of 33) with lower dose IV MTX protocols compared to 9.7% (n=13 of 134) with higher-dose MTX protocols respectively (P<0.001). Other potential differences, such as use of vincristine during the IV MTX phases on COG-protocols, compared to BFM-based protocols which do not use vincristine dur-

We assessed the incidence of several clinical outcomes: including rates of CNS relapse, LFS, recurrent MTX neurotoxicity and long-term neurological outcomes in children who experienced a first episode of MTX neurotoxicity. Out of 95 patients who experienced MTX neurotoxicity, 48 (50.5%) had subsequent IT MTX permanently discontinued, 34 (35.8%) had IT MTX continued, and one patient developed MTX neurotoxicity following their last scheduled IT MTX dose. Subsequent IT management could not be determined for 12 (12.6%) cases. In patients where IT MTX was permanently discontinued, subsequent IT therapy included cytarabine alone (n=23), cytarabine and hydrocortisone (n=17), no further IT agents (n=2, in maintenance and Protocol M respectively), and in six patients the alternative IT regime was unknown. We examined CNS relapse risk in patients who experienced MTX neurotoxicity, and assessed the impact of IT MTX continuation on CNS relapse rates. Cases were excluded if IT MTX management was unknown following neurotoxicity (n=12), where no further IT MTX doses were scheduled as per protocol (n=1), or where IT MTX was ceased due to seizures from a different etiology (n=2). In the overall cohort where IT MTX was continued, 49 of 1.178 children experienced a CNS-based relapse (4.2%).

Figure 1. Cumulative incidence of central nervous system relapse, according to intrathecal methotrexate strategy. Children who had intrathecal (IT) methotrexate (MTX) omitted permanently following symptomatic MTX neurotoxicity (n=48), had an increased risk of central nervous system (CNS) relapse compared to children who had IT MTX continued through ALL treatment (n=1,174) (P=0.047). Five-year CNS relapse-free survival was 95.4±0.6% when IT MTX was continued compared to 89.2±4.6%, when IT MTX was ceased.

638

haematologica | 2022; 107(3)

Methotrexate neurotoxicity: risk factors & outcome

Six relapses occurred among children who had IT MTX permanently ceased following MTX neurotoxicity (n=48), and five of 48 were CNS-based relapses (10.4%). In 34 patients where IT MTX was continued following MTX

neurotoxicity, there was one isolated CNS relapse (2.9%). Cumulative incidence of CNS relapse was increased in children where IT MTX was omitted permanently following symptomatic MTX neurotoxicity (n=48) compared to

Table 3. Top single-nucleotide polymorphisms for methotrexate neurotoxicity with significance levels P<1x10-6.

CHR

Position

Effect allele

MAF

20196934 rs4712462

0.35

2.54E-07 0.26

0.16

0.45

14590919 rs2241357

0.2

3.60E-07 4.18

2.41

7.24

195925355 rs1106479

0.16

4.08E-07 3.97

2.33

6.75

rs35307996

0.2

5.70E-07 0.11

0.03

0.34

14571966 rs74956940

0.23

6.19E-07 3.58

2.17

5.92

124286453 rs62576054

0.18

7.50E-07 16.7

3.92

71.13

95072136 rs9590003

0.11

9.73E-07 5.24

2.73

10.07

747700

OR OR 95% CI OR 95% CI Genea Location (lower) (upper)

Non effect allele

SNP

Gene function

MBOAT1 intron

Involved in fatty acid biomembrane synthesis. Possible role in neuronal growth and differentiation. siRNA-mediated knockdown of MBOAT1 leads to reduced neurite growth26 GIPC1 intron Encodes for GIPC1, a scaffolding protein that regulates cell surface receptor dynamics in endothelial cells. Microdeletionsat this locus (19p13.12) have been reported25 b ZDHHC19 intron Involved in post-translational protein modification that alters trafficking, activity and localization of lipidated proteins, such as R-Ras29 NXN intron Encoded protein acts as a redox-dependent regulator of the Wnt signaling pathway, exerting effects via dishevelled (dvl), and is involved in cell growth and differentiation27 PKN1 intron Located at 19p13.12. Encodes a protein that is involved in neuronal function and neuroprotection. siRNA-mediated knockdown of PKN1 led to neuron apoptosis and inhibition of neurite formation28 HMGB1P37 unknown HMGB1 family of genes contain an HMG-box domain that bends DNA, affects transcription and facilitates DNA-protein interactions50 -

This table shows the top seven single-nucleotide polymorphisms (SNP) that associate with methotrexate (MTX) central neurotoxicity at significance level P<1x10-6, ordered by P-value for significance. SNP with a minor allele frequency (MAF) <1% were excluded. aThe annotated gene was determined by cross-referencing relevant genomic databases (see the Online Supplementary Appendix). bintron (downstream variant/nc transcript variant). SNP with a minor allele frequency (MAF) <1% were excluded. Odds ratio (OR) 95% Confidence Interval (CI) (lower) and OR 95% CI (upper) refer to lower value and upper value for 95% CI for odds ratio. P-value from genome wide association study, where P<5x10-6 is significant. CHR: chromosome; siRNA: small interfering RNA.

haematologica | 2022; 107(3)

639

M.K. Mateos et al.

Figure 2. Timing of methotrexate neurotoxicity and risk of leukemic relapse. There is a significantly increased rate of leukemic relapse in children who experienced methotrexate (MTX) neurotoxicity early in therapy (induction/consolidation) compared to later in therapy (after consolidation, P=0.011). Five-year leukemia-free survival (LFS) in children who experienced early MTX neurotoxicity was 79.3±6.5% compared to 96±2.8% for children who experienced late MTX neurotoxicity.

children who had IT MTX continued through ALL treatment (n=1.174 with available LFS data) (P=0.047) (Figure 1). Five-year CNS relapse-free survival was 95.4±0.6% when IT MTX was continued compared to 89.2%±4.6% when IT MTX was ceased. For children who experienced MTX neurotoxicity and had IT MTX ceased (n=48) or continued (n=34), there was no difference in CNS status (P=0.536) or age (age <10 vs. ≥10 years, P=0.560) (Online Supplementary Table S4). There was a difference in risk group status, however regression analysis did not show an impact of risk group on incidence of CNS relapse among the whole cohort (P=0.291, HR for high-risk group 0.72, 95% CI: 0.39-1.32). Age ≥10 years was also not associated with risk of CNS relapse (P=0.56, HR 0.83, 95% CI: 0.45-1.55). Distribution of MTX neurotoxicity by protocols and cases evaluable for recurrent MTX neurotoxicity are outlined in the Online Supplementary Table S5. In the group where IT MTX was ceased, two of 48 had HD MTX ceased, three of 48 had HD MTX reduced, two of 48 had asparaginase ceased after neurotoxicity, and one of 48 had dexamethasone ceased in maintenance after development of osteonecrosis. Overall, there were ten relapses in patients who experienced MTX neurotoxicity, six of whom had IT MTX ceased (and one of these patients also had HD MTX ceased). There were no other treatment modifications in these patients who experienced relapse. Specifically, in HR patients who experienced relapse (n=4), two had IT MTX ceased, two had IT MTX continued, and there were no other treatment alterations. Radiotherapy details are contained in the Online Supplementary Appendix.

Recurrent neurotoxicity Six patients had recurrent neurotoxicity events, four of 640

whom had received further IT MTX and two where ongoing IT MTX exposure could not be determined. An additional three patients who had been re-exposed to MTX did not have enough clinical information available to determine whether a subsequent neurotoxicity episode occurred. Therefore, recurrence of MTX neurotoxicity in evaluable patients occurred in 12.9% (n=4 of 31) upon IT MTX re-exposure. Three of the four patients did not receive any further IT MTX following recurrent MTX neurotoxicity. Further information regarding recurrent neurotoxicity events is shown in the Online Supplementary Tables S5 and S6. For patients who were rechallenged with IT MTX, six patients were treated according to protocols that suggest oral leucovorin with IT rechallenge, and 28 were treated on protocols without that recommendation. Of those patients evaluable for recurrent MTX neurotoxicity, there was one recurrence out of five patients who were treated on protocols that recommended oral leucovorin; and three recurrences out of 26 patients who were treated on protocols that did not recommend oral leucovorin with IT rechallenge. Of the patients who had IT MTX ceased after a first MTX neurotoxicity event, 95.8% (n=46 of 48) did not experience further neurotoxicity episodes. One patient experienced unusual ataxic episodes during a phase where oral MTX was administered, however further clinical information was not available and in the other patient the information was unknown.

Methotrexate neurotoxicity: negative impact of occurrence during early therapy Those who experienced MTX neurotoxicity early (during induction/consolidation, n=42) had a higher subsequent risk of leukemia relapse (all sites) compared to those haematologica | 2022; 107(3)

Methotrexate neurotoxicity: risk factors & outcome

who experienced late MTX neurotoxicity after consolidation (n=53, P=0.011, Figure 2) (Cox regression P=0.036, HR 1.92 [95% CI: 1.04-3.54]). A similar number of patients had IT MTX ceased in both groups (21 of 42 in the group with early MTX toxicity; 27 of 53 who experienced late MTX toxicity). ALL risk-group did not vary significantly between those who experienced MTX neurotoxicity before or after consolidation (P=0.874).

Long-term neurological outcome Long-term neurological outcome was reported at last follow-up (median 94 months, range, 3-181 months) for 41.7% (n=522 of 1,251) of patients. There was no reported mortality due to long-term neurological problems following MTX neurotoxicity. A total of 1.3% (n=7 of 522) were diagnosed with epilepsy at last follow-up. Three out of these seven children were diagnosed with epilepsy following symptomatic MTX neurotoxicity, and all remained in CR1.

Genome-wide association study results There was no difference in the incidence of MTX neurotoxicity among the subgroup who were included and excluded from the GWAS, in relation to age, sex, timing and CTCAE grade of toxicity (Online Supplementary Table S7). We found seven intronic SNP that were associated with MTX neurotoxicity at a significance level of P<1x10-6 (Table 3), which mapped to six genes (MBOAT-1, GIPC1, ZDHHC19, NXN, PKN1, HMGB1P37). The rare alleles at SNP near MBOAT-1 and NXN were protective (OR <1) while those at the other SNP correlated with increased risk of MTX neurotoxicity. There were 53 additional SNP (minor allele frequency [MAF] >2%) associated with MTX neurotoxicity at a significance level of P<5x10-6, which mapped near to seven genes (Online Supplementary Table S8). Further understanding of SNP function and roles was determined using contemporary online data repositories (see the Online Supplementary Appendix). Most SNP (P<5x10-6) were intronic, except for rs76301301 (3’ untranslated region [UTR] variant, P=1.34x10-6) located within GIPC1 and rs7555699 located within 2 kb upstream of a 5’ end of BMP8A (P=4.54x10-6) (Online Supplementary Table S9).

Discussion This is the largest cohort of children treated for ALL/LBL mapped for the incidence, risk factors, and long-term impact of MTX toxicity. Independent risk factors for symptomatic MTX neurotoxicity were age ≥10 years at diagnosis and >grade 3 elevation of serum AST during early therapy. Discontinuation of IT MTX, in an attempt to minimize further neurotoxicity after a first MTX neurotoxic event, was associated with an increased incidence of CNS relapse. Patients continuing IT MTX had only a small risk of MTX neurotoxicity recurrence (<13%). Regardless of subsequent IT management strategy, children who developed MTX neurotoxicity had an increased risk of epilepsy. We report on independent risk factors in multivariable analysis that are associated with symptomatic methotrexate neurotoxicity: age ≥10 years at diagnosis and >grade 3 elevation of serum AST during induction/consolidation. haematologica | 2022; 107(3)

Importantly, these factors were determined in a combined cohort of Australian children treated for ALL on either BFM or COG-based therapy. AST elevation in the current series may reflect direct hepatotoxicity from methotrexate,17 systemic metabolic disturbances,18 release from nonhepatic sources such as erythrocytes,18 or a marker of longer serum exposure to MTX as has been shown for ALT elevation post HD MTX.17 Older age has been previously reported as a risk factor in univariate1 and recently in multivariable analysis in a smaller cohort.19 Potential reasons include reduced clearance of HD MTX20 or higher steady state MTX concentration following IV administration21 in older children. Protocols that used higher IT doses (15 mg) for children ≥9 years of age were not associated with increased neurotoxicity in our study. The observed increase in cumulative incidence of CNS relapse following permanent cessation of IT MTX therapy requires validation in contemporaneous cohorts. In study POG 9005 (1991-1995), which had the same number of cases of acute MTX neurotoxicity (n=95), 54 patients had intrathecal therapy modified without an increase in CNS relapse.6 CNS-relapse rates in our overall cohort (n=1,251) were consistent with published rates.22,23 Treatment of CNS relapse involves exposure to additional neurotoxic agents including CNS irradiation.24 Taken together, these data suggest that clinicians should not cease IT MTX therapy after a first episode of MTX neurotoxicity, especially when it occurs early in therapy. Out of the 95 patients identified with MTX central neurotoxicity, 53 fulfilled the Ponte di Legno Toxicity Working Group MTX stroke-like syndrome (SLS) definition that was published after our study commenced, i.e., symptoms within 21 days of IT/IV MTX, characteristic clinical course and/or imaging, with exclusion of other causes.4 All patients included in the GWAS experienced MTX neurotoxicity <21 days from IT/IV MTX. We did not identify any SNP associated with MTX neurotoxicity at genome-wide levels of significance, but report seven SNP at P<1x10-6 which require replication in larger, independent studies of MTX neurotoxicity. These seven SNP mapped near six genes, five of which have potential roles in neuronal cell growth, differentiation, development, or developmental delay phenotypes (MBOAT-1, GIPC1, ZDHHC19, NXN, PKN1).25-29 A 3’UTR variant in GIPC1 was also associated with MTX neurotoxicity (P=1.34x10-6). A prospective study by Bhojwani et al.1 reported 14 children with subacute symptomatic MTX neurotoxicity and did not identify any genome-wide significant SNP (top SNP significance level P=3.65x10-6). We were unable to replicate their top SNP for symptomatic leukoencephalopathy. There is a critical need to validate germline risk factors for MTX neurotoxicity in a larger international study, as larger numbers will increase power of the GWAS. There are several limitations inherent to systematic retrospective analyses. Collated clinical laboratory data were dependent on tests performed during routine clinical care, including AST values. Patients with missing data were censored and some variables were more complete than others. Clinical documentation regarding longer-term neurological function was not consistently available nor standardized, reducing the cohort for analysis. With respect to the MTX doses, analysis was performed 641

M.K. Mateos et al.

on the basis of protocols that administered low dose or high dose MTX, and age-based IT MTX dosing. Where there were significant deviations to administered chemotherapy, this was usually captured in the data extraction and maintained in the database as additional phenotypic descriptors. However, it was impossible to calculate the total administered dose of IV and IT MTX per patient, therefore this could be attempted in future prospective trials to assess the relationship with MTX neurotoxicity occurrence. While the omission of IT MTX appears to be an important factor in patients who experienced a subsequent relapse, it is possible that concurrent cessation of HD MTX in one patient could have also contributed to relapse. There were no other treatment modifications identified in this cohort which could further contribute to relapse risk, however prospective capture and analysis of all treatment modifications would be important for future studies. The main risk period in our study occurred when IT MTX was administered with cytarabine and cyclophosphamide (50 of 94 cases), which is consistent with prior publications.2,5 Surprisingly, for children treated for highrisk ALL, there was a higher incidence of MTX neurotoxicity when a lower total dose of IV MTX was used compared to protocols using higher-dose MTX. This finding requires prospective validation and may be related to the routine use of leucovorin rescue with HD MTX. Possible strategies for toxicity reduction include reintroducing IT MTX when cytarabine and cyclophosphamide are not co-administered, and/or additional doses of leucovorin following IT MTX. Administration of an additional leucovorin dose reduced acute neurotoxicity in a historical cohort;30 and this could be targeted to specific risk periods such as when cyclophosphamide and cytarabine are co-administered. Caution is advised however, due to evidence that higher leucovorin doses could be associated with increased relapse risk.31 While outside the scope of this paper, a systematic prospective study regarding secondary prophylaxis following IT MTX rechallenge would be important. Potential agents include leucovorin,1 aminophylline1 and the NMDA receptor antagonist dextromethorphan.32 In addition, drug-drug interactions that may potentiate MTX neurotoxicity or which could impact on efficacy of systemic chemotherapy such as anti-epileptic drugs should be systematically collected.33,34 Future laboratorybased modeling using neural cells derived from pluripotent stem cells35 could assess MTX transport, influx and efflux from the CNS including i) IT MTX pharmacokinetics and pharmacodynamics in the presence of cyclophos-

References 1. Bhojwani D, Sabin ND, Pei D, et al. Methotrexate-induced neurotoxicity andlLeukoencephalopathy in childhood acute lymphoblastic leukemia. J Clin Oncol. 2014;32(9):949-959. 2. Badke C, Fleming A, Iqbal A, et al. Rechallenging with intrathecal methotrexate after developing subacute neurotoxicity in children with hematologic malignancies. Pediatr Blood Cancer. 2016;63(4):723-726. 3. Rubnitz J, Relling M, Harrison P, et al. Transient encephalopathy following high-

642

phamide and cytarabine and ii) MTX transport in the presence of associated SNP. In summary, this study supports the continued use of IT MTX in patients with a first episode of MTX neurotoxicity. Continued attempts to identify and prospectively validate clinical risk factors and germline variants that indicate high risk of MTX neurotoxicity may provide an opportunity to prevent this debilitating toxicity in the future. Disclosures The authors declare no competing financial interests. Contributions MKM developed study materials, collected data, extracted patient DNA samples, wrote the manuscript, analyzed data and helped with interpretation of GWAS results; GMM and TNT wrote the study concept, supervised writing of the manuscript and assisted with interpretation of results; PMB, CG, TR, RSK collected data; MCJQ performed the GWAS; CM assisted with statistical analysis; RS, JG, DC helped with extraction of patient DNA samples; RS, DB, FA, FM, LDP assisted data collection processes; JAL assisted with data interpretation; GCT and SM provided assistance with the GWAS. All authors reviewed and approved the final version of the manuscript. Acknowledgments The authors would like to thank the Australian and New Zealand Children’s Hematology Oncology Group, ASSET study members, data managers and clinical research associates at each site for their contribution to this project. MKM has recently completed a PhD at the University of New South Wales and this work was submitted in partial fulfillment of the requirement for the PhD. The authors thank the Sydney Children’s Tumor Bank Network for providing samples for this study. Funding This work was supported by the Kids Cancer Alliance (a Translational Cancer Research Centre of Cancer Institute NSW), Cancer Institute NSW (ECF 181430, MKM), Royal Australasian College of Physicians - Kids Cancer Project Research Entry Scholarship (MKM), Cancer Therapeutics CRC (CTx) PhD Clinician Researcher Top-Up Scholarship (MKM), Anthony Rothe Memorial Trust (MKM, TNT), JGW Patterson Foundation United Kingdom (MKM), National Health and Medical Research Council of Australia (NHMRC Postgraduate Scholarship APP1056667, PMB; NHMRC Fellowship APP1142627, RSK). The authors thank the Sydney Children’s Tumour Bank Network for providing samples for this study, with support from the Cancer Council NSW, NHMRC Australia and Tour de Cure.

dose methotrexate treatment in childhood acute lymphoblastic leukemia. Leukemia. 1998;12(8):1176-1181. 4. Schmiegelow K, Attarbaschi A, Barzilai S, et al. Consensus definitions of 14 severe acute toxic effects for childhood lymphoblastic leukaemia treatment: a Delphi consensus. Lancet Oncol. 2016;17(6):e231e239. 5. Bond J, Hough R, Moppett J, Vora A, Mitchell C, Goulden N. ‘Stroke-like syndrome’caused by intrathecal methotrexate in patients treated during the UKALL 2003 trial. Leukemia. 2013;27(4):954-956.

6. Mahoney D, Shuster JJ, Nitschke R, et al. Acute neurotoxicity in children with B-precursor acute lymphoid leukemia: an association with intermediate-dose intravenous methotrexate and intrathecal triple therapy: a Pediatric Oncology Group study. J Clin Oncol. 1998;16(5):1712-1722. 7. Schmiegelow K. Advances in individual prediction of methotrexate toxicity: a review. Br J Haematol. 2009;146(5):489503. 8. Vezmar S, Becker A, Bode U, Jaehde U. Biochemical and clinical aspects of methotrexate neurotoxicity.

haematologica | 2022; 107(3)

Methotrexate neurotoxicity: risk factors & outcome

Chemotherapy. 2003;49(1-2):92-104. 9. Kishi S, Griener J, Cheng C, et al. Homocysteine, pharmacogenetics, and neurotoxicity in children with leukemia. J Clin Oncol. 2003;21(16):3084-3091. 10. Ramsey LB, Panetta JC, Smith C, et al. Genome-wide study of methotrexate clearance replicates SLCO1B1. Blood. 2013; 121(6):898-904. 11. Radtke S, Zolk O, Renner B, et al. Germline genetic variations in methotrexate candidate genes are associated with pharmacokinetics, toxicity, and outcome in childhood acute lymphoblastic leukemia. Blood. 2013; 121(26):5145-5153. 12. Mateos MK, Trahair TN, Mayoh C, et al. Risk factors for symptomatic venous thromboembolism during therapy for childhood acute lymphoblastic leukemia. Thromb Res. 2019;178:132-138. 13. NIH. National Cancer Institute Common Terminology Criteria for Adverse Events v4.0, NCI, NIH, DHHS. NIH publication # 09-7473. 2009. 14. Amos CI, Dennis J, Wang Z, et al. The OncoArray Consortium: a network for understanding the genetic architecture of common cancers. Cancer Epidemiol Biomarkers Prev. 2017;26(1):126-135. 15. Mateos MK, Tulstrup M, Quinn MC, et al. Genome-wide association meta-analysis of single-nucleotide polymorphisms and symptomatic venous thromboembolism during therapy for acute lymphoblastic leukemia and lymphoma in caucasian children. Cancers. 2020;12(5):1285. 16. Howie B, Marchini J, Stephens M. Genotype imputation with thousands of genomes. G3 (Bethesda). 2011;1(6):457470. 17. Holmboe L, Andersen AM, Mørkrid L, Slørdal L, Hall KS. High dose methotrexate chemotherapy: pharmacokinetics, folate and toxicity in osteosarcoma patients. Br J Clin Pharmacol. 2012;73(1):106-114. 18. Pratt DS, Kaplan MM. Evaluation of abnormal liver-enzyme results in asymptomatic patients. N Engl J Med. 2000;342(17):12661271. 19. Taylor OA, Brown AL, Brackett J, et al. Disparities in neurotoxicity risk and outcomes among pediatric acute lymphoblastic leukemia patients. Clin Cancer Res. 2018;24(20):5012-5017. 20. Csordas K, Hegyi M, Eipel OT, Muller J, Erdelyi DJ, Kovacs GT. Comparison of pharmacokinetics and toxicity after highdose methotrexate treatments in children with acute lymphoblastic leukemia. Anticancer Drugs. 2013;24(2):189-197. 21. Borsi JD, Moe PJ. A comparative study on the pharmacokinetics of methotrexate in a dose range of 0.5 g to 33.6 g/m2 in children with acute lymphoblastic leukemia. Cancer. 1987;60(1):5-13. 22. Pui C-H, Yang JJ, Hunger SP, et al. Childhood acute lymphoblastic leukemia: progress through collaboration. J Clin Oncol. 2015;33(27):2938-2948. 23. Moricke A, Zimmermann M, Reiter A, et al. Long-term results of five consecutive trials in childhood acute lymphoblastic leukemia performed by the ALL-BFM study group from 1981 to 2000. Leukemia. 2009;24(2):265-284. 24. Ochs JJ, Rivera G, Aur R, Hustu H, Berg R, Simone J. Central nervous system morbidi-

haematologica | 2022; 107(3)

ty following an initial isolated central nervous system relapse and its subsequent therapy in childhood acute lymphoblastic leukemia. J Clin Oncol. 1985;3(5):622-626. 25. Bonaglia MC, Marelli S, Novara F, et al. Genotype–phenotype relationship in three cases with overlapping 19p13. 12 microdeletions. Eu J Hum Genet. 2010; 18(12):1302-1309. 26. Tabe S, Hikiji H, Ariyoshi W, et al. Lysophosphatidylethanolamine acyltransferase 1/membrane-bound O-acyltransferase 1 regulates morphology and function of P19C6 cell–derived neurons. FASEB J. 2016;30(7):2591-2601. 27. Funato Y, Miki H. Nucleoredoxin, a novel thioredoxin family member involved in cell growth and differentiation. Antioxid Redox Signal. 2007;9(8):1035-1058. 28. Thauerer B, Zur Nedden S, Baier-Bitterlich G. Vital role of protein kinase C-related kinase in the formation and stability of neurites during hypoxia. J Neurochem. 2010;113(2):432-446. 29. Baumgart F, Corral-Escariz M, Pérez-Gil J, Rodríguez-Crespo I. Palmitoylation of R-Ras by human DHHC19, a palmitoyl transferase with a CaaX box. Biochim Biophys Acta. 2010;1798(3):592-604. 30. Winick NJ, Bowman WP, Kamen BA, et al. Unexpected acute neurologic toxicity in the treatment of children with acute lymphoblastic leukemia. J Natl Cancer Inst. 1992;84(4):252-256. 31. Skärby TC, Anderson H, Heldrup J, Kanerva J, Seidel H, Schmiegelow K. High leucovorin doses during high-dose methotrexate treatment may reduce the cure rate in childhood acute lymphoblastic leukemia. Leukemia. 2006;20(11):19551962. 32. Drachtman RA, Cole PD, Golden CB, et al. Dextromethorphan is effective in the treatment of subacute methotrexate neurotoxicity. Pediatr Hematol Oncol. 2002;19(5):319327. 33. Forster VJ, van Delft FW, Baird SF, Mair S, Skinner R, Halsey C. Drug interactions may be important risk factors for methotrexate neurotoxicity, particularly in pediatric leukemia patients. Cancer Chemother Pharmacol. 2016;78(5):1093-1096. 34. Relling MV, Pui C-H, Sandlund JT, et al. Adverse effect of anticonvulsants on efficacy of chemotherapy for acute lymphoblastic leukaemia. Lancet. 2000;356(9226):285290. 35. Diouf B, Crews KR, Lew G, et al. Association of an inherited genetic variant with vincristine-related peripheral neuropathy in children with acute lymphoblastic leukemia. JAMA. 2015;313(8):815-823. 36. Sutton R, Venn NC, Tolisano J, et al. Clinical significance of minimal residual disease at day 15 and at the end of therapy in childhood acute lymphoblastic leukaemia. Br J Haematol. 2009;146(3):292-299. 37. Möricke A, Reiter A, Zimmermann M, et al. Risk-adjusted therapy of acute lymphoblastic leukemia can decrease treatment burden and improve survival: treatment results of 2169 unselected pediatric and adolescent patients enrolled in the trial ALL-BFM 95. Blood. 2008;111(9):44774489. 38. Marshall G, Dalla Pozza L, Sutton R, et al. High-risk childhood acute lymphoblastic

leukemia in first remission treated with novel intensive chemotherapy and allogeneic transplantation. Leukemia. 2013; 27(7):1497-1503. 39. Termuhlen AM, Smith LM, Perkins SL, et al. Disseminated lymphoblastic lymphoma in children and adolescents: results of the COG A5971 trial: a report from the Children's Oncology Group. Br J Haematol. 2013;162(6):792-801. 40. Conter V, Valsecchi MG, Buldini B, et al. Early T-cell precursor acute lymphoblastic leukaemia in children treated in AIEOP centres with AIEOP-BFM protocols: a retrospective analysis. Lancet Haematol. 2016;3(2):e80-e86. 41. Nachman JB, Sather HN, Sensel MG, et al. Augmented post-induction therapy for children with high-risk acute lymphoblastic leukemia and a slow response to initial therapy. N Engl J Med. 1998;338(23):1663-1671. 42. Matloub Y, Lindemulder S, Gaynon PS, et al. Intrathecal triple therapy decreases central nervous system relapse but fails to improve event-free survival when compared with intrathecal methotrexate: results of the Children's Cancer Group (CCG) 1952 study for standard-risk acute lymphoblastic leukemia, reported by the Children's Oncology Group. Blood. 2006; 108(4):1165-1173. 43. Seibel NL, Steinherz PG, Sather HN, et al. Early postinduction intensification therapy improves survival for children and adolescents with high-risk acute lymphoblastic leukemia: a report from the Children's Oncology Group. Blood. 2008;111(5):25482555. 44. Matloub Y, Bostrom BC, Hunger SP, et al. Escalating intravenous methotrexate improves event-free survival in children with standard-risk acute lymphoblastic leukemia: a report from the Children's Oncology Group. Blood. 2011;118(2):243251. 45. Hunger SP, Loh ML, Whitlock JA, et al. Children's Oncology Group's 2013 blueprint for research: acute lymphoblastic leukemia. Pediatr Blood Cancer. 2013;60(6):957-963. 46. Schultz KR, Bowman WP, Aledo A, et al. Improved early event-free survival with imatinib in Philadelphia chromosome–positive acute lymphoblastic leukemia: a Children's Oncology Group study. J Clin Oncol. 2009;27(31):5175-5181. 47. Larsen EC, Devidas M, Chen S, et al. Dexamethasone and high-dose methotrexate improve outcome for children and young adults with high-risk B-acute lymphoblastic leukemia: a report from Children’s Oncology Group Study AALL0232. J Clin Oncol. 2016;34(20):23802388. 48. Hunger SP, Mullighan CG. Acute lymphoblastic leukemia in children. N Engl J Med. 2015;373(16):1541-1552. 49. Rodriguez V, Kairalla J, Salzer WL, et al. A pilot study of intensified PEG-asparaginase in high-risk acute lymphoblastic leukemia: Children’s Oncology Group Study AALL08P1. J Pediatr Hematol Oncol. 2016;38(6):409-417. 50. Strichman-Almashanu LZ, Bustin M, Landsman D. Retroposed copies of the HMG genes: a window to genome dynamics. Genome Res. 2003;13(5):800-812.

643

ARTICLE Ferrata Storti Foundation

Hematopoiesis

Ddx41 inhibition of DNA damage signaling permits erythroid progenitor expansion in zebrafish Joshua T. Weinreb,1,2 Varun Gupta,3 Elianna Sharvit,1 Rachel Weil,1 and Teresa V. Bowman1,2,4 Albert Einstein College of Medicine, Department of Developmental and Molecular Biology; 2Albert Einstein College of Medicine, Gottesman Institute for Stem Cell Biology and Regenerative Medicine; 3Albert Einstein College of Medicine, Department of Cell Biology and 4Albert Einstein College of Medicine and Montefiore Medical Center, Department of Medicine (Oncology), Bronx, NY, USA 1

Haematologica 2022 Volume 107(3):644-654

ABSTRACT

EAD-box Helicase 41 (DDX41) is a recently identified factor mutated in hematologic malignancies whose function in hematopoiesis is unknown. Using an in vivo model of Ddx41 deficiency, we unveiled a critical role for this helicase in regulating erythropoiesis. We demonstrated that loss of ddx41 leads to anemia caused by diminished proliferation and defective differentiation of erythroid progenitors. Mis-expression and alternative splicing of cell cycle genes is rampant in ddx41 mutant erythroid progenitors. We delineated that the DNA damage response is activated in mutant cells resulting in an Ataxiatelangiectasia mutated (ATM) and Ataxia-telangiectasia and Rad3-related (ATR)-triggered cell cycle arrest. Inhibition of these kinases partially suppressed ddx41 mutant anemia. These findings establish a critical function for Ddx41 in promoting healthy erythropoiesis via protection from genomic stress and delineate a mechanistic framework to explore a role for ATM and ATR signaling in DDX41-mutant hematopoietic pathologies.

Correspondence: TERESA V. BOWMAN teresa.bowman@einsteinmed.org Received: April 28, 2020. Accepted: March 16, 2021. Pre-published: March 25, 2021. https://doi.org/10.3324/haematol.2020.257246

644

Introduction Mutations in DEAD-box Helicase 41 (DDX41) were identified recently in hematologic malignancies including myelodysplastic syndrome (MDS), acute myeloid leukemia (AML), and acute erythroid leukemia (AEL).1,2 Germline DDX41 frameshift mutations are loss-of-function and somatic missense mutations are thought to result in hypomorphic helicase activity.1,3 The human genetics thus suggest that diminished function of this helicase is detrimental to hematopoiesis, but this has yet to be demonstrated in an animal model. In particular, a significant number of DDX41-mutated MDS patients experience mild cytopenia in the years preceding diagnosis, indicating that anemia may be one of the first warning signs of disease.4 Anemia in MDS is attributed to numerous cellular mechanisms including erythroid precursor apoptosis, defective progenitor expansion, and ineffective erythrocytic maturation.5-7 The clinical findings suggest DDX41 could be important in erythropoiesis, but the cellular and molecular underpinnings remain unclear. Roles for DDX41 have been implicated in genomic stability, inflammation, and splicing, all processes linked to hematopoietic health, but the current lack of DDX41 mutant animal models has slowed exploration of its function in the blood system.8-11 In order to uncover the in vivo role of DDX41 in erythropoiesis, we established a zebrafish ddx41 loss-of-function mutant. We demonstrated that ddx41 mutants develop anemia due to a decrease in erythroid progenitor expansion and defective differentiation. Mechanistically, the erythroid proliferative defect is due in part to ATM- and ATR-mediated cell cycle arrest induced by elevated DNA damage as well as mis-expression and alternative splicing of cell cycle regulators. Our data demonstrate that Ddx41 plays a critical role in hematopoiesis and provide a possible mechanism by which anemia may arise in DDX41-mutated hematopoietic pathologies.

haematologica | 2022; 107(3)

Ddx41 function in erythropoiesis

Methods

Single-cell immunofluorescence of zebrafish embryonic cells

Zebrafish

Single-cell suspensions were prepared, and cell staining was performed as described in Sorrells & Nik et al.22 and detailed in the Online Supplementary Appendix. Fluorescence intensity measurements of γH2AX were performed using FIJI.

Zebrafish were maintained as described12 and according to Institutional Animal Care and Use Committee-approved protocols in accordance with the Albert Einstein College of Medicine research guidelines. Genotyping was performed to confirm animal identity. Mutants for ddx41 (ddx41sa14887) were acquired from the Zebrafish International Resource Center.13 The mutation results in a premature stop codon at tyrosine 410. For all experiments, sibling controls are a mix of heterozygotes and wild types. Tg(gata1:dsred)14 transgenics were used. Genotyping details are in the Online Supplementary Appendix and the Online Supplementary Table S7.

Drug treatments All drugs were dissolved in dimethyl sulfoxide (DMSO). Dilutions were made in E3 embryo water. KU60019 (ATM inhibitor) and AZ20 (ATR inhibitor) were used with DMSO as the vehicle control.

May-Grunwald Giemsa staining of primitive erythroid cells May-Grunwald Giemsa staining was performed as previously described15 and as detailed in the Online Supplementary Appendix.

Statistics Experiments were performed with a minimum of three replicates. Statistical analyses were performed as indicated in each figure using unpaired Student’s t-test or a one-way ANOVA with Tukey’s multiple testing correction as appropriate; error bars indicate the standard deviation of mean, unless otherwise indicated.

Results Whole-mount in situ hybridization and o-dianisidine staining In situ hybridization was performed as previously described.15,16 After in situ, embryos were scored manually, imaged and genotyped. The βe3-globin,17 cmyb,18 and gata119 probes were used, and in situ levels were quantified using FIJI.20 O-dianisidine staining was performed as previously described.21

Flow cytometry Mutant and sibling embryos were binned based on morphological differences. For generation of single-cell suspensions, 1020 embryos were processed as previously described15 (also see the Online Supplementary Appendix). Quantification for the absolute number of cells was performed by acquiring all events in a tube on the flow cytometer to determine the total number of target cells. This number was then divided by the total number of embryos analyzed to calculate the number of target cells per embryo.

Cell cycle and apoptosis analyses For 5-ethynyl-2′-deoxyuridine (EdU) incorporation experiments, embryos were incubated with 20 mM EdU for 2 hours. Single-cell suspensions of embryos were generated. Click-IT EdU Flow Cytometry Assay Kit was used according to the manufacturer’s instructions. Flow cytometry analysis for active caspase-3 was performed as previously described.22 Samples were analyzed with a LSRII flow cytometer (BD Biosciences) and FlowJo software.

RNA sequencing and splicing analysis Erythroid progenitors from ddx41 mutants and siblings were isolated by fluorescently-activated cell sorting (FACS). RNA from these cells was subsequently isolated, DNAse-digested and library prepared for sequencing. Details on library preparation, sequencing and bioinformatic analyses can be found in the Online Supplementary Appendix. All data are deposited under GEO accession number GSE160979.

Reverse transcription quantitative polymerase chain reaction In order to validate the RNA sequencing (RNA-seq) data, we performed reverse transcription quantitative polymerase chain reaction (RT-qPCR). RNA was isolated from 40 hpf embryos. Details are listed in the Online Supplementary Appendix and Online Supplementary Table S7.

haematologica | 2022; 107(3)

Ddx41 regulates erythroid development In order to explore a function for Ddx41 in hematopoiesis, we examined erythrocyte formation and differentiation in zebrafish ddx41 homozygous loss-offunction mutants (ddx41sa14887). Maternally-deposited Ddx41 (data not shown) helps the mutants develop and survive until 3 days post fertilization (dpf) (Online Supplementary Figure S1A and B). Based on this, we consider the mutants to be functionally hypomorphic with greatly diminished but not completely absent Ddx41 levels. DDX41 is highly conserved between humans and zebrafish suggesting that lessons learned about the in vivo role of zebrafish Ddx41 function in hematopoiesis will be relevant to human DDX41. In zebrafish, primitive hematopoiesis begins ~12-24 hours post fertilization (hpf), producing embryonic erythrocytes and myeloid cells that constitute the hematopoietic system early on in development.23,24 Cells of the erythrocytic lineage first arise from the intermediate cell mass (ICM) within the posterior lateral mesoderm (PLM) (Figure 1A). These erythrocytes express factors such as the progenitor transcription factor c-myb and the erythroid-specific transcription factor gata1 starting during somitogenesis.19,25 Using in situ hybridization, we determined that these erythroid progenitor markers were expressed similarly in ddx41 mutants compared to siblings (mix of ddx41 heterozygotes + wild types) at 22 hpf, indicating initial erythroid specification is unaffected (Figures 1B to E). Oxygenated hemoglobinized erythrocytes are detectable beginning around 36 hpf using o-dianisidine staining.26 In ddx41 mutants, we observed little o-dianisidine-positive erythroid cells at 40 hpf (Figure 1F). We sorted gata1:dsRed+ erythrocytes at 40 hpf and found that the ddx41 mutant cells were larger than those from sibling controls (Figure 1G). This size difference could be indicative of delayed erythroid differentiation. As mutants display some developmental delay that becomes more severe as the embryos get older, it is possible that the erythroid delay is a side effect of the general developmental delay. In order to distinguish between these possibilities, we examined erythrocytes in ddx41 mutants and siblings at 48 hpf. Oxygenated hemoglobin levels remained low in mutants at 48 hpf (Figures 1H to I). In order to assess maturation, we also bled ddx41 mutants and sibling control embryos at 48 hpf and analyzed the morphology of isolated 645

J.T. Weinreb et al.

H I

Figure 1. Loss of ddx41 causes anemia. (A) Schema of primitive erythroid development, PLM: posterior lateral mesoderm; ProE: proerythroblasts; BasoE: basophilic erythroblasts; OrthoE: orthochromatophilic erythroblasts; func. Ery: functional erythrocytes. (B and D) In situ hybridization of the erythroid markers cmyb (B) (scale bars =200 μm) and gata1 (D) (scale bars =250 mm] at 22 hours post fertilization (hpf) in sibling controls (top) and ddx41 mutants (bottom). Arrowheads highlight the intermediate cell mass (ICM) region in the embryos. (C and E) Quantification of c-myb (C) and gata1 (E) in situ hybridization levels from (B) and (D), respectively. Quantification was done using Fiji. (F and H) Staining for o-dianisidine, marking functional hemoglobin in mature primitive erythrocytes, in sibling controls (left) and ddx41 mutants (right) at 40 hpf (F) (scale bars =350 mm) and 48 hpf (H) (scale bars =400 mm). Numbers on bottom left corner indicate the fraction of embryos with the same phenotype as the one depicted in the image. (G) Graph depicting size of erythrocytes in sibling controls and ddx41 mutants at 40 hpf. (I) Graph depicting frequency of designated odianisidine staining levels in sibling controls and ddx41 mutants at 48 hpf. (J) Representative images of orthochromatophilic erythroblasts stained with May–Grunwald– Giemsa from sibling controls (left) and ddx41 mutants (right) at 48 hpf (scale bars =5 mm). Graphs display means ± standard deviations (stds) with P-values calculated with unpaired Student’s t-test, ns=not significant (P>0.05), ****P>0.0001. For in situs and o-dianisidine staining n=6-72 embryos per experiment.

646

haematologica | 2022; 107(3)

Ddx41 function in erythropoiesis

erythrocytes using May-Grunwald-Giemsa staining (Figure 1J). The ddx41 mutant erythrocytes displayed a megaloblastoid-like phenotype, suggesting some abnormalities in erythrocyte maturation. In order to acquire enough erythrocytes to perform the morphological assessment, we needed to bleed four times as many ddx41 mutant embryos as compared to sibling control embryos, suggesting mutants had fewer erythrocytes than siblings. In order to test this hypothesis, we assessed the number of gata1:dsred+ erythroid progenitors in ddx41 mutants and siblings using flow cytometry quantification. We determined that the absolute number of gata1:dsred+ erythrocytes per embryo was significantly reduced in ddx41 mutants compared to siblings at both 28 and 40 hpf (Figures 2A to D). These data indicate that decreased erythrocyte number contributes to the development of anemia in ddx41 mutants. Erythroid progenitors arising from both primitive and definitive erythroid-myeloid progenitor (EMP)-derived waves are present during the developmental time points analyzed. The gene programs for the specification and differentiation of primitive and EMP-derived erythropoiesis are highly similar, but the developmental timings are distinct (Figure 2E). EMP specification begins around 26 hpf.27 In order to determine whether there were defects in EMPderived erythropoiesis, we performed in situ hybridization for the progenitor marker c-myb at 26 hpf and 36 hpf and gata1 at 26 hpf in siblings and ddx41 mutants (Figures 2F and G; Online Supplementary Figure S1C to F). Expression of both c-myb and gata1 within the posterior blood island (PBI) region where EMP form were not decreased and in fact were increased in ddx41 mutants as compared to siblings. As the gene programs are highly similar between these two waves of erythroid development, these data indicate that the reduction in erythrocytes in ddx41 mutants is occurring at an erythroid progenitor stage after c-myb and gata1 are both expressed, which is shortly after erythroid lineage specification. In order to further characterize the maturation state of the erythrocytes at 40 hpf in ddx41 mutants and siblings, we performed RT-qPCR for embryonic and larval globins. Expression of the embryonic globins αε1, αε3, βε1, and βε3 begins during somitogenesis with expression of all of these globins except βε3 persisting in primitive and EMP-derived erythrocytes throughout larval development.28 In contrast, levels of βε3 globin diminish dramatically from 24-48 hpf, somewhat concomitant with the increasing expression of the larval βε2 globin. The other larval globin αε5 is not expressed significantly until 14 dpf. In ddx41 mutants, we determined that while the levels of the embryonic/larval globins αε1 and βε1 were diminished, the levels of the embryonic-restricted βε3 globin remained high, consistent with a maturational defect in primitive erythrocytes. Additionally, expression of the larval βε2 globin was lower in mutants compared to sibling controls. Although ddx41 mutants die before there are expansive numbers of maturing erythrocytes derived from EMP, these data indicate that mutants have fewer definitive erythrocytes compared to siblings. This finding suggests that similar to primitive erythroid progenitors, EMP are specified normally, but there is a later stage defect, although the underlying cause (e.g., diminished expansion, maturation or differentiation) cannot be deciphered. Together, our findings establish that Ddx41 is critical for erythrocyte expansion and maturation.

haematologica | 2022; 107(3)

Cell cycle genes are mis-expressed and alternatively spliced in ddx41 mutant erythroid progenitors In order to mechanistically assess the underlying cause of the erythrocytic defect in ddx41 mutants, we conducted RNA-seq on gata1:dsred+ erythrocytes isolated from ddx41 mutants and siblings at 40 hpf. Over 1,800 genes were downregulated and more than 1,900 were upregulated in ddx41 mutants compared to siblings (Figure 3A; Online Supplementary Table S1, log fold-change ≥1, adjusted Pvalue <0.05). In order to understand if particular pathways were enriched in the differentially expressed genes, we performed gene set level analysis on the upregulated and downregulated gene lists by comparing each to the Molecular Signature Database (MSigDb), a platform that computes overlaps between classes of genes that are overor underrepresented in lists of genes in known pathways.29,30 In the downregulated gene list, mRNA splicing was the top gene set with DNA replication, cell cycle, and DNA repair also enriched (Figure 3B; Online Supplementary Table S2). In the upregulated gene list, genes associated with adaptive immunity, posttranslational modifications, innate immune system, and cell cycle were enriched (Figure 3C; Online Supplementary Table S3). We validated the expression changes in several cell cycle and DNA-damage-associated genes using RT-qPCR (Figure 3D). Ddx41 interacts with components of the spliceosome.1 Additionally, the top downregulated pathway in our gene set was pre-mRNA splicing, thus we examined how ddx41 loss affected mRNA splicing in erythrocytes. When comparing splicing between ddx41 mutants and siblings, a total of 370 alternative splicing events were observed (Figures 3E; Online Supplementary Table S4). The specific splicing defects detected included exon skipping (SE), which was the most frequently altered splicing event, intron retention (RI), alternative 5’-splice site usage, alternative 3’-splice site usage, and changes in mutually exclusive exon usage. Alternative splicing within protein coding regions of a transcript can result in the introduction of premature termination codon (PTC) or generation of a novel peptide. For all SE and RI events (comprising nearly 85% of all splicing changes), we determined how the alternative splicing event might alter the protein sequence (Figure 3F; Online Supplementary Table S5). More than 50% of SE events altered the protein sequence and are predicted to generate novel peptides. Approximately 43% of SE and 90% of RI events are predicted to target the alternatively spliced transcript for nonsense-mediated decay (NMD) due to the introduction of a PTC. For example, the retained intron variant for homologous repair-associated factor structural maintenance of chromosome 5 (smc5) identified in ddx41 mutants is predicted to result in NMD that could result in elevated DNA damage (Figure 3G). Another example of an NMD isoform expressed in ddx41 mutant is the exon 3 skipped isoform of signal transducer and activator of transcription 1a (stat1a) that would diminish signaling by numerous cytokine pathways. Pathway analysis of these alternatively spliced factors revealed that those resulting in novel peptide sequences are enriched in mRNA metabolism, morphogenesis, and cell cycle, and those predicted to result in NMD are enriched for mRNA processing, DNA replication, and gene expression (Figure 3H; Online Supplementary Table S6). These results depict that Ddx41 influences the expression and splicing of cell cycle, DNA repair, and mRNA processing genes in erythrocytes. 2

647

J.T. Weinreb et al.

Ddx41 deficiency triggers cell cycle arrest in erythroid progenitors The diminished number of erythroid progenitors and dysregulated expression of cell cycle genes suggest that defects in erythrocyte proliferation could contribute to

the anemia in ddx41 mutants. In order to examine proliferation, we analyzed cell cycle status of 30 hpf gata1:dsred+ erythroid progenitors by flow cytometry quantification of DNA synthesis via EdU incorporation and DNA content via DAPI incorporation. The ddx41

Figure 2. Ddx41 regulates erythroid progenitor numbers. (A and C) Flow cytometry plots of gata1:dsred+ erythroid cells from sibling controls (left) and ddx41 mutants (right) at 28 days post fertilization (dpf) (A) and 40 hpf (C). (B and D) Graphs depicting the absolute number of gata1:dsred+ erythroid cells per embryo from (A) and (C), respectively. n=5 pools of ~5-20 embryos per pool. (E) Schema of erythroid-myeloid progenitor (EMP) development. ProE: proerythroblasts; BasoE: basophilic erythroblasts; OrthoE: orthochromatophilic erythroblasts; func. Ery: functional erythrocytes. (F) In situ hybridization of cmyb at 26 hpf in sibling controls (left) and ddx41 mutants (right) (scale bars=150 μm). (G) Quantification of cmyb PBI in situ hybridization levels from (F). Quantification was done using Fiji; n=10-12 embryos. Graphs display means ± standard deviations (stds). (H) Graph of reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis of the expression of globin genes between sibling controls and ddx41 mutants. Expression levels were normalized to slc4a1 levels. Graph displays means ± standard error mean. The P-values were calculated with an unpaired t-test, *P<0.05, **P≤0.01, ***P≤0.001; n=3 replicates per genotype.

648

haematologica | 2022; 107(3)

Ddx41 function in erythropoiesis

Figure 3. Cell cycle genes are mis-expressed and alternatively spliced in ddx41 mutant erythroid progenitors. (A) Volcano plot displaying differentially expressed genes between gata1:dsred+ erythrocytes from ddx41 mutants and siblings. Significant differences are defined as false discovery rate (FDR) <0.05 and log fold-change ≥1. Black vertical lines denote the fold-change threshold and the black horizontal line denotes the FDR threshold. Five biological replicates for both ddx41 mutants and siblings were used to generate RNA sequencing data. (B and C). Representative charts of pathways significantly enriched in genes downregulated (B) or upregulated (C) in ddx41 mutant erythroid progenitors compared to sibling controls as determined by MSigDB analysis. (D) Graph of reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis of the expression of cell cycle and DNA damage-associated genes between sibling controls and ddx41 mutants. Expression levels were normalized to β-actin levels. Graph displays means ± standard error mean. The P-values were calculated with an unpaired t-test, *P<0.05, ****P≤0.0001; n=3 replicates per experiment. (E) Graph depicting the Dψ of individual splicing events between sibling controls and ddx41 mutants as detected by analysis with rMATS. Significant differences are defined as FDR ≤0.01 and Dψ ≥0.1. SE: skipped exons; RI: retained introns; A5SS: alternative 5’ splice site; A3SS: alternative 3’ splice site; MXE: mutually exclusive exons. (F) Graph depicting the frequency of alternatively spliced isoforms in ddx41 mutants that are predicted to result in nonsense-mediated (NMD) decay, protein sequence alterations (nonNMD), or changes in untranslated regions (UTR). (G) Sashimi plot for smc5 (exons 18-19) and stat1a (exons 2-4) in ddx41 mutant erythrocytes compared to sibling controls. RPKM: reads per kilobase of transcript per million mapped reads; Inc: inclusion. (H) Representative charts of pathways significantly enriched in alternatively spliced genes in ddx41 mutant erythrocytes compared to sibling controls as determined by pathway analysis. 2

haematologica | 2022; 107(3)

649

J.T. Weinreb et al.

mutant gata1:dsred+ progenitors showed a reduction of cells in S phase and an accumulation of cells in the G0/G1 and G2/M phases compared to sibling controls (Figures 4A and B). These results are in-line with a decrease in proliferation in ddx41 mutant erythrocytes caused by cell cycle arrests at the G0/G1-to-S phase and G2-to-M transitions. Prolonged cell cycle arrest can lead to apoptosis,31 thus we also assessed apoptosis in ddx41 mutants. We measured levels of active caspase-3, an essential executor of apoptosis, in ddx41 mutant and sibling gata1:dsred+ erythrocytes by flow cytometry. We observed a significant increase in active caspase-3 in ddx41 mutant gata1:dsred+ erythrocytes at 30 hpf (Figures 4C-D). These data indicate that both cell cycle

arrest and elevated apoptosis in ddx41 mutant erythrocytes may drive anemia.

Ddx41 regulation of ATM and ATR signaling contributes to proper erythropoiesis These molecular and cellular phenotypes in ddx41 mutants imply that loss of ddx41 could promote DNA damage. In order to address this question, we analyzed the DNA damage response (DDR) by performing immunofluorescence (IF) for γH2AX in ddx41 mutants and siblings. We showed that γH2AX levels were increased nearly two-fold in ddx41 mutants compared to sibling cells (Figures 4E and F). These data demonstrate that ddx41 deficiency triggers DDR in vivo.

Figure 4. Ddx41 deficiency triggers cell cycle arrest and DNA damage response in erythroid progenitors. (A) Cell cycle analysis of gata1:dsred+ erythroid cells from sibling controls (left) and ddx41 mutants (right) after a 2-hour pulse of 5-ethynyl-2′-deoxyuridine (EdU) at 28 hpf. EdU incorporation (y-axis) and DAPI content (x-axis) were measured by flow cytometry at 30 hours post fertilization (hpf). (B) Quantification of the percentage of cells in each cell cycle phase from (A). (C) Flow cytometry analysis of active-caspase 3 in gata1:dsred+ erythroid cells from sibling controls (left) and ddx41 mutants (right). (D) Quantification of the percentage of gata1:dsred+ erythroid cells that are active caspase-3-positive from (C). (E) Confocal images showing immunofluorescence of nuclei (DAPI) and γH2AX in cells isolated from 28 hpf siblings (top) and ddx41 mutants (bottom). scale bars =5 mm. (F) Quantification of γH2AX levels from (E). Graphs display means ± standard deviations (stds) with Pvalues calculated with unpaired Student’s t-test, *P<0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001. For flow cytometry, n=3-5 pools of ~5-20 embryos per pool. For immunofluorescence imaging, n=100-300 cells per genotype.

650

haematologica | 2022; 107(3)

Ddx41 function in erythropoiesis

Our model is that loss of Ddx41 contributes to excessive DDR signaling and subsequent cell cycle arrest in erythrocytes, leading to anemia in ddx41 mutants. If correct, then inhibiting components of the DDR pathway would i) reverse cell cycle defects and ii) increase erythrocyte levels. In order to test this model, we examined how the two primary mediators of DDR, Ataxia-telengiectasia-mutated (ATM) and Ataxia-telengiectasia and Rad3-related (ATR), affected erythrocytic cell cycle kinetics in ddx41 mutants. We assessed cell cycle status of 30 hpf ddx41-mutant gata1:dsred+ erythroid progenitors in embryos treated with DMSO vehicle control, the ATM inhibitor KU60019, or the ATR inhibitor AZ20. There was a significant increase of gata1:dsred+ ddx41-mutant cells in S phase when treated with either ATM or ATR inhibitors as compared to DMSO vehicle control (Figures 5A and B). Additionally, pharmacological inhibition of ATM or ATR increased erythropoietic output in ddx41 mutants, as measured by quantification of gata1:dsred+ erythrocyte numbers per embryo using flow cytometry (Figures 5C to F). Although there was a trend towards an increase in erythrocyte numbers in control siblings treated with ATM or ATR inhibitors these changes were not statistically significant. Taken together, these data indicate that DDR signaling triggers a G0/G1 cell cycle arrest in ddx41-mutant erythrocytes that results in a reduction of erythroid progenitor cell number. Finally, we wanted to assess if increasing the number of erythroid progenitors via ATM or ATR inhibition would increase the number of oxygenated erythrocytes in ddx41 mutants. Surprisingly, we only observed a significant increase in o-dianisidine-positive erythrocytes in ddx41 mutants treated with ATM inhibitor, but not ATR inhibitor (Figures 6A and B). These data indicate that Ddx41 regulation of ATM might have a broader impact on erythropoiesis than ATR signaling.

Discussion Although DDX41 mutations are found in numerous human hematologic diseases, its function in hematopoiesis is unknown. Our work is the first to establish Ddx41 as a critical mediator of erythropoiesis with ddx41 loss suppressing the expansion and maturation of erythrocytes. We showed a profound effect on the expression of cell cycle and DNA damage-associated genes in ddx41 mutant erythroid progenitors consistent with the observed cell cycle arrest. The DNA damage response is elevated in ddx41 mutant cells and triggers an ATM and ATR-triggered cell cycle arrest. Inhibition of ATM and ATR partially suppressed anemia in ddx41 mutants. These findings establish Ddx41 as a positive regulator of erythropoiesis in part by preventing genomic stress and promoting proper erythroid progenitor expansion. Patients with germline mutations in DDX41 do not develop hematologic symptoms until later in life,1 yet zebrafish ddx41 mutants show anemia within 40 hpf. We posit that the difference has to do with the extent of Ddx41 deficiency. Zebrafish homozygous mutants have maternally deposited Ddx41 that is naturally depleted over the first few days of life. When the levels drop below a certain threshold, the mutants die, demonstrating it is an essential factor. In contrast, zebrafish ddx41 heterozygous animals are phenotypically indistinguishable from wildtype animals during embryogenesis and in adulthood, sughaematologica | 2022; 107(3)

gesting a 50% decrease of Ddx41 alone is insufficient to alter hematopoiesis. This is in agreement with the clinical observation that patients with germline DDX41 mutations who develop hematologic malignancies often acquire somatic missense mutations in the second allele that are thought to diminish DDX41 ATPase activity.1 Combined, the data indicate that when DDX41 levels decrease to less than 50%, this leads to hematologic defects, but when critically too low, it leads to lethality. DDX41 was previously identified as a mediator of genomic stability in a cell line-based genome-wide siRNA screen.11 However, a role for DDX41 in genomic integrity as well as the downstream consequences of its loss were never demonstrated in vivo. Our current work revealed that Ddx41 regulates genomic integrity in vivo, and that loss of ddx41 leads to both cell cycle arrest and apoptosis in erythrocytes that contributes to anemia in ddx41 mutants. We established that ATM and ATR signaling contribute to these attributes, but only ATM inhibition significantly increased o-dianisidine-positive erythrocytes in ddx41 mutants. The differential impact on oxygenated erythrocyte output by inhibition of ATM and ATR might indicate that Ddx41-regulated ATM signaling is more critical for proper erythropoiesis. ATM has an additional role in apoptosis, especially during development that might explain some of the phenotypic differences when comparing ATM and ATR inhibition effects on erythropoiesis. However, it should be noted that although the ATM and ATR kinases respond uniquely, there exists an extensive ‘cross-talk’ between them, which can make determining which precise pathway is involved in a phenotype confusing.32 Further dissection of the role of DDX41 in ATM and ATR pathway regulation will need to be investigated. Splicing mutations are commonly found in hematologic malignancies.33,34 DDX41 interacts with multiple components of the spliceosome.1 Our work aligns with prior studies showing DDX41 insufficiency associates with numerous deleterious splicing outcomes. If and how these splicing events contribute to hematopoietic pathogenesis is unclear. We showed that components related to cell cycle and DNA repair are commonly mis-spliced in ddx41 mutants. Therefore, it is possible that loss of ddx41 may be mediating cell cycle arrest and activation of DDR via mis-splicing of crucial regulators of these pathways. The contribution of DDR pathway component mis-splicing in human cytopenias remains to be addressed. In addition to the effect on cell cycle, we delineated maturation defects in ddx41 erythrocytes marked by aberrant globin expression and a megaloblastoid-like morphology of mutant erythrocytes. Although we could not perform a complete analysis of definitive erythropoiesis as ddx41 mutants die before EMP-derived or HSC-derived erythrocytes fully mature, the diminished expression of the larval βe2 globin suggests a decrease in EMP-derived definitive erythrocytes. This finding combined with the elevated cmyb and gata1 levels in EMP cells suggests that this defect could be caused in part by maturation defects. As the treatment with the ATM inhibitor KU60019 only partially increased hemoglobinized erythrocytes in ddx41 mutants, it suggests that deregulation of another pathway underlies additional maturational defects in ddx41 mutants. In summary, our study unveils a critical role for Ddx41 as a key gatekeeper to maintain cell cycle progression, a necessary component for erythrocytic development. We 651

J.T. Weinreb et al. A

Figure 5. Ddx41 regulation of ATM/ATR signaling contributes to proper erythroid progenitor proliferation. (A) Cell cycle analysis of gata1:dsred+ erythroid cells from sibling controls (top) and ddx41 mutants (bottom) treated with dimethyl sulfoxide (DMSO) (left), 30 nM KU60019 (Ataxia-telengiectasia-mutated [ATMi] inhibitor, middle), and 30 nM AZ20 (Ataxia-telengiectasia and Rad3-related [ATRi] inhibitor, right) after a 2-hour pulse of 5-ethynyl-2′-deoxyuridine (EdU) at 28 hours post fertilization (hpf). EdU incorporation (y-axis) and DAPI content (x-axis) were measured by flow cytometry at 30 hpf. (B) Quantification of the percentage of cells in each cell cycle phase from (A). (C and E) Flow cytometry plots of gata1:dsred+ erythroid cells from sibling controls (C) and ddx41 mutants (E) treated with DMSO (left), 30 nM KU60019 (ATM inhibitor, middle), and 30 nM AZ20 (ATR inhibitor, right). (D and F). Graphs depicting the absolute number of gata1:dsred+ erythroid cells per embryo from (C) and (E). Graphs display means ± standard deviations (stds) with P-values calculated with a one-way ANOVA with Tukey’s multiple testing correction, *P<0.05, ***P≤0.001, ****P≤0.0001. For flow cytometry, n=3-5 pools of ~5-20 embryos per pool.

652

haematologica | 2022; 107(3)

Ddx41 function in erythropoiesis

Figure 6. ATM inhibition partially suppresses ddx41 mutant anemia. (A) Representative images of o-dianisidine staining and corresponding levels of staining from sibling controls (left) and ddx41 mutants (right) treated with dimethyl sulfoxide (DMSO) (top), 30 nM KU60019 (Ataxia-telengiectasia-mutated inhibitor [ATMi], middle), and 30 nM AZ20 (Ataxia-telengiectasia and Rad3-related inhibitor[ATRi], bottom). Numbers on bottom left corner indicate the fraction of embryos with the same phenotype as the one depicted in the image. scale bars =100 μm. (B) Graph depicting frequency of designated o-dianisidine staining levels in sibling controls and ddx41 mutants at 40 hours post fertilization (hpf) treated with DMSO vehicle control, 30 nM KU60019 (ATMi), and 30 nM AZ20 (ATRi). Graphs display frequency of embryos with each phenotype with P-values calculated with Chi-squared test, *P<0.05, ****P≤0.0001. For o-dianisidine staining, n=36-67 embryos per experiment.

demonstrated that deficiency of ddx41 triggers cell cycle arrest via activation of ATM and ATR, which ultimately mediates a decrease in proliferation and maturation of erythrocytic progenitors in ddx41 mutants. These findings establish a critical function for Ddx41 in promoting healthy erythropoiesis by suppressing genomic stress and present a potential role for ATM and ATR signaling in DDX41-mutant pathologies. Disclosures No conflicts of interest to disclose. Contributions JTW and TVB designed the project experimental approach; JTW, ES, RW, and TVB performed the experiments; VG per-

References 1. Polprasert C, Schulze I, Sekeres MA, et al. Inherited and somatic defects in DDX41 in myeloid neoplasms. Cancer Cell. 2015;27(5):658-670. 2. Iacobucci I, Wen J, Meggendorfer M, et al. Genomic subtyping and therapeutic targeting of acute erythroleukemia. Nat Genet. 2019;51(4):694-704. 3. Yoneyama-Hirozane M, Kondo M, Matsumoto SI, et al. High-throughput screening to identify inhibitors of DEAD box helicase DDX41. SLAS Discov. 2017;22(9):1084-1092. 4. Sebert M, Passet M, Raimbault A, et al. Germline DDX41 mutations define a significant entity within adult MDS/AML patients. Blood. 2019;134(17):1441-1444.

haematologica | 2022; 107(3)

formed bioinformatics analysis; JTW and TVB analyzed the data; JTW and TVB wrote and edited the manuscript; all authors reviewed and approved the manuscript. Acknowledgments This work was funded by American Cancer Society RSG129527-DDC, DOD BM180109, NIH 1R01DK12173801A1 and the Edward P. Evans Foundation (to TVB), NIH MSTP training grant T32GM007288-45 and F30 fellowship 1F30HL142161 (to JTW), and NIH 1R01GM057829-23 to Charles Query for support of VG. We also want to acknowledge the assistance of numerous core facilities at Albert Einstein College of Medicine including Flow Cytometry, Analytical Imaging, and Genomics Facilities (funded by NCI Cancer Grant P30CA013330), and the Zebrafish Core Facility.

5. Schneider RK, Schenone M, Ferreira MV, et al. Rps14 haploinsufficiency causes a block in erythroid differentiation mediated by S100A8 and S100A9. Nat Med. 2016;22(3): 288-297. 6. Danilova N, Sakamoto KM, Lin S. Ribosomal protein L11 mutation in zebrafish leads to haematopoietic and metabolic defects. Br J Haematol. 2011;152(2): 217-228. 7. Payne EM, Virgilio M, Narla A, et al. LLeucine improves the anemia and developmental defects associated with DiamondBlackfan anemia and del(5q) MDS by activating the mTOR pathway. Blood. 2012;120(11):2214-2224. 8. Parvatiyar K, Zhang Z, Teles RM, et al. The helicase DDX41 recognizes the bacterial secondary messengers cyclic di-GMP and

cyclic di-AMP to activate a type I interferon immune response. Nat Immunol. 2012;13 (12):1155-1161. 9. Zhang Z, Yuan B, Bao M, et al. The helicase DDX41 senses intracellular DNA mediated by the adaptor STING in dendritic cells. Nat Immunol. 2011;12(10):959-965. 10. Zhang Z, Bao M, Lu N, et al. The E3 ubiquitin ligase TRIM21 negatively regulates the innate immune response to intracellular double-stranded DNA. Nat Immunol. 2013;14(2):172-178. 11. Paulsen RD, Soni DV, Wollman R, et al. A genome-wide siRNA screen reveals diverse cellular processes and pathways that mediate genome stability. Mol Cell. 2009;35(2):228-239. 12. Lawrence C. Advances in zebrafish husbandry and management. Methods Cell

653

J.T. Weinreb et al. Biol. 2011;104:429-451. 13. Kettleborough RN, Busch-Nentwich EM, Harvey SA, et al. A systematic genome-wide analysis of zebrafish protein-coding gene function. Nature. 2013;496(7446):494-497. 14. Traver D, Paw BH, Poss KD, et al. Transplantation and in vivo imaging of multilineage engraftment in zebrafish bloodless mutants. Nat Immunol. 2003;4(12):12381246. 15. De La Garza A, Cameron RC, Nik S, Payne SG, Bowman TV. Spliceosomal component Sf3b1 is essential for hematopoietic differentiation in zebrafish. Exp Hematol. 2016;44 (9):826-837. 16. Thisse C, Thisse B. High-resolution in situ hybridization to whole-mount zebrafish embryos. Nat Protoc. 2008;3(1):59-69. 17. Brownlie A, Hersey C, Oates AC, et al. Characterization of embryonic globin genes of the zebrafish. Dev Biol. 2003;255(1):4861. 18. Liao EC, Paw BH, Oates AC, et al. SCL/Tal1 transcription factor acts downstream of cloche to specify hematopoietic and vascular progenitors in zebrafish. Genes Dev. 1998;12(5):621-626. 19. Detrich HW 3rd, Kieran MW, Chan FY, et al. Intraembryonic hematopoietic cell migration during vertebrate development. Proc Natl Acad Sci U S A. 1995;92(23):1071310717. 20. Dobrzycki T, Krecsmarik M, Bonkhofer F,

654

Patient R, Monteiro R. An optimised pipeline for parallel image-based quantification of gene expression and genotyping after in situ hybridisation. Biol Open. 2018;7(4):bio031096. 21. Lieschke GJ, Oates AC, Crowhurst MO, Ward AC, Layton JE. Morphologic and functional characterization of granulocytes and macrophages in embryonic and adult zebrafish. Blood. 2001;98(10):3087-3096. 22. Sorrells S, Nik S, Casey M, et al. Spliceosomal components protect embryonic neurons from R-loop-mediated DNA damage and apoptosis. Dis Model Mech. 2018;11(2):dmm031583. 23. Carroll KJ, North TE. Oceans of opportunity: exploring vertebrate hematopoiesis in zebrafish. Exp Hematol. 2014;42(8):684696. 24. Clements WK, Traver D. Signalling pathways that control vertebrate haematopoietic stem cell specification. Nat Rev Immunol. 2013;13(5):336-348. 25. Lyons SE, Lawson ND, Lei L, et al. A nonsense mutation in zebrafish gata1 causes the bloodless phenotype in vlad tepes. Proc Natl Acad Sci U S A. 2002;99(8):5454-5459. 26. Paffett-Lugassy NN, Zon LI. Analysis of hematopoietic development in the zebrafish. Methods Mol Med. 2005;105: 171-198. 27. Bertrand JY, Kim AD, Violette EP, et al. Definitive hematopoiesis initiates through a

committed erythromyeloid progenitor in the zebrafish embryo. Development. 2007;134(23):4147-4156. 28. Ganis JJ, Hsia N, Trompouki E, et al. Zebrafish globin switching occurs in two developmental stages and is controlled by the LCR. Dev Biol. 2012;366(2):185-194. 29. Liberzon A, Birger C, Thorvaldsdottir H, et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 2015;1(6):417-425. 30. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):1554515550. 31. Orth JD, Loewer A, Lahav G, Mitchison TJ. Prolonged mitotic arrest triggers partial activation of apoptosis, resulting in DNA damage and p53 induction. Mol Biol Cell. 2012;23(4):567-576. 32. Cimprich KA, Cortez D. ATR: an essential regulator of genome integrity. Nat Rev Mol Cell Biol. 2008;9(8):616-627. 33. Yoshida K, Sanada M, Shiraishi Y, et al. Frequent pathway mutations of splicing machinery in myelodysplasia Nature. 2011;478(7367):64-69. 34. Haferlach T, Nagata Y, Grossmann V, et al. Landscape of genetic lesions in 944 patients with myelodysplastic syndromes. Leukemia. 2014;28(2):241-247.

haematologica | 2022; 107(3)

ARTICLE

Hematopoiesis

Reduced frequencies and functional impairment of dendritic cell subsets and non-classical monocytes in myelodysplastic syndromes

Ferrata Storti Foundation

Nathalie van Leeuwen-Kerkhoff,1 Theresia M. Westers,1 Pino J. Poddighe,2 Giovanni A.M. Povoleri,3 Jessica A. Timms,4 Shahram Kordasti,4,5# Tanja D. de Gruijl6# and Arjan A. van de Loosdrecht1# 1 Department of Hematology, Amsterdam University Medical Centers, Vrije Universiteit Amsterdam, Cancer Center Amsterdam, the Netherlands; 2Department of Clinical Genetics, Amsterdam University Medical Centers, Vrije Universiteit Amsterdam, the Netherlands; 3Department Inflammation Biology, King's College London, Center for Inflammation Biology and Cancer Immunology, London, UK; 4Systems Cancer Immunology Lab, Comprehensive Cancer Center, King's College London, London, UK; 5 Dipartimento Scienze Cliniche e Molecolari, UNIVPM, Ancona, Italy and 6Department of Medical Oncology, Amsterdam University Medical Centers, Vrije Universiteit Amsterdam, Cancer Center Amsterdam, the Netherlands #

Haematologica 2022 Volume 107(3):655-667

SK, TDG and AAL contributed equally as co-senior authors.

ABSTRACT

n myelodysplastic syndromes (MDS) the immune system is involved in pathogenesis as well as in disease progression. Dendritic cells (DC) are key players of the immune system by serving as regulators of immune responses. Their function has been scarcely studied in MDS and most of the reported studies didn’t investigate naturally occurring DC subsets. Therefore, we here examined the frequency and function of DC subsets and slan+ non-classical monocytes in various MDS risk groups. Frequencies of DC as well as of slan+ monocytes were decreased in MDS bone marrow compared to normal bone marrow samples. Transcriptional profiling revealed down-regulation of transcripts related to pro-inflammatory pathways in MDS-derived cells as compared to normal bone marrow. Additionally, their capacity to induce T-cell proliferation was impaired. Multidimensional mass cytometry showed that whereas healthy donor-derived slan+ monocytes supported Th1/Th17/Treg differentiation/expansion their MDS-derived counterparts also mediated substantial Th2 expansion. Our findings point to a role for an impaired ability of DC subsets to adequately respond to cellular stress and DNA damage in the immune escape and progression of MDS. As such, it paves the way toward potential novel immunotherapeutic interventions.

Correspondence: ARJAN A. VAN DE LOOSDRECHT a.vandeloosdrecht@amsterdamumc.nl Received: July 26, 2020. Accepted: February 3, 2021. Pre-published: February 11, 2021. https://doi.org/10.3324/haematol.2020.268136

Introduction Development of human dendritic cells (DC) occurs in the bone marrow (BM), where they originate from common precursor cells and differentiate into specialized subsets: plasmacytoid DC (pDC) and conventional myeloid DC (cDC).1-3 These cDC are further separated in cDC1 (CD141+) and cDC2 (CD1c+) DC.4-6 Initially, a fourth DC subset, slanDC, was identified based on the expression of M-DC8 (6-sulfo LacNAc or slan) and CD16.7-9 Recent studies by our own group and others, showed that these cells are actually more closely related to monocytes than to DC and they were renamed slan+ non-classical monocytes.10-12 DC act as antigen presenting cells (APC) and orchestrate immune responses. Upon activation DC undergo a maturation process and up-regulate co-stimulatory molecules and secrete different types of cytokines, leading to antigen presentation and subsequent cellular immune responses.13 In several hematological malignancies immune dysregulation affecting the DC compartment has been reported, which

haematologica | 2022; 107(3)

655

N. van Leeuwen-Kerkhoff et al.

might contribute to the pathogenesis of these malignancies and could be an important target for therapy.14-16 In MDS different types of immune cells are believed to play a role in pathophysiology.17,18 Low-risk disease is often characterized by an activated immune system in which pro-inflammatory cells are numerically increased.19,20 In contrast, it has been demonstrated in high-risk disease that immunosuppressive cell types, e.g., Tregs and myeloid derived suppressor cells (MDSC), are expanded and eventually facilitate immune escape and disease progression.20-23 Limited data is available on the role of DC in different MDS risk groups. Thus far, the focus of most studies has been on the frequencies and function of either in vitro generated monocyte-derived DC (MoDC) or total DC rather than functionally distinct DC subsets.24-27 In this study, we have investigated the frequencies of pDC and myeloid subsets (cDC and slan+ non-classical monocytes) in the BM and peripheral blood (PB) of different MDS risk group patients (i.e., low- and high risk based on the International Prognostic Scoring System [IPSS] and Revised IPSS [IPSS-R] or using the 2016 World Health Organization [WHO] classification) and compared them to normal BM (NBM) samples. Furthermore, we performed fluorescence in situ hybridization (FISH) analysis to demonstrate clonal involvement. A genome wide transcriptional analysis was carried out to find differences between healthy donor (HD) and MDS-derived subsets. In functional assays, their maturation and cytokine secreting capacity as well as their ability to induce T-cell proliferation was assessed. Their reduced frequencies and a selective functional impairment, related to danger and tissue damage responsiveness, provide clues as to the role of these myeloid APC subsets in MDS progression.

Methods

subsequent FISH analysis (details are given in the Online Supplementary Appendix).

Functional assays and multidimensional mass cytometry The maturation capacity, the secretion of cytokines and the ability to induce T-cell proliferation was tested for MDS BM- and NBM-derived cDC2 and slan+ non-classical monocytes. See the Online Supplementary Table S1 for clinical data. A multi-parameter deep-phenotyping strategy, known as cytometry by time-of-flight (CyTOF), was used for T cells cultured in the presence of MDSderived or healthy PB-derived slan+ monocytes. DC subsets could not be included in this experiment because of low cell numbers. See the Online Supplementary Appendix file for technical details.

Microarray transcriptional analysis RNA was isolated from MDS BM- and NBM-derived cDC2 and slan+ monocytes (5.000-67.000 cells) and amplified using the Ovation Pico WTA System V2 (NuGen, San Carlos, CA) as previously described.11 RNA was labeled with the Encore Biotin Module Kit (NuGEN) and 5 mg of cDNA from each sample was hybridized to Human Transcriptome Arrays 2.0 microarrays (Affymetrix) and signals were scanned by Affymetrix GeneChip Scanner 3000 7G. See the Online Supplementary Appendix for details on data analysis. The microarray data have been deposited in the GEO public database under the accession number: GSE161058.

Statistical analysis Graphpad Prism 6 software (San Diego, USA) was used for flow cytometry and functional data analysis and graphic display. For two-group comparisons, differences were assessed by applying a non-parametric Mann-Whitney U test. Multi-group comparisons were analyzed with a Kruskal-Wallis with Dunn’s multiple comparisons test. The non-parametric Spearman’s correlation test was used for correlations. A P-value of <0.05 was considered significant.

Patient and control samples In this study, 30 NBM samples and 187 BM and 26 PB samples of newly diagnosed MDS patients were used. Risk scores according to IPSS28 and the IPSS-R29, were available for 150 and 136 patients, respectively. The 2016 WHO classification was available for 163 patients (details are given in the Online Supplementary Appendix and the Online Supplementary Table S1). NBM samples were obtained after written informed consent from patients who were undergoing cardiac surgery and were considered hematologically healthy (i.e., no cytopenia, normal morphology and normal flow cytometric profile). The study was approved by the local Institutional Review Board and was in accordance with the declaration of Helsinki.

Enumeration of antigen presenting cells subsets and fluorescence in situ hybridization PB and BM cells were analyzed on a flow cytometer (FACSCantoTM, BD Biosciences) after incubation with a panel of monoclonal antibodies (see the Online Supplementary Appendix for details). After debris and doublet exclusion, cell subtypes were identified in the CD45+ compartment. CD141, CD1c, CD303 and M-DC8/CD16 were used for the identification of cDC1, cDC2, pDC and slan+ monocytes, respectively (Figure 1A). Frequencies of all populations were calculated as percentages of CD45+ mononuclear cells. Three MDS samples with a known cytogenetic aberrancy were used for the isolation of cDC2 and slan+ monocytes and

656

Results Dendritic cell frequencies are reduced in the bone marrow of MDS patients Frequencies of different DC subsets and slan+ monocytes were analyzed in BM samples of 187 newly diagnosed myelodysplastic syndrome patients (detailed in Table 1) and compared to 30 NBM samples. An eightcolour flow cytometry panel was used for the detection of CD303+ pDC, CD141hi cDC1, CD1c+ cDC2 and M-DC8+/CD16+ non-classical monocytes (Figure 1A). Except for pDC, all subsets showed significantly lower frequencies in MDS-derived BM compared to NBM (Figure 1B; NBM vs. MDS BM: cDC1 0.048% vs. 0.030%, cDC2 0.67% vs. 0.54% and slan+ 0.36% vs. 0.24%). pDC rates were increased in MDS BM (NBM, 0.76%; MDS BM, 0.91%). This was mainly observed in cases that were associated with low blast counts (Figure 1C; NBM, 0.76%; (RS-)SLD/MLD, 1.11%; EB-1/EB-2, 0.63%). For cDC1 and cDC2, frequencies gradually decreased in classification groups associated with higher risk MDS (Figure 1C; NBM, 0.048%; (RS-)SLD/MLD, 0.038%; EB-1/EB-2, 0.015% and NBM, 0.67%; (RS-)SLD/MLD, 0.59%; EB-1/EB-2, 0.44%, respectively). Also, for slan+ monocytes lowest frequencies were found in the EB-1/EB-2 classification group (NBM, 0.36%; (RS-)SLD/MLD, 0.24%; EB-1/EB-2, haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

0.23%). Using the IPSS and IPSS-R risk stratification, similar results were observed. Patients within higher risk groups showed lower frequencies of cDC (Figure 1D and E). Furthermore, MDS-derived paired BM and PB samples showed strong correlations for frequencies of designated subsets, except for cDC1 (Figure 1F). In order to assess clonal involvement of isolated cDC2 and slan+ non-classical monocytes (due to limited cell availability cDC1 was not tested), three different MDS samples with a known cytogenetic aberration (del5q, trisomy 8 and monosomy 7) were selected for FISH analysis. In all tested cases, cDC2 were highly involved in the dysplastic clone. Due to the limited cell numbers of slan+ monocytes, a clear conclusion could only be drawn for the del5q analysis. In this patient slan+ monocytes were also highly involved in the dysplastic clone. As expected, CD34+ progenitor cells were also clonally involved whereas B cells were not (Figure 2).

Immunological gene sets related to danger response are under-represented in myelodysplastic syndrome-derived antigen presenting cells In order to screen for functional differences between HD- and MDS-derived cDC and slan+ monocytes a genome-wide transcriptional profiling study was performed. cDC2 and slan+ monocytes were sorted from previously stored HD (n=3) and MDS (n=4) BM samples. Because of low cell numbers, cDC1 were not included in this study. Patient samples were selected based on their 2016 WHO classification group in order to create a more homogenous set of samples. All patients were diagnosed with RS-MLD, were in very low to intermediate IPSS-R risk categories and, apart from one patient who had a 45,X,-Y [3] / 46,XY [7] karyotype, showed normal cytogenetics. In total, 135.750 genes were found to be expressed (including non-coding genes). Using ANOVA testing the number of differentially expressing genes (DEG) for coding transcripts between HD and MDS was 1.922 for cDC2 and 2.415 for slan+ monocytes. In cDC2 1.075 genes were under-expressed in MDS compared to HD and 847 genes were over-expressed. Slan+ monocytes showed 1.655 under-expressed and 760 overexpressed genes. Hierarchical clustering showed a clear separation between HD- and MDS-derived samples for both subsets (Figure 3A). Volcano plots were used to show over- and under-expressed genes in HD compared to MDS with fold change levels of <-2.5 or >2.5 and a gene level P-value <0.05 (Figure 3B). Next, DEGs were used for further pathway analysis in order to find biologically relevant differences between HD and MDS subsets. Transcripts that were under-expressed in MDS were imported in the STRING v10.5 database and a gene ontology (GO) term enrichment analysis was performed. Six of the most enriched pathways are shown for both subsets (Figure 3C). Pathways highly involved in proinflammatory processes and innate immune activation were under-represented in MDS as compared to HD. When uploading lists containing over-expressed genes in MDS, hardly any immune response-related pathways were enriched (not shown). Interestingly, for cDC2 pathways related to apoptosis were enriched in biological processes as well as in KEGG pathways. For slan+ monocytes this analysis mainly yielded pathways containing metabolic- and general cell biology-related processes. In order to confirm these analyses gene set enrichment haematologica | 2022; 107(3)

analysis (GSEA) was executed. Twenty gene sets involved in immunological processes were selected from the Broad Institute database and tested for enrichment in either HD samples or MDS samples. For cDC2 5 gene sets were significantly enriched in HD (with a nominal Pvalue <0.01) and none of them was enriched in MDS. Slan+ monocytes showed five significantly enriched gene sets in HD and two in MDS (Figure 3D and E). For each cell subset an example enrichment plot is shown. Genes that contribute to the core enrichment for this specified gene set are displayed in a heatmap (67 genes for cDC2, 75 of 280 are shown for slan+ monocytes). For cDC2, a great part of these genes consisted of pattern recognition receptors, such as the toll-like receptors (TLR 1, 2, 3, 5, 7, 8 and 10), C-type lectin receptors (CLEC4A, E, C and CLEC6A) and CD180. For slan+ monocytes also Fc-γ receptors (FCGR1A, FCGR2A/B and FCGR3A) and Table 1. Patient and control characteristics

Bone marrow Peripheral blood Number HD MDS Age - mean, y HD MDS Sex HD - male/female MDS - male/female IPSS Low risk Intermediate-1 Intermediate-2 High risk Missing IPSS score IPSS-R Very low risk Low risk Intermediate risk High risk Very high risk Missing IPSS-R score WHO MDS-SLD MDS-MLD MDS-RS-SLD MDS-RS-MLD MDS-EB-1 MDS-EB-2 Missing WHO 2016 classification % Blasts <5% ≥5%

217 30 187

26 26

62 69

20/10 134/53

17/9

53 71 21 5 37

10 10 6

29 48 32 15 12 51

5 8 4 1 8

11 65 8 31 24 24 24

5 4 2 7 3 5

116 51

13 5

MDS: myelodysplastic syndromes; EB: excess blasts; HD: healthy donor; IPSS(-R): (Revised) International Prognostic Scoring System; MLD: multilineage dysplasia; RS: ring sideroblasts; SLD: single lineage dysplasia; WHO: World Health Organisation; y: years.

657

N. van Leeuwen-Kerkhoff et al. A

Figure 1. Legend on following page.

658

haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

Figure 1. Cell subset enumeration in myelodysplastic syndrome- and healthy-derived bone marrow and peripheral blood. (A) Gating strategy of dendritic cells (DC) and slan+ non-classical monocytes in normal bone marrow (NBM) and myelodysplastic syndromes (MDS)-derived BM. After debris and doublet exclusion, CD45+ mononuclear cells were gated. Then plasmacytoid DC (pDC), myeloid DC (cDC1 and cDC2) and slan+ monocytes were identified based on the expression of CD141high, CD1c and M-DC8/CD16, respectively. (B) Frequencies of different cell subsets in normal bone marrow (NBM) compared to MDS BM. In total 30 NBM samples and 187 MDS BM samples were used. Percentages were calculated from the mononuclear cell fraction. Mean frequencies ± standard error of the mean (SEM) are given (NBM vs. MDS BM: pDC 0.76% SEM ± 0.09 vs. 0.91% SEM ± 0.11, cDC1 0.048% SEM ± 0.006 vs. 0.030% SEM ± 0.003, cDC2 0.67% SEM ± 0.05 vs. 0.54% SEM ± 0.04 and slan+ 0.36% SEM ± 0.07 vs. 0.24% SEM ± 0.02). (C) Cell frequencies in different classification groups according to the 2016 World Health Organization (WHO) classification. Patients having a higher blast count-related 2016 WHO classification (EB-1/EB-2) show lower percentages of DC and slan+ monocytes compared to NBM and lower risk groups (SLD/MLD/RS-SLD/RS-MLD). NBM (n=30) vs. (RS-)SLD/MLD (n=115) vs. EB-1/EB-2 (n=48): pDC 0.76% vs. 1.11% vs. 0.63%, cDC1 0.048% vs. 0.038% vs. 0.015%, cDC2 0.67% vs. 0.59% vs. 0.44%, slan+ 0.36% vs. 0.24% vs. 0.23%. (D) Cell frequencies in different risk groups within the International Prognostic Scoring System (IPSS). The percentages of myeloid DC subsets decrease gradually in higher risk groups. NBM (n=30) vs. low risk (n=49) vs. intermediate-1 (n=71) vs. intermediate-2 (n=21) vs. high risk (n=5): cDC1 0.048% vs. 0.035% vs. 0.031% vs. 0.010% vs. 0.005%, cDC2 0.67% vs. 0.57% vs. 0.52% vs. 0.42% vs. 0.26%. (E) Cell frequencies in different risk groups within the IPSS-R. Again, differences between subgroups are mainly seen in DC subsets. Higher risk groups show lower percentages of DC compared to NBM and lower risk groups. NBM (n=30) vs. very low/low risk (n=77) vs. intermediate risk (n=32) vs. high/very high risk (n=27): cDC1 0.048% vs. 0.038% vs. 0.019% vs. 0.013%, cDC2 0.67% vs. 0.62% vs. 0.41% vs. 0.35%. (F) Correlation of cell frequencies in MDS-derived peripheral blood (PB) and BM samples. In total, 26 paired MDS samples were included. The non-parametric Spearman’s correlation test was used to find significant correlations between frequencies in PB and BM. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. EB: excess blasts; MLD: multilineage dysplasia; RS-MLD: ring sideroblasts with multilineage dysplasia; RS-SLD: ring sideroblasts with single lineage dysplasia; SLD: single lineage dysplasia.

Figure 2. Clonal involvement of dendritic cells subsets and slan+ monocytes. Fluorescence in situ hybridization (FISH) analysis of sorted cells, including B cells and CD34+ blast cells, with a known cytogenetic aberrancy. In three tested cases (monosomy 7, del 5q and trisomy 8), isolated cDC2 and CD34+ blast cells were highly involved in the dysplastic clone, whereas B cells were not involved. Slan+ monocytes showed clonal involvement in del5q. Interphase FISH on whole bone marrow samples showed both an aberrant and a normal cell line. A representative FISH analysis is shown in which interphase cells are hybridized with the chromosome 5q probe displayed in red and 5p probe displayed in green (LSI EGR1(5q31)/D5S23,D5S721(5p15.2) Dual Colour Probe Set). Loss of 5q is seen in CD34+ blasts, cDC2 and slan+ monocytes (2G1R), but not in B cells (2G2R).

complement receptors (C3AR1) are at the top of this list. Further leading-edge analysis using all five gene sets was performed to identify the genes that highly account for the gene set’s enrichment signal (Figure 4A). There was a great overlap of genes that formed the leading-edge subset between all gene sets (Figure 4B). In total, 418 genes were found to form the leading-edge subset for cDC2 of which 32 were present in all five gene sets. For slan+ monocytes 353 genes formed the leading-edge subset. Of them, 39 genes were found in all five gene sets. These genes were considered most relevant because they form the core of the enrichment (Figure 4C). Again, both lists with lead targets consisted of multiple pattern recognition receptors, which suggests an overall diminished capacity for sensing pathogen/damage associated molecular patterns (PAMP/DAMP) by MDS-derived cells. This was further confirmed by the fact that also genes that were highly involved in subsequent down-stream cell signaling, such as BTK, CARD9, IRAK4, IRF3/7, MyD88, SYK, and usually lead to activation of pro-inflammatory processes, were down-regulated in MDS-derived APC. haematologica | 2022; 107(3)

Myelodysplastic syndrome-derived cells show reduced T-cell priming capacities and clear Th1/2-type T cell skewing Next, in order to confirm the hypothesis that was formed from the gene expression profiling data, functional capacities of cDC2 and slan+ monocytes were tested. Upon stimulation with LPS and R848, a combination of proven synergistically working TLR ligands,33 cDC2 showed upregulation of co-stimulatory molecules. In contrast, slan+ monocytes were unable to upregulate maturation markers (Figure 5A). Again, cDC1 were not tested because of low frequencies. There was no statistical difference in maturation capacity for cDC2 when they were compared to NBM-derived cDC2. Slan+ monocytes showed a significantly reduced ability to upregulate CD80 upon stimulation compared to their equivalents in NBM (Figure 5B). In order to investigate their cytokine secreting capacity, MDS-derived cDC2 and slan+ monocytes were isolated and either left non-stimulated or stimulated overnight with TLR ligands. Culture supernatants were tested for the presence of different cytokines. Compared 659

N. van Leeuwen-Kerkhoff et al.

Figure 3. Transcriptomic comparison between healthy donor- and myelodysplastic syndrome-derived cell subsets. Three healthy donor (HD)-derived samples and four myelodysplastic syndromes (MDS)-derived samples were used for the isolation of cDC2 and slan+ monocytes and subsequent microarray analysis. (A) Hierarchical clustering, based on differentially expressed genes, of replicate samples for cDC2 and slan+ monocytes. Heatmap visualization is used to show transcript clustering for the two different conditions (HD vs. MDS). (B) Volcano plots showing overand under-expressed genes in red and green, respectively, in HD compared to MDS. A cut-off of -2.5 / 2.5 for foldchange and a P-value < 0.05 were used to show results. (C) Pathway analyses for transcripts that are under-expressed in MDS compared to HD for cDC2 and slan+ monocytes. Coding differentially expressed genes (774 genes for cDC2 and 987 genes for slan+ monocytes) were selected and imported into the STRING v10.5 bioinformatics tool. Six enriched biological processes with lowest false discovery rate (FDR) are shown for cDC2 and slan+ monocytes (in total, eight enriched pathways were found for cDC2 and 383 for slan+ monocytes). (D and E) Gene set enrichment analysis for HD- and MDS-derived cDC2 (D) and slan+ monocytes (E). For both subsets five gene sets were enriched in HD. An enrichment plot is displayed for each subset. Heatmaps show the core enriched genes (67 for cDC2 and the top 75 of 280 for slan+ monocytes) with interesting genes highlighted by black stars. GO: gene ontology; NES: normalized enrichment score; Nom P-value, nominal P-value.

660

haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

Figure 4. Leading-Edge Analysis. (A) Leading-Edge Analysis using five gene sets for both cell subsets. The tables show total number of genes that are present in a specified gene set and the number and percentage of genes that were considered to form the leading-edge subset of that gene set. (B) A set-to-set analysis for cDC2 and slan+ monocytes. Overlap in leading-edge genes between gene sets are displayed using a color intensity graph. A dark green cell indicates that sets have the same leading-edge genes. (C) Heatmap of the leading-edge subset for cDC2 and slan+ monocytes. Genes displayed are present in the leading-edge subset of all five gene sets. The heatmaps show relative expression levels per gene between healthy donor (HD) and myelodysplastic syndrome (MDS) samples. FDR: false discovery rate; GO: gene ontology; NES: normalized enrichment score; Nom P-value: nominal P-value.

haematologica | 2022; 107(3)

661

N. van Leeuwen-Kerkhoff et al.

Figure 5. Functional capacities of dendritic cell subsets and slan+ monocytes. (A) Maturation capacity of cDC2 and slan+ monocytes in myelodysplastic syndrome (MDS) bone marrow (BM) upon toll-like receptor (TLR)-stimulation. Expression levels of CD80, CD86 and HLA-DR were assessed by flow cytometry at baseline (T=0) and after overnight stimulation with LPS + R848 (+). Mean fluorescence intensity (MFI) values of these three markers were measured. Median values of 4-7 experiments are shown. (B) Up-regulation of CD80, CD86 and HLA-DR after overnight TLR-stimulation in normal bone marrow (NBM)- (n=4) and MDS-derived (n=4-7) cDC2 and slan+ monocytes. Median values are shown. (C) Cytokine secretion assay. Culture supernatants of healthy (n=4) and MDS (n=10) BM-derived unstimulated and stimulated cDC2 and slan+ monocytes were analyzed for the presence of different cytokines by cytometric bead array. Median values are shown. (D) Allogeneic mixed leukocyte reaction (MLR). Peripheral blood lymphocytes (PBL) were labeled with carboxyfluorescein succinimidyl ester (CFSE) and co-cultured with healthy (n=2-4) or MDS-derived (n=1-5) stimulated cDC1, cDC2 or slan+ monocytes. The percentage of CFSE-diluted T cells was determined by flow cytometry. Median values of different experiments are shown. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. pg: picogram.

662

haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

Figure 6. Mass cytometry of T cells co-cultured in the presence of healty donor- or myelodysplastic syndrome-derived slan+ monocytes. Healthy donor (HD)-derived CD4+ T cells were co-cultured in the presence of slan+ non-classical monocytes from healthy donors (n=2) or from myelodysplastic syndrome (MDS) patients (n=2). T cells at the start of the experiment (named “day 0”) as well as T cells co-cultured for 5 days with slan+ non-classical monocytes were stained with a panel consisting of surface markers and intracellular markers, and markers for transcription factors and cytokines and analysed using mass cytometry (CyTOF). First, viable T cells were identified for each experiment. Then the FlowSOM algorithm was used to identify 15 metaclusters containing cells that express the same set of markers. (A) T cells are visualized using viSNE plots. The expression of a selection of markers are shown in the viSNE plots for cultures containing HD- or MDS-derived slan+ monocytes at day 0 and day 5. T-cell subsets were identified based on the expression of IFN-γ, Tbet, IL-4, GATA3, IL-17, CD25, CD127, IL-10 and FoxP3 (Th1 were considered to be IFN-γ and Tbet+, IL-17 and GATA3–; Th2 were GATA3+ or IL-4+; Th17 were IL-17+; Tregs were CD127- and FoxP3+CD25+). (B) FlowSOM-identified metaclusters were laid over day 0 and day 5 viSNE maps. Percentages of identified T-cell subsets at the start of the experiment and at day 5 are shown. Compared to day 0, HD-derived slan+ non-classical monocytes mainly induced pro-inflammatory T cells (Th1 and Th17), as well as collateral Tregs. In contrast, T cells cultured in the presence of slan+ non-classical monocytes from MDS patients showed Th1, and above all, Th2 skewing. In HD-derived cultures Th2 cells disappeared at day 5.

haematologica | 2022; 107(3)

663

N. van Leeuwen-Kerkhoff et al.

to NBM-derived cells, no statistical differences, except for IL-8 in cDC2 cultures, were found in the ability to secrete cytokines upon stimulation (Figure 5C). There was a wide variability between individual experiments indicating that cells from some MDS patients were able to secrete high amounts of cytokines, whereas others secreted hardly any. Next, isolated NBM- and MDS-derived cDC1, cDC2 and slan+ monocytes were co-cultured with HD-derived T cells. The capacity to induce CD4+ and CD8+ T-cell proliferation was clearly reduced for all MDS-derived subsets compared to NBM-derived subsets (Figure 5D). In order to further investigate the T-cell skewing capacity of slan+ monocytes, mass cytometry (CyTOF) was used. Both healthy PB-derived as well as MDS-derived slan+ monocytes were co-cultured with CD4+ T cells. Using the FlowSOM algorithm, 15 metaclusters were designated representing cells with similar marker expression profiles. Then, distinct T-cell subsets were identified and assigned to a specific metacluster using viSNE plots (Figure 6A and B). Compared to day 0 T cells co-cultured with HD-derived slan+ monocytes showed an increase of Th1, Th17 and Treg cells, whereas growth of Th2 cells was not supported (Figure 6B). This was in contrast with MDSderived slan+ monocytes. They induced a mixed T-cell response, including a clear Th2 cell differentiation, compared to day 0.

Discussion DC are important regulators of immune reactions and form a crucial bridge between the innate and adaptive immune system by directing T-cell responses. Alterations in DC frequency and function have been widely reported in the context of several diseases such as autoimmunity and cancer.34-38 In hematological malignancies the number of DC is often decreased and studies on their function mainly show an impaired induction of type-I immune responses.14,15,39-43 Slan+ monocytes have also been investigated in the context of disease. They are recruited to the site of inflammation in chronic inflammatory conditions.44-51 In cancer, including hematological malignancies, variable functional characteristics have been observed for slan+ monocytes.52-55 Enhanced stimulation of tumor-specific T-cell responses as well as differentiation into a more tolerogenic subtype have been described for this particular subset. For MDS, so far most studies that have been published describe DC in general without discriminating between subsets, or they describe in vitro generated MoDC.24-27,56-59 No reports have been published on MDSderived slan+ monocytes. Since the immune system plays an important role in MDS pathogenesis and is an attractive target for therapies, this study focused on in vivo circulating cDC and slan+ monocytes. Of note, in part because of their shared lineage ontogeny with MDS blasts, we decided to focus mainly on myeloid cDC subsets in this study and, hence, our data set lacks transcriptional and functional findings for pDC. Moreover, at the time of this study, the recently proposed subdivision of cDC2 into DC2 and DC3 had not been recognized yet; therefore, we did not analyze discriminating markers like CD32b, CD36 and CD163. Frequencies of all studied APC subsets, except for pDC, were lower in MDS BM compared to NBM. This decrease was most prominent in higher MDS risk groups (according to the IPSS(-R) or 2016 664

WHO classification). Rates of cell subsets strongly correlated between the PB and BM compartment in MDS. The finding of lower DC frequencies in MDS is important and may partly explain the poor immune responses seen in a subgroup of patients. Especially in high-risk groups, in which the dysplastic clone evades immune surveillance, restoration of the cDC lineage differentiation could be of benefit. Intact DC frequency is also relevant in vaccination strategies targeting in vivo circulating cDC. It was shown in a phase-I trial that MDS patients with higher numbers of cDC1 showed a more robust immune response to vaccination with the NY-ESO-1 antigen.60 Another relevant question in this context is whether lower frequencies result from clonal involvement of the DC compartment or not. Previously, Ma et al. showed clonal involvement of myeloid DC, which were characterized by the expression of CD33 and HLA-DR.25 Using the recommended phenotypic sub-division, we now clearly show that cDC2 are clonally involved. Unfortunately, cDC1 could not be tested because of low cell counts, but their shared lineage ontogeny with cDC2 suggests that they likely would be clonally involved as well.61 Although cell numbers were also low for slan+ non-classical monocytes, we were able to confirm mixed clonal involvement. Additional reasons for decreased frequencies of DC in MDS, such as increased apoptotic rates of hematopoietic stem cells, should be investigated in future studies. Our transcriptional profiling data showed enriched apoptotic pathways in cDC2 from MDS patients and indeed underlines this hypothesis. Very recently, Srivastava et al. showed that a decrease in DC progenitor cells could partly explain the decrease in DC frequency in MDS.62 Furthermore, they showed that higher frequencies of cDC1 in the BM of MDS patients correlated with better overall survival independent of risk categories whereas cDC2 frequencies did not. Reduced IRF8 expression, a crucial transcription factor for cDC1 differentiation, was associated with lower cDC1 numbers. Inhibition of LSD1, using therapeutically relevant compounds, enhanced the expression of IRF8 and subsequent differentiation to cDC1 and could therefore be of potential benefit in restoring DC frequencies. If this is also the case for e.g., IRF4 in cDC2 remains to be investigated. In our transcriptional analysis of cDC2 and slan+ monocytes, innate immunity and danger response-related transcripts were prominently under-represented in MDSderived subsets as compared to their HD-derived counterparts. Under-represented transcripts included pattern recognition receptors, Fc-γ receptors and down-stream signaling elements. Since these receptors and their signaling pathways form a crucial basis for normal DC/monocyte function, disruption of expression can lead to diminished immune responsiveness and possibly immune escape of dysplastic myeloid blasts and aberrant stem cells in the BM microenvironment. A wide range of TLR was found to be downregulated in MDS-derived APC. TLR are important receptors for both PAMP/DAMP-derived danger signals, which upon binding of their ligands trigger activation of downstream signaling pathways, involving, amongst others, NF-κB, MyD88 or IRAK1/4 kinases. This normally leads to the activation of pro-inflammatory transcriptional programs. Our transcriptional finding of lower PAMP/DAMP-sensing molecules and down-stream signaling genes in MDS-derived APC is therefore striking. It suggests a defective DC functionality in response to cellular haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

stress. And, in the context of MDS it could lead to higher vulnerability for infections, which is often seen in this patient group. Altered TLR profiles have been observed in MDS CD34+ hematopoietic stem cells (HSC) before.16,63-66 Sustained TLR activation and constitutively activated downstream molecules as well as a loss of TLR signaling repressors have both been described in this context. Furthermore, enhanced TLR-mediated signaling in CD14+ MDS BM cells has been shown.67 It has been suggested that together this would lead to chronic immune stimulation and subsequent DNA damage and increased cell death. In contrast, it has been shown in AML that TLR stimulation of AML blasts with TLR agonists has a positive effect and leads to differentiation.68 Thus, on one hand, increased TLR triggering can lead to excessive immune activation in hematopoietic progenitor cells, but on the other hand, decreased TLR function and defective down-stream signaling in immune effector cells (such as APC) can possibly lead to inadequate induction of immune responses and immune escape. This is important information for future studies that investigate the effect of therapeutic inhibition (by for example TLR antagonists) of these pathways in the complex MDS environment. Additionally, our gene expression data could form the basis for new research on expression levels of TLR on different MDS-derived immune cells. These expression levels can be correlated to immune cell function and therapeutic agents (e.g., TLR agonists/antagonists) should be tested in order to gain information on possible functional restoration of these cells. DC primarily function as APC and for effective DC activity three signals are required for the interaction with T cells ( i) antigen presentation; ii) co-stimulation and iii) cytokine secretion). MDS BM- and NBM-derived APC were tested for their maturation, cytokine secretion and induction of T-cell priming capacity. cDC2 from MDS patients were able to upregulate HLA-DR and co-stimulatory molecules to approximately the same extent as NBM cDC2. In contrast, slan+ monocytes showed impaired maturation capacity. For both subsets, the ability for cytokine secretion seemed largely unaffected. Although the capacity of upregulating co-stimulatory molecules and secreting cytokines was intact, cDC2 were unable to translate this into effective T-cell proliferation induction. Also cDC1 and slan+ monocytes showed a negatively affected T-cell stimulatory function. These results are part-

References 1. Breton G, Zheng S, Valieris R, et al. Human dendritic cells (DCs) are derived from distinct circulating precursors that are precommitted to become CD1c+ or CD141+ DCs. J Exp Med. 2016;213(13):2861-2870. 2. Breton G, Lee J, Zhou YJ, et al. Circulating precursors of human CD1c+ and CD141+ dendritic cells. J Exp Med. 2015;212(3):401413. 3. Schlitzer A, Sivakamasundari V, Chen J, et al. Identification of cDC1- and cDC2-committed DC progenitors reveals early lineage priming at the common DC progenitor stage in the bone marrow. Nat Immunol. 2015;16(7):718-728. 4. Dzionek A, Fuchs A, Schmidt P, et al. BDCA-2, BDCA-3, and BDCA-4: three markers for distinct subsets of dendritic

haematologica | 2022; 107(3)

ly in line with previously published data on in vitro generated MoDC from MDS patients.26,27,56,57 In these studies, MoDC were already affected in their maturation and cytokine secreting capacity, underlining the difference in naturally occurring DC and MoDC. Additionally, the effect of slan+ monocytes on T-cell polarization was assessed using mass cytometry. A clear induction of proinflammatory T cells (Th1 and Th17) as well as Tregs, combined with a considerable decrease in Th2 cells was observed in cultures with HD-derived slan+ monocytes. In contrast, cultures with MDS-derived slan+ monocytes showed similar Th1/Th17 skewing, but failed to expand Tregs and, above all, Th2 cells were induced, revealing altered T-effector cell induction. Thus, in addition to lower DC frequencies, we show altered functionality of MDS APC, which again can contribute to ineffective immune responses. Of note, there is high inter-patient variability and it would be highly interesting to correlate DC function with clinical parameters (e.g., molecular background, transfusion independence, better overall and leukemia-free survival) and frequencies of other immune cells by performing extensive immune profiling studies. Interestingly, Wang et al. have suggested an inhibitory role for MDS-derived mesenchymal stem cells (MSC) that may contribute to altered MoDC function.59 Whether this is also the case for different DC subsets has to be confirmed in future research. In conclusion, this study provides the first data on the frequency and functionality of cDC subsets and slan+ monocytes in the context of MDS. It shows a clearly affected pro-inflammatory status of MDS-derived APC which might contribute to the complex process of immune escape. A more comprehensive insight in the basal immune-pathogenesis of this disease is essential for future studies that focus on new immunotherapeutic intervention options. Disclosures No conflicts of interest to disclose. Contributions NvL-K performed experiments; NvL-K, PP, GP and JT analyzed data; NvL-K, TW, PP, GP, JT, SK, TdG, and AvdL interpreted data; NvL-K, TW, SK, TdG, and AvdL designed research; PP performed FISH experiments; NvL-K, TW, SK, PP, TdG and AvdL wrote the paper.

cells in human peripheral blood. J Immunol. 2000;165(11):6037-6046. 5. MacDonald KP, Munster DJ, Clark GJ, et al. Characterization of human blood dendritic cell subsets. Blood. 2002;100(13):45124520. 6. Robbins SH, Walzer T, Dembélé D, et al. Novel insights into the relationships between dendritic cell subsets in human and mouse revealed by genome-wide expression profiling. Genome Biol. 2008;9(1):R17. 7. Schäkel K, von Kietzell M, Hänsel A, et al. Human 6-sulfo LacNAc-expressing dendritic cells are principal producers of early interleukin-12 and are controlled by erythrocytes. Immunity. 2006;24(6):767-777. 8. Schäkel K, Mayer E, Federle C, et al. A novel dendritic cell population in human blood: one-step immunomagnetic isolation

by a specific mAb (M-DC8) and in vitro priming of cytotoxic T lymphocytes. Eur J Immunol. 1998;28(12):4084-4093. 9. Schäkel K, Kannagi R, Kniep B, et al. 6Sulfo LacNAc, a novel carbohydrate modification of PSGL-1, defines an inflammatory type of human dendritic cells. Immunity. 2002;17(3):289-301. 10. van Leeuwen-Kerkhoff N, Lundberg K, Westers TM, et al. Human bone marrowderived myeloid dendritic cells show an immature transcriptional and functional profile compared to their peripheral blood counterparts and separate from Slan+ nonclassical monocytes. Front Immunol. 2018;9(1619):1-18. 11. van Leeuwen-Kerkhoff N, Lundberg K, Westers TM, et al. Transcriptional profiling reveals functional dichotomy between human slan + non-classical monocytes and

665

N. van Leeuwen-Kerkhoff et al. myeloid dendritic cells. J Leukoc Biol. 2017;102(4):1055-1068. 12. Hofer TP, Zawada AM, Frankenberger M, et al. slan-defined subsets of CD16-positive monocytes: impact of granulomatous inflammation and M-CSF receptor mutation. Blood. 2015;126(24):2601-2611. 13. Banchereau J, Steinman RM. Dendritic cells and the control of immunity. Nature. 1998;392(6673):245-252. 14. Mohty M, Isnardon D, Vey N, et al. Low blood dendritic cells in chronic myeloid leukaemia patients correlates with loss of CD34+/CD38- primitive haematopoietic progenitors. Br J Haematol. 2002; 119(1):115-118. 15. Mohty, M, Jarrosay, D, Lafage-Pochitaloff L, et al. Circulating blood dendritic cells from myeloid leukemia patients display quantitative and cytogenetic abnormalities as well as functional impairment. Blood. 2001;98(13):3750-3756. 16. Barreyro L, Chlon TM, Starczynowski DT. Chronic immune response dysregulation in MDS pathogenesis. Blood. 2018;132(15): 1553-1560. 17. Epling-Burnette PK, Bai F, Painter JS, et al. Reduced natural killer (NK) function associated with high-risk myelodysplastic syndrome (MDS) and reduced expression of activating NK receptors. Blood. 2007;109 (11):4816-4824. 18. Kiladjian J-J, Bourgeois E, Lobe I, et al. Cytolytic function and survival of natural killer cells are severely altered in myelodysplastic syndromes. Leukemia. 2006; 20(3):463-470. 19. Kordasti SY, Afzali B, Lim Z, et al. IL-17producing CD4(+) T cells, pro-inflammatory cytokines and apoptosis are increased in low risk myelodysplastic syndrome. Br J Haematol. 2009;145(1):64-72. 20. Bouchliou I, Miltiades P, Nakou E, et al. Th17 and Foxp3(+) T regulatory cell dynamics and distribution in myelodysplastic syndromes. Clin Immunol. 2011;139(3):350-359. 21. Kordasti SY, Ingram W, Hayden J, et al. CD4+CD25high Foxp3+ regulatory T cells in myelodysplastic syndrome (MDS). Blood. 2007;110(3):847-850. 22. Kotsianidis I, Bouchliou I, Nakou E, et al. Kinetics, function and bone marrow trafficking of CD4+CD25+FOXP3+ regulatory T cells in myelodysplastic syndromes (MDS). Leukemia. 2009;23(3):510-518. 23. Kittang AO, Kordasti S, Sand KE, et al. Expansion of myeloid derived suppressor cells correlates with number of T regulatory cells and disease progression in myelodysplastic syndrome. Oncoimmunology. 2015;5(2):e1062208. 24. Saft L, Björklund E, Berg E, et al. Bone marrow dendritic cells are reduced in patients with high-risk myelodysplastic syndromes. Leuk Res. 2013;37(3):266-273. 25. Ma L, Delforge M, van Duppen V, et al. Circulating myeloid and lymphoid precursor dendritic cells are clonally involved in myelodysplastic syndromes. Leukemia. 2004;18(9):1451-1456. 26. Ma L, Ceuppens J, Kasran A, et al. Immature and mature monocyte-derived dendritic cells in myelodysplastic syndromes of subtypes refractory anemia or refractory anemia with ringed sideroblasts display an altered cytokine profile. Leuk Res. 2007;31(10):1373-1382. 27. Matteo Rigolin G, Howard J, Buggins A, et al. Phenotypic and functional characteristics of monocyte-derived dendritic cells

666

from patients with myelodysplastic syndromes. Br J Haematol. 1999;107(4):844850. 28. Greenberg P, Cox C, LeBeau MM, et al. International scoring system for evaluating prognosis in myelodysplastic syndromes. Blood. 1997;89(6):2079-2088. 29. Greenberg PL, Tuechler H, Schanz J, et al. Revised international prognostic scoring system for myelodysplastic syndromes. Blood. 2012;120(12):2454-2465. 30. Kotecha N, Krutzik PO, Irish JM. Webbased analysis and publication of flow cytometry experiments. Curr Protoc Cytom. 2010;Chapter 10:Unit10.17. 31. Van Der Maaten L, Hinton G. Visualizing Data using t-SNE. J Mach Learn Res. 2008;9:2579-2605. 32. Van Gassen S, Callebaut B, Van Helden MJ, et al. FlowSOM: Using self-organizing maps for visualization and interpretation of cytometry data. Cytometry A. 2015;87(7):636-645. 33. Nizzoli G, Krietsch J, Weick A, et al. Human CD1c+ dendritic cells secrete high levels of IL-12 and potently prime cytotoxic T cell responses. Blood. 2013;122(6):932942. 34. Perrot I, Blanchard D, Freymond N, et al. Dendritic cells infiltrating human non-small cell lung cancer are blocked at immature stage. J Immunol. 2007;178(5):2763-2769. 35. Della Bella S, Gennaro M, Vaccari M, et al. Altered maturation of peripheral blood dendritic cells in patients with breast cancer. Br J Cancer. 2003;89(8):1463-1472. 36. Tran Janco JM, Lamichhane P, Karyampudi L, Knutson KL. Tumor-infiltrating dendritic cells in cancer pathogenesis. J Immunol. 2015;194(7):2985-2991. 37. Chan VS-F, Nie Y-J, Shen N, et al. Distinct roles of myeloid and plasmacytoid dendritic cells in systemic lupus erythematosus. Autoimmun Rev. 2012;11(12):890-897. 38. Carvalheiro T, Rodrigues A, Lopes A, et al. Tolerogenic versus inflammatory activity of peripheral blood monocytes and dendritic cells subpopulations in systemic lupus erythematosus. Clin Dev Immunol. 2012;2012:934161. 39. Orsini E, Calabrese E, Maggio R, et al. Circulating myeloid dendritic cell directly isolated from patients with chronic myelogenous leukemia are functional and carry the bcr-abl translocation. Leuk Res. 2006;30(7):785-794. 40. Dong R, Cwynarski K, Entwistle A, et al. Dendritic cells from CML patients have altered actin organization, reduced antigen processing, and impaired migration. Blood. 2003;101(9):3560-3567. 41. Dhodapkar KM, Barbuto S, Matthews P, et al. Dendritic cells mediate the induction of polyfunctional human IL17-producing cells (Th17-1 cells) enriched in the bone marrow of patients with myeloma. Blood. 2008;112(7):2878-2885. 42. Leone P, Berardi S, Frassanito MA, et al. Dendritic cells accumulate in the bone marrow of myeloma patients where they protect tumor plasma cells from CD8+ T-cell killing. Blood. 2015;126(12):1443-1451. 43. Saulep-Easton D, Vincent FB, Le Page M, et al. Cytokine-driven loss of plasmacytoid dendritic cell function in chronic lymphocytic leukemia. Leukemia. 2014; 28(10):2005-2015. 44. Hänsel A, Günther C, Baran W, et al. Human 6-sulfo LacNAc (slan) dendritic cells have molecular and functional features of an important pro-inflammatory cell type

in lupus erythematosus. J Autoimmun. 2013;40:1-8. 45. Hänsel A, Günther C, Ingwersen J, et al. Human slan (6-sulfo LacNAc) dendritic cells are inflammatory dermal dendritic cells in psoriasis and drive strong TH17/TH1 T-cell responses. J Allergy Clin Immunol. 2011;127(3):787-794. 46. Thomas K, Dietze K, Wehner R, et al. Accumulation and therapeutic modulation of 6-sulfo LacNAc(+) dendritic cells in multiple sclerosis. Neurol Neuroimmunol Neuroinflamm. 2014;1(3):e33. 47. Bsat M, Chapuy L, Baba N, et al. Differential accumulation and function of proinflammatory 6-sulfo LacNAc dendritic cells in lymph node and colon of Crohn’s versus ulcerative colitis patients. J Leukoc Biol. 2015;98(4):671-681. 48. Ogino T, Nishimura J, Barman S, et al. Increased Th17-inducing activity of CD14+ CD163 low myeloid cells in intestinal lamina propria of patients with Crohn’s disease. Gastroenterology. 2013;145(6):13801391. 49. Olaru F, Döbel T, Lonsdorf AS, et al. Intracapillary immune complexes recruit and activate slan-expressing CD16+ monocytes in human lupus nephritis. JCI Insight. 2018;3(11):e96492. 50. Baran W, Oehrl S, Ahmad F, et al. Phenotype, function, and mobilization of 6-sulfo LacNAc-expressing monocytes in atopic dermatitis. Front Immunol. 2018;9:1352. 51. Ahmad F, Döbel T, Schmitz M, Schäkel K. Current Concepts on 6-sulfo LacNAc Expressing Monocytes (slanMo). Front Immunol. 2019;10:948. 52. Vermi W, Micheletti A, Lonardi S, et al. slanDCs selectively accumulate in carcinoma-draining lymph nodes and marginate metastatic cells. Nat Commun. 2014;5 (1):3029. 53. Vermi W, Micheletti A, Finotti G, et al. slan+ Monocytes and macrophages mediate CD20-dependent B-cell lymphoma elimination via ADCC and ADCP. Cancer Res. 2018;78(13):3544-3559. 54. Toma M, Wehner R, Kloß A, et al. Accumulation of tolerogenic human 6-sulfo LacNAc dendritic cells in renal cell carcinoma is associated with poor prognosis. Oncoimmunology. 2015;4(6):e1008342. 55. Lamarthée B, de Vassoigne F, Malard F, et al. Quantitative and functional alterations of 6-sulfo LacNac dendritic cells in multiple myeloma. Oncoimmunology. 2018;7(7):e1444411. 56. Micheva I, Thanopoulou E, Michalopoulou S, et al. Defective tumor necrosis factor alpha-induced maturation of monocytederived dendritic cells in patients with myelodysplastic syndromes. Clin Immunol. 2004;113(3):31031-31037. 57. Davison GM, Novitzky N, Abdulla R. Monocyte derived dendritic cells have reduced expression of co-stimulatory molecules but are able to stimulate autologous T-cells in patients with MDS. Hematol Oncol Stem Cell Ther. 2013;6(2):49-57. 58. Micheva I, Thanopoulou E, Michalopoulou S, et al. Impaired generation of bone marrow CD34-derived dendritic cells with low peripheral blood subsets in patients with myelodysplastic syndrome. Br J Haematol. 2004;126(6):806-814. 59. Wang Z, Tang X, Xu W, et al. The different immunoregulatory functions on dendritic cells between mesenchymal stem cells derived from bone marrow of patients with

haematologica | 2022; 107(3)

Dendritic cells in myelodysplastic syndromes

low-risk or high-risk myelodysplastic syndromes. PLoS One. 2013;8(3):e57470. 60. Griffiths EA, Srivastava P, Matsuzaki J, et al. Clinical NY-ESO-1 Vaccination in combination with decitabine induces antigenspecific T-lymphocyte responses in patients with myelodysplastic syndrome. Clin Cancer Res. 2018;24(5):1019-1029. 61. Srivastava P, Tzetzo SL, Gomez EC, et al. Inhibition of LSD1 in MDS progenitors restores differentiation of CD141Hi conventional dendritic cells. Leukemia. 2020;34(9):2460-2472. 62. See P, Dutertre CA, Chen J, et al. Mapping the human DC lineage through the integration of high-dimensional techniques.

haematologica | 2022; 107(3)

Science. 2017;356(6342):eaag3009. 63. Maratheftis CI, Andreakos E, Moutsopoulos HM, Voulgarelis M. Tolllike receptor-4 is up-regulated in hematopoietic progenitor cells and contributes to increased apoptosis in myelodysplastic syndromes. Clin Cancer Res. 2007;13(4):1154-1160. 64. Wei Y, Dimicoli S, Bueso-Ramos C, et al. Toll-like receptor alterations in myelodysplastic syndrome. Leukemia. 2013; 27(9):1832-1840. 65. Dimicoli S, Wei Y, Bueso-Ramos C, et al. Overexpression of the Toll-like receptor (TLR) signaling adaptor MYD88, but lack of genetic mutation, in myelodysplastic

syndromes. PLoS One. 2013;8(8):e71120. 66. Monlish DA, Bhatt ST, Schuettpelz LG. The role of Toll-like receptors in hematopoietic malignancies. Front Immunol. 2016;7:390. 67. Velegraki M, Papakonstanti E, Mavroudi I, et al. Impaired clearance of apoptotic cells leads to HMGB1 release in the bone marrow of patients with myelodysplastic syndromes and induces TLR4-mediated cytokine production. Haematologica. 2013;98(8):1206-1215. 68. Ignatz-Hoover JJ, Wang H, Moreton SA, et al. The role of TLR8 signaling in acute myeloid leukemia differentiation. Leukemia. 2015;29(4):918-926.

667

ARTICLE Ferrata Storti Foundation

Hemostasis

Sialylation on O-linked glycans protects von Willebrand factor from macrophage galactose lectin-mediated clearance Soracha E. Ward,1 Jamie M. O’Sullivan,1 Alan B. Moran,2,3 Daniel I. R. Spencer,2 Richard A. Gardner,2 Jyotika Sharma,4 Judicael Fazavana,1 Marco Monopoli,5 Thomas A. J. McKinnon,6 Alain Chion,1 Sandra Haberichter7 and James S. O’Donnell1,8,9

Haematologica 2022 Volume 107(3):668-679

Irish Centre for Vascular Biology, School of Pharmacy and Biomolecular Sciences, Royal College of Surgeons in Ireland, Dublin, Ireland; 2Ludger, Ltd., Culham Science Centre, Abingdon, Oxfordshire, UK; 3Leiden University Medical Centre, Centre for Proteomics and Metabolomics, Leiden, the Netherlands; 4Department of Basic Biomedical Sciences, University of North Dakota School of Medicine and Health Sciences, Grand Forks, ND, USA; 5Department of Chemistry, Royal College of Surgeons in Ireland, Dublin, Ireland; 6Faculty of Medicine, Imperial College, Hammersmith Hospital, London, UK; 7 Versiti, Blood Research Institute, Milwaukee, WI, USA; 8National Children's Research Centre, Our Lady's Children's Hospital, Dublin, Ireland and 9National Coagulation Centre, St James’s Hospital, Dublin, Ireland 1

ABSTRACT

Correspondence: JAMES O’DONNELL jamesodonnell@rcsi.ie Received: October 21, 2020. Accepted: March 12, 2021. Pre-published: March 25, 2021. https://doi.org/10.3324/haematol.2020.274720

668

erminal sialylation determines the plasma half-life of von Willebrand factor (VWF). A role for macrophage galactose lectin (MGL) in regulating hyposialylated VWF clearance has recently been proposed. In this study, we showed that MGL influences physiological plasma VWF clearance. MGL inhibition was associated with a significantly extended mean residence time and 3-fold increase in endogenous plasma VWF antigen levels (P<0.05). Using a series of VWF truncations, we further demonstrated that the A1 domain of VWF is predominantly responsible for enabling the MGL interaction. Binding of both full-length and VWF-A1-A2-A3 to MGL was significantly enhanced in the presence of ristocetin (P<0.05), suggesting that the MGL-binding site in A1 is not fully accessible in globular VWF. Additional studies using different VWF glycoforms demonstrated that VWF O-linked glycans, clustered at either end of the A1 domain, play a key role in protecting VWF against MGLmediated clearance. Reduced sialylation has been associated with pathological, increased clearance of VWF in patients with von Willebrand disease. Herein, we demonstrate that specific loss of α2-3 linked sialylation from O-glycans results in markedly increased MGL-binding in vitro, and markedly enhanced MGL-mediated clearance of VWF in vivo. Our data further show that the asialoglycoprotein receptor (ASGPR) does not have a significant role in mediating the increased clearance of VWF following loss of O-sialylation. Conversely however, we observed that loss of N-linked sialylation from VWF drives enhanced circulatory clearance predominantly via the ASGPR. Collectively, our data support the hypothesis that in addition to regulating physiological VWF clearance, the MGL receptor works in tandem with ASGPR to modulate enhanced clearance of aberrantly sialylated VWF in the pathogenesis of von Willebrand disease.

Introduction von Willebrand disease (VWD) is the commonest inherited human bleeding disorder and is caused by either quantitative or qualitative deficiency of plasma von Willebrand factor (VWF).1,2 Increased plasma clearance of VWF constitutes an important mechanism in the pathogenesis of VWD.3 The MCMDM-1VWD European study, US Zimmerman Program and Willebrand in the Netherlands (WIN) study have all found pathological, enhanced VWF clearance in approximately 45% of patients with type 1 VWD, leading to the proposal that patients

haematologica | 2022; 107(3)

O-linked glycans protect VWF

with a shortened VWF half-life should be considered as a distinct type 1C (1-Clearance) subgroup.4-8 Interestingly, subsequent studies have highlighted that enhanced VWF clearance also contributes to pathogenesis in patients with low VWF, as well as type 2 and type 3 VWD.7,9-12 Given the importance of enhanced clearance in the pathogenesis of VWD, significant research has focused on defining the cellular and molecular clearance pathways involved. Potential roles for macrophages, liver sinusoidal endothelial cells and hepatocytes have been proposed.13-17 A number of specific clearance receptors have also been described.3 These include the low-density lipoprotein receptor-related protein-1 (LRP1), the scavenger receptor class A member I (SR-A1), sialic-acid-bindingimmunoglobulin-like-lectins 5 (Siglec 5) and the macrophage galactose-type lectin (MGL) which are all expressed on macrophages.18-21 On liver sinusoidal endothelial cells, receptors that may play a role in VWF clearance include stabilin-2 (STAB2), scavenger receptor class A member 5 (SCARA 5) and C-type lectin domain family 4 member M (CLEC4M).17,22,23 Finally, the asialoglycoprotein receptor (ASGPR), predominantly expressed on hepatocytes and macrophages, has also been implicated.24 More than 30 different VWF sequence variations have been reported in patients with increased VWF clearance.3,25 The archetypal type 1C mutation is the VWD Vicenza variant which is characterized by an R1205H substitution in the D3 domain of VWF.15,26 VWF glycosylation also plays a critical role in determining the rate of clearance of the protein.27-30 For example, plasma VWF:antigen (VWF:Ag) levels are 20-30% lower in blood group O individuals compared to non-O subjects due to a significant reduction in plasma half-life.31,32 Enzymatic removal of terminal sialic acid residues from VWF also markedly enhances clearance.27,28 Moreover, genetic inactivation of the ST3Gal-IV sialyltransferase was associated with a significant reduction in plasma VWF half-life.33 These data are important from a clinical perspective because the majority of both the N- and O-linked glycans of VWF are normally capped by sialic acid residues.34-37 As glycoproteins age in plasma, there is a stepwise elimination of saccharides from the termini of complex glycan chains.38 Glycan remodeling begins with loss of capping sialic acid, catalyzed by plasma neuraminadases 1 (Neu1) and 3 (Neu3) respectively. This time-dependent desialylation is important in triggering clearance of senescent glycoproteins.38 Significantly reduced VWF sialylation levels have also been observed in a number of pathological conditions including sepsis, pulmonary hypertension and liver cirrhosis.39-41 Importantly, several groups have reported reduced VWF sialylation in patients with type 1 VWD.10,33,40,42 Together, these data suggest that quantitative sialylation plays a critical role in regulating both physiological and pathological clearance of VWF in vivo. Grewal et al. originally described a role for the ASGPR in regulating enhanced clearance of desialylated VWF (particularly in the context of sepsis).24 More recently, we identified MGL as another receptor involved in regulating hyposialylated VWF clearance.21 Critically, however, important questions remain unanswered regarding the roles played by VWF sialylation in regulating physiological and/or pathological clearance. These include: (i) the relative importance of N- versus O-linked sialylation in regulating VWF clearance; (ii) the relative contributions of the ASGPR and MGL clearance receptors; and (iii) the haematologica | 2022; 107(3)

molecular mechanisms through which hyposialylated VWF interacts with its clearance receptors.

Methods A detailed description of the materials and methods can be found in the associated Online Supplementary Material.

Isolation and purification of human plasma-derived von Willebrand factor Plasma-derived VWF (pdVWF) was purified from the VWFcontaining concentrate Fandhi® (Grifols, Barcelona, Spain) as previously described.21 Platelet-VWF was purified from lysed platelets as described elsewhere.43 Eluate fractions were then assessed for VWF antigen, multimer distribution, and purity.

Glycosidase digestion and quantitative analysis of glycan expression To generate VWF glycoforms, pdVWF was treated with α2-3 neuraminidase, α2-3,6,8,9 neuraminidase, β1-3 galactosidase, peptide N-glycosidase F (PNGase F) and/or O-glycosidase under non-denaturing conditions overnight at 37°C.29,44 Following glycosidase digestion, changes in VWF glycans were assessed using specific lectin enzyme-linked immunosorbent assays (ELISA) as previously described.10

Expression and purification of recombinant von Willebrand factor variants The expression vectors pcDNA-VWF encoding full length recombinant VWF, VWF-A1A2A3, VWF-A1, VWF-A2, VWF-A3, VWF-D’A3 or VWF-A3-CK fragments have previously been described.16 Additional VWF-A1 constructs containing either of the two O-linked glycan (OLG) clusters were also included; A1OLG cluster 1 (T1248A, T1255A, T1256A, S1263A), and A1-OLG cluster 2 (T1468A, T1477A, S1486A, T1487A). All recombinant VWF variants were transiently expressed in HEK293T cells. Conditioned serum-free medium was harvested 72 h after transfection and concentrated via anion exchange chromatography as described before.16

In vitro von Willebrand factor binding studies Solid phase plate-binding assays were used to evaluate VWF binding to MGL. Briefly, recombinant human MGL (Stratech, UK) was immobilized on a PolySorp™ 96-well plate (Nunc, Thermo Scientific™), the wells were blocked, and VWF was incubated at 37°C for 1 h. Bound VWF was detected using horseradish peroxidase (HRP)-conjugated polyclonal anti-VWF (Dako, Agilent Technologies), high-sensitivity streptavidin-HRP (ThermoScientific, UK) or anti-His-HRP antibody (Qiagen, UK) (see Online Supplementary Material for details).

von Willebrand factor clearance studies in MGL1-/-, VWF-/- and VWF-/-/Asgr1-/- mice All clearance experiments were performed on mice 6 to 8 weeks old. All animal studies were approved by the Health Product Regulatory Authority, Ireland and an internal ethics committee. VWF-/- and Asgr1-/- mice, both on a C57BL/6J background, were obtained from the Jackson Laboratory (Sacremento, CA, USA) and crossbred to obtain a dual VWF-/-/Asgr1-/- knockout model as previously described.21 MGL specific clearance studies were also performed after inhibition of murine MGL1/2 using a commercial polyclonal goat anti-mouse MGL1/2 antibody (2 mg/kg) (R&D Systems, UK) as previously described.21 For endogenous clearance studies, murine VWF was labeled with N-hydrox-

669

S.E. Ward et al.

ysuccinimide-biotin (10 mg/kg; Thermo-Scientific), residual biotinylated murine VWF was quantified using a modified VWF ELISA. All clearance data were fitted to monoexponential equations, based on analysis of the Akaike information criterion. The slope and intercept of the equation of the line were used to calculate pharmacokinetic parameters including mean residence time (MRT) and half-life (t ). 1/2

Data presentation and statistical analysis Experimental data were analyzed with GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA). Data were expressed as mean values ± standard error of the mean. Data were analyzed with the Student unpaired two-tailed t-test and P-values <0.05 were considered to be statistically significant.

Results Physiological importance of MGL in regulating von Willebrand factor clearance in vivo Mice have two distinct MGL homologs - murine MGL1 (mMGL1) and murine MGL2 (mMGL2).45 To gain insight into the biological importance of MGL, we first investigated murine MGL1 and MGL2 binding to VWF in vitro.

670

Similar to human MGL, dose-dependent binding of both mMGL1 and mMGL2 to VWF was observed (Figure 1A). VWF binding to MGL2 was significantly greater than to mMGL1 (P<0.001). Murine plasma VWF:Ag levels are significantly elevated (~ 1.5 fold) in MGL1-/- mice. Since both mMGL1 and mMGL2 bind VWF, we hypothesized that knocking down mMGL1 alone may underestimate the biological importance of MGL-mediated clearance. Dual mMGL1-/-/mMGL2-/- mice are not commercially available; thus, to address this hypothesis, in vivo clearance studies were repeated in mMGL1-/- mice in the presence or absence of dual anti-MGL1/2 inhibitory antibodies. Following treatment with anti-MGL2, murine VWF:Ag levels were significantly increased compared to those in mMGL1-/- controls (2.78±04 U/mL vs. 1.5±0.5 U/mL respectively; P<0.05) (Figure 1B). Thus, complete murine MGL inhibition was associated with an almost 3-fold increase in endogenous plasma VWF:Ag levels compared to the levels in wild-type (mMGL1+/+mMGL2+/+) controls. In the presence of combined mMGL1 and mMGL2 inhibition, endogenous VWF clearance was significantly attenuated compared to that in controls (P<0.05) (Figure 1C) and murine VWF MRT was increased 2.4-fold (Figure 1D). These data confirm that the observed increase in murine VWF levels associated with Figure 1. Physiological importance of MGL in regulating von Willebrand factor clearance. (A) In vitro binding of purified human plasma-derived (pd) von Willebrand factor (VWF) to murine MGL1 and MGL2 (mMGL1 and mMGL2) receptors was assessed using a plate binding assay as detailed in the ‘Methods’ section. (B) Plasma VWF levels were measured using a VWF:antigen (Ag) enzyme-linked immunosorbent assay (ELISA) in wildtype mice, MGL1-/- mice, and MGL1-/mice 24 h following infusion of antiMGL2 antibody. (C) NHS-biotin (10 mg/kg) was infused at t=0 h. Subsequently, residual biotinylated VWF clearance was quantified by a modified VWF ELISA. Clearance experiments were performed in MGL1-/- mice in the presence or absence of antiMGL1/2 antibody. (D) Mean residence time (MRT) for endogenous murine VWF was determined for wild-type mice, MGL1-/- mice and MGL1-/- mice following infusion of anti-MGL2 antibody. Three to five mice were studied per time point, and data are represented as mean ± standard error of mean. *P<0.05, **P<0.01.

haematologica | 2022; 107(3)

O-linked glycans protect VWF

inhibition of MGL-mediated clearance is attributable to an increase in VWF half-life. Importantly the magnitude of this effect of MGL on plasma VWF levels is also greater than that previously reported in Asgr1-/- mice (VWF:Ag levels increased ~1.5 fold).24 Collectively, these data demonstrate that mMGL2 constitutes another novel macrophage clearance receptor for VWF in mice. More importantly, the findings further suggest that MGL has a greater effect than ASGPR in regulating physiological VWF clearance.

The A domains of von Willebrand factor play a critical role in regulating MGL binding Previous studies have demonstrated that macrophages play a key role in VWF clearance, and have further shown that macrophage receptor-recognition site(s) are present within the VWF-A1A2A3 domains.13,16 To assess the specific role of VWF domains in modulating MGL binding a series of variants were expressed (Figure 2A) (Online Supplementary Table S1). Since MGL is expressed on macrophages, we investigated whether the A domains of VWF influence MGL binding. Dose-dependent binding of pdVWF to recombinant human MGL was observed (Figure 2B). In keeping with the fact that MGL is a C-type lectin, this binding was ablated in the presence of EDTA (Figure 2B). Conversely, VWF-MGL binding was significantly enhanced in the presence of ristocetin (1 mg/mL) (Figure 2B). No significant effect of VWF multimer distribution on MGL-binding was observed (Online Supplementary Figure S1). Binding studies confirmed dose-dependent binding of VWF-A1A2A3 to MGL, which was again ristocetin- and calcium-dependent (Figure 2C). Finally, the relative importance of the individual domains within A1A2A3 in determining MGL binding was assessed. Although significant binding of the VWF-A1 domain to MGL was seen, no binding for either VWF-A2 or VWF-A3 was observed (Figure 2D). Together, these data support the hypothesis that the A1 domain of VWF plays a critical role in determining VWF binding to the MGL surface receptor on macrophages.

O-linked glycans on von Willebrand factor modulate the MGL interaction Each VWF monomer contains 13 N-linked and ten Olinked glycan structures (Figure 3A).34,35 Whereas N-glycans are distributed across the VWF monomer, eight of the ten O-glycans are clustered in two groups around the VWF-A1 domain.35,36 To study the importance of specific N- and Oglycans in regulating MGL binding, pdVWF was treated with PNGase F and/or O-glycosidase respectively. Following each digestion, residual VWF glycan expression was assessed using lectin-binding ELISA (Online Supplementary Figure S2A, B). Although treatment with PNGase F successfully removed N-linked glycans from fulllength pdVWF, it did not have a significant effect on MGL binding (Figure 3B). Conversely, digestion with O-glycosidase was associated with a marked reduction in pdVWF binding to MGL (P<0.001) (Figure 3B). The VWF-A1A2A3 truncation contains two N-linked glycans and eight Olinked glycans (Figure 3A). In keeping with full-length VWF, PNGase treatment of A1A2A3-VWF (Online Supplementary Figure S2B, C) had no impact on the MGL interaction, whereas removal of O-glycans markedly attenuated binding (P<0.01) (Figure 3C). Digestion of isolated A1 domain with O-glycosidase (Online Supplementary Figure S2E) was also associated with a significant reduction haematologica | 2022; 107(3)

in MGL binding (P<0.001) (Figure 3D). Finally, to investigate the relative importance of the two OLG clusters at either side of the A1 domain, isolated A1-OLG cluster 1 and A1-OLG cluster 2 were expressed (Figure 2A). Although MGL-binding was observed for both of these VWF A1 domain truncations, significantly enhanced binding was observed for A1-OLG cluster 2 (Figure 3E). Cumulatively, these findings demonstrate that VWF OLG, particularly those clustered either side of the A1 domain, play a major role in regulating MGL interaction.

α2-3 sialylation on O-glycans protects von Willebrand factor against MGL-mediated clearance Recent mass spectrometry studies have characterized the O-glycan structures expressed on human pdVWF and highlighted significant heterogeneity (Figure 4A).35-37 Critically, however, a consistent feature of these O-glycan chains is that they generally terminate with sialic acid, which may be present in either α2-3 or α2-6 linkage.35,36 In contrast, most N-linked sialic acids are α2-6 linked.34 To further investigate the role of VWF O-glycans in determining MGL-mediated clearance, pdVWF was digested with a series of exoglycosidases to generate specific VWF glycoforms (Figure 4A). Treatment with α2-3 neuraminidase to remove α2-3 linked sialylation from O-glycans (Online Supplementary Figure S2F) significantly enhanced pdVWF binding to MGL (P=0.017) (Figure 4B). Similarly, digestion with α2-3,6,8,9 neuraminidase (which removes α2-3 linked sialylation from O-glycans and α2-6 linked sialylation from both N- and O-glycans) (Online Supplementary Figure S2G) was also associated with significantly increased MGL binding (P=0.006). Despite the fact that an estimated 80% of total sialylation on VWF is α2-6 linked, α2-3,6,8,9 Neu-VWF binding to MGL was not different to that observed following α2-3 neuraminidase digestion alone (Figure 4B). Significantly enhanced binding was observed for PNG-VWF following additional removal of α2-3 linked sialylation and exposure of the O-linked T antigen structure (Figure 4C). Finally, PNG-VWF was sequentially treated with α2-3 neuraminidase and β1-3 galactosidase to remove both terminal sialic acid and subterminal galactose (Gal) residues from VWF O-glycan chains (Online Supplementary Figure S2H). This combined digestion ablated the enhanced binding observed following α2-3 neuraminidase digestion alone (Figure 4C). These data demonstrate that α2-3 linked sialylation on VWF Oglycans specifically protects VWF against MGL-mediated clearance. Loss of this capping sialic acid results in Gal residue exposure on VWF O-glycans, which then triggers clearance through the MGL receptor. In order to consider whether other VWF domains/glycans may contribute to MGL-interaction, we compared binding for N-terminal D’A3-VWF and C-terminal A3-CK-VWF fragments. In keeping with a key role for the A1 domain, significant binding of D’A3-VWF to MGL was observed (Figure 4D). Interestingly, however, some A3-CK-VWF binding was also seen, suggesting that O-glycans (T1679 and/or T2298) downstream of the A1 domain may also play a role.

Role of MGL and ASGPR in modulating pathological enhanced clearance of desialylated von Willebrand factor Previous studies have demonstrated altered VWF sialylation in patients with VWD as well as in a number of other conditions.41 To investigate the role of MGL in mediating the enhanced clearance of pathologically desia671

S.E. Ward et al.

lylated VWF, pdVWF was treated ex vivo with α2-3 neuraminidase to remove α2-3 linked sialylation from O-glycans. In vivo clearance studies were then performed in VWF-/- mice in the presence or absence of combined mMGL1 and mMGL2 inhibition. Removal of α2-3 linked sialylation was associated with a marked reduction in VWF half-life compared to that of the wild-type control (Figure 5A). Importantly, however, this enhanced clearance was attenuated in the presence of MGL inhibition (Figure 5A). To assess the relative roles of MGL and ASGPR in modulating the pathological, enhanced clear-

ance following removal of α2-3 sialylation, in vivo clearance studies were also performed in dual VWF-/-Asgr1-/knockout mice in the presence or absence of combined mMGL1 and mMGL2 inhibition (Figure 5B). Critically, we observed that MGL inhibition was also able to block enhanced clearance of pdVWF after loss of α2-3 sialylation equally effectively in the presence or absence of ASGPR (Figure 5B). Terminal sialylation on VWF O-glycans can be either α2-3 or α2-6 linked. In contrast, sialylation on VWF Nglycan chains is predominantly α2-6 linked (Figure

D Figure 2. The A domains of von Willebrand factor play a critical role in regulating MGL binding. (A) Schematic of von Willebrand factor (VWF) variants used to characterize the VWF-MGL interaction. All VWF variants were expressed in and purified from HEK293T cells. (B, C) In vitro binding of purified human plasma-derived (pd)VWF (B) and truncated A1A2A3-VWF (C) was assessed using plate binding assays in the presence or absence of 10 mM EDTA or 1 mg/mL ristocetin. (D) Binding to human MGL was assessed for individual A domain proteins (A1-VWF, A2VWF and A3-VWF). Significant binding was observed for the A1-VWF domain compared with A2-VWF and A3-VWF. Bovine serum albumin (BSA) was used as a negative control. All data are presented as mean ± standard error of mean of three independent experiments. Percentage binding was calculated based on optical density at 450 nm obtained for 100 nM A1A2A3-VWF. *P<0.05, **P<0.01, ***P<0.001.

672

haematologica | 2022; 107(3)

O-linked glycans protect VWF

Figure 3. O-linked glycans on von Willebrand factor modulate the interaction with MGL. (A) Each von Willebrand factor (VWF) monomer contains 13 N-linked and ten O-linked glycan structures. The diagram also shows the most common VWF N-linked carbohydrate structure (a monosialylated, biantennary, core fucosylated complex glycan) and O-linked carbohydrate structure (core 1 sialylated T-antigen). (B) To investigate the role of VWF carbohydrate determinants in modulating the interaction with MGL, plasma-derived (pd)VWF (10 mg/mL) was treated with either PNGase F (PNGase VWF) to remove N-glycans or PNGase F and O glycosidase (PNGaseOGly VWF) to remove both N- and O-glycans. Binding of the pdVWF glycoforms to human MGL was then compared to binding to untreated pd-VWF as before (100% binding = OD450 obtained for 10 mg/mL pdVWF). (C) To study a potential role for glycans in the A domains of VWF in regulating MGL binding, A1A2A3-VWF (150 nM VWF) was treated with either PNGase F or O-glycosidase. Binding to human MGL was then assessed compared to to the binding to WT A1A2A3-VWF (100% binding = OD450 obtained for 150 nM A1A2A3-VWF). (D) Since A1-VWF does not contain any N-linked glycan determinants, MGL binding studies were examined for WT-A1VWF compared to O-glycosidase-treated VWF-A1 (100% binding = OD450 obtained for 150 nM A1-VWF). (E) Eight O-linked glycans are located in two clusters of four either side of the VWF A1 domain. To investigate the importance of these O-glycans in modulating the MGL interaction, two A1-VWF variants were generated each of which contained only one O-glycan cluster (A1-OLG cluster 1 contained T1248A, T1255A, T1256A, S1263A, while A1-OLG cluster 2 contained T1468A, T1477A, S1486A, T1487A). MGL-binding studies were compared for these two cluster variants as previously described (100% binding = OD450 obtained for A1-OLG cluster 1). All data are represented as mean ± standard error of mean of three independent experiments. *P<0.05, **P<0.01, ***P<0.001. PNGase F: peptide N-glycosidase F; OLG: O-linked glycans; OD450: absorbance at 450 nm.

haematologica | 2022; 107(3)

673

S.E. Ward et al.

4A).34,35 Since sepsis-related neuraminidases may target both the N- and O-glycans of VWF, we further investigated the role of MGL in clearing VWF from which both the N- and O-sialylation had been removed following digestion with α2-3,6,8,9 neuraminidase. In vivo clearance studies in VWF-/- mice demonstrated that combined mMGL1 and mMGL2 inhibition was not able to significantly reduce the pathological, enhanced clearance observed following loss of N-linked sialylation (Figure 6A). Interestingly, however, in mice deficient for the ASGPR clearance receptor, mMGL1/2 inhibition was associated with attenuation of the enhanced clearance of

α2-3,6,8,9 Neu-VWF (Figure 6B). Collectively, these findings further support the hypothesis that O-linked α2-3 sialylation on VWF plays a critical role in protecting against MGL-mediated clearance. Moreover, the data also suggest that loss of α2-6 sialylation (predominantly Nlinked) on VWF drives enhanced clearance in a predominantly MGL-independent manner, mediated through the ASGPR. Increased VWF clearance plays a key role in the pathogenesis of both type 1 and type 2B VWD.3,11,46 Previous studies have implicated macrophages, and in particular the LRP1 and SR-A1 receptors, in regulating this enanced

674

Figure 4. α2-3 sialylation on O-glycans protects von Willebrand factor against MGLmediated clearance. (A, B) To investigate the role of terminal sialylation in modulating the von Willebrand factor (VWF) interaction with MGL, plasma-derived (pd)VWF (10 mg/mL) was treated with either α2-3 neuraminidase (α2-3 Neu) to remove α2-3 linked sialylation from O-glycans, or α2-3,6,8,9 neuraminidase (α2-3,6,8,9 Neu) to remove N- and O-sialylation. Binding of the treated pdVWF sialo-glycoforms to human MGL was then compared to the binding to untreated pdVWF. (C) In order to specifically focus on O-linked sialylation, pdVWF (10 mg/mL) was first digested with PNGase F to remove N-glycans and then sequentially treated with α2-3 neuraminidase (PNGase, α2-3 Neu VWF) ± β1-3 galactosidase neuraminidase (PNGase, α2-3 Neu, β1-3 Gal VWF). MGL binding was then assessed for each of the VWF O-glycan variants compared to the binding of untreated pdVWF (100% binding = OD450 obtained for 10 mg/mL pdVWF). (D) To study whether other regions of VWF influence the interaction with MGL, binding studies were compared for N-terminal D’A3-VWF compared to C-terminal A3-CK-VWF fragments. All binding experiments were performed in the presence of 1 mg/mL ristocetin. All data are shown as mean ± standard error of mean of three independent experiments. ns: not significant, *P<0.05, **P<0.01, ***P<0.001. PNGase F: peptide N-glycosidase F; OD450: absorbance at 450 nm.

haematologica | 2022; 107(3)

O-linked glycans protect VWF

clearance.14,15,19,20 To examine whether MGL may also play a role, we investigated binding for a number of type 1C (VWF-R1205H, R1205C, R1205S, S2179F) and type 2B (VWF-V1316M and -R1450E) variants. No evidence of enhanced MGL binding was observed for VWF-V1316M or any of the type 1C variants (Online Supplementary Figure S3). Interestingly, significantly reduced MGL binding was seen for VWF-R1450E compared to wild-type recombinant VWF. We hypothesise that this change in binding is due to conformational effects within the A1 domain affecting O-linked glycosylation during posttranslational modification and/or accessibility of specific OLG for the MGL interaction.

Platelet-von Willebrand factor sialylation and MGL interaction Platelet α-granules contain approximatey 20% of the total VWF present in platelet-rich plasma.47,48 Previous studies have demonstrated that platelet-derived (plt)VWF has altered glycosylation compared to pdVWF.48 In particular, plt-VWF does not express ABO blood group determinants and is hypo-sialylated.49,50 Importantly, these glycosylation differences influence susceptibility to ADAMTS-13 cleavage.43 Using lectin-binding ELISA, we confirmed that the quantitative reduction in plt-VWF sialylation was predominantly attributable to a specific reduction in N-linked sialylation (Figure 7A, B). As a result of this decreased N-sialylation, terminal galactose

expression was significantly increased on plt-VWF compared to pd-VWF (Figure 7C). Critically, despite the significant reduction in N-linked sialylation, we observed no increase in MGL binding for plt-VWF (Figure 7D). Moreover, in vivo clearance of plt-VWF in VWF-/- mice was similar to that of pd-VWF (Figure 7E). Cumulatively, these novel data further support our hypothesis that Olinked sialylation on VWF plays a key role in protecting VWF against MGL-mediated clearance.

Discussion Recent studies have demonstrated that complex glycan structures, which account for 20% of total VWF monomeric mass, play a key role in regulating the halflife of VWF in vivo.3,25,51 In addition, a number of lectin receptors have been shown to bind VWF.3 Critically, however, the relative importance of these receptors in modulating physiological and pathological VWF clearance has not been defined. Moreover, the particular VWF glycan determinants involved in modulating interactions with specific lectin receptors remain unclear. In this study, using a series of in vivo and in vitro methodologies, we demonstrated that both murine homologs of the MGL receptor bind to VWF and contribute to the physiological clearance of endogenous murine VWF. Consequently, combined inhibition of both mMGL1 and

Figure 5. Role of MGL in modulating pathological enhanced clearance of α2-3 Neu-VWF. (A) To investigate the importance of MGL in regulating the enhanced clearance of von Willebrand factor (VWF) with reduced O-linked sialylation, purified human plasma-derived (pd)VWF was treated with α2-3 neuraminidase (α2-3 Neu-VWF). In vivo clearance was then assessed in VWF-/- mice for α2-3 Neu-VWF in the presence or absence of combined mMGL1 and mMGL2 inhibition and compared to that of wild type pdVWF. At each time point, residual circulating VWF concentration was determined by an enzyme-linked immunosorbent assay for VWF:antigen (VWF:Ag). All results are plotted as percentage residual VWF:Ag levels relative to the amount injected. Data are represented as mean ± standard error of mean (SEM). In some cases, the SEM cannot be seen because of its small size. (B) To assess the relative roles of MGL and ASGPR in modulating the pathological increased clearance following removal of α2-3 sialylation, in vivo clearance studies were also performed in dual VWF-/-Asgr1-/- knockout mice in the presence or absence of combined mMGL1 and mMGL2 inhibition.

haematologica | 2022; 107(3)

675

S.E. Ward et al.

mMGL2 resulted in a 3-fold increase in murine plasma VWF levels which was attributable to a significant decrease in clearance rate. Importantly, the magnitude of the increased in vivo VWF levels associated with combined MGL inhibition was greater than that reported following inhibition of other VWF clearance receptors in mice (~2.5 fold vs. ~1.5 fold), suggesting that MGL plays an important role in regulating physiological clearance of VWF. To further investigate how MGL interacts with VWF, we first investigated the roles of specific VWF domains. Our data demonstrate that the A1A2A3 domains of VWF are predominantly responsible for modulating MGL binding. Furthermore, studies using isolated A domains showed that the A1 domain plays a critical role in regulating the MGL interaction. Interestingly, the binding of both full-length and A1A2A3-VWF to MGL was markedly enhanced in the presence of ristocetin, suggesting that the MGL-binding site in A1 may not be fully accessible in normal globular VWF. This concept is in keeping with findings of previous studies that showed significantly increased VWF binding to macrophages in the presence of ristocetin, botrecetin or shear stress.14,16 From a biological perspective, these data suggest that any VWF circulating in an ‘active’ GpIb binding conformation will be cleared rapidly by macrophage MGL, which may be important in minimizing thrombotic risk. Importantly, our data further show that C-terminal A3-CK-VWF also

binds MGL. Although the binding was less than that observed with N-terminal D’A3-VWF, this observation suggests that additional MGL-recognition sites beyond the A1 domain may contribute to the MGL interaction. Mass spectrometry studies have demonstrated significant and site-specific heterogeneity in the carbohydrate structures expressed on human pdVWF.34-37 Nevertheless, the majority of both the N- and O-linked glycans are capped with negatively-charged sialic acid residues. In this study, we demonstrated that specific loss of α2-3 linked sialylation from the O-linked glycans of VWF causes enhanced MGL binding in vitro, and causes markedly enhanced MGL-mediated clearance in vivo. In contrast, removal of α2-6 linked sialylation, which constitutes most of the total sialic acid expressed on human VWF and, in particular, the vast majority of the sialylation on N-glycans, has minimal effect on MGL binding and/or clearance. Our data further suggest that the two O-linked glycan clusters located either side of the A1 domain play a key role in regulating binding to MGL. Previous studies have demonstrated that these O-glycan clusters have significant effects on local VWF conformation.52,53 Further studies will be required to determine the molecular mechanisms through which these specific Oglycans regulate MGL-mediated VWF binding and clearance. Nevertheless, our findings demonstrate that MGL contributes to physiological VWF clearance by binding to exposed Gal residues on O-linked carbohydrate strucFigure 6. ASGPR in combination with MGL modulates the increased clearance α2-3,6,8,9 Neu-VWF. (A) To investigate whether MGL plays a role in the enhanced clearance of von Willebrand factor (VWF) from which both the N- and O-sialylation had been removed purified human plasmderived (pd)VWF was treated with α2-3,6,8,9 neuraminidase (α2-3,6,8,9 Neu-VWF). In vivo clearance of α23,6,8,9 Neu-VWF was then assessed in VWF-/- mice in the presence or absence of combined mMGL1 and mMGL2 inhibition and compared to that of wild-type pdVWF. At each time point, residual circulating VWF concentration was determined by an enzyme-linked immunosorbent assay for VWF:antigen (VWF:Ag). All results are plotted as percentage residual VWF:Ag levels relative to the amount injected. Data are represented as mean ± standard error of mean (SEM). In some cases, the SEM cannot be seen because of its small size. (B) To assess the relative roles of MGL and ASGPR in modulating the pathological, increased clearance following removal of α2-3,6,8,9 sialylation, in vivo clearance studies were also performed in dual VWF-/Asgr1-/- knockout mice in the presence or absence of combined mMGL1 and mMGL2 inhibition. *P<0.05, **P<0.01, ns: not significant.

676

haematologica | 2022; 107(3)

O-linked glycans protect VWF

tures. Importantly, glycoprotein aging in plasma is associated with progressive loss of capping sialic acid, and thus increased exposure of these sub-terminal Gal residues.38 There are previous reports of significantly increased binding of RCA-I lectin to plasma VWF in patients with VWD.10,33,40,42 This lectin binds preferentially to Gal or GalNAc sugars which are typically present as sub-terminal residues on the O- and N-glycans of pdVWF, but become exposed following loss of capping sialic acid. Increased RCA-I binding has also been correlated with enhanced VWF clearance in VWD patients.10,33,40 Our data suggest that the shortened half-life associated with increased Gal exposure (and hence RCA-I binding) in VWD patients is mediated in large part through enhanced MGL-mediated clearance. Importantly, van Schooten et al. previously reported significantly increased binding of peanut agglutinin (PNA) lectin to VWF in a cohort of VWD patients.40 This lectin preferentially binds to the T antigen structure which is exposed following loss of Olinked sialylation. The authors further showed that increased PNA-binding (T antigen exposure) was associated with a significant increase in the VWFpp/VWF:Ag ratio, consistent with enhanced VWF clearance.40 In keeping with these results, we have demonstrated that α2-3

linked sialylation on O-linked glycan structures plays a particular role in protecting VWF against MGL-mediated clearance. Consequently, our findings suggest that the enhanced clearance associated with T antigen exposure on VWF previously reported by van Schooten et al. is attributable to enhanced clearance via MGL. Besides VWD, abnormal VWF glycosylation has also been reported in a number of other disease states.24,39-41 For example, reduced PNA-binding to VWF has been reported in patients with liver cirrhosis who have significantly elevated plasma VWF:Ag levels. The biological mechanisms underlying reduced T antigen exposure on VWF in patients with cirrhosis have not been defined. Nonetheless, our findings build upon these previous observations and in particular suggest that the altered Oglycosylation associated with cirrhosis will cause increased plasma VWF levels as a result of decreased MGL-mediated clearance. Conversely, a number of different pathogens including Streptococcus pneumoniae, Haemophilus influenzae and Pseudomonas aeruginosa express neuraminidase enzymes that can cause desialylation of host glycoproteins.24,54 VWF desialylation associated with pathological, enhanced clearance has been observed in mice infected with S. pneumoniae.24 Our data further suggest that increased MGL-mediated clearance

Figure 7. α2-3 linked sialic acid on platelet von Willebrand factor protects from enhanced circulatory clearance. (A-C) Platelet-derived (plt) von Willebrand factor (VWF) sialylation was assessed using lectin binding assays with Sambucus nigra (A), Maackia amuresis (B) and Ricinus communis (C). Plasma-derived (pd) VWF was used as a control. (D) Solid phase binding assay was used to assess the binding of plt-VWF to immobilized human MGL and again compared to human pd-VWF. (E) In vivo pharmacokinetic experiments were performed in VWF-/- mice to compare the clearance rates or plt-VWF compared to pd-VWF. At each time point, residual circulating VWF:antigen (VWF:Ag) concentration was determined by enzyme-linked immunosorbent assay. All results are plotted as percentage residual VWF:Ag levels relative to the amount injected. Data are represented as mean ± standard error of mean (SEM). In some cases, the SEM cannot be seen because of its small size. *P<0.05, **P<0.01, ns: not significant.

haematologica | 2022; 107(3)

677

S.E. Ward et al.

will play a key role in mediating this pathogen-associated enhanced VWF clearance. Interestingly, two previous studies have demonstrated that complete loss of Olinked carbohydrate structures is associated with significantly increased VWF clearance in vivo.28,55 Given that Oglycans are known to influence protein conformation, the observation that complete removal triggers enhanced clearance is likely attributable to conformational changes in VWF. In addition to MGL, other macrophage receptors that can also interact with VWF include LRP1, SR-A1, Siglec5, Gal-1 and Gal-3.3,56,57 Some of these receptors have also been shown to bind with enhanced affinity to hyposialylated VWF (ASGPR, Gal-1 and Gal-3). Additional studies will be necessary to fully elucidate the relative roles of these other macrophage receptors in regulating the physiological and/or pathological clearance of hyposialylated VWF. Although it remains unclear whether these receptors may function synergistically in regulating desialylated VWF clearance, recent studies have demonstrated that LRP1 can form heterologous functional complexes with other macrophage receptors including β2-integrins. Importantly, Deppermann et al recently demonstrated that MGL on hepatic Küpffer cells plays a significant role in the removal of desialylated platelets, and that MGL and ASGPR appear to function collaboratively in physiological platelet clearance.58

References 1. Lenting PJ, Christophe OD, Denis CV. von Willebrand factor biosynthesis, secretion, and clearance: connecting the far ends. Blood. 2015;125(13):2019-2028. 2. Leebeek FWG, Eikenboom JCJ. von Willebrand's disease. N Engl J Med. 2017; 376(7):701-702. 3. O'Sullivan JM, Ward S, Lavin M, O'Donnell JS. von Willebrand factor clearance - biological mechanisms and clinical significance. Br J Haematol. 2018;183(2):185-195. 4. Castaman G, Lethagen S, Federici AB, et al. Response to desmopressin is influenced by the genotype and phenotype in type 1 von Willebrand disease (VWD): results from the European Study MCMDM-1VWD. Blood. 2008;111(7):3531-3539. 5. Flood VH, Christopherson PA, Gill JC, et al. Clinical and laboratory variability in a cohort of patients diagnosed with type 1 VWD in the United States. Blood. 2016;127(20):2481-2488. 6. Eikenboom J, Federici AB, Dirven RJ, et al. VWF propeptide and ratios between VWF, VWF propeptide, and FVIII in the characterization of type 1 von Willebrand disease. Blood. 2013;121(12):2336-2339. 7. Sanders YV, Groeneveld D, Meijer K, et al. von Willebrand factor propeptide and the phenotypic classification of von Willebrand disease. Blood. 2015;125(19):3006-3013. 8. Haberichter SL, Castaman G, Budde U, et al. Identification of type 1 von Willebrand disease patients with reduced von Willebrand factor survival by assay of the VWF propeptide in the European study: molecular and clinical markers for the diagnosis and management of type 1 VWD (MCMDM1VWD). Blood. 2008;111(10):4979-4985. 9. Lavin M, Aguila S, Schneppenheim S, et al. Novel insights into the clinical phenotype

678

Disclosures JSO’D has served on speakers’ bureaux for Baxter, Bayer, Novo Nordisk, Boehringer Ingelheim, Leo Pharma, Takeda and Octapharma. He has also served on advisory boards for Baxter, Bayer, Octapharma CSL Behring, Daiichi Sankyo, Boehringer Ingelheim, Takeda and Pfizer and has received research grant funding awards from Baxter, Bayer, Pfizer, Shire, Takeda and Novo Nordisk. JMO’S has received research grant funding from LEO Pharma and Grifols Contributions SEW, ABM, JF and AC performed experiments; SEW, JMO’S, ABM, DS, RG, JS, JF, MM, TAJM, AC, SH and JSO’D designed the research and analyzed the data. All authors were involved in writing and reviewing the paper. Funding This work was supported by funds from the NIH for the Zimmerman Program (HL081588); a Science Foundation Ireland Principal Investigator Award (11/PI/1066); a Health Research Board Investigator Lead Project Award (ILP-POR2017-008) and a National Children’s Research Centre Project Award (C/18/1). ABM is supported by the European Union (GlySign, grant n. 722095) Data-sharing statement All original data and protocols can be made available to other investigators upon request.

and pathophysiology underlying low VWF levels. Blood. 2017;130(21):2344-2353. 10. Aguila S, Lavin M, Dalton N, et al. Increased galactose expression and enhanced clearance in patients with low von Willebrand factor. Blood. 2019;133(14):1585-1596. 11. Casari C, Du V, Wu YP, et al. Accelerated uptake of VWF/platelet complexes in macrophages contributes to VWD type 2Bassociated thrombocytopenia. Blood. 2013;122(16):2893-2902. 12. O'Donnell JS. Low VWF: insights into pathogenesis, diagnosis, and clinical management. Blood Adv. 2020;4(13):3191-3199. 13. Lenting PJ, Westein E, Terraube V, et al. An experimental model to study the in vivo survival of von Willebrand factor. Basic aspects and application to the R1205H mutation. J Biol Chem. 2004;279(13):12102-12109. 14. van Schooten CJ, Shahbazi S, Groot E, et al. Macrophages contribute to the cellular uptake of von Willebrand factor and factor VIII in vivo. Blood. 2008;112(5):1704-1712. 15. Rawley O, O'Sullivan JM, Chion A, et al. von Willebrand factor arginine 1205 substitution results in accelerated macrophagedependent clearance in vivo. J Thromb Haemost. 2015;13(5):821-826. 16. Chion A, O'Sullivan JM, Drakeford C, et al. N-linked glycans within the A2 domain of von Willebrand factor modulate macrophage-mediated clearance. Blood. 2016;128(15):1959-1968. 17. Swystun LL, Lai JD, Notley C, et al. The endothelial cell receptor stabilin-2 regulates VWF-FVIII complex half-life and immunogenicity. J Clin Invest. 2018;128(9):40574073. 18. Pegon JN, Kurdi M, Casari C, et al. Factor VIII and von Willebrand factor are ligands for the carbohydrate-receptor Siglec-5. Haematologica. 2012;97(12):1855-1863. 19. Rastegarlari G, Pegon JN, Casari C, et al.

Macrophage LRP1 contributes to the clearance of von Willebrand factor. Blood. 2012;119(9):2126-2134. 20. Wohner N, Muczynski V, Mohamadi A, et al. Macrophage scavenger receptor SR-AI contributes to the clearance of von Willebrand factor. Haematologica. 2018;103 (4):728-737. 21. Ward SE, O'Sullivan JM, Drakeford C, et al. A novel role for the macrophage galactosetype lectin receptor in mediating von Willebrand factor clearance. Blood. 2018;131 (8):911-916. 22. Rydz N, Swystun LL, Notley C, et al. The Ctype lectin receptor CLEC4M binds, internalizes, and clears von Willebrand factor and contributes to the variation in plasma von Willebrand factor levels. Blood. 2013;121 (26):5228-5237. 23. Swystun LL, Ogiwara K, Lai JD, et al. The scavenger receptor SCARA5 is an endocytic receptor for von Willebrand factor expressed by littoral cells in the human spleen. J Thromb Haemost. 2019;17(8):1384-1396. 24. Grewal PK, Uchiyama S, Ditto D, et al. The Ashwell receptor mitigates the lethal coagulopathy of sepsis. Nat Med. 2008;14(6):648655. 25. Casari C, Lenting PJ, Wohner N, Christophe OD, Denis CV. Clearance of von Willebrand factor. J Thromb Haemost. 2013;11 Suppl 1:202-211. 26. Casonato A, Pontara E, Sartorello F, et al. Reduced von Willebrand factor survival in type Vicenza von Willebrand disease. Blood. 2002;99(1):180-184. 27. Sodetz JM, Pizzo SV, McKee PA. Relationship of sialic acid to function and in vivo survival of human factor VIII/von Willebrand factor protein. J Biol Chem. 1977;252(15):5538-5546. 28. Stoddart JH, Jr., Andersen J, Lynch DC. Clearance of normal and type 2A von

haematologica | 2022; 107(3)

O-linked glycans protect VWF

Willebrand factor in the rat. Blood. 1996;88(5):1692-1699. 29. O'Sullivan JM, Aguila S, McRae E, et al. Nlinked glycan truncation causes enhanced clearance of plasma-derived von Willebrand factor. J Thromb Haemost. 2016;14(12): 2446-2457. 30. Preston RJ, Rawley O, Gleeson EM, O'Donnell JS. Elucidating the role of carbohydrate determinants in regulating hemostasis: insights and opportunities. Blood. 2013;121(19):3801-3810. 31. Gallinaro L, Cattini MG, Sztukowska M, et al. A shorter von Willebrand factor survival in O blood group subjects explains how ABO determinants influence plasma von Willebrand factor. Blood. 2008;111(7):35403545. 32. Ward SE, O'Sullivan JM, O'Donnell JS. The relationship between ABO blood group, von Willebrand factor, and primary hemostasis. Blood. 2020;136(25):2864-2874. 33. Ellies LG, Ditto D, Levy GG, et al. Sialyltransferase ST3Gal-IV operates as a dominant modifier of hemostasis by concealing asialoglycoprotein receptor ligands. Proc Natl Acad Sci U S A. 2002;99(15): 10042-10047. 34. Canis K, McKinnon TA, Nowak A, et al. Mapping the N-glycome of human von Willebrand factor. Biochem J. 2012;447(2): 217-228. 35. Canis K, McKinnon TA, Nowak A, et al. The plasma von Willebrand factor O-glycome comprises a surprising variety of structures including ABH antigens and disialosyl motifs. J Thromb Haemost. 2010;8(1):137145. 36. Solecka BA, Weise C, Laffan MA, Kannicht C. Site-specific analysis of von Willebrand factor O-glycosylation. J Thromb Haemost. 2016;14(4):733-746. 37. Gashash EA, Aloor A, Li D, et al. An Insight into glyco-microheterogeneity of plasma von Willebrand factor by mass spectrometry. J Proteome Res. 2017;16(9):3348-3362. 38. Yang WH, Aziz PV, Heithoff DM, Mahan MJ, Smith JW, Marth JD. An intrinsic mechanism of secreted protein aging and

haematologica | 2022; 107(3)

turnover. Proc Natl Acad Sci U S A. 2015;112(44):13657-13662. 39. Lopes AA, Ferraz de Souza B, Maeda NY. Decreased sialic acid content of plasma von Willebrand factor in precapillary pulmonary hypertension. Thromb Haemost. 2000;83 (5):683-687. 40. van Schooten CJ, Denis CV, Lisman T, et al. Variations in glycosylation of von Willebrand factor with O-linked sialylated T antigen are associated with its plasma levels. Blood. 2007;109(6):2430-2437. 41. Ward S, O'Sullivan JM, O'Donnell JS. von Willebrand factor sialylation-A critical regulator of biological function. J Thromb Haemost. 2019;17(7):1018-1029. 42. Millar CM, Riddell AF, Brown SA, et al. Survival of von Willebrand factor released following DDAVP in a type 1 von Willebrand disease cohort: influence of glycosylation, proteolysis and gene mutations. Thromb Haemost. 2008;99(5):916-924. 43. McGrath RT, van den Biggelaar M, Byrne B, et al. Altered glycosylation of plateletderived von Willebrand factor confers resistance to ADAMTS13 proteolysis. Blood. 2013;122(25):4107-4110. 44. McGrath RT, McKinnon TA, Byrne B, et al. Expression of terminal alpha2-6-linked sialic acid on von Willebrand factor specifically enhances proteolysis by ADAMTS13. Blood. 2010;115(13):2666-2673. 45. Tsuiji M, Fujimori M, Ohashi Y, et al. Molecular cloning and characterization of a novel mouse macrophage C-type lectin, mMGL2, which has a distinct carbohydrate specificity from mMGL1. J Biol Chem. 2002;277(32):28892-28901. 46. Wohner N, Legendre P, Casari C, Christophe OD, Lenting PJ, Denis CV. Shear stress-independent binding of von Willebrand factortype 2B mutants p.R1306Q & p.V1316M to LRP1 explains their increased clearance. J Thromb Haemost. 2015;13(5):815-820. 47. Mannucci PM. Platelet von Willebrand factor in inherited and acquired bleeding disorders. Proc Natl Acad Sci U S A. 1995;92(7): 2428-2432. 48. McGrath RT, McRae E, Smith OP,

O'Donnell JS. Platelet von Willebrand factor--structure, function and biological importance. Br J Haematol. 2010;148 (6):834-843. 49. Williams SB, McKeown LP, Krutzsch H, Hansmann K, Gralnick HR. Purification and characterization of human platelet von Willebrand factor. Br J Haematol. 1994;88(3): 582-591. 50. Brown SA, Collins PW, Bowen DJ. Heterogeneous detection of A-antigen on von Willebrand factor derived from platelets, endothelial cells and plasma. Thromb Haemost. 2002;87(6):990-996. 51. Lenting PJ, Pegon JN, Christophe OD, Denis CV. Factor VIII and von Willebrand factor-too sweet for their own good. Haemophilia. 2010;16 Suppl 5:194-199. 52. Tischer A, Machha VR, Moon-Tasson L, Benson LM, Auton M. Glycosylation sterically inhibits platelet adhesion to von Willebrand factor without altering intrinsic conformational dynamics. J Thromb Haemost. 2020;18(1):79-90. 53. Deng W, Wang Y, Druzak SA, et al. A discontinuous autoinhibitory module masks the A1 domain of von Willebrand factor. J Thromb Haemost. 2017;15(9):1867-1877. 54. Soong G, Muir A, Gomez MI, et al. Bacterial neuraminidase facilitates mucosal infection by participating in biofilm production. J Clin Invest. 2006;116(8):2297-2305. 55. Badirou I, Kurdi M, Legendre P, et al. In vivo analysis of the role of O-glycosylations of von Willebrand factor. PLoS One. 2012;7(5): e37508. 56. Saint-Lu N, Oortwijn BD, Pegon JN, et al. Identification of galectin-1 and galectin-3 as novel partners for von Willebrand factor. Arterioscler Thromb Vasc Biol. 2012;32(4): 894-901. 57. O'Sullivan JM, Jenkins PV, Rawley O, et al. Galectin-1 and galectin-3 constitute novelbinding partners for factor VIII. Arterioscler Thromb Vasc Biol. 2016;36(5):855-863. 58. Deppermann C, Kratofil RM, Peiseler M, et al. Macrophage galactose lectin is critical for Kupffer cells to clear aged platelets. J Exp Med. 2020;217(4):e20190723.

679

ARTICLE Ferrata Storti Foundation

Myelodysplastic Syndromes

ZRSR1 co-operates with ZRSR2 in regulating splicing of U12-type introns in murine hematopoietic cells Vikas Madan,1* Zeya Cao,1,2* Weoi Woon Teoh,1 Pushkar Dakle,1 Lin Han,1,2 Pavithra Shyamsunder,1,3 Maya Jeitany,1,4 Siqin Zhou,1 Jia Li,1 Hazimah Binte Mohd Nordin,1 Jizhong Shi,1 Shuizhou Yu,1 Henry Yang,1 Md Zakir Hossain,1 Wee Joo Chng,1,2,5# and H. Phillip Koeffler1,6,7#

Haematologica 2022 Volume 107(3):680-689

Cancer Science Institute of Singapore, National University of Singapore, Singapore; Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore; 3Programme in Cancer and Stem Cell Biology, Duke–NUS Medical School, Singapore; 4School of Biological Sciences, Nanyang Technological University, Singapore; 5Hematology-Oncology, National University Cancer Institute, National University Hospital Singapore, Singapore; 6Cedars-Sinai Medical Center, Division of Hematology/Oncology, UCLA School of Medicine, Los Angeles, CA, USA and 7National University Cancer Institute, National University Hospital Singapore, Singapore. 1 2

*VM and ZC contributed equally as co-first authors. #

WJC and HPK contributed equally as co-senior authors.

ABSTRACT

R Correspondence: VIKAS MADAN vikasmadan@aol.com Received: May 24, 2020. Accepted: March 1, 2021. Pre-published: March 11, 2021. https://doi.org/10.3324/haematol.2020.260562

ecurrent loss-of-function mutations of spliceosome gene, ZRSR2, occur in myelodysplastic syndromes (MDS). Mutation/loss of ZRSR2 in human myeloid cells primarily causes impaired splicing of the U12-type introns. In order to further investigate the role of this splice factor in RNA splicing and hematopoietic development, we generated mice lacking ZRSR2. Unexpectedly, Zrsr2-deficient mice developed normal hematopoiesis with no abnormalities in myeloid differentiation evident in either young or ≥1-year old knockout mice. Repopulation ability of Zrsr2-deficient hematopoietic stem cells was also unaffected in both competitive and non-competitive reconstitution assays. Myeloid progenitors lacking ZRSR2 exhibited mis-splicing of U12-type introns, however, this phenotype was moderate compared to the ZRSR2-deficient human cells. Our investigations revealed that a closely related homolog, Zrsr1, expressed in the murine hematopoietic cells, but not in human cells contributes to splicing of U12-type introns. Depletion of Zrsr1 in Zrsr2 KO myeloid cells exacerbated retention of the U12-type introns, thus highlighting a collective role of ZRSR1 and ZRSR2 in murine U12-spliceosome. We also demonstrate that aberrant retention of U12-type introns of MAPK9 and MAPK14 leads to their reduced protein expression. Overall, our findings highlight that both ZRSR1 and ZRSR2 are functional components of the murine U12-spliceosome, and depletion of both proteins is required to accurately model ZRSR2-mutant MDS in mice.

680

Introduction Mutations in RNA splicing factors constitute the leading class of genetic alterations in myelodysplastic syndromes (MDS).1-4 Somatic mutations in spliceosome genes, SF3B1, SRSF2, U2AF1 and ZRSR2 are observed in >50% of MDS.2,3 These mutations are early events during disease development and largely occur mutually exclusive of each other.1-5 Intense efforts in the last few years have enhanced our understanding of the impact of spliceosome mutations on RNA splicing and defined a mis-splicing pattern for each mutation.6-17 Animal models expressing mutant hotspots of splice factor genes - SRSF2 (P95H), SF3B1 (K700E) and U2AF1 (S34F) have enabled elucidation of consequences of these genetic lesions on RNA splicing and hematopoietic development.6,7,12,15,16,18-20 Unlike mutational hotspots observed in

haematologica | 2022; 107(3)

ZRSR2 & ZRSR1 co-operate in splicing of U12 introns

SF3B1, SRSF2 and U2AF1, alterations of ZRSR2 are truncating mutations spread throughout the transcript. ZRSR2 is located on human chromosome X and somatic mutations are primarily observed in males, suggesting its lossof-function in MDS.1 It is involved in 3’ splice site recognition and interacts with the U2AF2/U2AF1 heterodimer and SRSF2 during pre-spliceosome assembly.21 ZRSR2 is recruited in an ATP-dependent fashion to the U12-type intron splice site, and is required for the formation of the spliceosome complex.22 We have previously illustrated that either truncating mutations of ZRSR2 in MDS or its silencing in acute myeloid leukemia (AML) cells impair predominantly splicing of U12-type introns, excision of which is mediated by minor spliceosome assembly.8 However, animal models of ZRSR2 deficiency have not been reported and an in-depth understanding of its function is lacking. In this study, we have generated the first mouse model of ZRSR2 deficiency to further uncover its function in splicing and to evaluate the effect of its loss on normal and malignant hematopoiesis. Although deletion of Zrsr2 induces aberrant splicing of U12-type introns in mouse myeloid precursors, myeloid differentiation is largely unaffected in young, as well as ≥1-year old Zrsr2 knockout (KO) male mice. ZRSR2-deficient hematopoietic stem cells (HSC) retain multilineage reconstitution ability, suggesting limited function of ZRSR2 in mouse hematopoiesis. We further investigated its closely related homolog, ZRSR1, and uncovered that it compensates for the loss of ZRSR2 in mouse hematopoietic cells, and concurrent deficiency of both proteins leads to more profound defects in splicing of U12-type introns. Hence, murine models lacking both ZRSR2 and ZRSR1 are necessary to replicate faithfully mis-splicing caused by lossof-function mutations of ZRSR2 in MDS.

Methods Generation of Zrsr2 knockout mice C57BL/6N embryonic stem (ES) cells (Zrsr2tm1(KOMP)Vlcg) with targeted deletion of mouse Zrsr2 were obtained from UC Davis KOMP Repository. Mice with a constitutive deletion of Zrsr2 were generated at the transgenic facility of CSI Singapore. Briefly, ES cells were microinjected into BALB/cJInv blastocysts and resulting chimera were mated with C57BL/6 mice. Black offspring were genotyped, and mice carrying the Zrsr2 null allele were used to establish a colony of Zrsr2 knockout (KO) mice. Subsequently, mice were crossed with CMV-Cre strain to remove the neomycin selection cassette. In this study, Zrsr2 KO mice refer to those before or after Cre-mediated excision, as they were phenotypically identical in all experiments. Primers used for genotyping are listed in the Online Supplementary Table S1. All animal experiments were approved by the Institutional Animal Care and Use Committee of the National University of Singapore, Singapore.

Flow cytometric analysis and fluorescence-activated cell sorting Single cell suspensions from bone marrow (BM), spleen and thymus were stained with fluorochrome-conjugated antibodies for 30 minutes. The lineage cocktail was comprised of antibodies targeting murine CD19, CD3ε, CD11b, Gr1 and TER119. Cells were washed with 2% fetal bovine serum in phosphate buffered saline and resuspended in SYTOX Blue Dead Cell Stain (ThermoFisher Scientific). Cells were acquired on FACS LSR II

haematologica | 2022; 107(3)

(BD Biosciences) and data were analyzed using FACSDIVA software (BD Biosciences). Cells were sorted on FACSAria cell sorter (BD Biosciences).

RNA sequencing Total RNA was extracted from sorted myeloid precursors (common myeloid precursors [CMP], granulocyte monocyte precursors [GMP] and megakaryocyte erythroid precursors [MEP]), murine embryonic fibroblasts [MEF] and ex vivo cultured Lin−Kit+ BM cells using either RNeasy Micro or Mini Kits (Qiagen). Libraries of polyA-selected RNA were prepared using either TruSeq sample preparation kit (CMP, GMP, MEP and MEF) or NEBNext Ultra RNA Library Prep Kit (Lin−Kit+ BM cells). Libraries were sequenced on HiSeq 4000, and paired-end reads were mapped to either mouse reference transcriptome (GRCm38/mm10; Ensemble version 84) or human hg38 reference genome using the STAR aligner23 with params ‘-outSAMstrandField intronMotif --alignSJDBoverhangMin 6 -alignIntronMax 299999 --outFilterMultimapNmax 4 -scoreGapATAC -4’.

Differential splicing analysis A list of valid introns was extracted from the Gencode24 gene transfer format file post removing transcripts with biotype ‘retained_intron’. Introns overlapping with an exon at their junctions were removed using pybedtools25,26 and gffutils (https://github.com/daler/gffutils). They were subsequently classified as either U2-type or U12-type based on position weight matrices from splicerack using gimmemotifs.27,28 MSI (mis-splicing index) values were calculated for each intron as described before.8 Only those introns with a coverage of at least one read at each of their junctions and with total coverage of the two junctions higher than four were considered. Also, introns which did not have a coverage of at least one for 95% of their length were filtered out. In order to identify differentially spliced introns, differences in MSI values (DMSI) were calculated as DMSI=MSIknockout−MSIwildtype. Significance of this difference was computed using a Fisher’s test, and the obtained Pvalues were corrected for multiple testing. The python and R scripts used for analyses are available at https://github.com/pd321/intron-retention-scripts.

Accession codes RNA sequencing data were deposited in Gene Expression Omnibus database repository under accession numbers GSE151470 and GSE152432.

Results Aberrant retention of U12-type introns in ZRSR2-deficient murine hematopoietic cells Mice lacking the entire Zrsr2 coding sequence were obtained by germline transmission of the targeted allele (Figure 1A). Following this, neomycin selection cassette was removed through mating with CMV-Cre transgenic mice, in effect replacing the Zrsr2 coding sequence with βgalactosidase gene (Figure 1A). Complete lack of Zrsr2 transcripts was evident in quantitative polymerase chain reaction (qPCR) analysis of BM, spleen and thymus cells, as well as RNA sequencing of targeted ES cells (Online Supplementary Figure S1A and B). Loss of Zrsr2 did not alter expression of its homolog, Zrsr1, in hematopoietic cells (Online Supplementary Figure S1B). In order to assess the effect of ZRSR2 deficiency on 681

V. Madan et al. A

682

Figure 1. Deficiency of ZRSR2 causes aberrant retention of U12type introns in murine myeloid cells. (A) Generation of Zrsr2 knockout (KO) mice. Mice carrying the targeted Zrsr2 allele were crossed with CMV-Cre mice to excise the neomycin resistance cassette (post-Cre allele). Polymerase chain reaction analysis (right) shows genotyping of deleted Zrsr2 alleles in male and female mice. (B) Dot plots show intron retention in Zrsr2-deficent murine myeloid precursors (CMP, GMP and MEP) and MEF compared to wild-type (WT) cells. U12type introns are depicted as red circles. (C) Range of DMSI values for retention of U12-type introns in murine cells. Outliers were removed from the plot using Gout method (Q=1). (D) Intron retention in ZRSR2 knockdown K562 cells (expressing either short hairpin RNA [shRNA] shRNA1 or shRNA2) compared to control transduced cells. (E) Range of DMSI values for intron retention (U12-type introns) in TF1, K562 and myelodysplastic syndromes (MDS) bone marrow cells. Outliers were removed from the plot using Gout method (Q=1). RNA sequencing data of ZRSR2deficient TF1 cells and ZRSR2 mutant MDS used in this analysis has been previously published.8 (F) Number of U12-type introns retained in murine and human cells lacking ZRSR2 (P<0.05; Fisher's exact test). CMP: common myeloid precursors; GMP: granulocyte monocyte precursors; MEP: megakaryocyte erythroid precursors; MEF: murine embryonic fibroblasts; MSI: mis-splicing index. Difference in MSI values (DMSI) was calculated as DMSI=MSIknockout−MSIwild-type.

haematologica | 2022; 107(3)

ZRSR2 & ZRSR1 co-operate in splicing of U12 introns

RNA splicing, we performed RNA sequencing on sorted wild-type (WT) and KO myeloid precursor populations from BM: CMP (Lin−Kit+Sca1−CD34+FcγRII/IIIlo), GMP MEP (Lin−Kit+Sca1−CD34+FcγRII/IIIhi), (Lin−Kit+Sca1−CD34−FcγRII/III−), as well as MEF from male WT and Zrsr2 KO embryos. Aberrant retention of U12type introns was observed in all three ZRSR2-deficient myeloid progenitors, but was not evident in MEF (Figure 1B and C; Online Supplementary Table S2). Despite a trend towards mis-splicing of U12-type introns in murine hematopoietic cells, we noted that the effect of ZRSR2 deficiency on splicing (as indicated by DMSI values) was significantly lower than that previously observed by us in ZRSR2 mutant human MDS BM and ZRSR2-deficient human AML cell lines, TF1 and K562 (Figure 1B and E; Online Supplementary Table S3).8 The number of aberrantly retained U12-type introns in Zrsr2 KO murine myeloid precursors was notably lower than ZRSR2 knockdown K562 and TF1 cells (Figure 1F). Only a modest effect on splicing was unexpected, especially given a complete loss of ZRSR2 expression in our mouse model. Moreover, U12-type introns are highly conserved between human and mouse genomes (Online Supplementary Figure S2), so a similar effect of ZRSR2 deficiency would be expected in the two species. Overall, our findings indicated reduced dependence of the U12-spliceosome on ZRSR2 in murine myeloid cells.

Normal myeloid development in mice lacking ZRSR2 We analyzed hematopoietic compartment in ZRSR2deficient (Zrsr2D/Y) compared to WT (Zrsr2+/Y) male mice. No significant difference was observed in peripheral blood counts of young mice of both genotypes; and age-dependent defects were also not apparent (Figure 2A; Online Supplementary Figure S3). Initial analysis of young males (710 weeks old) showed that deficiency of ZRSR2 neither affected the BM cellularity nor the frequency of Lin−Sca1+Kit+ (LSK) cells (Figure 2B and C). Proportion of HSC (CD34–Flt3– LSK cells) and progenitors also remained unaltered in young male mice (Figure 2D). MDS is primarily a disease of elderly, and in order to understand if loss of ZRSR2 manifested its effect with aging, BM of ≥1-year old male mice were examined. Surprisingly, old ZRSR2deficient males also exhibited normal BM cellularity and frequencies of HSC and multipotent progenitors (Figure 2B to D). Moreover, proportion of myeloid precursors, CMP, GMP and MEP, and stages of erythroid development were also largely unaltered in the BM of young and old ZRSR2deficient male mice (Figure 2E; Online Supplementary Figure S4A). Normal myeloid differentiation was also evident by unchanged frequencies of granulocytes in spleen and BM of Zrsr2 KO mice (Online Supplementary Figure S4B). In order to evaluate the repopulation ability of Zrsr2 KO HSC, competitive repopulation assays were performed. ZRSR2-deficient HSC reconstituted both myeloid and lymphoid lineages in recipient mice as efficiently as WT cells (Figure 2F), suggesting that repopulation potential of HSC is maintained in the absence of ZRSR2. Further, in non-competitive repopulation assays, loss of ZRSR2 did not affect the peripheral blood cell counts in recipient mice even 1 year after transplantation (Online Supplementary Figure S5). Collectively, our comprehensive analyses of hematopoietic development in Zrsr2 KO mice and reconstitution ability of ZRSR2-deficient HSC demonstrated that ZRSR2 is not essential for hematopoietic development in mice. haematologica | 2022; 107(3)

Murine Zrsr1 is a putative functional copy of Zrsr2 Given our unexpected observations that deletion of Zrsr2 in mice did not impact hematopoietic development and modestly affected splicing of U12-type introns, we postulated that other spliceosome protein(s) might compensate for its absence. ZRSR1, a closely-related homolog, is a single exon gene formed by retrotransposition of ZRSR2 cDNA sequence. This autosomal copy of X-linked ZRSR2 gene is highly similar to the parent gene in coding sequence (95% and 77% identical in human and mouse, respectively) and amino acid sequences (92% and 75% identical in human and mouse, respectively) with a conserved open reading frame. While ZRSR2 is ubiquitously expressed, human ZRSR1 is designated as a pseudogene with negligible transcript levels in human tissues (https://gtexportal.org/home/), including the hematopoietic cells (Figure 3A). In contrast, the mouse Zrsr1 gene is expressed in hematopoietic cells, albeit at levels lower than Zrsr2 (Figure 3B). This difference in transcriptional regulation of murine and human ZRSR1 genes is possibly caused by their location in distinct, nonorthologous genomic loci (Figure 3C). Human ZRSR1 is located in the intron of the REEP5 gene on chromosome 5q while the mouse gene localizes to the first intron of the Commd1 gene on chromosome 11 (Figure 3C), a genomic locus not syntenic to human chromosome 5. This demonstrates that retrotransposition of ZRSR2 gene occurred independently in rodents and primates, after the evolutionary divergence between the two mammalian orders. Furthermore, we inspected the chromatin structure and presence of histone marks associated with transcriptional activation at the ZRSR1 locus in human and murine cells, using available DNase-seq/ATAC-seq and ChIP-Seq data, respectively. While no ZRSR1 locus-specific signal was observed for histone marks associated with gene activation (H3K4me3, H3K4me1 and H3K27ac) in human common myeloid progenitors, significant ChIP-Seq peaks for all three histone modifications were detected in murine hematopoietic cells (Figure 3D). Correspondingly, accessible chromatin was evident the murine but not human ZRSR1 locus (Figure 3D). Based on epigenetic profiles and expression levels, we hypothesized that murine ZRSR1 encodes for a functional protein that could possibly regulate splicing of the U12-type introns.

Loss of ZRSR1 impairs splicing of U12-type introns in the absence of ZRSR2 In order to investigate if ZRSR1 functionally compensates for deficiency of ZRSR2 in splicing of U12-type introns, we silenced expression of Zrsr1 in murine myeloid precursors using short hairpin RNA (shRNA) (Online Supplementary Figure S6A). Stable knockdown of Zrsr1 in WT Lin−Kit+ BM cells did not notably alter splicing of the U12-type introns, although it resulted in a reduced number of myeloid colonies in the methylcellulose medium (Figure 4A; Online Supplementary Figures S6B and S7A; Online Supplementary Table S4). Zrsr2 KO myeloid cells exhibited sizable retention of the U12-type introns (Figure 4A and B; Online Supplementary Figure S7A; Online Supplementary Table S4), similar to that previously described for sorted CMP, GMP and MEP populations (Figure 1B and C). Notably, knockdown of Zrsr1 in ZRSR2-deficient myeloid cells further exacerbated missplicing of the U12-type introns (Figure 4A and B; Online 683

V. Madan et al.

Supplementary Figure; Online Supplementary Table S4), suggesting that ZRSR1 can also regulate their splicing. Comparison of Zrsr2 single KO with Zrsr2/Zrsr1 doubledeficient Lin−Kit+ BM cells showed two subclasses of retained introns in Zrsr2/Zrsr1 double deficient cells. The first subclass comprises those introns which are retained in Zrsr2 KO cells, and their DMSI increases significantly in double-deficient cells; while the second category includes introns which are retained only when both ZRSR1 and ZRSR2 are lacking (Online Supplementary Figure S7B). Retention of U12-type introns in the Zrsr2/Zrsr1 double-deficient cells was validated by quantitative real-time PCR (qRT-PCR) analysis of 10 introns, which illustrated a modest effect of loss of ZRSR2 alone on splicing of U12type introns, while concomitant deficiency of ZRSR1 and

ZRSR2 significantly enhanced intron retention compared to either WT or single KO cells (Figure 4C). Further, in order to verify the compensatory role of murine ZRSR1 in splicing of U12-type introns, CRISPR/Cas9 technique was used to generate myeloid cells (32D) lacking either one or both ZRSR proteins. Firstly, the whole Zrsr2 coding sequence was deleted using two guide RNA targeting exons 2 and 11 of Zrsr2 gene (Online Supplementary Figure S8A and B). Two single cell clones of Zrsr2 KO 32D cells were generated (Online Supplementary Figure S8B and C). Subsequently, we deleted the N-terminus portion of the Zrsr1 gene to generate single or double KO 32D cells (Online Supplementary Figure S8A, B, D and E). qRT-PCR analysis demonstrated aberrant retention of U12-type introns in Zrsr1/Zrsr2 double-

Figure 2. Hematopoietic development and reconstitution potential are unperturbed upon deletion of Zrsr2. (A) Peripheral blood white blood cell (WBC) counts in wild-type (WT) and Zrsr2 knockout (KO) mice. (B) Total bone marrow (BM) leukocyte counts (femurs + tibias) in young (7-10 weeks) and old (≥1-year old) WT and Zrsr2-deficient mice. (C) Proportion of LSK cells in the BM of young and old WT and Zrsr2 KO mice. (D and E) Frequencies of LT-HSC, ST-HSC and MPP (D) (Young mice: five WT and five KO; Old mice: five WT and three KO) and myeloid precursors (CMP, GMP, MEP) (E) (Young mice: five WT and five KO; Old mice: six WT and four KO). (F) Donor-derived B cells, T cells, granulocytes and monocytes in peripheral blood of recipient mice in competitive repopulation assays (six recipient mice/group). HSC: hematopoietic stem cells; CMP: common myeloid precursors; GMP: granulocyte monocyte precursors; MEP: megakaryocyte erythroid precursors.

684

haematologica | 2022; 107(3)

ZRSR2 & ZRSR1 co-operate in splicing of U12 introns

Figure 3. Human and mouse ZRSR1 are located in non-orthologous genomic loci. (A) Levels of ZRSR1 and ZRSR2 transcripts in human whole blood and spleen cells. Expression data are collated from Genotype-Tissue Expression (GTEx) portal. (B) Expression levels of Zrsr1 and Zrsr2 transcripts in murine spleen cells (Bgee database) and myeloid precursors (in-house RNA sequencing of sorted CMP, GMP and MEP cells). TPM: transcripts per million reads. (C) Genomic location of human and murine ZRSR1 genes. Orange arrows denote the direction of transcription for each gene. (D) H3K4me3, H3K4me1 and H3K27ac ChIP-seq signals and regions of open chromatin at human and murine ZRSR1 locus. All data were downloaded from the ENCODE Consortium. Human ChIP-seq and DNase-seq data are for common myeloid progenitors, while murine data are from following sources - H3K4me3: CD1 embryonic erythroblasts; H3K4me1: CD1 embryonic megakaryocytes; H3K27ac: bone marrow macrophages; ATAC-seq: adult erythroblasts; CMP: common myeloid precursors; GMP: granulocyte monocyte precursors; MEP: megakaryocyte erythroid precursors.

haematologica | 2022; 107(3)

685

V. Madan et al. A

Figure 4. Loss of ZRSR1 exacerbates retention of U12-type introns in ZRSR2-deficient murine hematopoietic cells. (A) Distribution of DMSI values for retention of U12-type introns in Lin−Kit+ bone marrow (BM) cells lacking either ZRSR1 (sh1 or sh10) or ZRSR2 or both compared to control cells. (B) Number of U12-type introns retained (P<0.05; Fisher's exact test) in various pair-wise comparisons of Lin−Kit+ BM cells deficient in either one or both ZRSR proteins. (C) Normalized expression of representative U12-type introns in Lin−Kit+ BM cells determined using quantitative real-time polymerase chain reaction. Expression of U12-type introns was measured relative to expression of flanking exons. Data are from at least three replicates and represented as mean ± standard error of the mean. P-values for each group compared to the ‘ZRSR2 WT; con sh’ group are depicted in the plot. Statistical difference between the Zrsr2/Zrsr1-double deficient and Zrsr2 KO cells are shown below the graph. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; ns: not significant; MSI: mis-splicing index: Difference in MSI values (DMSI) was calculated as DMSI=MSIknockout−MSIwildtype.

deficient cells compared to control cells (Online Supplementary Figure S8F). In order to verify that human ZRSR1 is a non-functional copy of ZRSR2, we generated K562 cells lacking either ZRSR1 or ZRSR2 or both proteins together. DZRSR1 cells were first generated using CRISPR/Cas9 technique followed by silencing of ZRSR2 using shRNA. As expected, ZRSR2 deficiency alone impaired splicing of U12-type introns. However, loss of ZRSR1, either alone or when combined with ZRSR2, did not significantly alter the splicing of U12-type introns in K562 cells (Online Supplementary Figure S9). Taken together, our experiments highlight a functional role of murine ZRSR1 in splicing of the U12-type introns.

Mis-splicing of MAPK9 and MAPK14 impacts their protein expression in murine and human cells Several crucial genes harbour U12-type introns including members of the mitogen-activated protein kinase 686

(MAPK) family of serine-threonine kinases. MAPK proteins play a crucial role in a wide range of biological processes including hematopoiesis.29 We focused on splicing of two members, MAPK9 (JNK2) and MAPK14 (p38α), as aberrant retention of their U12-type introns was verified in murine myeloid precursors deficient in both Zrsr1 and Zrsr2 (Figure 4C). Intron retention of both MAPK9 and MAPK14 resulted in the generation of aberrant transcripts containing a premature stop codon. In order to investigate if the retention of U12-type introns in these two genes affects their protein expression, western blotting was performed with anti-MAPK9 and antiMAPK14 antibodies. Protein levels of both MAPK9 and MAPK14 were reduced, albeit modestly, in murine myeloid precursors, and MAPK14 levels were also reduced in 32D cells lacking ZRSR2 and ZRSR1 (Online Supplementary Figure S10A and B). Retention of U12-type introns of MAPK9 and MAPK14 was also observed in ZRSR2 knockdown K562 and TF1 cells (Figure 5A and B). haematologica | 2022; 107(3)

ZRSR2 & ZRSR1 co-operate in splicing of U12 introns

Figure 5. Aberrant retention of U12-type introns in human MAPK9 and MAPK14 genes impacts their protein expression. (A and B) Normalized expression of U12-type introns of MAPK9 (A) and MAPK14 (B) in TF1 and K562 cells stably expressing short hairpin RNA (shRNA) targeting human ZRSR2. The expression of U12-type introns was measured relative to the expression of flanking exons using quantitative polymerase chain reaction (qPCR). Data are from five PCR experiments and represented as mean ± standard error of the mean. *P<0.05, **P<0.01, ***P<0.001, ns: not significant. (C) Protein levels of human MAPK9 and MAPK14 in TF1 and K562 cells transduced with shRNA targeting ZRSR2.

This was accompanied by significantly reduced MAPK9 and MAPK14 protein expression in human cells with silencing of ZRSR2 (Figure 5C). Amongst the two cell lines, K562 cells exhibited increased mis-splicing of both MAPK9 and MAPK14, which correlated with a more profound effect on their protein levels, compared to the TF1 cells (Figure 5A to C). These observations illustrate that mis-splicing of MAPK9 and MAPK14 results in lower levels of these proteins, potentially impacting their function in hematopoietic development.

Discussion Mouse models have been effective in deciphering key splicing and hematopoietic defects caused by mutations of spliceosome genes. Hematopoietic cells from mice expressing either mutant SF3B1, SRSF2 or U2AF1 recapitulated the patterns of splicing changes observed in mutant MDS samples.6,12,15,16,18-20 Although the splice factors are evolutionary conserved, intronic sequences are divergent between human and mouse. These differences in intronic splicing elements and regulatory motifs between the two species possibly lead to a divergent set of mis-spliced transcripts, which can contribute to phenotypic differences in hematopoietic development. This presents a conundrum about the utility of mouse models to identify genes whose splicing is altered by spliceosome mutation in MDS. We have previously demonstrated that ZRSR2 mutations/deficiency impairs splicing of the U12haematologica | 2022; 107(3)

type intron in human cells.8,17 U12-type introns are highly conserved during evolution, and the tissue-specific expression of transcripts harboring this class of introns is largely preserved between mouse and human.30,31 Therefore, we envisaged that a murine model would faithfully replicate the loss of ZRSR2 in leukemic human cells. Indeed, murine ZRSR2 KO myeloid cells exhibited aberrant retention of U12-type introns. Unexpectedly, splicing of the U12-type introns was unaffected in ZRSR2-deficient MEF, suggesting tissue-specific effects of ZRSR2-deficiency on splicing. A previous study also demonstrated a more pronounced mis-splicing of U12type introns in blood cells compared to fibroblast and amniocytes in Taybi–Linder syndrome (TALS) cases harboring germline mutations of a U12 spliceosome-specific component, RNU4ATAC.32 One limitation of our constitutive KO mouse model is that complete absence of ZRSR2 caused by germline loss of ZRSR2 might promote functional compensation during development. Therefore, findings from our KO mice may not reflect the consequences of acute loss of ZRSR2 in myeloid cells. Hematopoietic cell-specific inducible models of ZRSR2 deficiency are needed to further elucidate the effect of conditional depletion of ZRSR2 during adult hematopoiesis, akin to what is observed in ZRSR2 mutant MDS. Mice with constitutive ZRSR2 deletion developed normally with no overt hematopoietic defects and functionally competent hematopoietic stem cells. This is in contrast with mice expressing mutant SF3B1, SRSF2 or 687

V. Madan et al.

U2AF1, which displayed a range of hematopoietic phenotypes.6,12,15,16,18-20,33 Isolated loss-of-function mutations of ZRSR2 have been associated with macrocytic anaemia in patients with myeloid diseases.34 However, our Zrsr2 KO mice did not display any significant changes in size and counts of blood cells. Moreover, while ZRSR2 deficiency affected splicing of U12-type introns in murine hematopoietic cells, magnitude of splicing defects was modest compared to either MDS BM cells or human leukemia cell lines. This led us to investigate whether ZRSR1 can functionally compensate for the lack of ZRSR2. ZRSR1 is a retrotransposed copy of ZRSR2, which originated via independent transposition events in rodents and primates. Murine Zrsr1 is an imprinted gene and expressed from the unmethylated, paternal allele.35-40 Unlike the human counterpart, which is designated as a pseudogene, the murine gene is expressed in hematopoietic cells, albeit at lower levels compared with Zrsr2. Our analyses revealed open chromatin and enrichment of epigenetic marks associated with transcriptional activation only at the murine Zrsr1 locus, supporting evidence for transcription of the murine retrogene. In our study, silencing of Zrsr1 had no evident effect on the splicing of U12-type introns in Zrsr2 WT myeloid cells. In contrast, deficiency of ZRSR1 clearly exacerbated the mis-splicing of U12-type introns in cells lacking ZRSR2, thereby underlining that both ZRSR1 and ZRSR2 collectively contribute to splicing of U12-type introns in mouse hematopoietic cells. While RNA splicing was not studied in Zrsr1 KO mice,41 recent studies have shown that expression of truncated Zrsr1 impacts splicing of the U12-type introns in testis and hypothalamus.42,43 Interestingly, Zrsr1-mutant mice exhibit defects in erythrocyte maturation and fewer peripheral red blood cells, with apparent morphological abnormalities.42 The differences observed in impact of ZRSR1 deficiency on splicing of U12-type introns in our study (myeloid precursors) versus studies investigating mutant Zrsr1-expressing spermatocytes/hypothalamus could possibly arise because of relative levels of ZRSR2 and ZRSR1 in different cell types. Another possibility is that the truncated ZRSR1 can impair recruitment of ZRSR2 to U12-type intron splice site, thereby perturbing its splicing function. Signal transduction mediated by MAP kinase pathway plays a vital role in numerous biological processes via phosphorylation of several downstream substrates.44 Genes encoding several members of MAP kinase family harbor U12-type introns, hence their splicing is dependent on ZRSR2 activity. We identified two candidates, MAPK9 and MAPK14, where aberrant splicing resulted in decreased protein levels both in human and murine cells. The effect of ZRSR2/ZRSR1 double deficiency on MAPK9 and MAPK14 protein levels in murine cells was less pronounced compared to human cells lacking ZRSR2, which corresponds with milder effect on splicing of U12type introns in murine cells. While MAPK9 regulates Tcell apoptosis and proliferation, and MAPK14 is indispensable for definitive erythropoiesis in mice,45,46 these proteins have not been directly implicated in pathogenesis of MDS. Although we experimentally validated mis-splicing of just two of the MAPK members, a broader effect on splicing of multiple components can be envisaged. Moreover, in murine myeloid precursors, we also validated aberrant retention of U12-type intron of MCTS1, 688

which modulates MAPK pathway by promoting phosphorylation of MAPK1 and MAPK3.47 Given an essential role of MAPK proteins in regulating hematopoiesis,29 collective decrease in their protein levels can potentially be detrimental to myeloid/erythroid differentiation and expansion, thereby contributing to the disease phenotype in MDS. Taken together, unlike human ZRSR1 pseudogene, Zrsr1 in mice is a functional autosomal copy of Zrsr2 and contributes to splicing of U12-type introns. This is also supported by a recent study which demonstrated that expression of at least one copy of either maternal Zrsr2 or paternal Zrsr1 is necessary for viability of murine embryos.48 Additionally, our study highlights that splicing of U12-type introns in murine cells depends conceivably on the balance between expression levels of ZRSR2 and ZRSR1. Hence, deficiency of ZRSR2 alone is insufficient to impact extensively RNA splicing in mice, and further studies with concurrent deficiency of ZRSR1 and ZRSR2 are warranted to replicate complete loss of ZRSR activity. Notably, germline expression of truncated Zrsr1 and Zrsr2 alleles showed that double mutant mice are non-viable, with Zrsr1/Zrsr2 double mutant embryos exhibiting defects in early preimplantation development.48 Hence, conditional KO alleles of both Zrsr1 and Zrsr2 are required to investigate their combined loss in adult mice. Disclosures No conflicts of interest to disclose. Contributions VM conceived the study, designed and performed research, analysed data and wrote the manuscript; ZC designed and performed research, analysed data and wrote the manuscript; WWT, LH, PS and MJ performed research and analysed data; PD performed bioinformatics and statistical analyses and wrote the manuscript; SZ, JL and HY performed and supervised bioinformatics and statistical analyses; SJ, YS and MZH performed blastocyst injections to generate chimeras from targeted ES cells; WJC supervised the study and wrote the manuscript; HPK conceived and supervised the study, interpreted the data and wrote the manuscript. All authors reviewed and approved the manuscript. Acknowledgements We thank the staff of Comparative Medicine, NUS for their support in maintaining mouse colonies. We also acknowledge expert help and support from the FACS facility at CSI, Singapore. Funding This work was funded by the Leukemia and Lymphoma Society, the Singapore Ministry of Health’s National Medical Research Council (NMRC) under its Singapore Translational Research (STaR) Investigator Award to HPK (NMRC/STaR/0021/2014), the NMRC Center Grant awarded to the National University Cancer Institute of Singapore (NMRC/CG/012/2013) and the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centers of Excellence initiatives. This research is also supported by the RNA Biology Center at the Cancer Science Institute of Singapore, NUS, as part of funding under the Singapore Ministry of Education’s Tier 3 grants, grant number MOE2014-T3-1-006. We thank the Melamed Family for their generous support. haematologica | 2022; 107(3)

ZRSR2 & ZRSR1 co-operate in splicing of U12 introns

References 1. Yoshida K, Sanada M, Shiraishi Y, et al. Frequent pathway mutations of splicing machinery in myelodysplasia. Nature. 2011;478(7367):64-69. 2. Haferlach T, Nagata Y, Grossmann V, et al. Landscape of genetic lesions in 944 patients with myelodysplastic syndromes. Leukemia. 2014;28(2):241-247. 3. Papaemmanuil E, Gerstung M, Malcovati L, et al. Clinical and biological implications of driver mutations in myelodysplastic syndromes. Blood. 2013;122(22):3616-3627. 4. Papaemmanuil E, Cazzola M, Boultwood J, et al. Somatic SF3B1 mutation in myelodysplasia with ring sideroblasts. N Engl J Med. 2011;365(15):1384-1395. 5. Damm F, Kosmider O, Gelsi-Boyer V, et al. Mutations affecting mRNA splicing define distinct clinical phenotypes and correlate with patient outcome in myelodysplastic syndromes. Blood. 2012;119(14):3211-3218. 6. Kim E, Ilagan JO, Liang Y, et al. SRSF2 Mutations contribute to myelodysplasia by mutant-specific effects on exon recognition. Cancer Cell. 2015;27(5):617-630. 7. Komeno Y, Huang YJ, Qiu J, et al. SRSF2 Is Essential for hematopoiesis, and its myelodysplastic syndrome-related mutations dysregulate alternative pre-mRNA splicing. Mol Cell Biol. 2015;35(17):30713082. 8. Madan V, Kanojia D, Li J, et al. Aberrant splicing of U12-type introns is the hallmark of ZRSR2 mutant myelodysplastic syndrome. Nat Commun. 2015;6:6042. 9. Zhang J, Lieu YK, Ali AM, et al. Diseaseassociated mutation in SRSF2 misregulates splicing by altering RNA-binding affinities. Proc Natl Acad Sci U S A. 2015;112 (34):E4726-4734. 10. Ilagan JO, Ramakrishnan A, Hayes B, et al. U2AF1 mutations alter splice site recognition in hematological malignancies. Genome Res. 2015;25(1):14-26. 11. Okeyo-Owuor T, White BS, Chatrikhi R, et al. U2AF1 mutations alter sequence specificity of pre-mRNA binding and splicing. Leukemia. 2015;29(4):909-917. 12. Shirai CL, Ley JN, White BS, et al. Mutant U2AF1 expression alters hematopoiesis and pre-mRNA splicing in vivo. Cancer Cell. 2015;27(5):631-643. 13. Alsafadi S, Houy A, Battistella A, et al. Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage. Nat Commun. 2016;7:10615. 14. Darman RB, Seiler M, Agrawal AA, et al. Cancer-Associated SF3B1 hotspot mutations induce cryptic 3' splice site selection through use of a different branch point. Cell Rep. 2015;13(5):1033-1045. 15. Mupo A, Seiler M, Sathiaseelan V, et al. Hemopoietic-specific Sf3b1-K700E knock-in mice display the splicing defect seen in human MDS but develop anemia without ring sideroblasts. Leukemia. 2017;31(3):720727. 16. Obeng EA, Chappell RJ, Seiler M, et al. Physiologic expression of Sf3b1(K700E) causes impaired erythropoiesis, aberrant splicing, and sensitivity to therapeutic spliceosome modulation. Cancer Cell. 2016;30(3):404-417.

haematologica | 2022; 107(3)

17. Madan V, Li J, Zhou S, et al. Distinct and convergent consequences of splice factor mutations in myelodysplastic syndromes. Am J Hematol. 2020;95(2):133-143. 18. Fei DL, Zhen T, Durham B, et al. Impaired hematopoiesis and leukemia development in mice with a conditional knock-in allele of a mutant splicing factor gene U2af1. Proc Natl Acad Sci U S A. 2018;115(44):E10437E10446. 19. Kon A, Yamazaki S, Nannya Y, et al. Physiological Srsf2 P95H expression causes impaired hematopoietic stem cell functions and aberrant RNA splicing in mice. Blood. 2018;131(6):621-635. 20. Smeets MF, Tan SY, Xu JJ, et al. Srsf2(P95H) initiates myeloid bias and myelodysplastic/myeloproliferative syndrome from hemopoietic stem cells. Blood. 2018;132(6): 608-621. 21. Tronchere H, Wang J, Fu XD. A protein related to splicing factor U2AF35 that interacts with U2AF65 and SR proteins in splicing of pre-mRNA. Nature. 1997;388(6640): 397-400. 22. Shen H, Zheng X, Luecke S, Green MR. The U2AF35-related protein Urp contacts the 3' splice site to promote U12-type intron splicing and the second step of U2type intron splicing. Genes Dev. 2010;24 (21):2389-2394. 23. Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29(1):15-21. 24. Frankish A, Diekhans M, Ferreira AM, et al. GENCODE reference annotation for the human and mouse genomes. Nucleic Acids Res. 2019;47(D1):D766-D773. 25. Dale RK, Pedersen BS, Quinlan AR. Pybedtools: a flexible Python library for manipulating genomic datasets and annotations. Bioinformatics. 2011;27(24):34233424. 26. Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841-842. 27. Sheth N, Roca X, Hastings ML, Roeder T, Krainer AR, Sachidanandam R. Comprehensive splice-site analysis using comparative genomics. Nucleic Acids Res. 2006;34(14):3955-3967. 28. van Heeringen SJ, Veenstra GJ. GimmeMotifs: a de novo motif prediction pipeline for ChIP-sequencing experiments. Bioinformatics. 2011;27(2):270-271. 29. Geest CR, Coffer PJ. MAPK signaling pathways in the regulation of hematopoiesis. J Leukoc Biol. 2009;86(2):237-250. 30. Lin CF, Mount SM, Jarmolowski A, Makalowski W. Evolutionary dynamics of U12-type spliceosomal introns. BMC Evol Biol. 2010;10:47. 31. Olthof AM, Hyatt KC, Kanadia RN. Minor intron splicing revisited: identification of new minor intron-containing genes and tissue-dependent retention and alternative splicing of minor introns. BMC Genomics. 2019;20(1):686. 32. Cologne A, Benoit-Pilven C, Besson A, et al. New insights into minor splicing-a transcriptomic analysis of cells derived from TALS patients. RNA. 2019;25(9):1130-1149. 33. Xu JJ, Smeets MF, Tan SY, Wall M, Purton LE, Walkley CR. Modeling human RNA spliceosome mutations in the mouse: not all mice were created equal. Exp Hematol.

2019;70:10-23. 34. Fleischman RA, Stockton SS, Cogle CR. Refractory macrocytic anemias in patients with clonal hematopoietic disorders and isolated mutations of the spliceosome gene ZRSR2. Leuk Res. 2017;61:104-107. 35. Hatada I, Sugama T, Mukai T. A new imprinted gene cloned by a methylationsensitive genome scanning method. Nucleic Acids Res. 1993;21(24):5577-5582. 36. Hayashizaki Y, Shibata H, Hirotsune S, et al. Identification of an imprinted U2af binding protein related sequence on mouse chromosome 11 using the RLGS method. Nat Genet. 1994;6(1):33-40. 37. Hatada I, Kitagawa K, Yamaoka T, et al. Allele-specific methylation and expression of an imprinted U2af1-rs1 (SP2) gene. Nucleic Acids Res. 1995;23(1):36-41. 38. Shibata H, Yoshino K, Sunahara S, et al. Inactive allele-specific methylation and chromatin structure of the imprinted gene U2af1-rs1 on mouse chromosome 11. Genomics. 1996;35(1):248-252. 39. Feil R, Boyano MD, Allen ND, Kelsey G. Parental chromosome-specific chromatin conformation in the imprinted U2af1-rs1 gene in the mouse. J Biol Chem. 1997;272 (33):20893-20900. 40. Nabetani A, Hatada I, Morisaki H, Oshimura M, Mukai T. Mouse U2af1-rs1 is a neomorphic imprinted gene. Mol Cell Biol. 1997;17(2):789-798. 41. Sunahara S, Nakamura K, Nakao K, Gondo Y, Nagata Y, Katsuki M. The oocyte-specific methylated region of the U2afbp-rs/U2af1rs1 gene is dispensable for its imprinted methylation. Biochem Biophys Res Commun. 2000;268(2):590-595. 42. Horiuchi K, Perez-Cerezales S, Papasaikas P, et al. Impaired spermatogenesis, muscle, and erythrocyte function in U12 intron splicingdefective Zrsr1 mutant mice. Cell Rep. 2018;23(1):143-155. 43. Alen F, Gomez-Redondo I, Rivera P, et al. Sex-dimorphic behavioral alterations and altered neurogenesis in U12 intron splicingdefective Zrsr1 mutant mice. Int J Mol Sci. 2019;20(14):3543. 44. Cargnello M, Roux PP. Activation and function of the MAPKs and their substrates, the MAPK-activated protein kinases. Microbiol Mol Biol Rev. 2011;75(1):50-83. 45. Tamura K, Sudo T, Senftleben U, Dadak AM, Johnson R, Karin M. Requirement for p38alpha in erythropoietin expression: a role for stress kinases in erythropoiesis. Cell. 2000;102(2):221-231. 46. Sabapathy K, Kallunki T, David JP, Graef I, Karin M, Wagner EF. c-Jun NH2-terminal kinase (JNK)1 and JNK2 have similar and stage-dependent roles in regulating T cell apoptosis and proliferation. J Exp Med. 2001;193(3):317-328. 47. Hsu HL, Choy CO, Kasiappan R, et al. MCT-1 oncogene downregulates p53 and destabilizes genome structure in the response to DNA double-strand damage. DNA Repair (Amst). 2007;6(9):1319-1332. 48. Gomez-Redondo I, Ramos-Ibeas P, Pericuesta E, Fernandez-Gonzalez R, Laguna-Barraza R, Gutierrez-Adan A. Minor splicing factors Zrsr1 and Zrsr2 are essential for early embryo development and 2-celllike conversion. Int J Mol Sci. 2020;21 (11):4115.

689

ARTICLE Ferrata Storti Foundation

Haematologica 2022 Volume 107(3):690-701

Non-Hodgkin Lymphoma

Subtype-specific and co-occurring genetic alterations in B-cell non-Hodgkin lymphoma Man Chun John Ma,1* Saber Tadros,1* Alyssa Bouska,2 Tayla Heavican,2 Haopeng Yang,1 Qing Deng,1 Dalia Moore,3 Ariz Akhter,4 Keenan Hartert,3 Neeraj Jain,1 Jordan Showell,1 Sreejoyee Ghosh,1 Lesley Street,5 Marta Davidson,5 Christopher Carey,6 Joshua Tobin,7 Deepak Perumal,8 Julie M. Vose,9 Matthew A. Lunning,9 Aliyah R. Sohani,10 Benjamin J. Chen,11 Shannon Buckley,12 Loretta J. Nastoupil,1 R. Eric Davis,1 Jason R. Westin,1 Nathan H. Fowler,1 Samir Parekh,8 Maher Gandhi,7 Sattva Neelapu,1 Douglas Stewart,5 Kapil Bhalla,13 Javeed Iqbal,2 Timothy Greiner,2 Scott J. Rodig,14 Adnan Mansoor5 and Michael R. Green1,14,15 Department of Lymphoma and Myeloma, Division of Cancer Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, USA; 2Department of Pathology and Microbiology, University of Nebraska Medical Center, Omaha, NE, USA; 3Eppley Institute for Research in Cancer and Allied Diseases, University of Nebraska Medical Center, Omaha, NE, USA; 4Department of Pathology and Laboratory Medicine, University of Calgary, Calgary, Alberta, Canada; 5Section of Hematology, Department of Medicine, University of Calgary, Calgary, Alberta, Canada; 6Northern Institute for Research, Newcastle University, Newcastle upon Tyne, UK; 7Diamantina Institute, University of Queensland, Queensland, Australia; 8Division of Hematology and Medical Oncology, Icahn School of Medicine at Mount Sinai, New York, NY, USA; 9Department of Internal Medicine, Division of Hematology-Oncology, University of Nebraska Medical Center, Omaha, NE, USA; 10Department of Pathology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA; 11Department of Pathology, University of Massachusetts Medical School, UMass Memorial Medical Center, Worcester, MA, USA; 12 Department of Genetics, Cell Biology and Anatomy, University of Nebraska Medical Center, Omaha, NE, USA; 13Department of Pathology, Brigham and Womens Hospital, Boston, MA, USA; 14Department of Genomic Medicine, University of Texas MD Anderson Cancer Center, Houston, TX, USA and 15Center for Cancer Epigenetics, University of Texas MD Anderson Cancer Center, Houston, TX, USA. 1

*MCJM and ST contributed equally as co-first authors.

ABSTRACT

Correspondence: MICHAEL R. GREEN mgreen5@mdanderson.org Received: October 16, 2020. Accepted: March 15, 2021. Pre-published: April 1, 2021. https://doi.org/10.3324/haematol.2020.274258

690

-cell non-Hodgkin lymphoma (B-NHL) encompasses multiple clinically and phenotypically distinct subtypes of malignancy with unique molecular etiologies. Common subtypes of B-NHL, such as diffuse large B-cell lymphoma, have been comprehensively interrogated at the genomic level, but rarer subtypes, such as mantle cell lymphoma, remain less extensively characterized. Furthermore, multiple B-NHL subtypes have thus far not been comprehensively compared using the same methodology to identify conserved or subtype-specific patterns of genomic alterations. Here, we employed a large targeted hybrid-capture sequencing approach encompassing 380 genes to interrogate the genomic landscapes of 685 B-NHL tumors at high depth, including diffuse large B-cell lymphoma, mantle cell lymphoma, follicular lymphoma, and Burkitt lymphoma. We identified conserved hallmarks of B-NHL that were deregulated in the majority of tumors from each subtype, including frequent genetic deregulation of the ubiquitin proteasome system. In addition, we identified subtype-specific patterns of genetic alterations, including clusters of co-occurring mutations and DNA copy number alterations. The cumulative burden of mutations within a single cluster were more discriminatory of B-NHL subtypes than individual mutations, implicating likely patterns of genetic cooperation that contribute to disease etiology. We therefore provide the first cross-sectional analysis of mutations and DNA copy number alterations across major B-NHL subtypes and a framework of co-occurring genetic alterations that deregulate genetic hallmarks and likely cooperate in lymphomagenesis.

haematologica | 2022; 107(3)

Patterns of genetic alterations in B-NHL

Introduction Non-Hodgkin lymphomas (NHL) are a heterogeneous group of lymphoid malignancies that predominantly arise from mature B cells (B-NHL). Although mature B-cell neoplasms encompass 38 unique diagnostic subtypes, over 85% of cases fall within only seven histologies.1,2 Recent next-generation sequencing studies have shed light onto the key driver mutations in many of these B-NHL subtypes; for example, large studies of diffuse large B-cell lymphoma (DLBCL) have led to proposed genomic subtypes that have unique etiologies.3-5 However, many less common NHL subtypes such as mantle cell lymphoma (MCL) have not been as extensively characterized.6,7 Furthermore, until recently3,4 genetic alterations have been considered in a binary fashion as either driver events, which directly promote disease genesis or progression, or passenger events, which have little or no impact on disease biology. In contrast to this principle, most B-NHL do not result from a single dominant driver but instead result from the serial acquisition of genetic alterations that cooperate in lymphomagenesis.8 The genetic context of each mutation likely determines its oncogenic potential, and groups of mutations should therefore be considered collectively rather than as singular events. For example, the ‘C5’ and ‘MCD’ clusters identified in DLBCL by Chapuy et al. and Schmitz et al., respectively, are characterized by the co-occurrence of CD79B and MYD88 mutations.3,4 In animal models, the Myd88 L252P mutation (equivalent to human L265P) was found to promote downregulation of surface IgM and a phenotype resembling B-cell anergy.9 However, this effect could be rescued by the Cd79b mutation, showing that these co-occurring mutations cooperate.9 The characterization of other significantly co-occurring genetic alterations are therefore likely to reveal additional important cooperative relationships. We approached this challenge by performing genomic profiling of 685 B-NHL across different subtypes. Through this cross-sectional analysis we characterized genomic hallmarks of B-NHL and sets of significantly co-associated events that likely represent subtype-specific cooperating genetic alterations. This study therefore provides new insight into how co-occurring clusters of genetic alterations may contribute to molecularly and phenotypically distinct subtypes of B-NHL.

microarrays10-12 (n=284). An additional series of 223 formalin-fixed paraffin-embedded tumors were provided by other centers. Samples were de-identified and accompanied by the patients’ diagnosis from the medical records, plus overall survival time and status when available. Medical record diagnosis was used in all cases except for those with fluorescence in situ hybridization (FISH) showing translocations in MYC and BCL2 and/or BCL6, which were amended to DHL. Sequencing results for a subset of 52 BL tumors were described previously.13 All MCL samples were either positive for CCND1 translocation by FISH or positive for CCND1 protein expression by immunohistochemistry, depending on the diagnostic practices of the contributing institution.

Next-generation sequencing A total of 500-1000 ng of genomic DNA was sonicated using a Covaris S2 Ultrasonicator, and libraries prepared using KAPA Hyper Prep Kits (Roche) with TruSeq Adapters (Bioo Scientific) and a maximum of eight cycles of polymerase chain reaction (average of 4 cycles). Libraries were qualified by TapeStation 4200, quantified by Qubit and 10- to 12-plexed for hybrid capture. Each multiplexed library was enriched using our custom LymphoSeq panel encompassing the full coding sequences of 380 genes that were determined to be somatically mutated in B-cell lymphoma (Online Supplementary Table S2, Online Supplementary Methods), as well as tiling recurrent translocation breakpoints. Enrichments were amplified with four to eight cycles of polymerase chain reaction and sequenced on a single lane of an Illumina HiSeq 4000 with 100PE reads in high-output mode at the Hudson Alpha Institute for Biotechnology or the MD Anderson Sequencing and Microarray Facility. Variants were called using our previously validated ensemble approach,13,14 germline polymorphisms were filtered using dbSNP annotation and the EXAC dataset containing 60,706 healthy individuals,15 and significantly mutated genes were defined by MutSig2CV.16 Copy number alterations (CNA) identified by CopyWriteR,17 which was validated using three FL tumors with matched Affymetrix 250K SNP array (Online Supplementary Figure S2), and significant DNA CNA were determined by GISTIC2.18 Translocations were called using FACTERA,19 which we previously validated against MYC translocation status determined by FISH.20 Mutation and CNA data are publicly viewable through cBioPortal: https://www.cbioportal.org/study/summary?id=mbn_mdacc_2013. Matched gene expression microarray data are available through the Gene Expression Omnibus, accession GSE132929. For further details, refer to the Online Supplementary Methods.

Methods Results An overview of our approach is shown in Online Supplementary Figure S1. For detailed methods, please refer to the Online Supplementary Information.

Tumor DNA samples We collected DNA from 685 B-NHL tumors, including 199 FL, 196 MCL, 148 DLBCL, 107 BL, 21 high-grade B-cell lymphoma not otherwise specified (HGBL-NOS), and 14 high-grade B-cell lymphoma with MYC, BCL2 and/or BCL6 rearrangement (DHL) (Online Supplementary Table S1). All samples were archival and deidentified. The study was approved by the institutional review board of the University of Nebraska Medical Center (203-15-EP) and performed in accordance with the Declaration of Helsinki. A total of 462 samples were obtained from the University of Nebraska Medical Center, and were prioritized for inclusion in this study if they had previously undergone pathology review and been interrogated by Affymetrix U133 Plus 2.0 gene expression

haematologica | 2022; 107(3)

Identification of significantly mutated genes and structural alterations We used a 380-gene custom targeted sequencing approach, LymphoSeq, to interrogate the genomes of 685 mature B-NHL, sequencing to an average depth of 578X (minimum, 101X; maximum, 1785X) (Online Supplementary Table S1) for a total yield of 1.81 Tbp. Somatic nucleotide variants and small insertions/deletions were identified using an ensemble approach that we have previously validated14 (Online Supplementary Table S3) and significantly mutated genes were identified using MutSig2CV (Online Supplementary Table S4). Matched germline DNA was available from purified T cells of 20 tumors (11 FL and 9 MCL) and sequenced to validate the filtering of germline variants; 0/632 variants called within these tumors were identified in the matched germline samples, which indicates that the fil691

M.C.J. Ma et al.

tering of germline variants was effective. Genes that were significantly mutated in the full cohort or in any one of the four subtypes with more than 100 tumors (BL, DLBCL, FL, and MCL) were included, as well as frequently mutated genes that are targets of activation-induced cytidine deaminase (AID) (Figure 1, Online Supplementary Table S5). Predictably, the frequency of AID-associated mutations was higher among germinal center-derived lymphomas (BL, DLBCL, FL), but also accounted for 7.6% of all coding and non-coding mutations in MCL (Online Supplementary Table S6). The mutational burden calculated from our targeted region correlated significantly with that from the whole exome (Online Supplementary Figure S3A) and was significantly higher in DLBCL and other high-grade tumors than in FL and MCL (Figure 1, Online Supplementary Figure S3B). The hybrid capture probes utilized in our design also targeted recurrent breakpoint regions in the immunoglobulin heavy- and light-chain loci, and recurrent breakpoints in or near the BCL2, MYC and BCL6 genes, and translocations were called using a method that detects discordantly mapped reads19 (Figures 1 and 2A). Our prior validation of this approach in cases with matched FISH data for MYC showed that it is 100% specific, but only ~40% sensitive for translocation detection.13 This limit of sensitivity likely varies for different genes depending on how well the breakpoints are clustered into hotspots that are targeted by our capture probes. Nonetheless, we observed a significantly higher fraction of BCL6 translocations (57% [27/47]) partnered to non-immunoglobulin loci (e.g., CIITA, RHOH, EIF4A2, and ST6GAL1) (Online Supplementary Table S7) compared to BCL2 (1% [1/114]) and MYC (5% [2/38]) translocations (Figure 2A; Fisher P-value <0.001). These were more frequent in FL (88% [15/17] of BCL6 translocations) than in DLBCL (39% [9/23] of BCL6 translocations), presumably because the two immunoglobulin loci in FL are either translocated with the BCL2 gene or functioning in immunoglobulin expression.21 We also employed off-target reads to detect DNA CNA in a manner akin to low-pass whole genome sequencing, identified significant peaks of copy gain and losses using GISTIC218 (Figures 1 and 2A, Online Supplementary Figure S4, Online Supplementary Tables S8 and S9), and defined the likely targets of these CNA by integrative analysis of matched gene expression profiling data from 290 tumors (Figure 2B, C, Online Supplementary Figure S4, Online Supplementary Tables S10 and S11). This identified known CNA targets, including but not limited to deletion of TNFAIP3 (6q24.2),22 ATM (11q22.3),23 B2M (15q15.5),24 and PTEN (10q23.21),25 and copy gain of REL and BCL11A (2p15), and TCF4 (18q23).26 In addition, we identified novel targets such as deletion of IBTK (6q14.1), UBE3A (11q22.1) and FBXO25 (8p23.3), and copy gain of ATF7 (12q13.13), UCHL5 (1q31.3), and KMT2A (11q23.3). Importantly, the frequency of DNA CNA in the target genes identified by next-generation sequencing-based analysis significantly correlated with those derived from single nucleotide polymorphism microarray-based measurements in independent cohorts of BL, DLBCL, FL and MCL tumors from previously published studies6,20,26-30 (Online Supplementary Figure S5), providing validation for the accuracy of this approach. The CNA peaks, defined as the smallest and most statistically significant region, included multiple genes that were significantly mutated (Figure 2D) as well as other genes for which we detected mutations at lower frequencies that were not significant by 692

MutSig2CV (POU2AF1, TP53BP1, FAS, PTEN). Deletions of ATM, B2M, BIRC3 and TNFRSF14 significantly co-associated with mutations of these genes, suggesting that these are complementary mechanisms contributing to biallelic inactivation.

Conserved functional hallmarks of B-cell non-Hodgkin lymphoma To understand key hallmarks that are deregulated by genetic alterations, we performed hypergeometric enrichment analysis of genes targeted by recurrent mutations and DNA CNA using DAVID31 (Online Supplementary Table S12). This revealed a significant enrichment of multiple overlapping gene sets that could be summarized into hallmark processes associated with epigenetics and transcription (Figure 3A), apoptosis and proliferation (Figure 3B), signaling (Figure 3C), and ubiquitination (Figure 3D). One or more genes within these hallmarks was altered in the majority (>50%) of tumors from each of the four major histologies included in this study. Genes annotated in epigenetic-associated gene sets were altered in 72%, 70%, 93% and 50% of BL, DLBCL, FL, and MCL, respectively, whereas genes annotated in transcription-associated gene sets were altered in 94%, 91%, 95% and 88% of BL, DLBCL, FL, and MCL, respectively. However, there is an extremely high degree of functional overlap between epigenetics and transcriptional regulation, as well as overlapping gene set annotations for many genes, leading us to consider these categories collectively as a single hallmark. Collectively, genes involved in epigenetics and transcription were mutated in 94% of BL, 92% of DLBCL, 96% of FL and 89% of MCL, and included those that encode proteins that catalyze post-translational modifications of histones (KMT2D, CREBBP, EZH2, EP300, WHSC1, ASHL1L, KMT2A), components of the BAF chromatin remodeling complex (ARID1A, SMARCA4, BCL7A, BCL11A), linker histones (HIST1H1E, HIST1H1C, HIST1H1B), and transcription factors (BCL6, IRF4, IRF8, TCF3, TCF4, MYC, REL, PAX5, POU2AF2, FOXO1, CIITA). Genes with a role in signaling included those involved in B-cell receptor signaling (CD79B, ID3, TCF3, TCF4, RFTN1), NFκB (TNFAIP3, CARD11, NFKBIE), NOTCH (NOTCH1, NOTCH2), JAK/STAT (SOCS1, STAT6), PI3K/mTOR (FOXO1, ATP6V1B2, APT6AP1) and G-protein signaling (GNA13, GNAI2). The CD79A and BCL10 genes were also mutated at a lower frequency that was not significant by MutSig2CV (Online Supplementary Figure S6A, B). Among these, the RFTN1 gene (Online Supplementary Figure S6C) is a novel recurrently mutated gene that was mutated in 7.4% of DLBCL and encodes a lipid raft protein that is critical for B-cell receptor signaling.32 Deregulation of the ubiquitin proteasome system is important in many cancers,33 but is not a well-defined hallmark of B-NHL. However, one or more genes with a role in regulating ubiquitination was genetically altered in 61% of BL, 79% of DLBCL, 61% of FL and 82% of MCL (Figure 3D). These included previously described genetic alterations such as amplification of MDM2,34 deletions of TNFAIP3,35 CUL4A,36 and RPL5,36 and mutations of KLHL6,37 DTX1,38 UBR5,39 SOCS1,40 and BIRC3.6 In addition, we identified novel targets such as recurrent deletions of IBTK, a negative regulator of Bruton tyrosine kinase,41 and somatic mutation of CDC27 in 14% of MCL, which encodes an E3 ligase for CCND1.42 Therefore, common hallmark processes are targeted by genetic alterations in the majority of haematologica | 2022; 107(3)

Patterns of genetic alterations in B-NHL

Figure 1. Recurrently mutated genes in subtypes of B-cell non-Hodgkin lymphoma. An oncoplot shows significantly mutated genes, DNA copy number alterations (CNA) and translocations (Tx.) across our cohort of 685 B-cell non-Hodgkin lymphoma tumors. Mutation types and frequencies are summarized for each gene/CNA on the right, and the mutational burden for each case is shown at the top. DHL: double-hit lymphoma; THL: triple-hit lymphoma; HGBL-NOS: high-grade B-cell lymphoma not otherwise specified; BL: Burkitt lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; MCL: mantle cell lymphoma.

haematologica | 2022; 107(3)

693

M.C.J. Ma et al.

Figure 2. Structural alterations in subtypes of B-cell non-Hodgkin lymphoma. (A) A circos plot shows translocations of MYC (purple), BCL2 (orange) and BCL6 (green) genes, and GISTIC tracks of DNA copy number gains (red) and losses (blue). (B, C) Volcano plots of integrative analysis results showing the changes in gene expression of genes within peaks of DNA copy number gain (B) or loss (C). A positive T-test score indicates increased expression in tumors with a given copy number alteration, and vice versa. Significantly expressed genes with the correct directionality are highlighted in the shaded areas. (D) Oncoplots show the overlap of structural alterations and mutations that target the same genes. P-values are derived from a Fisher exact test (ns: not significant).

major B-NHL subtypes, including genes with a role in regulating protein ubiquitination.

Subtype-specific patterns of genetic alterations We formally tested the over- or under-representation of recurrent genetic alterations in each of the four subtypes with more than 100 samples (BL, DLBCL, FL, MCL), compared to all other tumors in the study (Figure 4, Online Supplementary Table S13). We observed some interesting patterns within hallmark characteristics that differ between subtypes. An illustrative example of this is the alternative avenues for BAF complex perturbation between different histologies (Figure 5). Specifically, mutations of the SMARCA4 (aka. BRG1) component of the ATPase module were significantly enriched in BL (24%) compared to other subtypes (4%, Q-value <0.001), while mutations of the BCL7A component of the ATPase module were significantly enriched in FL (11%) compared to other subtypes (4%, Qvalue=0.007). In contrast, mutations of ARID1A were frequent in both BL (19%) and FL (15%), and DNA copy number gains of BCL11A were frequent in both DLBCL (28%) and FL (22%). The BAF complex is therefore a target of recurrent genetic alterations, as previously suggested,43 but the manner in which this complex is perturbed varies between B-NHL subtypes (Figure 5). Similar disease-specific patterns were also observed for signaling genes; for example, TCF3 and ID3 have important functions in normal germinal center B cells (GCB),44 but mutations of these genes are specifically enriched within BL and are rarely found in the other GCB-derived malignancies, DLBCL and FL. Similarly, the ATP6AP1 and ATP6V1B2 genes that function in mTOR signaling45,46 are specifically mutated in FL, and the DUSP2 gene which inactivates ERK1/247 and 694

STAT348 is specifically mutated in DLBCL. The disease-specific patterns of genetic alterations therefore reveal subtle but important differences in how each subtype of B-NHL perturbs hallmark features.

Clusters of co-associated genomic alterations in subtypes of B-cell non-Hodgkin lymphoma We next defined how each genetic alteration co-associated with or mutually excluded other genetic alterations by pairwise assessments using false-discovery rate (FDR)-corrected Fisher tests (Online Supplementary Table S14). A matrix of the transformed FDR Q-values (-logQ) was used for unsupervised hierarchical clustering to identify clusters of co-associated genetic alterations. Together with patterns of disease-specificity, unsupervised clustering revealed clear groupings of co-associated events for BL, DLBCL, FL and MCL (Figure 4). We identified a single cluster of significantly co-associated genetic alterations that was specifically enriched in BL (Cluster 1), including mutations and translocations of MYC, and mutations of CCND3, SMARCA4, TCF3 and ID3 which have been previously reported in BL.4 A single cluster was significantly enriched in MCL (Cluster 7), with a high frequency of ATM mutations and deletions, as well as other DNA CNA. Other mutations that were not significantly co-associated were also enriched in MCL (Cluster 6), such as those in WHSC1, NOTCH1, NOTCH2, BCOR and UBR5, although statistical assessment of coassociation may be hampered in this context by the low frequencies of mutations within these genes. A single cluster was also enriched in FL (Cluster 4), with a high prevalence of KMT2D, BCL2, CREBBP, EZH2 and TNFRSF14 mutations and BCL2 translocations. The genes within Cluster 4 also significantly overlapped with the previously reported haematologica | 2022; 107(3)

Patterns of genetic alterations in B-NHL

C3, EZB and BCL2 clusters from prior whole exome sequencing studies of DLBCL3,49,50 (Fisher test P-values: P=0.0006, P=0.0148 and P=0.0173, respectively). Two clusters (Clusters 2 and 3) were enriched in DLBCL, with lower frequencies of mutations in a larger number of genes, in line with the genetic heterogeneity of this disease.3,4 Cluster 2 includes co-associated genetic alterations that overlapped with the previously described C5, MCD, and MYD88 clusters3,4 (Fisher P-values: P=0.0004, P=0.0002 and P=0.0007, respectively) including CD79B, MYD88 and TBL1XR1 mutations. Genes within Cluster 3 overlapped in a statistically significant manner with those in the previously

described C4 and SOCS1/SGK1 clusters (Fisher P-values: P=0.0002 and P=0.0074, respectively), including SGK1, TET2, SOCS1 and histone H1 genes. We also identified a cluster consisting of TP53 mutations and multiple CNA (Cluster 5) similar to the genetically complex C2/A53 subtype reported in DLBCL;3,49 however, the overlap of features within these clusters could not be formally assessed due to differing annotations. The CNA captured in this cluster were variably represented across B-NHL subtypes, but were most frequent in DLBCL. B-NHL subtypes therefore harbor characteristic clusters of co-associated genetic alterations that likely cooperate in disease etiology.

A Figure 3. Figure continued on the following page.

haematologica | 2022; 107(3)

695

M.C.J. Ma et al.

Figure 3. Functional enrichment of targets of somatic mutations and DNA copy number alterations. Genes targeted by somatic mutation and/or DNA copy number alteration were evaluated for enrichment in curated gene sets, and significant gene sets subsequently grouped according to overlapping gene set membership and functional similarity. In addition to genes assigned by DAVID (purple), some genes were manually curated into hallmark processes by literature review of their function (pink). Enriched gene sets could be summarized into four major hallmark processes, including (A) epigenetic and transcriptional control of gene expression, (B) regulation of apoptosis and proliferation, (C) regulation of signaling pathway activity, and (D) regulation of protein ubiquitination. The frequency of each genetic alteration is shown for each of the four major histologies included in this study, and the fraction of tumors in each histology bearing genetic alterations of one or more of the genes is summarized by a pie graph at the bottom for each hallmark. BL: Burkitt lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; MCL: mantle cell lymphoma; HMT, histone methyltransferase. HAT, histone acetyltransferase. DDR, DNA damage response. BCR, B-cell receptor.

Combinations of genetic alterations define molecular subtypes of B-cell non-Hodgkin lymphoma Our data have revealed statistical enrichment of individual genetic alterations in subtypes of B-NHL, and pairwise relationships between different genetic alterations that define clusters of subtype-specific events. To validate and expand upon these observations we leveraged gene expression microarray data from 284 tumors that underwent pathology review and were profiled as part of prior studies.10-12 We utilized BL, DHL, HGBL-NOS and DLBCL tumors to perform classification into molecularly-defined BL (mBL) and non-mBL using a Bayesian classifier with previously described marker genes,51 and subclassified nonmBL into activated B-cell (ABC)-like and GCB-like subtypes as we have described previously26 (Online Supplementary Figure S7). We evaluated the frequency of cumulative (≥2) genetic alterations within each cluster among mBL, ABClike DLBCL, GCB-like DLBCL, FL and MCL (Figure 6). This showed that Cluster 1 genetic alterations that were individually enriched in BL are cumulatively acquired in mBL, with 87% of tumors having ≥2 of these alterations compared to only 22% of GCB-like DLBCL. Similarly, Cluster 4 and Cluster 7 alterations were cumulatively acquired in 77% and 72% of molecularly-annotated FL and MCL, respectively. Cluster 4 mutations were also cumulatively acquired in 51% of GCB-like DLBCL, likely capturing the C3/EZB/BCL2 subtype that has genetic similarities to FL.3,4,50 696

Furthermore, Cluster 2 and Cluster 4 alterations were cumulatively acquired in 58% of ABC-like DLBCL and 60% of GCB-like DLBCL, respectively, further supporting their respective overlap with the C5/MCD/MYD88 and C4/ST2 subtypes of DLBCL. CNA within Cluster 5 were cumulatively acquired at high but variable frequencies in all of the subtypes, but showed subtype-specific patterns within this cluster such as higher frequencies of 18q21 and 18q23 gains in ABC-like DLBCL, and higher frequencies of chromosome 7 gains in GCB-like DLBCL and FL. B-NHL tumors therefore cumulatively acquire co-associated sets of genetic alterations in a manner that is characteristically associated with histologically- and molecularly-defined subsets of disease.

Discussion By performing cross-sectional genomic profiling of a large cohort of tumors, we have developed a resource of genes and functional hallmarks that are recurrently targeted by genetic alterations in B-NHL, and have shown that the cumulative acquisition of combinations of genetic alterations are characteristic of histological and molecular subtypes of disease. Some of the functional hallmarks that we identified have been previously appreciated, with a few exceptions. For example, the mutation of genes with roles haematologica | 2022; 107(3)

Patterns of genetic alterations in B-NHL

Figure 4. Subtype-specific clusters of co-occurring genetic alterations. The frequency (bar graph) and over- or under-representation (blue to red scale) of mutations and structural alterations is shown on the left for Burkitt lymphoma, diffuse large B-cell lymphoma, follicular lymphoma and mantle cell lymphoma. The correlation matrix of co-associated (green) and mutually exclusive (purple) relationships was clustered to identify seven groups of co-occurring genetic alterations that were predominantly over-represented in a single subtype of B-cell non-Hodgkin lymphoma. BL: Burkitt lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; MCL: mantle cell lymphoma

haematologica | 2022; 107(3)

697

M.C.J. Ma et al.

Figure 5. Subtype-specific patterns of BAF complex mutations. (A) An oncoplot shows the frequency of genetic alterations in genes that encode components of the BAF complex. (B) A schematic of the BAF complex shows recurrently mutated genes, ARID1A, SMARCA4 and BCL7A, and the BCL11A gene that is targeted by 2p15 DNA copy number gains. (C-E) Lollipop plots show the distribution of mutations in the BAF components ARID1A (C), SMARCA4 (D), and BCL7A (E). (F) A heatplot shows the location of chromosome 2p DNA copy number gains (red) ordered from highest DNA copy number (top) to lowest (bottom, copy number = 2.2). The BCL11A gene is in the peak focal copy gain. BL: Burkitt lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; MCL: mantle cell lymphoma.

in epigenetic and transcriptional control of gene expression are known to be a hallmark of FL52 and we observed that 96% of FL tumors possessed mutations in one or more of the genes in this category. However, mutations within these genes were also observed in the majority of BL, DLBCL and MCL tumors, highlighting the conservation of this functional hallmark across B-NHL subtypes. There are subtype-specific patterns of chromatin-modifying gene alterations, such as those that we highlighted for BAF complex mutations, but we suggest that the genetic deregulation of epigenetic and transcriptional control of gene expression should be considered a general hallmark of BNHL. In addition, we suggest that the deregulation of the ubiquitin proteasome system is a hallmark of B-NHL that requires further investigation. Mutations in genes such as KLHL637 and UBR539 have recently been shown to play an important role in B-cell lymphoma, while the roles of other frequently mutated genes such as DTX1 and SOCS1 have not yet been functionally dissected. Furthermore, while the nature of AID-driven mutations in genes such as DTX1 and SOCS1 remain to be defined, other genes that are recurrently mutated by AID such as BCL7A53 and linker histone genes54 have been shown to play driving roles in lymphomagenesis. Genetic deregulation of the ubiquitin proteasome system has the potential to influence the activity or abundance of a range of substrate proteins, and represents a current gap in our knowledge of B-NHL etiology. 698

The role of cooperative interactions between co-occurring genetic alterations is also an emerging field that requires further investigation. These interactions are not uncommon in cancer,55 and have been recently highlighted in DLBCL,3,4 but our data show that they are pervasive and characteristic features of the B-NHL genetic landscape. Cooperation between co-associated genetic alterations identified in this study requires formal validation in cell lines and/or animal models. However, there are many instances in which co-occurring genetic alterations that we observed have already been shown to cooperate in lymphomagenesis. In addition to the aforementioned example of MYD88 and CD79B mutations, transgenic mouse models of Ezh2 activating mutations or conditional deletion of Crebbp or Kmt2d have shown that these events are not alone sufficient for lymphomagenesis.56-61 We and others have observed a co-association between mutation of these genes and BCL2 translocations,14,62 and the addition of a Bcl2 transgene to these murine models indeed promoted lymphoma at a significantly higher rate than that observed with the Bcl2 transgene alone.56-61 These genetic alterations are therefore significantly more lymphomagenic in combination than they are alone, which provides proof of principle that a cooperative relationship exists between these cooccurring genetic alterations. Future studies focusing on other co-occurring mutations, such as MYC translocation and SMARCA4 mutation in BL, CREBBP and KMT2D haematologica | 2022; 107(3)

haematologica | 2022; 107(3)

Figure 6. Cumulative acquisition of co-occurring genetic alterations. (A) An oncoplot shows the presence or absence of genetic alterations according to their clusters of co-association in molecularly-defined Burkitt lymphoma, activated Bcell-like diffuse large B-cell lymphoma (DLBCL), germinal center B-cell-like DLBCL, follicular lymphoma and mantle cell lymphoma with available gene expression microarray data. Shading shows histological or molecular subtypes with ≥50% of tumors bearing ≥2 genetic alterations within a given cluster. (B) Bar plots shows the frequency of tumors with ≥2 genetic alterations from each cluster. mBL: molecularly-defined Burkitt lymphoma; ABC: activated B-cell-like DLBCL; COO: cell of origin; GCB: germinal center B-cell-like DLBCL; FL: follicular lymphoma; MCL: mantle cell lymphoma.

Patterns of genetic alterations in B-NHL

699

M.C.J. Ma et al.

mutation in FL, TCF4 copy gain and MYD88 mutation in DLBCL, and ATM mutation and RPL5 deletion in MCL, should therefore be performed to further explore these concepts and define their underlying functional relationship. We suggest that combinations of genetic alterations are likely to recapitulate the biology of B-NHL more accurately than are single gene models, and may reveal contextually different functional roles of genetic alterations depending on the co-occurring events. The caveats regarding this study include the targeted nature of the LymphoSeq platform which may preclude consideration of a subset of important genes, the lack of germline DNA for the majority of samples that may lead to a small number of germ-line variants being falsely assigned as somatic, and the sample size for any given histological subtype being below that required to identify genes that are mutated at low frequency. Nonetheless, these data represent the first broad cross-sectional analysis of multiple histological and molecular subtypes of B-NHL using the same methodology and provide a framework of functional hallmarks and co-occurring genetic alterations that are enriched within these subtypes of B-NHL. These functional hallmarks are genetically perturbed in the majority of B-NHL, but our cross-sectional approach enabled us to elucidate subtype-specific preferences for genetic alterations within each functional hallmark. Furthermore, the subtype-specific clusters of co-occurring genetic alterations likely represent cooperative interactions that underpin the biology of different subtypes of B-NHL. These combinations identify opportunities for moving from single-allele to multiallele designs in cell line or animal models to better understand the molecular etiology of B-NHL subtypes. Together,

References 1. Swerdlow SH, Campo E, Pileri SA, et al. The 2016 revision of the World Health Organization classification of lymphoid neoplasms. Blood. 2016;127(20):2375-2390. 2. Armitage JO, Gascoyne RD, Lunning MA, Cavalli F. Non-Hodgkin lymphoma. Lancet. 2017;390(10091):298-310. 3. Chapuy B, Stewart C, Dunford AJ, et al. Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat Med. 2018;24(5):679-690. 4. Schmitz R, Wright GW, Huang DW, et al. Genetics and pathogenesis of diffuse large B-cell lymphoma. N Engl J Med. 2018;378(15):1396-1407. 5. Reddy A, Zhang J, Davis NS, et al. Genetic and functional drivers of diffuse large B cell lymphoma. Cell. 2017;171(2):481-494. 6. Bea S, Valdes-Mas R, Navarro A, et al. Landscape of somatic mutations and clonal evolution in mantle cell lymphoma. Proc Natl Acad Sci U S A. 2013;110(45):1825018255. 7. Zhang J, Jima D, Moffitt AB, et al. The genomic landscape of mantle cell lymphoma is related to the epigenetically determined chromatin state of normal B cells. Blood. 2014;123(19):2988-2996. 8. Green MR, Alizadeh AA. Common progenitor cells in mature B-cell malignancies: implications for therapy. Curr Opin Hematol. 2014;21(4):333-340. 9. Wang JQ, Jeelall YS, Humburg P, et al. Synergistic cooperation and crosstalk between MYD88L265P and mutations that

700

these hallmarks and clusters of co-associated genetic alterations represent processes that are potentially druggable with targeted therapies,63-66 but that are likely influenced in a non-binary fashion by different combinations of genetic alterations. Deciphering the relationships between complex sets of genetic alterations and targetable dependencies will be a next step towards developing new rationally targeted therapeutic strategies in B-NHL. Disclosures No conflicts of interest to disclose. Contributions MCJM and ST performed experiments, analyzed data and wrote the manuscript. AB analyzed data. TH, HY, QD, DM, KH, NJ, JS, and SG performed experiments. AA, LS, MD, CC, JT, DP, KMV, MAL, ARS, BJC, RB, SN, LN, RED, JW, SP, MG, DS, KB, JI, SR, and AM provided samples and/or clinical data. MRG conceived and supervised the study, performed experiments, analyzed the data and wrote the manuscript. All authors reviewed and approved the manuscript. Funding This research was supported by NCI R01CA201380 (MRG), the Nebraska Department of Health and Human Services (LB506 2016-17; MRG), and NCI cancer center support grants to the University of Texas MD Anderson Cancer Center (P30 CA016672) and the Fred & Pamela Buffet Cancer Center (P30 CA036727). HY is supported by a Fellow award from the Leukemia and Lymphoma Society. MRG is supported by a Scholar award from the Leukemia and Lymphoma Society and an Andrew Sabin Family Foundation Fellow award.

dysregulate CD79B and surface IgM. J Exp Med. 2017;214(9):2759-2776. 10. Lenz G, Wright G, Dave SS, et al. Stromal gene signatures in large-B-cell lymphomas. N Engl J Med. 2008;359(22):2313-2323. 11. Dave SS, Fu K, Wright GW, et al. Molecular diagnosis of Burkitt's lymphoma. N Engl J Med. 2006;354(23):2431-2442. 12. Iqbal J, Shen Y, Liu Y, et al. Genome-wide miRNA profiling of mantle cell lymphoma reveals a distinct subgroup with poor prognosis. Blood. 2012;119(21):4939-4948. 13. Bouska A, Bi C, Lone W, et al. Adult high grade B-cell lymphoma with Burkitt lymphoma signature: genomic features and potential therapeutic targets. Blood. 2017;130(16):1819-1831. 14. Green MR, Kihira S, Liu CL, et al. Mutations in early follicular lymphoma progenitors are associated with suppressed antigen presentation. Proc Natl Acad Sci U S A. 2015;112(10):E1116-1125. 15. Lek M, Karczewski KJ, Minikel EV, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016;536(7616):285-291. 16. Lawrence MS, Stojanov P, Polak P, et al. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature. 2013;499(7457):214-218. 17. Kuilman T, Velds A, Kemper K, et al. CopywriteR: DNA copy number detection from off-target sequence data. Genome Biol. 2015;16(1):49. 18. Mermel CH, Schumacher SE, Hill B, Meyerson ML, Beroukhim R, Getz G. GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic

copy-number alteration in human cancers. Genome Biol. 2011;12(4):R41. 19. Newman AM, Bratman SV, Stehr H, et al. FACTERA: a practical method for the discovery of genomic rearrangements at breakpoint resolution. Bioinformatics. 2014;30(23):3390-3393. 20. Bouska A, Bi C, Lone W, et al. Adult highgrade B-cell lymphoma with Burkitt lymphoma signature: genomic features and potential therapeutic targets. Blood. 2017;130(16):1819-1831. 21. Akasaka T, Lossos IS, Levy R. BCL6 gene translocation in follicular lymphoma: a harbinger of eventual transformation to diffuse aggressive lymphoma. Blood. 2003;102(4): 1443-1448. 22. Kato M, Sanada M, Kato I, et al. Frequent inactivation of A20 in B-cell lymphomas. Nature. 2009;459(7247):712-716. 23. Greiner TC, Dasgupta C, Ho VV, et al. Mutation and genomic deletion status of ataxia telangiectasia mutated (ATM) and p53 confer specific gene expression profiles in mantle cell lymphoma. Proc Natl Acad Sci U S A. 2006;103(7):2352-2357. 24. Challa-Malladi M, Lieu YK, Califano O, et al. Combined genetic inactivation of beta2Microglobulin and CD58 reveals frequent escape from immune recognition in diffuse large B cell lymphoma. Cancer Cell. 2011;20(6):728-740. 25. Pfeifer M, Grau M, Lenze D, et al. PTEN loss defines a PI3K/AKT pathway-dependent germinal center subtype of diffuse large B-cell lymphoma. Proc Natl Acad Sci U S A. 2013;110(30):12420-12425. 26. Jain N, Hartert K, Tadros S, et al. Targetable

haematologica | 2022; 107(3)

Patterns of genetic alterations in B-NHL

genetic alterations of TCF4 (E2-2) drive immunoglobulin expression in the activated B-cell subtype of diffuse large B-cell lymphoma. Sci Transl Med. 2019;11(497): eeav5599. 27. Kim D, Fiske BP, Birsoy K, et al. SHMT2 drives glioma cell survival in ischaemia but imposes a dependence on glycine clearance. Nature. 2015;520(7547):363-367. 28. Rosenwald A, Wright G, Wiestner A, et al. The proliferation gene expression signature is a quantitative integrator of oncogenic events that predicts survival in mantle cell lymphoma. Cancer Cell. 2003;3(2):185-197. 29. Salaverria I, Royo C, Carvajal-Cuenca A, et al. CCND2 rearrangements are the most frequent genetic events in cyclin D1(-) mantle cell lymphoma. Blood. 2013;121(8): 1394-1402. 30. Green MR, Vicente-Duenas C, RomeroCamarero I, et al. Transient expression of Bcl6 is sufficient for oncogenic function and induction of mature B-cell lymphoma. Nat Commun. 2014;5:3904. 31. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009;4(1):44-57. 32. Saeki K, Miura Y, Aki D, Kurosaki T, Yoshimura A. The B cell-specific major raft protein, Raftlin, is necessary for the integrity of lipid raft and BCR signal transduction. EMBO J. 2003;22(12):3015-3026. 33. Senft D, Qi J, Ronai ZA. Ubiquitin ligases in oncogenic transformation and cancer therapy. Nat Rev Cancer. 2018;18(2):69-88. 34. Monti S, Chapuy B, Takeyama K, et al. Integrative analysis reveals an outcomeassociated and targetable pattern of p53 and cell cycle deregulation in diffuse large B cell lymphoma. Cancer Cell. 2012;22(3): 359-372. 35. Honma K, Tsuzuki S, Nakagawa M, et al. TNFAIP3/A20 functions as a novel tumor suppressor gene in several subtypes of nonHodgkin lymphomas. Blood. 2009;114(12): 2467-2475. 36. Hartmann EM, Campo E, Wright G, et al. Pathway discovery in mantle cell lymphoma by integrated analysis of high-resolution gene expression and copy number profiling. Blood. 2010;116(6):953-961. 37. Choi J, Lee K, Ingvarsdottir K, et al. Loss of KLHL6 promotes diffuse large B-cell lymphoma growth and survival by stabilizing the mRNA decay factor roquin2. Nat Cell Biol. 2018;20(5):586-596. 38. Meriranta L, Pasanen A, Louhimo R, et al. Deltex-1 mutations predict poor survival in diffuse large B-cell lymphoma. Haematologica. 2017;102(5):e195-e198. 39. Swenson SA, Gilbreath TJ, Vahle H, et al.

haematologica | 2022; 107(3)

UBR5 HECT domain mutations identified in mantle cell lymphoma control maturation of B cells. Blood. 2020;136(3):299-312. 40. Mottok A, Renne C, Seifert M, et al. Inactivating SOCS1 mutations are caused by aberrant somatic hypermutation and restricted to a subset of B-cell lymphoma entities. Blood. 2009;114(20):4503-4506. 41. Liu W, Quinto I, Chen X, et al. Direct inhibition of Bruton's tyrosine kinase by IBtk, a Btk-binding protein. Nat Immunol. 2001;2 (10):939-946. 42. Pawar SA, Sarkar TR, Balamurugan K, et al. C/EBP{delta} targets cyclin D1 for proteasome-mediated degradation via induction of CDC27/APC3 expression. Proc Natl Acad Sci U S A. 2010;107(20):9210-9215. 43. Krysiak K, Gomez F, White BS, et al. Recurrent somatic mutations affecting Bcell receptor signaling pathway genes in follicular lymphoma. Blood. 2017;129(4): 473-483. 44. Gloury R, Zotos D, Zuidscherwoude M, et al. Dynamic changes in Id3 and E-protein activity orchestrate germinal center and plasma cell development. J Exp Med. 2016;213(6):1095-1111. 45. Okosun J, Wolfson RL, Wang J, et al. Recurrent mTORC1-activating RRAGC mutations in follicular lymphoma. Nat Genet. 2016;48(2):183-188. 46. Wang F, Gatica D, Ying ZX, et al. Follicular lymphoma-associated mutations in vacuolar ATPase ATP6V1B2 activate autophagic flux and mTOR. J Clin Invest. 2019;130: 1626-1640. 47. Hu J, Li L, Chen H, et al. MiR-361-3p regulates ERK1/2-induced EMT via DUSP2 mRNA degradation in pancreatic ductal adenocarcinoma. Cell Death Dis. 2018;9 (8):807. 48. Lu D, Liu L, Ji X, et al. The phosphatase DUSP2 controls the activity of the transcription activator STAT3 and regulates TH17 differentiation. Nat Immunol. 2015;16(12):1263-1273. 49. Wright GW, Huang DW, Phelan JD, et al. A probabilistic classification tool for genetic subtypes of diffuse large B cell lymphoma with therapeutic implications. Cancer Cell. 2020;37(4):551-568. 50. Lacy SE, Barrans SL, Beer PA, et al. Targeted sequencing in DLBCL, molecular subtypes, and outcomes: a Haematological Malignancy Research Network report. Blood. 2020;135(20):1759-1771. 51. Hummel M, Bentink S, Berger H, et al. A biologic definition of Burkitt's lymphoma from transcriptional and genomic profiling. N Engl J Med. 2006;354(23):2419-2430. 52. Green MR. Chromatin modifying gene mutations in follicular lymphoma. Blood.

2018;131(6):595-604. 53. Balinas-Gavira C, Rodriguez MI, Andrades A, et al. Frequent mutations in the aminoterminal domain of BCL7A impair its tumor suppressor role in DLBCL. Leukemia. 2020;34(10):2722-2735. 54. Yusufova N, Kloetgen A, Teater M, et al. Histone H1 loss drives lymphoma by disrupting 3D chromatin architecture. Nature. 2021;589(7841):299-305. 55. Ashworth A, Lord CJ, Reis-Filho JS. Genetic interactions in cancer progression and treatment. Cell. 2011;145(1):30-38. 56. Beguelin W, Popovic R, Teater M, et al. EZH2 is required for germinal center formation and somatic EZH2 mutations promote lymphoid transformation. Cancer Cell. 2013;23(5):677-692. 57. Garcia-Ramirez I, Tadros S, GonzalezHerrero I, et al. Crebbp loss cooperates with Bcl2 over-expression to promote lymphoma in mice. Blood. 2017;129(19):26452656. 58. Zhang J, Vlasevska S, Wells VA, et al. The Crebbp acetyltransferase is a haploinsufficient tumor suppressor in B cell lymphoma. Cancer Discov. 2017;7(3):322-337. 59. Jiang Y, Ortega-Molina A, Geng H, et al. CREBBP inactivation promotes the development of HDAC3-dependent lymphomas. Cancer Discov. 2017;7(1):38-53. 60. 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. 61. 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):11991208. 62. 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. 63. Sermer D, Pasqualucci L, Wendel HG, Melnick A, Younes A. Emerging epigeneticmodulating therapies in lymphoma. Nat Rev Clin Oncol. 2019;16(8):494-507. 64. Shen M, Schmitt S, Buac D, Dou QP. Targeting the ubiquitin-proteasome system for cancer therapy. Expert Opin Ther Targets. 2013;17(9):1091-1108. 65. Merino D, Kelly GL, Lessene G, Wei AH, Roberts AW, Strasser A. BH3-mimetic drugs: blazing the trail for new cancer medicines. Cancer Cell. 2018;34(6):879-891. 66. Roschewski M, Staudt LM, Wilson WH. Diffuse large B-cell lymphoma-treatment approaches in the molecular era. Nat Rev Clin Oncol. 2014;11(1):12-23.

701

ARTICLE Ferrata Storti Foundation

Non-Hodgkin Lymphoma

Deregulation of JAK2 signaling underlies primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma Armando N. Bastidas Torres,1 Davy Cats,2 Jacoba J. Out-Luiting,1 Daniele Fanoni,3 Hailiang Mei,2 Luigia Venegoni,3 Rein Willemze,1 Maarten H. Vermeer,1 Emilio Berti4 and Cornelis P. Tensen1 Department of Dermatology, Leiden University Medical Center, Leiden, the Netherlands; 2Sequencing Analysis Support Core, Leiden University Medical Center, Leiden, the Netherlands; 3Department of Pathophysiology and Transplantation, University of Milan, Milan, Italy and 4Department of Dermatology, Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Milan, Italy 1

Haematologica 2022 Volume 107(3):702-714

ABSTRACT

Correspondence: CORNELIS P. TENSEN c.p.tensen@lumc.nl Received: October 19, 2020. Accepted: March 24, 2021. Pre-published: April 1, 2021. https://doi.org/10.3324/haematol.2020.274506

702

rimary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL) is a rare variant of cutaneous T-cell lymphoma with an aggressive clinical course and a very poor prognosis. Until now, neither a systematic characterization of genetic alterations driving pcAECyTCL has been performed, nor effective therapeutic regimes for patients have been defined. Here, we present the first highresolution genetic characterization of pcAECyTCL by using wholegenome and RNA sequencing. Our study provides a comprehensive description of genetic alterations (i.e., genomic rearrangements, copy number alterations and small-scale mutations) with pathogenic relevance in this lymphoma, including events that recurrently impact genes with important roles in the cell cycle, chromatin regulation and the JAKSTAT pathway. In particular, we show that mutually exclusive structural alterations involving JAK2 and SH2B3 predominantly underlie pcAECyTCL. In line with the genomic data, transcriptome analysis uncovered upregulation of the cell cycle, JAK2 signaling, NF-κB signaling and a high inflammatory response in this cancer. Functional studies confirmed oncogenicity of JAK2 fusions identified in pcAECyTCL and their sensitivity to JAK inhibitor treatment. Our findings strongly suggest that overactive JAK2 signaling is a central driver of pcAECyTCL, and consequently, patients with this neoplasm would likely benefit from therapy with JAK2 inhibitors such as Food and Drug Adminstration-approved ruxolitinib.

Introduction Primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL) is a rare variant of cutaneous T-cell lymphoma (CTCL) still regarded as a provisional entity by the World Health Organization (WHO) and characterized by an abrupt onset and a highly aggressive clinical course.1,2 pcAECyTCL presents primarily in the skin with widespread plaques and tumors, often with hemorrhagic ulcerations and necrosis; however, dissemination to extracutaneous sites (especially the central nervous system, lungs, oral cavity and testes) is not uncommon.3,4 Malignant T cells causing pcAECyTCL typically express CD3, CD7, CD8, CD45RA, TCR-βF1, T-BET and one or more cytotoxic markers (e.g., granzyme B, perforin, TIA-1), which strongly suggests that neoplastic cells in this lymphoma derive from CD8+ T cells.2,3 Effective therapeutic regimes for pcAECyTCL are currently lacking, and consequently, patients have a poor prognosis with a median overall survival of 12 months.1 Thus far the study of the pathogenetic basis of this malignancy has been marginal due to its rarity. Recently, a study performed on tumors from 20 patients defined the copy number alteration (CNA) profile of pcAECyTCL by using array-based comparative genomic hybridization,5 and before this, two clinical case reports

haematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL

included the evaluation of CNA in single patients by using array-based methods as well.6,7 Recurrent CNA uncovered by these studies include losses within 1p, 9p, 13q and 16p as well as gains within 7q, 8q and 17q, with loss of the region containing CDKN2A/B being the most frequent CNA.5 However, aside from the aforementioned chromosomal imbalances, causative genetic changes in pcAECyTCL remain unknown. Here, we present the first high-resolution genomic analysis of pcAECyTCL using whole-genome sequencing (WGS) and RNA sequencing (RNA-seq). We describe for the first time a number of genomic rearrangements, CNA and smallscale mutations with pathogenic relevance in this lymphoma. In particular, our results suggest that overactivation of JAK2 signaling due to oncogenic changes in JAK2 and SH2B3, two genes with key roles in this signaling pathway, underlie predominantly pcAECyTCL. These findings have important implications for patient standard of care.

Methods Patient selection and sequencing Frozen tumor biopsies (≥70% tumor cells) from 12 patients with pcAECyTCL (Online Supplementary Figure S1; Online Supplementary Table S1) were subjected to WGS. Six samples of this cohort (i.e., AEC2-4/6/8/12) were additionally subjected to RNA-seq. Sequencing, data processing and DNA/RNA analyses are described in the Online Supplementary Appendix (Online Supplementary Figures S2 and S3; Online Supplementary Tables S2 to S9). Diagnosis was performed by an expert panel of dermatologists/pathologists in accordance with the WHO-EORTC classification for primary cutaneous lymphomas.1,2 Patient material was approved by the Institutional Review Boards of the Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico and Leiden University Medical Center. Informed consent was obtained from patients in accordance with the declaration of Helsinki.

Validation of structural genomic alterations and small-scale mutations Select rearrangements, interstitial deletions and single nucleotide variants (SNV) were validated by Sanger sequencing, droplet digital polymerase chain reaction (ddPCR) and/or fluorescence in situ hybridization (FISH). Details of the validation experiments are included in the Online Supplementary Appendix (Online Supplementary Figures S4 to S9; Online Supplementary Table S10).

Immunohistochemistry Formalin-fixed paraffin-embedded (FFPE) tissue sections were immunohistochemically stained with primary antibodies against phospho-STAT3 (Cell Signaling Technology, Cat.No. 9145) or phospho-STAT5 (Cell Signaling Technology, Cat.No. 9359) using Dako REAL detect system (Dako, Cat.No. K5005), counterstained in Mayer’s hematoxylin solution and coverslipped using Vectamount (Vector Laboratories, Cat.No. H5000).

Primers, templates and vectors used for fusion gene construction are detailed in the Online Supplementary Appendix (Online Supplementary Tables S11 to S13). Lentiviral particles were produced in HEK-293T cells, quantified by p24 enzyme-linked immunosorbent assay (ELISA), and transduced into Ba/F3 cells at MOI-9 with lipofectamine. Successfully transduced Ba/F3 cells were selected with puromycin (2.5 mg/mL) for 3 days.

Ba/F3 cell viability and inhibitor assays Cell viability of parental and transduced Ba/F3 cells was determined 7 days after IL3 withdrawal by MTT assay (Promega, Cat.No. G4000). Inhibitor assays were performed by treating IL3-independent Ba/F3 cells expressing fusion genes with ruxolitinib or AZD1480 at seven different concentrations for 72 hours and measuring cell viability by MTT assay.

Western blots The effect of JAK1/2 inhibitors ruxolitinib and AZD1480 on JAK2 and STAT5 phosphorylation was evaluated by western blotting. Cells were washed to remove traces of serum and incubated with inhibitor for 90 minutes. Cells were lysed in SDS lysis buffer containing protease inhibitors and separated by SDSPAGE. Antibodies employed were anti-JAK2 (Abcam, Cat.No. ab108596), anti-phospho-JAK2 (Cell Signaling Technology, Cat.No. 3776), anti-STAT5 (Cell Signaling Technology, Cat.No. 94205), anti-phospho-STAT5 (Cell Signaling Technology, Cat.No. 9351) and anti-GAPDH (Cell Signaling Technology, Cat.No. 2118).

Results JAK2 fusions are prominent in a complex landscape of rearrangements The analysis revealed a heterogeneous and complex landscape of genomic rearrangements (total events, 426; range, 10-65; mean/patient ± standard deviation [SD], 36±21) (Figure 1; Figure 2A; Online Supplementary Figure S3). Fifty-three percent of events were interchromosomal (range/patient, 27-80%) (Figure 2B). The majority of rearrangements (77%) disrupted either one or two annotated genes, while the rest (23%) disrupted nongenic regions (Figure 2C). Four patients, AEC6, AEC7, AEC10 and AEC12, displayed complex rearrangements (chromothripsis/chromoplexy-like) affecting chromosomes 13, 10, 1/9/12 and 4, respectively (Figure 2D; Online Supplementary Figure S3 and S10). We observed a total of 305 rearranged genes, 59 of which are implicated in neoplasms at present (Online Supplementary Table S14). Gene ontology analysis revealed that rearranged genes encode principally (n =91 of 305) proteins with roles in signal transduction (i.e., hydrolases, transferases, enzyme modulators, receptors) and transcriptional regulation (i.e., transcription factors, chromatin regulators) (Figure 2E; Online Supplementary Tables S15 and S16). Out of seventeen recurrently rearranged genes detected in our cohort (n =2 or 3) (Online Supplementary Table S17), six are established cancer genes with important functions in the regulation of the cell cycle (i.e., MYC, RB1), chromatin remodeling (i.e., BAZ1A) and the JAK-STAT pathway (i.e., JAK2, PTPRC, SH2B3) (Figure 1; Online Supplementary Figures S4 and S5). The JAK-STAT pathway, a frequent driver of hematological neoplasms, was the only cytokine-elicited signal transduction pathway impacted by rearrangements in pcAECyTCL. Fusion genes involving JAK2 were detected genes

patients

Cell culture, fusion gene construction and viral transduction Ba/F3 cells (DSMZ, Cat.No. ACC-300) were used for functional experiments. Parental Ba/F3 cells were cultured in RPMI-1640 (10% heat-inactivated fetal bovine serum, 10 ng/mL interleukin3 [IL3]) at 37° C with 5% CO in a humidified atmosphere. JAK2 fusions (i.e., TFG-JAK2, PCM1-JAK2, KHDRBS1-JAK2) and control genes (i.e., eGFP, TFG-MET) were constructed and inserted into a lentiviral vector using the method described by Lu et al.8 2

haematologica | 2022; 107(3)

703

A.N. Bastidas Torres et al.

Figure 1. Landscape of genomic rearrangements in pcAECyTCL. Circos plot showing 426 genomic rearrangements detected in twelve pcAECyTCL genomes by wholegenome sequencing (WGS). The outer ring shows rearranged genes with established roles in cancer. The area at the center of the plot contains arcs representing interchromosomal (blue) and intrachromosomal (red) events. The ring between the gene labels and the arcs contains human chromosome ideograms arranged circularly end to end. pcAECyTCL: primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma.

in three of 12 patients (i.e., AEC4: KHDRBS1-JAK2; AEC9: PCM1-JAK2; AEC11: TFG-JAK2) (Figure 2F to H). These events fused the tyrosine kinase domain of JAK2 with one or more oligo/dimerization domains from the fusion partner (i.e., AEC4: Qua1 domain, AEC9: coiled-coil domains, AEC11: PB1 domain) (Figure 2F to H). The resulting chimeric proteins are predicted to self-oligo/dimerize and become activated without the need of cytokine-mediated receptor stimulation, ultimately overactivating JAK2 signaling. Of note, two of three patients carrying JAK2 fusions carried MYC fusions as well (i.e., AEC4: ACTBMYC, AEC9: NPM1-MYC) (Online Supplementary Figure S4). Interestingly, apart from acquiring the ability to selfactivate, JAK2 fusions under the transcriptional control of their partner’s promoter may also experience augmented expression in comparison to wild-type JAK2, as evidenced in patient AEC4 (Figure 2I). In contrast, rearrangements involving PTPRC and SH2B3, each observed in two of 12 patients, disrupted these two negative regulators of the JAK-STAT pathway.

JAK2 signaling inhibitor SH2B3 is focally deleted in pcAECyTCL The most frequent broad chromosomal imbalances (npatients≥4; >3 Mb) were deletions within 1p, 8p, 9q, 10p, 11q and 13q and gains within 7q, 8q, 17q and 21q (Figure 3A; Figure 4A). We identified 24 recurrent focal (≤ 3 Mb) minimal common regions (MCR) shared by CNA between patients (npatients≥3; deletions: 19, gains: 5) (Online Supplementary Table S5), 12 of which contained cancer genes predominantly involved in the cell cycle, chromatin regulation and the JAK-STAT pathway (Figure 4A). The most common focal MCR involving cancer genes 704

was deletion at 9p21.3 (10 patients), which included cell cycle regulators CDKN2A/B (Online Supplementary Figure S7). Of note, CDKN2A/B were found to be inactivated by interstitial deletions, unbalanced rearrangements, SNV and presumably even the action of long non-coding RNA ANRIL (CDKN2B-AS1)9 (Figure 4A; Online Supplementary Figure S11). Five of 12 patients had deletions at 1p36.11 and 13q14.11, which contained chromatin remodeler ARID1A and candidate cancer gene ELF1,10 respectively. Deletions at 1p36.32-p36.33, 1p36.22 and 12q24.12, observed in four of 12 patients, involved tumor suppressors TNFRSF14, MIR34AHG and SH2B3, respectively. Finally, three of 12 patients had deletions at 4q13.1-q13.2, 10p11.22, 11q14.2, 16p13.13 and 19p13.3, which contained tumor suppressors EPHA5, EPC1 (alongside ZEB1), EED, SOCS1 and STK11, respectively. On the other hand, gain at 17q21.31 (four patients), which enclosed ETV4, was the only recurrent (npatients ≥3) focal gain containing a cancer gene. Remarkably, deletions at 12q24.12 were strikingly focal in all affected patients (20 Kb – 457 Kb), leading to the loss of one or more functional domains of SH2B3 (i.e., DD, PH, SH2 domains) in these individuals (Figure 3B and C; Online Supplementary Figure S5). SH2B3 (LNK) encodes an adaptor protein that antagonizes JAK2 signaling as part of a negative feedback loop in various hematopoietic cell types (e.g., erythroid progenitors, hematopoietic stem cells, megakaryocytes, pre-B cells, etc.) by suppressing the kinase activity of JAK2 through its SH2 domain.11 Of note, structural alterations involving JAK2 and SH2B3 were mutually exclusive in our cohort, affecting altogether seven of 12 patients. In addition, we investigated the possibility of SH2B3 silencing by promoter hypermethylation haematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL

Fii

Gii

Hii

Fiii

Giii

Hiii

Iii

Iiii

Figure 2. Legend on following page.

haematologica | 2022; 107(3)

705

A.N. Bastidas Torres et al.

Figure 2. JAK2 fusions are recurrent in a complex landscape of rearrangements. (A) Number of genomic rearrangements per patient. The distribution of inter- and intrachromosomal rearrangements per patient is shown too. (B) Distribution of inter- and intrachromosomal rearrangements (cohort). (C) Distribution of genomic rearrangements based on the type of DNA sequences (genic, nongenic) involved in the event (cohort). (D) Circos plot showing a chromoplexy-like event in patient AEC10 that mediated the loss of multiple genomic regions in chromosomes 1, 9 and 12, several of which enclosed established tumor suppressor genes. (E) Distribution of rearranged genes according to the protein class their encoded proteins belong to. (F, G and H) Genomic rearrangements generated self-activating JAK2 fusions in pcAECyTCL as evidenced in patients (F) AEC4, (G) AEC9 and (H) AEC11. (i) Circos plots showing interchromosomal rearrangements involving chromosome 9 in patients with pcAECyTCL. JAK2 rearrangements were the common denominator between chromosome 9 events observed in these individuals. (ii) Validation of translocation breakpoints at JAK2 by Sanger sequencing in pcAECyTCL patients. Breakpoints occurred between exon 16 and exon 17 in all cases. (iii) Rearrangements involving JAK2 led to the formation of fusion genes encoding the tyrosine kinase domain of JAK2 and the oligo/dimerization domains of the fusion partners (KHDRBS1: Qua1 domain, PCM1: coiled-coil domains, TFG: PB1 domain), conferring the resulting chimeric protein the ability to self-activate. (I) In addition to acquiring self-activation ability, JAK2 fusions can also experience increased expression in comparison to wild-type JAK2. (i) Image of break-apart fluorescence in situ hybridization (FISH) analysis showing a JAK2 rearrangement in patient AEC4. Scale bar, 10 mm. (ii) Active expression of fusion gene KHDRBS1-JAK2 in patient AEC4 was detected by RNA sequencing (chimeric reads shown in diagram). (iii) Plot showing mean read coverage across all exons of JAK2 in patient AEC4. RNA expression between exon 17 and exon 25, the part of JAK2 under the transcriptional control of KHDRBS1’s promoter and encoding its tyrosine kinase domain, is considerably higher compared to RNA expression between exon 1 and exon 16. The red line indicates the breakpoint position. pcAECyTCL: primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma.

Table 1. Identical variants in JAK and STAT proteins reported in other hematological malignancies.

Gene

Variant

Type

JAK2 JAK3 JAK3 STAT5B STAT5B STAT5B STAT5B

p.L393V p.M511I p.R657W p.S434L p.N642H p.Y665F p.P702S

Germline Somatic Somatic Somatic Somatic Somatic Somatic

Neoplasm PV15,28 T-PLL, AML, JMML, NKTCL13,27,28 T-ALL12,28 T-ALL17 T-ALL, T-LGL, T-PLL, NKTCL, EATL14,27,28 T-LGL, T-PLL, ALCL ALK–, NKTCL14,27,28 T-PLL16

Effect

Functionally validated

Affected patient

Slightly HS GoF GoF Unknown GoF GoF Possibly GoF

Yes15 Yes13 Yes12 No Yes14 Yes14 No

AEC3 AEC12 AEC5 AEC8 AEC1, AEC6 AEC7 AEC2

GoF: gain-of-function; HS: hypersensitive; ALCL ALK–: ALK– anaplastic large cell lymphoma; AML: acute myeloid leukemia; EATL: enteropathy-associated T-cell lymphoma; JMML: juvenile myelomonocytic leukemia; PV: polycythemia vera; T-ALL: T-cell acute lymphoblastic leukemia; T-LGL: T-cell large granular lymphocytic leukemia; T-PLL: T-cell prolymphocytic leukemia; NKTCL: extranodal natural killer T-cell lymphoma.

in our patients using methylation-specific melting curve analysis (MS-MCA) and found no evidence of this inactivation mechanism (Online Supplementary Figure S9).

Pathogenic small-scale mutations in JAK-STAT pathway genes predominate in pcAECyTCL The discovery of recurrent structural alterations affecting principally genes involved in the cell cycle, chromatin regulation and the JAK-STAT pathway (via JAK2) prompted us to search for pathogenic indels and SNV in exonic sequences of genes with roles in the aforesaid cellular processes and additional signal transduction pathways (i.e., MAPK, NF-κB, PI-3-K/Akt and T-cell receptor [TCR] pathways) (Online Supplementary Table S7). Besides the seven patients with structural alterations impacting the JAK2-SH2B3 signaling axis, four additional patients were found to carry bona fide gain-of-function SNV either in JAK3 (i.e., AEC5: p.R657W12; AEC12: p.M511I13) or STAT5B (i.e., AEC1 and AEC6: p.N642H14). Also, patient AEC3 bore a germline SNV in JAK2 (p.L393V15) which has been reported to render JAK2 slightly hypersensitive to cytokine stimulation (EPO ligand) (Figures 4A, 5A and B; Table 1). Moreover, two patients with JAK2 fusions and three patients with SH2B3 deletions also carried SNV affecting conserved residues in STAT3 (i.e., AEC4: p.H19R; AEC9: p.G604A) and STAT5B (i.e., AEC2: p.P702S16; AEC7: p.Y665F14; AEC8: p.S434L17), respectively (Online Supplementary Figure S6). Similarly, three patients carrying (putative) gain-of-function SNV in JAK or STAT genes also had indels leading to premature stop codons either in SH2B3 (i.e., AEC2: p.L201Sfs*78; AEC6: p.V35Afs*154) or SOCS1 (i.e., AEC5: p.S71Rfs*14) (Figures 4A and B, 5A). Overall, nine of 12 patients had either structural or small-scale genetic alterations impacting the JAK2-SH2B3 signaling axis whereas the remaining three patients carried 706

pathogenic indels/SNV in other JAK-STAT pathway genes (Figures 4A and B, 5C). In addition, cancer genes involved in the cell cycle (i.e., TP53) and chromatin regulation (i.e., ARID1A, KMT2D, NCOR1) were found to be recurrently impacted either by truncating mutations (i.e., nonsense, frameshift) or SNV predicted as deleterious (Figure 4A). We also observed 34 additional patient-specific small-scale mutations of unknown significance in reputable cancer genes (Online Supplementary Table S6).

Transcriptome analysis uncovers upregulation of JAK2 signaling in pcAECyTCL At present it is widely accepted that malignant T cells in pcAECyTCL derive from CD8+ T cells;1 however, to date no specific CD8+ T-cell subtype has been proposed as the cell of origin of this lymphoma. Since a hallmark of malignant T cells in pcAECyTCL is their distinctive epidermotropism,3,4 we compared gene expression in pcAECyTCL with gene expression in normal skin-resident CD8+ T cells, which are characterized by a marked preferential tropism to the epidermal layer of the skin.18 This analysis identified 1,603 differentially expressed (DE) genes (1,076 upregulated, 527 downregulated, false discovery rate [FDR] <0.01) in the disease (Figure 6A; Online Supplementary Table S8). We next performed gene set enrichment analysis (GSEA) using annotated gene sets from MSigDB to search for deregulated pathways/ processes. Upregulated canonical signaling profiles included the JAK-STAT pathway (via STAT3, and to a lesser extent, via STAT5) and the TNF-α/NF-κB pathway. In addition, pcAECyTCL was characterized by the upregulation of the cell cycle (i.e., E2F targets, G2/M checkpoint, mitotic spindle) and high inflammatory response (Figure 6B; Online Supplementary Table S18). Further examination of DE genes involved in the JAKSTAT pathway revealed that JAK2 signaling was specifihaematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL

cally deregulated in pcAECyTCL. Upregulated genes included among others JAK2 itself, components of type I and II cytokine receptors that signal predominantly via JAK2 (i.e., IFNGR2, IL12RB2) and established enhancers of JAK2 signaling (i.e., PTPN11, SH2B1). In contrast, downregulated JAK-STAT pathway genes included PTPRC and genes encoding receptors exclusively associated with signal transduction via JAK1, JAK3 or TYK2 (Figure 6C). In order to validate JAK-STAT pathway activation in pcAECyTCL, we investigated the presence of activated STAT proteins (pSTAT3 and pSTAT5) by immunohistochemistry (IHC) in eight sequenced patients with available tumor tissue. Robust activation of JAK-STAT signal-

ing (via STAT3, STAT5 or both) was confirmed in all evaluated patients (Figure 6D).

JAK2 fusions identified in pcAECyTCL confer cytokineindependent survival ability to cells In order to validate the predicted effects of the JAK2 fusions found in pcAECyTCL (i.e., PCM1-JAK2, KHDRBS1-JAK2, TFG-JAK2) on cell survival, we engineered these fusion genes into murine pro-B Ba/F3 cells which die in the absence of exogenous IL3 (Figure 7A). Because self-oligo/dimerizing JAK2 fusions were predicted to activate downstream STAT proteins without the need of upstream cues elicited by cytokine stimulation,

Figure 3. Landscape of copy number alterations reveals focal SH2B3 inactivation in pcAECyTCL. (A) Human chromosome ideograms showing regions of gain and loss detected through whole-genome sequencing (WGS) in twelve primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL) genomes. Blue bars to the right of the chromosomes depict regions of loss whereas red bars to the left of the chromosomes depict regions of gain. (B) Deletions at 12q24.12 (blue bars), where SH2B3 resides, were the most focal (<500 Kb) copy number alteration (CNA) events in pcAECyTCL. Inactivation of SH2B3 was mediated by interstitial deletions and unbalanced rearrangements. Breakpoints of structural alterations at 12q24.12 in all affected patients were validated by Sanger sequencing. Genomic coordinates of breakpoints according to reference genome GRCh38. Arrows indicate the direction towards which genomic coordinate numbers increase. Plus (+) and minus (-) signs specify strand polarity. CTX: interchromosomal rearrangement; ITX: intrachromosomal rearrangement; iDel: interstitial deletion. (C) Copy number losses involving SH2B3 in patients with pcAECyTCL were validated by droplet digital polymerase chain reaction. Ctrl: control CD8+ T cells.

haematologica | 2022; 107(3)

707

A.N. Bastidas Torres et al.

Figure 4. Distribution of recurrent chromosomal rearrangements, copy number alterations and deleterious indels/single nucleotide variants in pcAECyTCL. (A) First panel: recurrent chromosomal rearrangements impacting cancer genes. Second panel: recurrent large-scale copy number alterations (CNA) (>3 Mb). Third panel: focal minimal common regions (MCR) (≤ 3 Mb) shared by CNA; bona fide cancer genes residing within focal MCR are specified. Fourth panel: Indels and single nucleotide variants (SNV) in cancer genes leading to protein truncations, reported as pathogenic in literature or predicted as disease-causing (SIFT and PolyPhen-2). Only genes altered in more than one patient are indicated. CTX: interchromosomal rearrangement; ITX: intrachromosomal rearrangement. (B) Circos plots showing genetic alterations in patients AEC6 and AEC9. Despite inter-patient heterogeneity, molecular abnormalities affecting genes with roles in the cell cycle, chromatin regulation and JAK2 signaling (genes in light purple) were recurrent in primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL).

these chimeric proteins were expected to increase survival of Ba/F3 cells in the absence of IL3. Seven days after IL3 withdrawal, survival of Ba/F3 cells expressing each of the three engineered JAK2 fusions was noticeably higher (P<0.05, student’s t-test) than survival of the parental Ba/F3 cells (wild-type control) and Ba/F3 cells expressing eGFP (negative control) (Figure 7B). We next evaluated the effect of FDA-approved JAK1/2 inhibitor ruxolitinib on each of the three IL3-independent cell lines carrying JAK2 fusions. Ruxolitinib inhibited the growth of all cell lines in a dose-dependent manner (Figure 7C and D) with half maximal inhibitory concentration (IC ) values in the low nanomolar range (9–15 nM), in concordance with the reported inhibitory activity of this drug.19 Since fusion partners PCM1 and TFG have extensively been proven by others to confer chimeric kinases (including JAK2) the ability to trans-autophosphorylate via self-oligo/dimerization,20-25 we carried on further validation with JAK2 fusion containing novel kinase fusion partner KHDRBS1. For extra verification, we treated Ba/F3 cells expressing KHDRBS1-JAK2 with inhibitor AZD1480, which has higher specificity for JAK2 than ruxolitinib,26 and confirmed that cytokine-independent survival of 50

708

these cells depends on JAK2 signaling (Figure 7D). Finally, we corroborated by western blotting that growth inhibition exerted by ruxolitinib and AZD1480 was accompanied by a dose-dependent inhibition of JAK2 and STAT5 phosphorylation in Ba/F3 cells driven by KHDRBS1-JAK2 (Figure 7D).

Discussion This study describes the first high-resolution genetic profiling of pcAECyTCL using next-generation sequencing. The landscape of structural genomic alterations of pcAECyTCL was characterized by considerable genomic instability and inter-patient heterogeneity. Most rearrangements (328 of 426) identified in pcAECyTCL disrupted annotated genes, and approximately one-third of all rearranged genes (91 of 305) were found to play roles in signal transduction and transcriptional regulation. In addition, four of 12 patients experienced chromothripsis/chromoplexy-like events which mediated the deletion of relevant tumor suppressors (e.g., CDKN2C, CHD5, FAS, PTEN, etc.). In full agreement with previously published haematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL

Figure 5. Small-scale mutations in genes of the JAK-STAT pathway are predominant in pcAECyTCL. (A) Diagrams showing deleterious indels and single nucleotide variants (SNV) in JAK2, JAK3, STAT3, STAT5B, SH2B3 and SOCS1 detected in primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL) by whole-genome sequencing (Table 1). (B) Sanger chromatograms confirming presence of bona fide pathogenic SNV in patients with pcAECyTCL. (C) Summary of genetic alterations affecting members of the JAK-STAT pathway in pcAECyTCL.

haematologica | 2022; 107(3)

709

A.N. Bastidas Torres et al.

Figure 6. Legend on following page.

710

haematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL Figure 6. RNA sequencing supports upregulation of JAK2 signaling in pcAECyTCL. (A) Heat map showing 1,603 differentially expressed genes (1,076 upregulated, 527 downregulated, false discovery rate [FDR] <0.01) in pcAECyTCL when compared to skin-resident CD8+ T cells. (B) Gene set enrichment analysis (GSEA) uncovered upregulation of the JAK-STAT pathway, the cell cycle (E2F targets, G2/M checkpoint, mitotic spindle), the NF-κB pathway and high inflammatory response in pcAECyTCL. NES: normalized enrichment score; FDR q-value: false discovery rate q-value. (See the Online Supplementary Table S18 for a complete list of GSEA signatures) (C) Examination of differentially expressed genes involved in the JAK-STAT pathway revealed that JAK2 itself, enhancers of JAK2 signaling and components of cytokine receptors that signal predominantly via JAK2 are upregulated in primary cutaneous CD8+ aggressive epidermotropic cytotoxic T-cell lymphoma (pcAECyTCL). (D) Activation of the JAK-STAT pathway (via STAT3 and/or STAT5) in pcAECyTCL was confirmed by immunohistochemistry (IHC) on tumor tissue from sequenced patients (i.e., AEC1/3/5-10). Neoplastic cells exhibited activated STAT3 and/or STAT5 in the nucleus. Normal skin (control) displayed STAT3 activation in keratinocytes and endothelial cells as well as STAT5 activation in melanocytes and endothelial cells. Scale bar, 50 mm.

Dii

Figure 7. Oncogenicity validation of JAK2 fusions identified in pcAECyTCL. (A) Expression of JAK2 fusions in transduced Ba/F3 cells was verified by reverse transcriptase polymerase chain reaction. M: molecular-weight marker; B: Fusion DNA in backbone (positive control); F: cDNA from Ba/F3 cells transduced with JAK2 fusion gene; N: cDNA from Ba/F3 cells transduced with eGFP gene (negative control); NTC: non-template control (H O). (B) Violin plots showing viability of Ba/F3 cells expressing fusion genes KHDRBS1-JAK2, PCM1-JAK2 or TFG-JAK2 seven days after interleukin-3 (IL3) withdrawal (mean OD, n=3). Viability of all cell lines expressing JAK2 fusions was noticeably higher (P<0.05, student’s t-test) compared to wild-type and negative control cells. Control samples: parental Ba/F3 cells (wild-type control), Ba/F3 cells expressing eGFP (negative control), Ba/F3 cells expressing fusion gene TFG-MET (positive control). *P<0.05; ****P<0.0001. (C) Dose-response curves of Ba/F3 cells expressing PCM1-JAK2 (half maximal inhibitory concentration [IC ]=11 nM) or TFG-JAK2 (IC =9 nM) when exposed to various concentrations of JAK1/2 inhibitor ruxolitinib (mean OD; error bars, standard deviation [SD], n=3). (D) Validation experiments with JAK2 fusion containing novel kinase fusion partner KHDRBS1. (i) Dose-response curves of Ba/F3 cells expressing KHDRBS1-JAK2 when exposed to various concentrations of JAK1/2 inhibitors ruxolitinib (IC =15 nM) or AZD1480 (IC =40 nM) (mean OD; error bars, SD, n=3). (ii) Western blot analysis of Ba/F3 cells expressing KHDRBS1-JAK2 showed a dose-response reduction in phosphorylation of JAK2 and STAT5 with increasing concentrations of ruxolitinib or AZD1480. 2

haematologica | 2022; 107(3)

711

A.N. Bastidas Torres et al.

data,5 we found that gains within 7q and 17q as well as losses within 1p and 13q were the most common largescale chromosomal imbalances. Our analysis identified a group of bona fide oncogenes and tumor suppressors with central roles in the cell cycle (i.e., CDKN2A/B, MIR34AHG, MYC, RB1, TP53), chromatin regulation (i.e., ARID1A, BAZ1A, EED, EPC1, KMT2D, NCOR1, ZEB1) and the JAK-STAT pathway (i.e., JAK2, JAK3, PTPRC, SH2B3, SOCS1, STAT3, STAT5B) whose copy number, sequence organization and/or nucleotide composition were found to be recurrently altered in our pcAECyTCL cohort. Genetic alterations involving JAK-STAT pathway genes were the most notable due to their predominance, likely proliferationpromoting effects and known causative roles in hematological cancers. A subset of SNV affecting JAK-STAT pathway genes in pcAECyTCL have confirmed oncogenic activity in other T-cell lymphomas (Table 1).27,28 JAK2 and SH2B3, which govern the activation and termination of JAK2 signaling in normal hematopoietic cells, respectively, underwent mutually exclusive alterations in nine of 12 patients from our cohort. Mutations in these two genes are associated with BCR-ABL1– myeloproliferative neoplasms (MPN), a group of myeloid malignancies driven by overactive JAK2 signaling.11,29 However, unlike BCR-ABL1– MPN where JAK2 and SH2B3 are mainly affected by pathogenic SNV and/or indels, these two genes experienced predominantly structural alterations in pcAECyTCL. On one hand, JAK2 formed fusion genes encoding self-activating chimeras. On the other hand, SH2B3 was inactivated by focal interstitial deletions and unbalanced rearrangements. The previous suggests that pcAECyTCL is mainly driven by aberrant JAK2 signaling resulting from oncogenic changes leading to JAK2 overactivation or SH2B3 deficiency. Moreover, we demonstrated that JAK2 fusions found in pcAECyTCL promote cytokine-independent cell survival and their oncogenic activity was shown to be successfully inhibited by ruxolitinib. Of note, JAK2 fusions functionally analogous to the ones identified in pcAECyTCL have been previously described and confirmed as oncogenic in other hematological malignancies (e.g., B- and T-cell acute leukemias, MPN).20 Also, recurrent deletion of SH2B3 has been reported in an aggressive subtype of B-cell precursor acute lymphoblastic leukemia.30 We found that genetic alterations involving JAK2 and SH2B3 co-existed with SNV predicted or confirmed as pathogenic in STAT3 or STAT5B in six of nine affected patients. Previous functional in vitro studies with cell lines have suggested that mutations in STAT proteins (especially dimerization-enhancing SNV) observed in T-cell lymphomas operate as aberrant amplifiers of upstream signals from cytokines, overactive receptors or deregulated JAK proteins, rather than as initiators of deregulated JAK-STAT signaling themselves.27 In this scenario, mutations in STAT3/5B would contribute to pcAECyTCL progression by making the pre-existing overactive JAK2 signaling more robust and severe. However, recent evidence derived from a murine model suggests that at least gainof-function mutation STAT5B (p.N642H), one of the most common pathogenic SNV in human T-cell lymphomas,31 is sufficient by itself to promote the development of neoplasms primarily derived from mature CD8+ T cells.32 Remarkably, malignant CD8+ T cells in these animals showed preferential migration to the skin, lung 712

and the central nervous system, all of which are commonly affected body sites in pcAECyTCL.32 Consistent with this evidence, patient AEC1, the only individual in our cohort who had a single JAK-STAT pathway gene mutated, carried the STAT5B (p.N642H) mutation biallelically. Several pathogenetic features found in pcAECyTCL have also been reported in mycosis fungoides (MF) and/or Sézary syndrome (SS). Genetic alterations common to pcAECyTCL, MF and SS include recurrent inactivation of ARID1A, CDKN2A, CDKN2B, NCOR1, PTPRC, TP53 and ZEB1 as well as occasional activating mutations in JAK3, MYC and STAT3.33-39 Other genetic alterations observed in pcAECyTCL have been found before either in MF (e.g., SOCS1 and STK11 inactivation) or SS (e.g., RB1 inactivation, STAT5B mutations).33,34,39 By contrast, JAK2 fusions and SH2B3 inactivation have not been reported in other CTCL variants to the best of our knowledge and appear to be characteristic features of pcAECyTCL. In agreement with the recurrent genetic alterations involving the JAK2-SH2B3 signaling axis observed in pcAECyTCL, transcriptome analysis revealed upregulation of JAK2 signaling. SH2B1 and PTPN11, which encode two proteins with the ability to enhance JAK2 signaling,40,41 stood out among upregulated JAK-STAT pathway genes. Adaptor protein SH2B1 has been proven to bind to JAK2 and stimulate its kinase activity.42 Similarly, phosphatase PTPN11 (SHP-2) has been shown to positively regulate JAK2-mediated STAT5 phosphorylation.43 In contrast, phosphatase PTPRC (CD45), whose expression has been shown to attenuate JAK2 signaling in hematopoietic and lymphoma cells,44,45 was downregulated in pcAECyTCL. Yet, the exact molecular interactions underlying the action of these three regulators of JAK2 signaling remain to be fully elucidated. Transcriptome analysis also revealed upregulation of the cell cycle, the TNF-α/NF-κB pathway and a high inflammatory response in pcAECyTCL. Notably, the co-activation (crosstalk) of JAK-STAT signaling (especially via STAT3) and NF-κB signaling is a well-documented phenomenon in cancer, and it has been shown to promote a pro-oncogenic inflammatory microenvironment in the tumor.46 For instance, aberrant JAK2 signaling (via STAT3) in MPN promotes chromatin changes that induce NF-κB signaling; and the resulting combined action of these two pathways, appear to drive the characteristic chronic inflammatory state observed in these neoplasms.47 Our data, in line with the previous, suggest that co-activation of JAK2 signaling and NF-κB signaling operates in pcAECyTCL as well, and their joint action might be responsible for the inflammatory state detected in pcAECyTCL tumors. Taken together, our findings strongly suggest that overactivation of JAK2 signaling plays a pivotal role in the pathogenesis of pcAECyTCL. Therefore, patients with this lymphoma would likely benefit from treatment with JAK2 inhibitors (e.g., FDA-approved ruxolitinib). In addition, the potential combination of JAK2 inhibitors with NF-κB inhibitors (e.g. bortezomib,48 dimethyl fumarate49) represents an attractive possibility since targeting both pathways might have a synergistic effect and reduce the chance of resistance acquisition. Disclosures No conflicts of interest to disclose. haematologica | 2022; 107(3)

Deregulation of JAK2 signaling underlies pcAECyTCL

Contributions ANBT, DF, RW, MV, EB and CPT conceptualized and designed the project; ANBT and CPT wrote the manuscript; ANBT, DC and HM performed the bioinformatic analyses; ANBT, JO, DF and LV performed the experiments; ANBT and DC produced figures and tables; ANBT analyzed the results and interpreted the data; DF, LV, RW, MV and EB provided valuable biological specimens; ANBT, DC, JO, DF, HM, LV, RW, MV, EB, and CPT revised and approved the final manuscript.

References 1. Berti E, Gaulard P, Willemze R, et al. WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues 4th ed. Lyon: IARC; 2017. 2. Willemze R, Jaffe ES, Burg G, et al. WHOEORTC classification for cutaneous lymphomas. Blood. 2005;105(10):3768-3785. 3. Berti E, Tomasini D, Vermeer MH, Meijer CJLM, Alessi E, Willemze R. Primary cutaneous CD8-positive epidermotropic cytotoxic T cell lymphomas. Am J Pathol. 1999;155(2):483-492. 4. Guitart J, Martinez-Escala ME, Subtil A, et al. Primary cutaneous aggressive epidermotropic cytotoxic T-cell lymphomas: reappraisal of a provisional entity in the 2016 WHO classification of cutaneous lymphomas. Mod Pathol. 2017;30(5):761-772. 5. Fanoni D, Corti L, Alberti-Violetti S, et al. Array-based CGH of primary cutaneous CD8+ aggressive EPIDERMO-tropic cytotoxic T-cell lymphoma. Genes Chromosomes Cancer. 2018;57(12):622629. 6. Kato K, Oh Y, Takita J, et al. Molecular genetic and cytogenetic analysis of a primary cutaneous CD8-positive aggressive epidermotropic cytotoxic T-cell lymphoma. Int J Hematol. 2016;103(2):196201. 7. Tomasini C, Novelli M, Fanoni D, Berti EF. Erythema multiforme-like lesions in primary cutaneous aggressive cytotoxic epidermotropic CD8+ T-cell lymphoma: A diagnostic and therapeutic challenge. J Cutan Pathol. 2017;44(10):867-873. 8. Lu H, Villafane N, Dogruluk T, et al. Engineering and functional characterization of fusion genes identifies novel oncogenic drivers of cancer. Cancer Res. 2017;77(13):3502-3512. 9. Kotake Y, Nakagawa T, Kitagawa K, et al. Long non-coding RNA ANRIL is required for the PRC2 recruitment to and silencing of p15(INK4B) tumor suppressor gene. Oncogene. 2011;30(16):1956-1962. 10. Paczkowska J, Soloch N, Bodnar M, et al. Expression of ELF1, a lymphoid ETS domain-containing transcription factor, is recurrently lost in classical Hodgkin lymphoma. Br J Haematol. 2019;185(1):79-88. 11. Maslah N, Cassinat B, Verger E, Kiladjian JJ, Velazquez L. The role of LNK/SH2B3 genetic alterations in myeloproliferative neoplasms and other hematological disorders. Leukemia. 2017;31(8):1661-1670. 12. Degryse S, de Bock CE, Cox L, et al. JAK3 mutants transform hematopoietic cells through JAK1 activation, causing T-cell acute lymphoblastic leukemia in a mouse model. Blood. 2014;124(20):3092-3100. 13. Yamashita Y, Yuan J, Suetake I, et al. Arraybased genomic resequencing of human leukemia. Oncogene. 2010;29(25):37233731.

haematologica | 2022; 107(3)

Acknowledgements The authors thank Tim van Groningen and Yixin Luo for providing valuable technical support. Funding This study was funded by the Dutch Cancer Society (KWF, grant UL2013-6104) and Associazione Amici di Sabrina Fadini Onlus (A.S.F.O., grant 0800000-PR-LUMC).

14. Rajala HL, Eldfors S, Kuusanmaki H, et al. Discovery of somatic STAT5b mutations in large granular lymphocytic leukemia. Blood. 2013;121(22):4541-4550. 15. Lanikova L, Babosova O, Swierczek S, et al. Coexistence of gain-of-function JAK2 germ line mutations with JAK2V617F in polycythemia vera. Blood. 2016;128(18):22662270. 16. Andersson EI, Putzer S, Yadav B, et al. Discovery of novel drug sensitivities in TPLL by high-throughput ex vivo drug testing and mutation profiling. Leukemia. 2018;32(3):774-787. 17. Bandapalli OR, Schuessele S, Kunz JB, et al. The activating STAT5B N642H mutation is a common abnormality in pediatric T-cell acute lymphoblastic leukemia and confers a higher risk of relapse. Haematologica. 2014;99(10):e188-192. 18. Takamura S. Niches for the Long-term maintenance of tissue-resident memory T cells. Front Immunol. 2018;9:1214. 19. Quintas-Cardama A, Vaddi K, Liu P, et al. Preclinical characterization of the selective JAK1/2 inhibitor INCB018424: therapeutic implications for the treatment of myeloproliferative neoplasms. Blood. 2010;115(15): 3109-3117. 20. Smith CA, Fan G. The saga of JAK2 mutations and translocations in hematologic disorders: pathogenesis, diagnostic and therapeutic prospects, and revised World Health Organization diagnostic criteria for myeloproliferative neoplasms. Hum Pathol. 2008;39(6):795-810. 21. Reiter A, Walz C, Watmore A, et al. The t(8;9)(p22;p24) is a recurrent abnormality in chronic and acute leukemia that fuses PCM1 to JAK2. Cancer Res. 2005;65(7): 2662-2667. 22. Lierman E, Selleslag D, Smits S, Billiet J, Vandenberghe P. Ruxolitinib inhibits transforming JAK2 fusion proteins in vitro and induces complete cytogenetic remission in t(8;9)(p22;p24)/PCM1-JAK2-positive chronic eosinophilic leukemia. Blood. 2012;120(7):1529-1531. 23. Stransky N, Cerami E, Schalm S, Kim JL, Lengauer C. The landscape of kinase fusions in cancer. Nat Commun. 2014;5:4846. 24. Greco A, Fusetti L, Miranda C, et al. Role of the TFG N-terminus and coiled-coil domain in the transforming activity of the thyroid TRK-T3 oncogene. Oncogene. 1998;16(6):809-816. 25. Hernandez L, Bea S, Bellosillo B, et al. Diversity of genomic breakpoints in TFGALK translocations in anaplastic large cell lymphomas: identification of a new TFGALK(XL) chimeric gene with transforming activity. Am J Pathol. 2002;160(4):14871494. 26. Hedvat M, Huszar D, Herrmann A, et al. The JAK2 inhibitor AZD1480 potently blocks Stat3 signaling and oncogenesis in solid tumors. Cancer Cell. 2009;16(6):487-497.

27. Waldmann TA, Chen J. Disorders of the JAK/STAT pathway in T cell lymphoma pathogenesis: implications for immunotherapy. Annu Rev Immunol. 2017;35:533-550. 28. Hammarén HM, Virtanen AT, Raivola J, Silvennoinen O. The regulation of JAKs in cytokine signaling and its breakdown in disease. Cytokine. 2019;118:48-63. 29. Levine RL, Pardanani A, Tefferi A, Gilliland DG. Role of JAK2 in the pathogenesis and therapy of myeloproliferative disorders. Nat Rev Cancer. 2007;7(9):673-683. 30. Baughn LB, Meredith MM, Oseth L, Smolarek TA, Hirsch B. SH2B3 aberrations enriched in iAMP21 B lymphoblastic leukemia. Cancer Genet. 2018;226-227:3035. 31. Pham HTT, Maurer B, Prchal-Murphy M, et al. STAT5BN642H is a driver mutation for T cell neoplasia. J Clin Invest. 2018;128 (1):387-401. 32. de Araujo ED, Erdogan F, Neubauer HA, et al. Structural and functional consequences of the STAT5B(N642H) driver mutation. Nat Commun. 2019;10(1):2517. 33. Woollard WJ, Pullabhatla V, Lorenc A, et al. Candidate driver genes involved in genome maintenance and DNA repair in Sezary syndrome. Blood. 2016;127(26):3387-3397. 34. Wang L, Ni X, Covington KR, et al. Genomic profiling of Sezary syndrome identifies alterations of key T cell signaling and differentiation genes. Nat Genet. 2015;47(12):1426-1434. 35. McGirt LY, Jia P, Baerenwald DA, et al. Whole-genome sequencing reveals oncogenic mutations in mycosis fungoides. Blood. 2015;126(4):508-519. 36. Kiel MJ, Sahasrabuddhe AA, Rolland DC, et al. Genomic analyses reveal recurrent mutations in epigenetic modifiers and the JAK-STAT pathway in Sezary syndrome. Nat Commun. 2015;6:8470. 37. da Silva Almeida AC, Abate F, Khiabanian H, et al. The mutational landscape of cutaneous T cell lymphoma and Sezary syndrome. Nat Genet. 2015;47(12):14651470. 38. Choi J, Goh G, Walradt T, et al. Genomic landscape of cutaneous T cell lymphoma. Nat Genet. 2015;47(9):1011-1019. 39. Bastidas Torres AN, Cats D, Mei H, et al. Genomic analysis reveals recurrent deletion of JAK-STAT signaling inhibitors HNRNPK and SOCS1 in mycosis fungoides. Genes Chromosomes Cancer. 2018;57(12):653664. 40. Maures TJ, Kurzer JH, Carter-Su C. SH2B1 (SH2-B) and JAK2: a multifunctional adaptor protein and kinase made for each other. Trends Endocrinol Metab. 2007;18(1):38-45. 41. Liu X, Qu CK. Protein tyrosine phosphatase SHP-2 (PTPN11) in hematopoiesis and leukemogenesis. J Signal Transduct. 2011;2011:195239. 42. Rui L, Carter-Su C. Identification of SH2-

713

A.N. Bastidas Torres et al. bbeta as a potent cytoplasmic activator of the tyrosine kinase Janus kinase 2. Proc Natl Acad Sci U S A. 1999;96(13):7172-7177. 43. Ali S, Nouhi Z, Chughtai N, Ali S. SHP-2 regulates SOCS-1-mediated Janus kinase-2 ubiquitination/degradation downstream of the prolactin receptor. J Biol Chem. 2003;278(52):52021-52031. 44. Irie-Sasaki J, Sasaki T, Matsumoto W, et al. CD45 is a JAK phosphatase and negatively regulates cytokine receptor signalling. Nature. 2001;409(6818):349-354.

714

45. Wu L, Bijian K, Shen SH. CD45 recruits adapter protein DOK-1 and negatively regulates JAK-STAT signaling in hematopoietic cells. Mol Immunol. 2009;46(1112):2167-2177. 46. Yu H, Pardoll D, Jove R. STATs in cancer inflammation and immunity: a leading role for STAT3. Nat Rev Cancer. 2009;9(11): 798-809. 47. Kleppe M, Koche R, Zou L, et al. Dual targeting of oncogenic activation and inflammatory signaling increases therapeutic effi-

cacy in myeloproliferative neoplasms. Cancer Cell. 2018;33(1):29-43. 48. Zinzani PL, Musuraca G, Tani M, et al. Phase II trial of proteasome inhibitor bortezomib in patients with relapsed or refractory cutaneous T-cell lymphoma. J Clin Oncol. 2007;25(27):4293-4297. 49. Nicolay JP, Muller-Decker K, Schroeder A, et al. Dimethyl fumarate restores apoptosis sensitivity and inhibits tumor growth and metastasis in CTCL by targeting NFkappaB. Blood. 2016;128(6):805-815.

haematologica | 2022; 107(3)

ARTICLE

Non-Hodgkin Lymphoma

Humoral serological response to the BNT162b2 vaccine is abrogated in lymphoma patients within the first 12 months following treatment with anti-CD2O antibodies Ronit Gurion,1,2 Uri Rozovski,1,2 Gilad Itchaki,1,2 Anat Gafter-Gvili,1,2,3 Chiya Leibovitch,1,2 Pia Raanani,1,2 Haim Ben-Zvi,4 Moran Szwarcwort,5 Mor Taylor-Abigadol,6 Eldad J. Dann,6,7 Nurit Horesh,6 Tsofia Inbar,6 Inna Tzoran,6,7 Noa Lavi,6,7 Riva Fineman,6 Shimrit Ringelstein-Harlev6,7# and Netanel A. Horowitz6,7# Institute of Hematology, Davidoff Center, Rabin Medical Center, Beilinson Hospital, Petach Tikva; 2Sackler School of Medicine, Tel Aviv University, Tel Aviv; 3Medicine A, Rabin Medical Center, Beilinson Hospital, Petach Tikva; 4Microbiology Laboratory, Rabin Medical Center, Beilinson Hospital, Petach Tikva; 5Virology Laboratory, Rambam Health Care Campus, Haifa; 6Department of Hematology and Bone Marrow Transplantation, Rambam Health Care Campus, Haifa and 7The Ruth and Bruce Rappaport Faculty of Medicine, Technion – Israel Institute of Technology, Haifa, Israel 1

Ferrata Storti Foundation

Haematologica 2022 Volume 107(3):715-720

SR-H and NAH contributed equally as co-senior authors.

ABSTRACT

atients with lymphoma, especially those treated with anti-CD20 monoclonal antibodies, suffer high COVID-19-associated morbidity and mortality. The goal of this study was to assess the ability of lymphoma patients to generate a sufficient humoral response after two injections of BNT162b2 Pfizer vaccine and to identify factors influencing the response. Antibody titers were measured with the SARS-CoV-2 IgG II Quant (Abbott©) assay in blood samples drawn from lymphoma patients 4±2 weeks after the second dose of vaccine. The cutoff for a positive response was set at 50 AU/mL. Positive serological responses were observed in 51% of the 162 patients enrolled in this cross-sectional study. In a multivariate analysis, an interval of <12 months between the last anti-CD20 monoclonal antibody dose and the second vaccine dose (odds ratio=31.3 [95% confidence interval: 8.4-116.9], P<0.001) and presence of active lymphoma (odds ratio=4.2 (95% confidence interval: 2.18.2), P=0.006) were identified as negative response predictors. The rate of seropositivity increased from 3% in patients vaccinated within 45 days after the last monoclonal antibody administration to 80% in patients vaccinated >1 year after this therapy. The latter percentage was equal to that of patients never exposed to monoclonal antibodies. In conclusion, lymphoma patients, especially those recently treated with antiCD20 monoclonal antibodies, fail to develop sufficient humoral response to BNT162b2 vaccine. While a serological response is not the only predictor of immunity, its low level could make this population more vulnerable to COVID-19, which implies the need for a different vaccination schedule for such patients.

Introduction The global pandemic of coronavirus disease 2019 (COVID-19) had resulted in about 3.85 million deaths world-wide as of June 2021, with the estimated fatality rate among infected patients being between 1.5% and 2.1%.1 Emerging data demonstrate higher mortality rates among certain high-risk populations with significant co-morbidities, such as organ transplant recipients2 and cancer patients.3-5 There is evidence showing that patients with hematologic malignancies are the most vulnerable cancer population,3-7 with a higher risk of hospitalization and mortality following exposure to the virus.7 Estimated odds ratios (OR) for mortality are reported to vary between

haematologica | 2022; 107(3)

Correspondence: NETANEL A. HOROWITZ n_horowitz@rambam.health.gov.il Received: May 12, 2021. Accepted: June 22, 2021 Pre-published: July 29, 2021. https://doi.org/10.3324/haematol.2021.279216

715

R. Gurion et al.

2.09 and 12.16, depending on the type of malignancy and whether the disease has been actively treated within the months preceding the infection.5,6,8,9 Both non-Hodgkin lymphoma per se and prior chemotherapy with or without antiCD20 monoclonal antibodies have been suggested to contribute to patients’ reduced survival and prolonged hospitalization following infection with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2).10-13 The damage the pandemic inflicted on multiple healthcare systems which collapsed as a result of the high incidence of respiratory illness and intensive care demand, mostly due to the severity of COVID-19, led to an accelerated Food and Drug Administration approval of several anti-SARS-CoV-2 vaccines, following the successful completion of phase III studies. Among them was the BNT162b2 mRNA vaccine, which was demonstrated to have an efficacy of 95% in disease prevention in the pivotal phase III study. While the trial included approximately 40,000 volunteers, patients with active cancer were not enrolled into the study.14 Promptly after the Food and Drug Administration approval, this vaccine was approved by the Israeli Ministry of Health (December 2020), and vaccination was initiated at a large scale nation-wide level, with around 70% of the population aged 16 years and above having been fully vaccinated by April 2021. In addition, vaccination of potentially immunocompromised populations was started, including patients with hematologic conditions, despite the lack of good quality efficacy data for these patients, but in accordance with recommendations by hematologic and infectious disease agencies around the world.15-17 The rationale for this action had been the emerging data regarding the high infection-related morbidity and mortality among these patients, especially during the periods of peak virus spread, along with the probable low risk of vaccine-induced complications. However, at the physiological level, it is unclear whether patients with lymphoma will be able to generate good quality immune responses to this vaccine, since the response to any vaccine requires interactions between various compartments of the immune system, many of which are compromised by the lymphoproliferative disease itself,11 but even more so, by the chemoimmunotherapy regimens used for the treatment of these diseases.18,19 The lower prevalence and slower evolution of a humoral response to SARS-CoV-2 infection observed in this population of patients20,21 insinuate that this might be the case with humoral responses to the vaccine as well22. The objectives of this study were to evaluate the rates of anti-spike (anti-S) antibody responses to the BNT162b2 vaccine among lymphoma patients and to identify patientand treatment-related factors influencing the antibody responses.

lymphoproliferative disease, including Hodgkin and non-Hodgkin lymphoma according to the World Health Organziation 2016 classification23 and no known history of COVID-19 infection. Study participants were divided into the following two groups: (i) patients who received treatment, including chemotherapy or immunochemotherapy, i.e., monoclonal antibodies, tyrosine kinase inhibitors or immunomodulatory drugs, within 12 months prior to anti-COVID-19 vaccination; and (ii) patients with indolent lymphoma who were under "watch-and-wait" management before anti-COVID-19 vaccination. All patients were vaccinated with two doses of BNT162b2 vaccine, 21 days apart, and were followed at hematology clinics. Blood samples were drawn 4±2 weeks after the second dose of vaccine and were evaluated for anti-spike SARS-CoV-2 antibodies. The SARS-CoV-2 IgG II Quant (Abbott©) assay was performed as per manufacturer’s instructions for quantitative measurement of IgG antibodies against the spike protein of SARS-CoV-2. The test result was considered positive if the IgG level was ≥50 AU/mL. The patients’ baseline characteristics, collected from institutional electronic medical records, included each patient's demographics, comorbidities, lymphoma characteristics, duration, type and the first and last dates of anti-cancer treatment as well as disease activity before vaccination. Laboratory data such as complete blood count and serum protein electrophoresis before vaccination were also documented. The primary outcome was the rate of seropositivity for anti-spike antibodies.

Statistical considerations We analyzed patients’ characteristics using frequencies (percentages) for categorical variables and median (range) for continuous variables. A logistic regression model with the exp(β) was applied as an estimator of an OR and the 95% confidence interval (95% CI) around it to define the baseline variables that predict negativity of a serological response to SARS-CoV-2 vaccine. We used the likelihood ratio of the receiver operator characteristics curves and area under the curve to define the optimal cutoff for continuous variables. Univariate and multivariate logistic regression analyses were performed to evaluate potential predictors of seronegativity. To predict anti-spike IgG levels, we fitted a multiple-variable linear regression model based on: age, gender, lymphoma type, absolute lymphocyte count and time from the last anti-CD20 monoclonal antibody treatment to vaccination. Stepping method criteria for entry and removal were 0.05 and 0.2, respectively. The Kruskal-Wallis test was used to compare medians of antibody titers. To generate 95% CI around proportions, we used the binomial approximation of the normal distribution. Statistical analyses were performed using SPSS software (version 27, SPSS inc. Chicago, IL, USA) and GraphPad Prism version 6.0 software (GraphPad Software, San Diego, CA, USA).

Results Patients’ characteristics

Methods This was a non-interventional cross-sectional study conducted at two medical centers in Israel: Rambam Health Care Campus, Haifa (RMB) and Rabin Medical Center, Petach Tikva (RMC). All the procedures involved in this study were in accordance with the ethical standards of the institutional review boards of the two centers (approvals: # 0883-20-RMB; 1087-20-RMC) and with the 1964 Helsinki Declaration and its later amendments. All patients signed the informed consent form. The inclusion criteria were: age ≥18 years, the diagnosis of a

716

A total of 162 lymphoma patients who received two doses of the BNT162b2 vaccine between January and April, 2021 were included in the study. The median age of participants was 65 years (interquartile range, 52-73), 55% were males, 142 (88%) had non-Hodgkin lymphoma, including indolent and aggressive disease and the remaining 20 (12%) had Hodgkin disease. Reported comorbidities included diabetes mellitus (19%), ischemic heart disease (11%), and other malignancies (17%). Most (55%) of the patients received first-line anti-lymphoma therapy, while about 17% were under "watch-and-wait" management. The most haematologica | 2022; 107(3)

Response to BNT162b2 vaccine in lymphoma patients

common treatment protocols included CHOP (cyclophosphamide, vincristine, adriamycin and prednisone) or bendamustine with or without anti-CD20 monoclonal antibodies, either rituximab or obinutuzumab. Few patients received other therapies, such as Bruton tyrosine kinase (BTK) inhibitors, lenalidomide or antiPD1 antibodies. The patients’ characteristics are presented in Table 1.

Serological response to vaccination Eighty-three patients (51%) were seropositive (IgG levels ≥50 AU/mL) and 49% had negative serology. In univariate Table 1. Characteristics of the patients with lymphoma.

Characteristics Age in years, median (IQR) Males

N (%) 65 (52-73) 89 (55%)

Comorbidities Diabetes mellitus Ischemic heart disease Hypertension Chronic renal failure Chronic obstructive pulmonary disease Other malignancy

30 (19%) 17 (11%) 54 (34%) 12 (7.5%) 7 (4%) 27 (17%)

Type of lymphoma Diffuse large B-cell lymphoma Follicular lymphoma Marginal zone lymphoma Hodgkin lymphoma Peripheral T-cell lymphoma Other lymphomas*

32 64 24 20 8 14

Line of treatment Watch & wait First line Second line Third line and beyond

30 89 20 23

Type of treatment Non-chemotherapy Anti-CD20 monoclonal antibodies Rituximab Obinutuzumab Anti-PD1 monoclonal antibodies BTK inhibitors Lenalidomide

98 (60%) 68 30 5 (3.5%) 4 (3%) 6 (4%)

Chemotherapy CHOP Bendamustine ABVD/BEACOPP COP Other treatments**

36 (25%) 32 (22%) 10 (7%) 9 (6%) 19 (13%)

*Other lymphomas: Waldenström macroglobulinemia, mantle cell lymphoma, primary mediastinal B-cell lymphoma. **Other treatments: platinum-based chemotherapy, gemcitabine, brentuximab vedotin, polatuzumab vedotin, pralatrexate, romidepsin, phosphoinositide 3-kinase inhibitors. IQR: interquartile range; PD-1: programmed death; BTK: Bruton tyrosine kinase; CHOP: cyclophosphamide, adriamycin, vincristine and prednisone; ABVD: adriamycin, bleomycin, vinblastine and dacarbazine; BEACOPP: bleomycin, etopside, adriamycin, cyclophosphamide, vinblastine, procarbazine and prednisone; COP: cyclophosphamide, vincristine and prednisone.

haematologica | 2022; 107(3)

analysis, the following variables were found to be significantly associated with a lack of serological response: age >80 years (OR=4.3, 95% CI: 1.1-1.6), absolute lymphocyte count <1.2x109/L (OR=2.3, 95% CI: 1.1-4.4), IgG levels <630 g/L (OR=15.8, 95% CI: 1.9-129.9), active disease (defined as being under treatment for remission induction or by a positive positron emission tomography/computed tomography result) at vaccination (OR=4.2, 95% CI: 2.18.2), a time period of <12 months between the last antiCD20 treatment and vaccination (OR=31.3, 95% CI: 8.4116.9), the use of obinutuzumab versus rituximab (OR >4.54), aggressive non-Hodgkin lymphoma versus Hodgkin lymphoma (OR=15.4, 95% CI: 3.1-76.6) (Table 2). Lack of seroconversion was most frequent among patients suffering from aggressive lymphoma (63%) followed by those with indolent lymphoma (54%) and was lowest in patients with Hodgkin disease (10%). With the negative response rate in the last group used as a reference, the OR of this variable for patients with indolent disease was 1.5 (not statistically significant), while it was as high as 15 for patients with aggressive lymphoma (statistically significant, P<0.01). The rates of negative serological responses in patients receiving CHOP relative to those treated with bendamustine, with or without anti-CD20 monoclonal antibodies, were 63% and 84%, respectively (P=0.056). In multivariate analysis, two variables remained statistically significant: a time period <12 months between the last anti-CD20 treatment and the second dose of vaccine, and presence of active lymphoma (Table 3).

The effect of anti-CD20 treatment on vaccination results Among 98 patients who received anti-CD20 monoclonal antibodies, as the time period between the last dose of this treatment and vaccination became longer, the likelihood of seropositivity increased. The seropositivity rate was 80% in patients vaccinated at least 12 months after administration Table 2. Univariate analysis of factors associated with a lack of serological response.

Variable

Reference (95% CI)

Odds ratio

Age ≥80 years Age < 80 years 4.3 (1.1-1.6) Gender: female Male 0.8 (0.42-1.5) ALC ≤1.2 x109/L ALC >1.2 G/L 2.3 (1.1-4.4) IgG ≤630 g/L IgG >630 g/L 15.8 (1.9-129.9) Active disease Disease in remission 4.2 (2.1-8.2) Time between the last >12 months or 31.3 (8.4-116.9) anti-CD20 treatment and non-exposure to vaccination <12 months anti-CD20 Type of anti-CD20 MoAb: obinutuzumab Rituximab >4.54 (NA) Type of lymphoma – indolent lymphoma* Hodgkin lymphoma 1.46 (0.67-3.1) Aggressive lymphoma** Hodgkin lymphoma 15.4 (3.1-76.6) Time between the last chemotherapy administration and vaccination: <19 days >19 days 1.75 (0.32-9.4)

P-value 0.031 0.9 0.02 0.001 <0.001 <0.001

0.04 0.34 <0.01

0.515

*Indolent lymphoma included follicular lymphoma, marginal zone lymphoma, mantle cell lymphoma and Waldenström macroglobulinemia. **Aggressive lymphoma included diffuse large B-cell lymphoma, primary mediastinal large B-cell lymphoma and peripheral T-cell lymphoma. 95% CI: 95% confidence interval; ALC: absolute lymphocyte count; NA: not applicable; MoAb: monoclonal antibody.

717

R. Gurion et al.

of anti-CD20 monoclonal antibodies, while this rate was only 3% in patients vaccinated within 45 days after antiCD20 therapy (Table 4). It is noteworthy that the seropositivity rate in the former group was similar to that observed in lymphoma patients who had not received this treatment (i.e., were treated with chemotherapy only or were under “watch-and-wait” management). None of the 28 patients treated with obinutuzumab developed a serological response in comparison to 62% seronegativity demonstrated in patients treated with rituximab within the same time frame.

Levels of SARS-CoV-2 IgG In a linear regression model, a shorter time period between anti-CD20 therapy and vaccination predicted lower levels of anti-spike IgG and explained the 18% variance in antibody titers, while all other evaluable variables, such as age, gender, lymphoma type and absolute lymphocyte count had no predictive power. A correlation was revealed between the levels of circulating anti-spike IgG antibodies and the time between the last anti-CD20 treatment and vaccination. A significant difference was found between patients never exposed to anti-CD20 therapy (median of 1161 AU/mL; range, 0-15,567) or those receiving these agents more than 12 months prior to vaccination (median of 661 AU/mL; range, 0-15,220), relative to patients treated with these drugs within 12 months before vaccination: 0-45 days (median of 0 AU/mL; range, 0-225); 46-120 days (median of 0.7 AU/mL; range, 0-1575); 121-180 days (median of 0.5 AU/mL; range, 0-234); and 181-365 days (median of 0 AU/mL; range, 0-373 AU/mL) (Figure 1). In a model taking into account age, gender, absolute lymphocyte count, disease activity and the time from the last anti-CD20 treatment to vaccination, only the last variable was statistically significant and predicted the titers of IgG antibodies.

Discussion The current study, evaluating the antibody-mediated response in lymphoma patients who received two doses of BNT162b2 vaccine, showed that only 51% of these individuals developed seropositivity. These findings are in line with results of the studies assessing the efficacy of other anti-viral vaccines in the lymphoma setting. Indeed, studies assessing the efficacy of the influenza vaccine demonstrated insufficient humoral immunity and higher rates of overt

Table 3. Multivariate analysis of factors associated with a lack of serological response.

Variable

P-value

Odds Ratio (95% CI)

Age ≥80 years ALC ≤1.2x109/L Active disease Time between the last anti-CD20 treatment and vaccination <12 months Type of lymphoma

0.5 0.4 0.006 <0.001

2.8 (0.13-61.9) 2.1 (0.4-10.4) 11.8 (2-67.6) 93 (12.3-704.4)

0.8

1.2 (0.25-6.1)

95% CI: 95% confidence interval; ALC: absolute lymphocyte count. The IgG variable was removed due to missing data

clinical disease in patients treated with chemotherapy, with only 10% of patients developing a sufficient antibody titer to at least one of the influenza A antigens, as compared to 45% in the control group.24 Moreover, lymphoma patients vaccinated within a randomized trial of the recombinant zoster vaccine administered during or a maximum of 6 months after anti-lymphoma therapy, also showed low levels of seropositivity, varying between 20% and 50%.25 Currently available data point to the vital importance of COVID-19 prevention in cancer patients in general and in those with hematologic malignancies in particular.7 Evidence-based prophylactic approaches such as vaccination, have become the top priority measures significantly contributing to infection control. Nevertheless, the pivotal study, demonstrating 95% efficacy of the BNT162b2 vaccine in COVID-19 prevention, did not include patients with lymphoma. In a single-center Israeli study, examining antibody-mediated response rates with the Elecsys anti-SARSCoV-2 S assay in patients with chronic lymphocytic leukemia, positive humoral responses were observed in 52% of patients, compared to 100% in an age- and sexmatched control cohort.26 Notably, the assay used in the latter study differed from the one employed in our analysis. In the current study, treatment with anti-CD20 monoclonal antibodies as well as active disease at the time of vaccination emerged as significant predictors of a lack of serological response to BNT162b2. Likewise, the impact of anti-CD20 therapy was evident in the observed titers of anti-spike IgG antibodies, which increased as the time

Table 4. Serological response compared between patients treated with anti-CD20 monoclonal antibodies and those who did not receive this treatment.

Time from antiCD20 therapy to vaccination (days)

N. of N. of % of patients patients patients with positive with positive serology (CI) serology

0-45 46-120 121-180 181-365 >366 No anti-CD20 therapy

34 21 4 7 21 56

N.: number; CI: confidence interval.

718

1 5 1 1 17 45

3 (1-15) 24 (8-47) 25 (1-81) 14 (1-58) 81 (58-95) 80 (68-90)

% of patients with negative serology (CI) 97 (85-99) 76 (53-92) 75 (19-99) 86 (42-99) 19 (5-42) 20 (10-32)

Figure 1. Correlation between the levels of circulating anti-spike IgG antibodies and the time from the last anti-CD20 treatment to vaccination. Dots represent antibody titer values in arbitrary units (AU); red lines represent medians. ***P<0.0001; **P<0.001; *P<0.01; NS: not statistically significant.

haematologica | 2022; 107(3)

Response to BNT162b2 vaccine in lymphoma patients

period between the last anti-CD20 administration and vaccination became longer. With a cutoff of 12 months, our findings demonstrated a significant difference in the antibody titers between patients vaccinated less than or more than 12 months after anti-CD20 therapy, while the impact of exposure to this therapy became negligible after this time point. Actually, the titers became similar to those found in naïve (untreated) lymphoma patients. A plausible explanation could be that rituximab and other anti-CD20 monoclonal antibodies commonly used for the treatment of B-cell lymphoma lead to prolonged Bcell depletion and subsequent hypogammaglobulinemia. Consistent with our results, several studies reported data suggesting lower likelihoods of developing a serological response following anti-CD20 treatment in immunocompromised patients. For instance, patients with rheumatoid arthritis were reported to have lower titers of antiinfluenza antibodies upon treatment with rituximab compared to patients with rheumatoid arthritis not receiving such therapy.27 In another study, none of the 67 lymphoma patients vaccinated against influenza A (H1N1) within 6 months of receiving rituximab-containing regimens developed an antibody-mediated response compared to 82% in the control group.28 Finally, in the recently published study including a small cohort of antiCD20-treated chronic lymphocytic leukemia patients, none of those treated with rituximab within a year prior to vaccination developed anti-spike antibodies against SARS-CoV-2.26 Remarkably, in our study, patients receiving rituximab demonstrated an attenuated serological response to the vaccine, whereas patients treated with obinutuzumab failed to generate any anti-spike antibodies during the study period. This could be attributed to differences in pharmacodynamic properties between these two agents, as observed in in vitro studies, showing enhanced direct cell death and antibody-dependent cellular cytotoxicity for obinutuzumab compared to rituximab.29 In the present study, active disease emerged as an additional factor negatively affecting the humoral response to vaccine. While this could reflect the time-wise proximity to anti-CD20 treatment in these patients, it could also be associated with the effect of chemotherapeutic agents and corticosteroids commonly used during induction therapy in this clinical setting. Since humoral immunity requires functional T cells for the development of memory B cells and plasma cells,30,31 agents such as bendamustine and high-dose steroids, applied in lymphoma, might impede the serological response. This study has several limitations, the lack of a control

References 1. WHO Coronavirus (COVID-19) Dashboard. Available at https://covid19.who.int/. Last accessed in June, 2021. 2. Azzi Y, Bartash R, Scalea J, Loarte-Campos P, Akalin E. COVID-19 and solid organ transplantation: a review article. Transplantation. 2021;105(1):37-55. 3. Lee LYW, Cazier JB, Starkey T, et al. COVID-19 prevalence and mortality in patients with cancer and the effect of primary tumour subtype and patient demographics: a prospective cohort study. Lancet Oncol. 2020;21(10):1309-1316. 4. Erdal GS, Polat O, Erdem GU, et al. The

haematologica | 2022; 107(3)

group being one of them. Nevertheless, Grupper et al., utilizing the same assay as in our study, showed that all healthy individuals included in the control group developed a serological response to the BNT162b2 vaccine.32 In addition, nucleocapsid antibody assessment was not part of the current analysis, since only patients with no documented febrile or respiratory events within months prior to vaccination were included in the study. Hence, the generation of anti-spike antibodies in response to subclinical COVID-19, while being possible, was unlikely in this population of patients. A potential relationship between a weak serological response and the true protection from clinical COVID-19 will only become evident with longer follow-up. However, these data might never mature, as presently the pandemic has significantly subsided in Israel. Moreover, there are several newly validated assays capable of examining cellular immune responses to the vaccine, which will be included in future studies aimed at better understanding the true extent of protective immunity achieved with this vaccine. In conclusion, the current study has shown that a heterogeneous group of lymphoma patients has developed attenuated serological responses to the BNT621b2 vaccine. Patients recently treated with anti-CD20 monoclonal antibodies (<12 months since the last anti-CD20 treatment) are less likely to develop a serological response to this vaccine. Disclosures No conflicts of interest to disclose. Contributions RG, SR-H and NAH designed and performed research, interpreted the data, and wrote the paper. UR designed research, analyzed the data, and wrote the paper. MS and HB-Z performed serology assays. GL, AG-G, CL, PR, MT-A, EJD, NH, TI, IT, NL and RF collected data. All authors approved the final version of the paper. Aknowledgments We would like to thank Mrs. Sonia Kamenetsky for her advice and assistance in the preparation of this manuscript. Funding This research was supported by grants from Janssen, Takeda, Gilead Sciences and Abbvie. The Abbot test kits were kindly provided by the Israeli Ministry of Health. Data-sharing statement: The data that support the findings of this study are available from the corresponding author upon reasonable request.

mortality rate of COVID-19 was high in cancer patients: a retrospective single-center study. Int J Clin Oncol. 2021;26(5):826834. 5. Rubinstein SM, Steinharter JA, Warner J, Rini BI, Peters S, Choueiri TK. The COVID19 and Cancer Consortium: a collaborative effort to understand the effects of COVID19 on patients with cancer. Cancer Cell. 2020;37(6):738-741. 6. Robilotti EV, Babady NE, Mead PA, et al. Determinants of COVID-19 disease severity in patients with cancer. Nat Med. 2020;26(8):1218-1223. 7. Vijenthira A, Gong IY, Fox TA, et al. Outcomes of patients with hematologic

malignancies and COVID-19: a systematic review and meta-analysis of 3377 patients. Blood. 2020;136(25):2881-2892. 8. Baker D, Roberts CAK, Pryce G, et al. COVID-19 vaccine-readiness for antiCD20-depleting therapy in autoimmune diseases. Clin Exp Immunol. 2020;202(2): 149-161. 9. Morrison VA. Infections in patients with leukemia and lymphoma. Cancer Treat Res. 2014;161:319-349. 10. Dai M, Liu D, Liu M, et al. Patients with cancer appear more vulnerable to SARSCoV-2: a multicenter study during the COVID-19 outbreak. Cancer Discov. 2020;10(6):783-791.

719

R. Gurion et al. 11. Shah V, Ko Ko T, Zuckerman M, et al. Poor outcome and prolonged persistence of SARS-CoV-2 RNA in COVID-19 patients with haematological malignancies; King's College Hospital experience. Br J Haematol. 2020;190(5):e279-e282. 12. Vabret N, Britton GJ, Gruber C, et al. Immunology of COVID-19: current state of the science. Immunity. 2020;52(6):910-941. 13. Dulery R, Lamure S, Delord M, et al. Prolonged in-hospital stay and higher mortality after Covid-19 among patients with non-Hodgkin lymphoma treated with Bcell depleting immunotherapy. Am J Hematol. 2021;96(8):934-944. 14. Polack FP, Thomas SJ, Kitchin N, et al. Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine. N Engl J Med. 2020;383(27):2603-2615. 15. Cheuk DK, Chiang AK, Lee TL, Chan GC, Ha SY. Vaccines for prophylaxis of viral infections in patients with hematological malignancies. Cochrane Database Syst Rev. 2011(3):CD006505. 16. Auletta J, Chemaly R, Khawaja F, et al. General principles of COVID-19 vaccines for immunocompromised patients. Available at https://www.hematology.org/ covid-19/ash-astct-covid-19-and-vaccines. Last accessed in April 2021 17. Interim clinical considerations for use of COVID-19 vaccines currently authorized in the United States. Available at https://www.cdc.gov/vaccines/covid19/info-by-product/clinical-considerations.html#underlying-conditions. Last

720

accessed in April 2021 18. He W, Chen L, Chen L, et al. COVID-19 in persons with haematological cancers. Leukemia. 2020;34(6):1637-1645. 19. Pleyer C, Ali MA, Cohen JI, et al. Effect of Bruton tyrosine kinase inhibitor on efficacy of adjuvanted recombinant hepatitis B and zoster vaccines. Blood. 2021;137(2):185-189. 20. Shah GL, DeWolf S, Lee YJ, et al. Favorable outcomes of COVID-19 in recipients of hematopoietic cell transplantation. J Clin Invest. 2020;130(12):6656-6667. 21. O'Nions J, Muir L, Zheng J, et al. SARSCoV-2 antibody responses in patients with acute leukaemia. Leukemia. 2021;35(1): 289-292. 22. Sun C, Pleyer C, Wiestner A. COVID-19 vaccines for patients with haematological conditions. Lancet Haematol. 2021;8(5): e312-e314. 23. Swerdlow SH, Campo E, Pileri SA, et al. The 2016 revision of the World Health Organization classification of lymphoid neoplasms. Blood. 2016;127(20):2375-2390. 24. Mazza JJ, Yale SH, Arrowood JR, et al. Efficacy of the influenza vaccine in patients with malignant lymphoma. Clin Med Res. 2005;3(4):214-220. 25. Dagnew AF, Ilhan O, Lee WS, et al. Immunogenicity and safety of the adjuvanted recombinant zoster vaccine in adults with haematological malignancies: a phase 3, randomised, clinical trial and posthoc efficacy analysis. Lancet Infect Dis. 2019;19(9):988-1000. 26. Herishanu Y, Avivi I, Aharon A, et al.

Efficacy of the BNT162b2 mRNA COVID19 vaccine in patients with chronic lymphocytic leukemia. Blood. 2021;137(23): 3165-3173. 27. Gelinck LB, Teng YK, Rimmelzwaan GF, van den Bemt BJ, Kroon FP, van Laar JM. Poor serological responses upon influenza vaccination in patients with rheumatoid arthritis treated with rituximab. Ann Rheum Dis. 2007;66(10):1402-1403. 28. Yri OE, Torfoss D, Hungnes O, et al. Rituximab blocks protective serologic response to influenza A (H1N1) 2009 vaccination in lymphoma patients during or within 6 months after treatment. Blood. 2011;118(26):6769-6771. 29. Tobinai K, Klein C, Oya N, FingerleRowson G. A review of obinutuzumab (GA101), a novel type II anti-CD20 monoclonal antibody, for the treatment of patients with B-cell malignancies. Adv Ther. 2017;34(2):324-356. 30. Haberman AM, Gonzalez DG, Wong P, Zhang TT, Kerfoot SM. Germinal center B cell initiation, GC maturation, and the coevolution of its stromal cell niches. Immunol Rev. 2019;288(1):10-27. 31. Nguyen DC, Joyner CJ, Sanz I, Lee FE. Factors affecting early antibody secreting cell maturation into long-lived plasma cells. Front Immunol. 2019;10:2138. 32. Grupper A, Sharon N, Finn T, et al. Humoral response to the Pfizer BNT162b2 vaccine in patients undergoing maintenance hemodialysis. Clin J Am Soc Nephrol. 2021;16(7):1037-1042.

haematologica | 2022; 107(3)

ARTICLE

Plasma Cell Disorders

The innate sensor ZBP1-IRF3 axis regulates cell proliferation in multiple myeloma

Ferrata Storti Foundation

Kanagaraju Ponnusamy,1 Maria Myrsini Tzioni,1 Murshida Begum,1 Mark E. Robinson,1 Valentina S. Caputo,1 Alexia Katsarou,1,2 Nikolaos Trasanidis,1 Xiaolin Xiao,1 Ioannis V. Kostopoulos,1,3 Deena Iskander,1,2 Irene Roberts,4 Pritesh Trivedi,5 Holger W. Auner,1,2 Kikkeri Naresh,5 Aristeidis Chaidos1,2 and Anastasios Karadimitris1,2 1 Hugh & Josseline Langmuir Centre for Myeloma Research, Centre for Haematology, Department of Immunology & Inflammation, Imperial College London, London, UK; 2 Department of Haematology, Hammersmith Hospital, Imperial College Healthcare NHS Foundation Trust, London, UK; 3Section of Animal and Human Physiology, National and Kapodestrian University of Athens, Department of Biology, School of Science, Athens, Greece; 4Department of Paediatrics and MRC Molecular Haematology Unit, Weatherall Institute of Molecular Medicine, University of Oxford and BRC Blood Theme, NIHR Oxford Biomedical Centre, Oxford, UK and 5Department of Cellular & Molecular Pathology, Northwest London Pathology, Imperial College Healthcare NHS Trust, London, UK

Haematologica 2022 Volume 107(3):721-732

ABSTRACT

ultiple myeloma is a malignancy of plasma cells initiated and driven by primary and secondary genetic events. However, myeloma plasma cell survival and proliferation might be sustained by non-genetic drivers. Z-DNA-binding protein 1 (ZBP1; also known as DAI) is an interferon-inducible, Z-nucleic acid sensor that triggers RIPK3-MLKL-mediated necroptosis in mice. ZBP1 also interacts with TBK1 and the transcription factor IRF3 but the function of this interaction is unclear, and the role of the ZBP1-IRF3 axis in cancer is not known. Here we show that ZBP1 is selectively expressed in late B-cell development in both human and murine cells and it is required for optimal T-cell-dependent humoral immune responses. In myeloma plasma cells, the interaction of constitutively expressed ZBP1 with TBK1 and IRF3 results in IRF3 phosphorylation. IRF3 directly binds and activates cell cycle genes, in part through co-operation with the plasma cell lineage-defining transcription factor IRF4, thereby promoting myeloma cell proliferation. This generates a novel, potentially therapeutically targetable and relatively selective myeloma cell addiction to the ZBP1-IRF3 axis. Our data also show a noncanonical function of constitutive ZBP1 in human cells and expand our knowledge of the role of cellular immune sensors in cancer biology.

Introduction Multiple myeloma (MM) is a common incurable blood cancer of the bone marrow plasma cells (PC), the immunoglobulin-secreting terminally differentiated B lineage cells.1-3 Primary and secondary somatic genetic events comprising copy number and single nucleotide variants shape a genomic landscape of extensive, in time and space, genetic heterogeneity and diversification rendering targeted therapies for MM a challenging task.1-3 In this regard, there is a need for identification of biological pathways that are involved in myelomagenesis independently of genetic status. Previous studies of murine late B lineage development identified ZBP1 as one of the genes that define the transcriptional signature of follicular B-cell transition to plasmablasts and mature PC.4 ZBP1 is an inducible cellular DNA/RNA sensor with two Zα domains that bind pathogen-derived or cellular Z-DNA5,6 or Z-RNA.7,8 Zαdependent nucleic acid sensing induces a RHIM-RHIM domain interaction of ZBP1 with receptor-interacting protein kinase 3 (RIPK3) which ultimately triggers mixed lineage kinase domain like pseudokinase (MLKL)-mediated necroptosis.7-10 This process is counteracted by RIPK1, preventing for example ZBP1-dependent cell death and inflammation in the developing skin.11-13

haematologica | 2022; 107(3)

Correspondence: ANASTASIOS KARADIMITRIS a.karadimitris@imperial.ac.uk Received: October 20, 2020. Accepted: February 2, 2021. Pre-published: February 18, 2021. https://doi.org/10.3324/haematol.2020.274480

721

K. Ponnusamy et al.

In addition, like other nucleic acid sensors, such as cGas/STING, ZBP1 has been shown to associate with serine/threonine kinase TBK1 with subsequent phosphorylation of the transcription factor IRF3 (pIRF3). Activation of STING and other nucleic acid sensors triggers translocation of pIRF3 to the nucleus where it directly activates transcription of interferon (IFN) type I response genes;14-16 however, this has been disputed in the case of ZBP1.17,18 Therefore, the role of ZBP1-TBK1-IRF3 in innate sensing or in the context of cancer remains unclear. Nevertheless, it is clear that ZBP1 itself is an IFN-inducible gene as its expression is induced by type I and II IFN which leads to RIPK3-dependent necroptosis7,13,19 and IFNAR1-/- cells fail to upregulate virus-induced ZBP1 expression.20 Since recent work demonstrated an active type I IFN versus proliferative transcriptional signature prevailing in early diagnosis versus relapsed patients’ myeloma PC or myeloma cell lines, respectively,21 we hypothesized that ZBP1 might regulate myeloma PC biology.

achieved, using the NEBNext poly(A) mRNA Magnetic Isolation Module, and libraries were prepared using NEBNext Ultra II RNA library prep kits (New England Biolabs). A 2 nM DNA library (350400 bp) was sequenced using Illumina HiSeq 2500 (for shZBP1) or NextSeq500 (for shIRF3) for paired-end 150 bp reads.

Chromatin immunoprecipitation (ChIP)-sequencing, ChIP-re-ChIP MM.1S cells were cross-linked with 1% formaldehyde and lysed with hypotonic lysis buffer followed by nuclear lysis buffer. The lysate was sonicated to shear the chromatin up to 500 bp followed by pre-clear with protein A/G magnetic beads and immunoprecipitation with IRF3 antibody or equivalent isotype control. ChIP DNA was collected with AMPure XP-beads after reverse cross-linking. ChIP DNA (1 ng) was used for library preparation with an NEBNext kit. The 2 nM DNA library (400-500 bp) was sequenced using Illumina NextSeq500. For ChIP-re-ChIP, pulled chromatin was eluted in 1% sodium dodecylsulfate (SDS) and diluted 10 times with elution buffer and repeated re-ChIP with the appropriate antibody. The detailed protocols are provided in the Online Supplementary Methods.

Methods Co-immunoprecipitation, immunoblotting Cell culture, primary samples U266 and NCI-H929 (Deutsche Sammlung von Mikroorganismen und Zellkulturen, Germany), MM.1R and MM.1S (American Type Culture Collection, Manassas, VA, USA), and HeLa, HEK293T, DU145, MCF7, HCC95, SF295, LNCAP, K562, Jurkat and C1R cells were cultured in either 10% fetal bovine serum, RPMI-1640 or Dulbecco modified Eagle medium. Primary myeloma cells were collected under ethical committee approval (REC n. 11/H0308/9) and cultured with IL-6 (10 ng/mL).

Co-immunoprecipitation was performed as described previously12 with 5% anti-strep-tagII-magnetic bead slurry (IBA) or 2 mg anti-ZBP1 or anti-V5-Tag or anti-IRF3 or their equivalent isotype control antibodies conjugated with protein A/G magnetic beads. For immunoblotting, total cell proteins denatured in 1x LDS Sample Buffer (ThermoFisher Scientific), were resolved in 10% SDS polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes followed by probing with appropriate antibodies shown in the Online Supplementary Methods.

Cell cycle, proliferation and apoptosis Cell cycle status was assessed using flow cytometry after staining cells with 10 mM Hoechst-33342 in RPMI-1640 medium for 1 h. The cytostatic effects of constitutive or inducible shRNA-transduced cells were measured as shown in Online Supplementary Figure S3C. Annexin V (BioLegend) staining was performed according to the manufacturer’s instructions.

Subcutaneous tumor model, immunization The subcutaneous tumor study was performed in NOD/SCIDγ mice (license PPL70/8586), by subcutaneously injecting 107 doxycycline-inducible shRNA- transduced cells with Matrigel (Corning). Tumor volume was calculated by the formula (1/2[length×width2]). Animals were administered 100 mg/mL doxycycline in drinking water and 0.2 mg/kg intraperitoneally. Zbp1-/- animals18 were obtained from Manolis Pasparakis, Institute of Genetics (Cologne, Germany). Ten- to 12-week-old littermates were immunized by intraperitoneal injection of 4 mg/kg NP-KLH (Santacruz Biotech) with Alum Adjuvant (Thermoscientific) at a 3:1 ratio followed by a booster dose of NPKLH on day 4. Serum NP-KLH-specific antibodies were measured.

Quantitative polymerase chain reaction analysis, RNA-sequencing Total RNA was isolated using a Nucleospin RNA kit (MachereyNagel) followed by cDNA synthesis using a RevertAid cDNA synthesis kit (Thermoscientific). Indicated cDNA were quantified using the primers shown in the Online Supplementary Methods. Poly(A)-tail mRNA was isolated, from total RNA of fluorescenceactivated cell sorting (FACS)-purified green fluorescent protein (GFP)+ cells with either scrambled or shRNA targeting ZBP1 or IRF3 on day 4 after transduction when >80% knockdown was

722

Data analysis The details of the data analysis are provided in the Online Supplementary Methods. In brief, RNA-sequencing and ChIPsequencing reads were aligned using STAR or Salmon and BWA MEM,22,23 respectively, to the GRCh38 genome. Differential expression analysis was performed using DESeq2 or limmavoom.22,24,25 ChIP-sequencing peaks were called with MACS226 and tracks with Deeptools27 and visualized in IGV.28 Homer was used for Motif analysis29 and the BETA-plus package30 to integrate RNA-sequencing and ChIP-sequencing data. Gene set enrichment analysis (GSEA)31 and the Enrichr web tool32 were used for pathway analysis.

Data availability All next-generation sequencing data for RNA-sequencing and ChIP-sequencing experiments can be accessed via the Gene Expression Omnibus (GSE163497).

Additional methods The details of shRNA sequences, cloning strategies, reagents, antibodies and additional methodologies are provided in the Online Supplementary Methods.

Results Restricted and constitutive expression of ZBP1 in normal and myeloma plasma cells By searching for genes selectively expressed in MM we identified ZBP1 as highly and selectively expressed in MM cell lines (MMCL) but not in other cancer cells (CCLE haematologica | 2022; 107(3)

ZBP1 regulates myeloma cell proliferation via IRF3/IRF4

Figure 1. Restricted and constitutive ZBP1 expression in normal and myeloma plasma cells. (A) ZBP1 mRNA expression as assessed in four multiple myeloma cell lines (MMCL) and in the erythromyeloid cell line K562 by quantitative polymerase chain reaction (n=3; data shown as mean ± standard error of mean). (B) ZBP1 expression in indicated MMCL as assessed by immunoblotting using GAPDH as the loading control. Two main isoforms ~48 and ~40 kDa are detected in MMCL. (C, D) ZBP1 mRNA and protein expression in six myeloma patient-derived bone marrow plasma cells purified for CD138+ marker using CD138 immunomagnetic microbeads, and U266 MMCL shown as control. (E) Immunohistochemistry on paraffin-embedded tonsil tissue section. The panel on the left depicts low and high magnification of a germinal center in which expression of PAX5 and ZBP1 is mutually exclusive in most parts indicating ZBP1 expression in plasma cells that are negative for PAX5. The panel on the right depicts subepithelial tonsillar plasma cells showing co-expression of IRF4 (MUM.1) and ZBP1. (F) Immunohistochemistry for ZBP1 expression in paraffin-embedded bone marrow tissue sections from two MM patients.

dataset) (Online Supplementary Figure S1A, B). While there is no difference in ZBP1 expression between normal PC and the whole cohort of myeloma PC, ZBP1 expression is significantly higher in the hyperdiploid subgroup of myeloma PC (Arkansas GSE4581 microarray dataset) (Online Supplementary Figure S1C). ZBP1 is universally expressed in primary myeloma PC (MMRF CoMMpass dataset; n=767 patients) (Online Supplementary Figure S1D) at similar levels as the PC-defining transcription factor PRDM1 (BLIMP1). In the human B-cell lineage, ZBP1 is expressed at a very low level in germinal center B cells (GCB) and a moderate level in naïve and memory B cells, but PC show by far the highest expression (Blueprint DCC Portal data) (Online Supplementary Figure S1E). We confirmed expression of ZBP1 mRNA and protein in MMCL (Figure 1A, B) and also in primary myeloma PC (Figure 1C, D) but not in other hematopoietic or epithelial cancer cells (Online Supplementary Figure S1F, G), normal blood lineage cells (Online Supplementary Figure S1H, I) or healthy nonhematopoietic tissues (Online Supplementary Figure S1J). haematologica | 2022; 107(3)

Immunohistochemistry of human tonsils and lymph nodes showed expression of ZBP1 primarily in a group of PAX5–IRF4+ GCB cells i.e., those committed to PC differentiation33,34 and in interfollicular and subepithelial IRF4+ PC, with low-level expression in mantle zone B cells (Figure 1E and Online Supplementary Figure S1K,L). In bone marrow, expression of ZBP1 is mostly restricted to myeloma and normal PC and is not present in other blood lineage cells (Figure 1F and Online Supplementary Figure S1K,M). These results show constitutive and restricted expression of ZBP1 in myeloma PC as well as in late GCB cells and normal PC.

ZBP1 is required for optimal T-cell-dependent humoral immune responses in mice Consistent with a conserved expression pattern, a previous transcriptome analysis of murine B-cell development identified Zbp1 as one of the signature genes that define transition from follicular B cells to plasmablasts and mature PC4 (Online Supplementary Figure S2). To determine 723

K. Ponnusamy et al.

Figure 2. ZBP1 is required for optimal T cell-dependent humoral immune responses in mice. (A) Zbp1 mRNA levels as assessed by quantitative polymerase chain reaction in FACS-purified splenic germinal center B (GCB) cells (B220+CD19+GL7+CD95+) and plasma cells (PC) (B220loCD138+) from Zbp1-/- (KO) mice or their wildtype (WT) littermates immunized with just alum (control) or alum-NP-KLH (NP-KLH) on day 10 after immunization. Lack of Zbp1 mRNA confirms Zbp1-deficiency in Zbp1-/- mice. (n=3). (B, C) Flow-cytometric identification of splenic GCB cells as B220+CD19+GL7+CD95+ and GCB frequency in Zbp1-/- mice and their WT littermates immunized with NP-KLH or alum-only control on day 10 after immunization (B) and GCB cell fold-difference between WT and Zbp1-/- mice after NP-KLH immunization normalized to the median frequency of control animals (C). Numbers in the flow cytometry plots represent % frequency. (n=11 mice/group). (D, E) Flow-cytometric identification of splenic PC as B220loCD138+ and PC frequency in Zbp1-/- mice and their WT littermates in control and NP-KLH-immunized animals on day 10 after immunization (D) and PC fold-difference between WT and Zbp1-/- mice after NP-KLH immunization normalized to median frequency of control animals (E). Numbers in the flow cytometry plots represent % frequency (n=11 mice/group). (F) NP-KLH-specific IgG responses in Zbp1-/- mice and their WT littermates in control- and NPKLH-immunized animals on day 10 after immunization. IgG antibody relative levels (left panels) after NP-KLH immunization and their fold-differences normalized to median antibody levels of control animals (right panels). (n=6 mice/group). (G) NP-KLH-specific IgM responses in Zbp1-/- mice and their WT littermates in controland NP-KLH-immunized animals on day 10 after immunization. IgM antibody relative levels (left panels) after NP-KLH immunization and their fold-differences normalized to median antibody levels of control animals (right panels) (n=6 mice/group). The error bars of all the cumulative data indicate mean ± standard error of mean. A two-tailed unpaired t-test was applied to determine the P values. *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001, ns- not significant (P>0.05). The number of experiments performed or animals used for the study are indicated separately in each panel legend.

724

haematologica | 2022; 107(3)

ZBP1 regulates myeloma cell proliferation via IRF3/IRF4

Legend continued on following page

Figure 3. ZBP1 is required for myeloma cell proliferation and survival. (A) Percent green fluorescent protein-positive (%GFP+) cells after transduction with GFP-encoding, ZBP1-targeting shRNA1 (sh1), shRNA2 (sh2), shRNA3 (sh3) or appropriate scrambled control (scr) lentiviral constructs in U266 and H929 cells. All the time points were normalized to day 3 %GFP expression levels for each shRNA or scr control shown. (n=3). (B) %GFP+ cells after transduction with ZBP1-targeting sh1, sh2 or appropriate scr control lentiviral constructs in dexamethasone-sensitive MM.1S and its resistant derivative MM.1R cell line. All the time points were normalized to day 3 %GFP expression levels for each shRNA or scr control shown (n=3). (C) %GFP+ cells after the transduction of K562 and HeLa cells, which do not express ZBP1, with ZBP1-targeting sh1 or sh2 or appropriate scr control. The %GFP+ cells were normalized to day 3 %GFP expression levels for all the time points for each shRNA or scr control shown (n=3). (D, E) Photographs of tumors explanted at sacrifice from control, i.e., non-doxycycline (dox)-treated or dox-treated animals engrafted with H929 myeloma cells transduced with dox-inducible sh1 or sh2 targeting ZBP1 (D). Subcutaneous tumor volume of H929 or MM.1S over a period of up to 4

haematologica | 2022; 107(3)

725

K. Ponnusamy et al. weeks in control or dox-treated animals (D, E) (n=4-5 mice/group). (F) RNA-sequencing was performed with poly(A) tail-enriched RNA from FACS-purified GFP+ live cells of scr or shRNA-transduced cells. Venn diagram showing the number of commonly and differentially (up- and down-regulated) expressed genes based on log2 fold-change to scr control with a cut-off adjusted P-value (Padj) <0.05 among the two ZBP1-depleted transcriptomes of the multiple myeloma cell lines (MMCL) H929 and MM.1S and both anti-ZBP1 sh1 or sh2 (n=2). (G) Heatmap showing the expression patterns of the top 289 commonly and differentially expressed genes, based on log2 fold-change to scr control with a cut off Padj <0.05, shared by the two ZBP1-depleted MMCL H929 and MM.1S and both anti-ZBP1 sh1 or sh2. (H) Enrichr pathway enrichment analysis of the shared 270 genes that are commonly downregulated between two MMCL treated with anti-ZBP1 sh1 or sh2 as compared to the scr control. (I) A representative histogram shows depletion of ZBP1 by anti-ZBP1 sh1 or sh2 induces cell cycle arrest in MMCL H929 as compared to the scr control, and its cumulative data shown for MMCL H929 and MM.1S cells. The analysis was performed on GFP+ cells on day 4 after transduction (n=3). (J) A representative histogram showing that anti-ZBP1 sh1- or sh2-mediated ZBP1 depletion induces cell cycle arrest as compared to the scr control in multiple myeloma (MM) patientderived bone marrow myeloma cells purified for the CD138+ marker, and the quantitative data. The flow cytometry plot at the top shows the gating strategy for the GFP+ population in transduced cells. The analysis was performed on GFP+ cells on day 4 after transduction (n=3 MM bone marrow samples). The error bars of all the cumulative data indicate the mean ± standard error of mean. A two-tailed unpaired t-test was applied to determine the P values. *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001, ns: not significant (P>0.05). The number of experiments performed is indicated separately in each panel legend.

the role of ZBP1 in GCB and PC development, we assessed the humoral immune responses against the T-cell-dependent antigen NP-KLH using alum as an adjuvant. We found that while Zbp1 mRNA is undetectable in FACS-purified GCB cells and PC from Zbp1-/- mice in response to either NP-KLH-alum or alum-only control, it increased from GCB cells to PC in wild-type (WT) littermates and this increase was more pronounced upon immunization with NP-KLH-alum (Figure 2A). Splenic frequencies of B220+CD19+GL7+CD95+ GCB cells (Figure 2B, C) and B220loCD138+ PC (Figure 2D, E) were not different at baseline between Zbp1-/- and their WT littermates. Similarly GCB-cell frequencies were not significantly different between WT and Zbp1-/- animals after NP-KLH-alum immunization; however, the increase in PC frequency and NP-KLH-specific IgG (but not IgM) serum levels in immunized Zbp1-/- animals was significantly lower compared to that in immunized WT littermates (Figure 2D-G). These findings suggest that although Zbp1 is not required for GCB cell and PC development under steady-state, it is required for optimal T-cell-dependent humoral immune responses. Whether cellular nucleic acids in complex with Zbp1 play a role in this process remains to be addressed.

ZBP1 is required for myeloma cell proliferation and survival To investigate the functional role of ZBP1 in MM, we depleted ZBP1 expression in MMCL by targeting both of its main isoforms, i.e., isoform 1 comprising Zα1 and Zα2 domains and isoform 2 which lacks Zα1. Depletion of either isoform 1 or both isoforms 1 and 2 by shRNA1- or shRNA2-mediated knockdown, respectively, was toxic to H929 and U266 cells while shRNA3 did not deplete ZBP1 and behaved like the scrambled control without affecting cell viability (Figure 3A and Online Supplementary Figure S3A-C). Depletion of ZBP1 by shRNA1/2 was also toxic to MMCL MM.1S and its dexamethasone-resistant derivative MM.1R (Figure 3B and Online Supplementary Figure S3D). These findings suggest that the observed effect is mediated by depletion of isoform 1 and cannot be rescued by isoform 2. This effect was specific because the antiproliferative function was not observed in the shRNAtransduced erythromyeloid K562 or epithelial HeLa cells (Figure 3C) which lack ZBP1 expression (Online Supplementary Figures S1F and 3E). Furthermure, depletion of shRNA1-transduced myeloma cells was at least in part rescued by overexpression of ZBP1 cDNA with appropriate silent mutations (Online Supplementary Figure S3F). Mutating the seed region of shRNA1, aimed at eliminating off-target effects,35 did not alter either the expression of ZBP1 or the cytotoxic effects in MM.1S cells (Online Supplementary Figure S3G,H). Using a doxycycline726

inducible shRNA,36 we found that ZBP1 depletion inhibited myeloma cell growth in vitro (Online Supplementary Figure S3I,J) and also subcutaneous myeloma tumor growth in vivo (Figure 3D, E and Online Supplementary Figure S3K,L). Together, these findings suggest an important role of ZBP1 in myeloma cell biology. In line with these observations, transcriptome analysis of two ZBP1-depleted MMCL, in which oncogenic transcriptomes are driven by MAF (MM.1S) or MMSET (H929) oncogenes, revealed 270 genes that are significantly downregulated in both cells and by both shRNA (Figure 3F, G and Online Supplementary Table S1). These genes were highly enriched for cell cycle control pathways (Figure 3H and Online Supplementary Table S2). We also validated reduction of the mRNA expression levels of the cell cycle regulators Ki-67, FOXM1 and E2F1 upon ZBP1-depletion by quantitative polymerase chain reaction analysis (Online Supplementary Figure S4A, B) and confirmed the decrease in proteins by immunoblotting (FOXM1 and E2F1) and flow-cytometry (Ki-67) (Online Supplementary Figure S4C-E). Flow-cytometric analysis of the cell cycle in ZBP1-depleted cells revealed arrest at the G0/G1 phase (Figure 3I) in conjunction with increased apoptosis as assessed by annexin V staining in MM.1S and H929 cells (Online Supplementary Figure S4F). Notably, we also confirmed that both anti-ZBP1 shRNA induced cell cycle arrest in MM patient-derived, bone marrow myeloma CD138+ PC (Figure 3J and Online Supplementary Figure S4G), thus confirming the role of ZBP1 in cell cycle regulation in primary myeloma PC as well as MMCL. In addition to downregulation of cell cycle pathways, GSEA also showed significant enrichment for the IFN type I pathway in upregulated genes induced by both shRNA1 and shRNA2 in MM.1S cells but in downregulated genes by only shRNA1 in H929 cells (Online Supplementary Figure S4H). This disparate effect of ZBP1 depletion on IFN type I response genes might reflect the distinct transcriptomes of MM.1S and H929 MMCL imposed by their primary driver oncogenes. Previous work demonstrated that a transcriptional proliferative signature identifies a minority of MM patients with adverse prognosis.37,38 Accordingly, GSEA of myeloma PC transcriptomes with the top 5% highest versus 90% lowest ZBP1 expression revealed significant enrichment in the former for cell cycle regulation pathways among overexpressed genes (Online Supplementary Figure S4I, J and Online Supplementary Table S3). Interestingly, among these overexpressed genes in the subgroup of ZBP1hi patients, we also observed significant enrichment for IFN type I signaling consistent with the role of ZBP1 as an IFNresponse gene (Online Supplementary Figure S4I, J and Online Supplementary Table S3).

haematologica | 2022; 107(3)

ZBP1 regulates myeloma cell proliferation via IRF3/IRF4

Figure 4. ZBP1 as a scaffold for IRF3 constitutive activation by TBK1. (A) Immunoblotting analysis of pIRF3/IRF3 expression in myeloma (MM.1S and H929) and non-myeloma cell lines (K562; Jurkat; HL60: acute myeloid leukemia; DG75: B lineage; AR230: chronic myeloid leukemia; GM1271: Epstein-Barr virus-transformed B-cell lineage). (B) Immunoblotting for pIRF3/IRF3 expression in MM patient-derived bone marrow plasma cells (PC) purified for CD138+ using CD138 immunomagnetic microbeads. (C, D) Immunoblotting against IRF3 and ZBP1 in MM.1S cells following co-immunoprecipitation with anti-ZBP1 (C), or anti-IRF3 (D) or corresponding isotype control antibodies. (E) Immunoblotting against TBK1 and ZBP1 in MM.1S cells following co-immunoprecipitation with anti-ZBP1 or its corresponding isotype control antibodies. (F) Immunoblotting for pIRF3/IRF3 expression in anti-ZBP1 shRNA1 (sh1) or shRNA2 (sh2) or scrambled (scr) control RNA transduced MM.1S cells on day 4 after transduction (left). ImageJ quantification shows profound reductions in pIRF3 but not in total IRF3 levels in anti-ZBP1 sh1- or sh2-transduced cells as compared to scr control cells (right). The protein lysates were prepared from cells with >90% transduction efficiency. (G) Immunoblotting for pIRF3/IRF3 expression on day 4 following anti-TBK1 sh1, sh2 or scr transduction in MM.1S cells. The protein lysates were prepared from the cells with >90% transduction efficiency. (H) Percentage green fluorescent protein-positive (%GFP+) cells after transduction with IRF3-targeting sh1, sh2 or scr control in MMCL H929 and MM.1S cells. All the time points were normalized to day 3 %GFP expression levels for each shRNA shown (n=3). (I) A representative flow-cytometric histogram of cell cycle in MMCL transduced with anti-IRF3 sh1 or sh2 or scr control. Analysis was performed on GFP+ cells day 4 after transduction and the cumulative data for H929 show cell cycle arrest in anti-IRF3 sh1- or sh2-transduced H929 cells (n=3). The error bars of all the cumulative data indicate mean ± standard error of mean. A two-tailed unpaired t-test was applied to determine the P values. *P≤0.05, **P≤0.01, ***P≤0.001. The number of experiments performed for the study is indicated separately in each panel legend.

haematologica | 2022; 107(3)

727

K. Ponnusamy et al.

Figure 5. IRF3 regulates cell cycle genes in myeloma cells. (A) RNA-sequencing was performed with poly(A) tail-enriched RNA from FACS-purified green fluorescent protein (GFP)+ live cells of scrambled (scr) or shRNA-transduced cells. The Venn diagram shows the numbers of commonly up- and down-regulated genes among the top 50% differentially expressed genes, based on log2 fold-change to the scr control with cut-off Padj <0.05, of ZBP1-depleted and IRF3-depleted transcriptomes in MM.1S cells (n=2). (B) Heatmap showing the expression patterns of the top 132 commonly expressed genes among the top 50% differentially expressed genes of ZBP1-depleted and IRF3-depleted transcriptomes in MM.1S cells. (C) Enrichr pathway enrichment analysis for the shared 109 genes that are downregulated in common upon ZBP1- or IRF3-depletion in MM.1S cells. (D) Transcription factor motif analysis of IRF3 chromatin immunoprecipitation (ChIP)-sequencing shows enrichment for IRF3-bound genomic regions in MM.1S cells (IRF3 ChIP-sequencing; n=2). (E) Metagene (top) and heatmap (bottom) representation of IRF3 genome-wide binding as assessed by IRF3 ChIP-sequencing in MM.1S cells along with indicated histone marks, RNA polymerase II (Pol II) binding and chromatin accessibility as assessed by ATAC-sequencing. Genomic feature annotation for each peak in the heatmap is shown on the left and the small color bar (below) indicates the genomic regions for 27,868 binding regions annotated. IRF3 binding is observed in genomic regions marked for active transcription i.e., with increased chromatin accessibility, activating chromatin marks (H3K27ac, H3K4me1/3) and Pol II binding. (F) Venn diagram showing numbers of genes predicted to be directly regulated (activated or repressed) by IRF3 in MM.1S cells as assessed by integration of IRF3 cistrome, IRF3 binding within 2 kb distance of the transcription start site, and IRF3-depleted transcriptome with cut-off Padj <0.05 using BETA-plus software. (G) IGV browser snapshots of IRF3 and Pol II binding, chromatin accessibility and histone mark enrichment at regulatory areas of several genes promoting cell cycle progression and cell proliferation. The red block on the top indicates 5 kb genome size.

728

haematologica | 2022; 107(3)

ZBP1 regulates myeloma cell proliferation via IRF3/IRF4

Figure 6. IRF3 co-operates with IRF4 and regulates cell cycle genes in myeloma cells. (A, B) IGV browser snapshots of IRF3 and IRF4 co-binding at the promoter and super-enhancer of IRF4 as assessed by chromatin immunoprecipitation (ChIP)-sequencing (A) and log2 fold-change of IRF4 expression (Padj <0.05) assessed by RNA-sequencing after depletion of indicated mRNA/protein in relation to the scrambled (scr)control in MM.1S cells (B). (C) Heatmaps of IRF3 and IRF4 genome-wide binding, and common binding regions of IRF3 and IRF4 (intersection) as their binding regions are intersected by Bedtools Intersect.50 Numbers of binding regions are shown in brackets. (D) Venn diagram showing the numbers of genomic regions of IRF3 and IRF4 co-binding with gene regulatory potential as assessed by integration of the whole transcriptome of IRF3-depleted MM.1S cells with IRF3- or IRF4-cistrome alone or IRF3 and IRF4 co-binding regions (intersection) in MM.1S cells. Here IRF4 genome-wide binding regions with very low scores were omitted and only the top 50% binding regions with highest scores were used. (E) Numbers of genes predicted to be directly co-regulated (repressed or activated) by IRF3 and IRF4 binding. The co-binding (intersection) regions were integrated with the IRF3-depleted transcriptome by shRNA1 and shRNA2 in MM.1S cells. (F) Enrichr pathway enrichment analysis for common genes predicted to be activated by IRF3-IRF4 co-binding. (G) Primary ChIP quantitative polymerase chain reaction (qPCR) against IRF3 (left) followed by re-ChIP-qPCR against IRF4 (right). The position of amplicons is shown as horizontal colored lines in Figure 6A. (H) Primary ChIP-qPCR against IRF4 (left) followed by re-ChIP-qPCR against IRF3 (right). The position of amplicons is shown as horizontal colored lines in Figure 6A.

ZBP1 interaction with IRF3 and TBK1 We next investigated the downstream processes that might link constitutive ZBP1 expression in myeloma cells with regulation of cell cycle. Unlike in non-malignant cells in which IRF3 phosphorylation/activation requires activation of sensors such as cGAS-STING,14,16 we found that IRF3 was constitutively phosphorylated (pIRF3) in myeloma cell lines (Figure 4A). In line with previous reports,39 pIRF3 was also detected in other non-myeloma cancer cells (Figure 4A). Importantly, pIRF3 was also detected in primary BM myeloma CD138+ PC (Figure 4B) and thus haematologica | 2022; 107(3)

establishing that IRF3 is constitutively phosphorylated in MM. Although the functional relationship of the ZBP1-IRF3 interaction in the context of cellular innate immune responses is a matter of debate, the physical interaction of ZBP1-IRF3 was previously demonstrated by their ectopic expression.17,40,40 Using protein co-immunoprecipitation assays we found that endogenous ZBP1 interacts with endogenous IRF3 (Figure 4C, D) and TBK1 (Figure 4E) in MM.1S cells. By co-transfection of IRF3 cDNA with a full length or C-terminus deleted mutant of ZBP1 (Online Supplementary Figure S5A), we found that the 729

K. Ponnusamy et al.

ZBP1-IRF3 interaction requires the RHIM domain-containing C-terminus of ZBP1 (Online Supplementary Figure S5B, C). Furthermore, while total IRF3 and TBK1 levels were not appreciably altered in ZBP1-depleted cells, pIRF3 and pTBK1 levels markedly decreased in both constitutive shRNA-transduced (Figure 4F) and doxycycline-induced shRNA targeting ZBP1 (Online Supplementary Figure S5D) in MM.1S cells. These findings suggest a post-translational dependency of IRF3 and TBK1 constitutive phosphorylation on ZBP1. Finally, shRNA-mediated depletion of TBK1 resulted in a decrease of IRF3 phosphorylation (Figure 4G). Together these data support a model whereby ZBP1 serves as a scaffold for TBK1-dependent constitutive phosphorylation of IRF3 in MMCL.

IRF3 regulates the cell cycle in myeloma cells As observed for ZBP1, IRF3 depletion also induces cell cycle arrest and apoptosis and thereby inhibits myeloma cell growth (Figure 4H, I and Online Supplementary Figure S6A-C). Of note, a similar effect was observed after depletion of TBK1 (Online Supplementary Figure S6D-G). In addition, transcriptome analysis of IRF3-depleted MM.1S cells revealed that among 185 genes downregulated by both anti-IRF3 shRNA1 and shRNA2, 109 genes were also commonly downregulated upon ZBP1 depletion (Figure 5A, B and Online Supplementary Table S4) and these are also enriched for cell cycle regulation (Figure 5C and Online Supplementary Table S5). However, only 23 genes were shared among those upregulated upon depletion of ZBP1 or IRF3. To identify candidate transcriptional targets of IRF3 in myeloma cells we generated and mapped its genomewide binding by IRF3 ChIP-sequencing in MM.1S cells. IRF3-bound regions (promoter, intergenic and intronic) (Online Supplementary Figure S7A) were highly enriched for IRF3-binding motifs (Figure 5D). In the same cells, we correlated genome-wide IRF3 binding with chromatin accessibility as assessed by an assay for transposase-accessible chromatin (ATAC)-sequencing, RNA polymerase II binding, and activating (H3K27ac and H3K4me1/2/3) and repressive (H3K27me3) histone marks (Figure 5E). This showed that IRF3 binding occurs in nearly 28,000 highly accessible chromatin regions with activating transcriptional potential as revealed by Pol II binding and the presence of activating histone marks. Thus, constitutively phosphorylated IRF3 in myeloma cells is highly transcriptionally active in the nucleus. Next, to obtain the compendium of genes directly regulated by IRF3 in MM.1S cells, we integrated the IRF3depleted transcriptome for each anti-IRF3 shRNA with the IRF3 cistrome using BETA-plus software (Figure 5F and Online Supplementary Figure S7D). After intersection of anti-IRF3 shRNA1 and shRNA2 data, we found that the 770 genes predicted to be directly activated by IRF3, included the key cell cycle regulators E2F1, E2F2, AURKB, CCNE1, MKI67 and MCM2–7 complex (Figure 5G, Online Supplementary Figure S7E, F and Online Supplementary Table S6), and were significantly enriched for cell cycle regulation pathways (Online Supplementary Figure S7B and Online Supplementary Table S7). Notably, the 339 genes predicted to be repressed by IRF3 were not enriched for IFN type I response genes (Figure 5F, Online Supplementary Figure S7C and Online Supplementary Tables S6 and S7). Consistent with this, while IRF3 binds in the regulatory regions of IFNA1 (but not of IFNB1) and ISG15, which are hallmark 730

type I IFN response genes, expression of these genes was not altered upon IRF3 depletion (Online Supplementary Figure S7G, H). IRF3 regulates transcription of and co-operates with IRF4 to regulate the cell cycle in myeloma cells. IRF4 is a transcription factor critical for normal PC development34 while in myeloma PC it co-operates with MYC to establish a transcriptional circuitry to which myeloma cells are highly addicted.41 Since the IRF4 motif was among the top-most enriched regions bound by IRF3 (Figure 5D), we explored potential synergy between IRF3 and IRF4 by overlaying our in-house-generated genome-wide binding profile of IRF3 with that of previously published IRF4 ChIP-sequencing in MM.1S cells.42 First, we observed cobinding of the two transcription factors at the promoter and the previously established super-enhancer of IRF443 (Figure 6A). This observation and the fact that IRF4 expression is significantly downregulated following IRF3 as well as ZBP1 depletion (Figure 6B) suggested that IRF4 transcriptional regulation is, at least in part, under control of the ZBP1-IRF3 axis. At a genome-wide level, 21,614 IRF3-bound chromatin regions were co-bound by IRF4 (Figure 6C, D). Correlating these with transcriptome changes following IRF3 depletion, we identified 612 and 267 genes predicted to be coactivated or co-repressed, respectively, by both IRF3 and IRF4 (Figure 6D, E) with the former highly enriched in cell cycle regulators (Figure 6F and Online Supplementary Tables S8 and S9). To validate this on-chromatin association of IRF3 with IRF4, we performed ChIP-re-ChIP assays at IRF3-IRF4 co-binding regions using a region upstream of IRF4 in which no binding of either transcription factor was observed as a negative control. In all tested regions we found specific co-occupancy of IRF3 and IRF4 including in the IRF4 promoter and super-enhancer regions (Figure 6G, H) and also at genes regulating the cell cycle including E2F1, E2F2, MCM2 and AURKB (Online Supplementary Figure S8A, B).

Discussion Here we demonstrate that the Z-nucleic acid sensor ZBP1 is an important and novel determinant of MM biology. We link the myeloma-selective constitutive expression of ZBP1 to constitutive IRF3 activation and regulation of IRF3-dependent IRF4 expression and myeloma cell proliferation. While other nucleic acid sensors, e.g., cGAS-STING, are expressed constitutively and are activated upon nucleic acid binding,14,16 ZBP1 differs in that its expression is only detected in response to nucleic acids, viral pathogens or inflammatory stimuli including interferons.17,8,20,44 Our extensive analysis confirmed that ZBP1 expression is low or not detected in all human normal and cancer cells tested with the striking exception of cells in the late B-cell development trajectory and in particular PC. Reflecting their cell of origin, we found constitutive ZBP1 expression also in MMCL and primary myeloma PC. While our data do not address the molecular role of Zbp1 in late B-cell development, at a cellular level, they identify suboptimal humoral immune response to a T-celldependent antigen in Zbp1-/- mice. Of note, in contrast to our results, antibody levels in Zbp1-/- mice were previously found to be intact in response to DNA vaccination;18 the haematologica | 2022; 107(3)

ZBP1 regulates myeloma cell proliferation via IRF3/IRF4

different routes and process of immunization might account for these differences. We found that ZBP1 depletion had a profound and selective effect on MMCL proliferation and survival in vitro and in vivo. Similarly, in primary myeloma PC which are less proliferative than MMCL, depletion of ZBP1 also induced cell cycle arrest. Based on appropriate design of ZBP1-targeting shRNA we could determine that myeloma cell proliferation is sustained by the isoform 1, which retains both Zα1 and Zα2 domains, but not by the isoform 2, which retains only the Zα2 domain. Future research will explore the nature and origin of nucleic acids that are bound by the Zα domain and their impact on the pro-proliferative function of ZBP1 in myeloma cells. Transcriptomes of ZBP1-depleted myeloma cells, which are driven by distinct primary oncogenes, i.e., MAF (MM.1S cells) and MMSET (H929 cells), highlighted cell cycle regulation as one of the main pathways regulated by ZBP1. This novel pro-proliferative function of ZBP1 contrasts with the anti-proliferative potential of IFNβ which can induce and sustain expression of ZBP1.45,20 However, although GSEA suggested that ZBP1 mediates repression of IFN type I response in ZBP1-depleted MM.1S cells, study of a large number of primary myeloma PC transcriptomes revealed a strong IFN type I response transcriptional signature as well as enrichment for cell cycle pathways among upregulated genes in ZBP1hi myeloma PC. Together, these findings are consistent with a model whereby an active IFN type I response restrains myeloma PC proliferation in the low proliferative early phase myeloma PC but sustains ZBP1 expression, which exerts a limited pro-proliferative function. In contrast, while the IFN type I transcriptional program is attenuated or even repressed in the highly proliferative MMCL such as MM.1S cells, which are representative of advanced MM,21,46 persistent ZBP1 expression regulates proliferation which is not constrained by the IFN type I response. In line with this model, a recent comparison of primary myeloma PC and MMCL transcriptomes demonstrated enrichment for the IFN type I response gene signature in myeloma PC in more than 700 MM patients at diagnosis while proliferative but not IFN type I gene signatures were dominant in relapsed disease myeloma PC and in MMCL.21 The physiological role of ZBP1 is to promote necroptosis and inflammation through interaction with RIPK3 in response to pathogens or cellular dsRNA.47,10,12 While interaction of ZBP1 with IRF3 and TBK1 has been shown previously in an ectopic expression system,17,40 whether it regulates the type I IFN response has been disputed.18 Here we confirmed direct and functional interactions of endogenous ZBP1-IRF3-TBK1 in myeloma cells which highlights ZBP1 as a physical platform that directs activation of TBK1 and IRF3. While transcriptional activation by phosphorylation of IRF3 in response to inflammatory stimuli is expected to be transient, we found that IRF3 is constitutively phosphorylated in both primary myeloma cells and cell lines. This is not unique to MM since constitutively phosphorylated IRF3 has been reported in several ZBP1-negative cancer lines39 but not functionally investigated although a proproliferative effect of IRF3 has been reported in acute myeloid leukemia cells at a cellular level.48 Importantly, we

haematologica | 2022; 107(3)

demonstrate that IRF3 binds to transcriptionally active regions of the genome and it directly regulates genes that promote cell cycle progression in myeloma cells. Accordingly, IRF3 depletion in myeloma cells leads to cell cycle arrest and apoptosis. We also show that TBK1 depletion results in cell cycle arrest and apoptosis in myeloma cells. Although this cellular effect is likely linked to downstream regulation of cell cycle genes by pIRF3, other mechanisms are also possible since TBK1 is a pleiotropic kinase.49 IRF4, the lineage-defining transcription factor in PC development,34 establishes an aberrant transcriptional circuity with MYC that renders myeloma PC highly dependent on an oncogenic program that includes activation of the cell cycle among other pathways.41 Here, based on IRF3 binding to the super-enhancer and promoter of IRF4 and the fact that depletion of IRF3 (also ZBP1) results in significant IRF4 downregulation, we demonstrate direct transcriptional activation of IRF4 by IRF3. Indeed, since as little as 50% reduction in IRF4 expression levels is toxic to myeloma cells,41 the greater than 50% reduction in IRF4 mRNA induced by IRF3 depletion would be expected to contribute significantly to myeloma cell death. Our genome-wide and sequential ChIP assays demonstrated and validated extensive co-occupancy of IRF3 and IRF4 in the myeloma regulatory genome including at the super-enhancer of IRF4, with genes involved in cell cycle control being among the targets of the IRF3-IRF4 synergy. In summary, our data show that like other nucleic acid sensors, ZBP1 can regulate cellular pathways critical for cancer biology. We show a constitutively active ZBP1-IRF3 axis that is co-opted into promoting proliferative pathways in myeloma PC and regulating expression of the critical myeloma oncogene IRF4. Guided by our initial delineation of the structural requirements of the ZBP1-IRF3 interaction in myeloma cells, disruption of the ZBP1-IRF3 axis will offer an opportunity for targeted and relatively selective therapeutic intervention in MM.

Disclosures No conflicts of interest to disclose. Contributions KP and AK conceived and designed the study; KP and MMT performed the co-immunoprecipitation and doxycycline-inducible study in vitro; KP and MB performed the shTBK1 study. PT and KN performed immunohistochemistry; AC processed RNAsequencing data for ZBP1; MER processed RNA-sequencing and ChIP-sequencing data for ZBP1 and IRF3; KP performed all other experiments in vitro, in vivo and integrated all the ChIP-sequencing and RNA-sequencing data and performed all other bioinformatics analysis and created all the figures; DI provided erythroblast cells; VSC, AK, NT, XX, IVK, IR, and HWA provided reagents; AK supervised the study. KP and AK wrote the manuscript. Funding We acknowledge funding from Blood Cancer UK (to KP, NT, XX, and VC), KKLF (to AK), and the Imperial NIHR Biomedical Research Centre, LMS/NIHR Imperial Biomedical Research Centre Flow Cytometry Facility, Imperial BRC Genomics Facility and the MRC/LMS Sequencing Facility for support.

731

K. Ponnusamy et al.

References 1. Palumbo A, Anderson K. Multiple myeloma. N Engl J Med. 2011;364(11):1046-1060. 2. Morgan GJ, Walker BA, Davies FE. The genetic architecture of multiple myeloma. Nat Rev Cancer. 2012;12(5):335-348. 3. 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):100-113. 4. Shi W, Liao Y, Willis SN, et al. Transcriptional profiling of mouse B cell terminal differentiation defines a signature for antibody-secreting plasma cells. Nat Immunol. 2015;16(6):663-673. 5. Ha SC, Kim D, Hwang H-Y, Rich A, Kim Y-G, Kim KK. The crystal structure of the second Z-DNA binding domain of human DAI (ZBP1) in complex with Z-DNA reveals an unusual binding mode to ZDNA. Proc Natl Acad Sci U S A. 2008;105 (52):20671-20676. 6. Schwartz T, Behlke J, Lowenhaupt K, Heinemann U, Rich A. Structure of the DLM-1–Z-DNA complex reveals a conserved family of Z-DNA-binding proteins. Nat Struct Biol. 2001;8(9):761-765. 7. Maelfait J, Liverpool L, Bridgeman A, Ragan KB, Upton JW, Rehwinkel J. Sensing of viral and endogenous RNA by ZBP1/DAI induces necroptosis. EMBO J. 2017; 36(17):2529-2543. 8. Thapa RJ, Ingram JP, Ragan KB, et al. DAI senses influenza A virus genomic RNA and activates RIPK3-dependent cell death. Cell Host Microbe. 2016;20(5):674-681. 9. Sridharan H, Ragan KB, Guo H, et al. Murine cytomegalovirus IE3-dependent transcription is required for DAI/ZBP1mediated necroptosis. EMBO Rep. 2017;18(8):1429-1441. 10. Upton JW, Kaiser WJ, Mocarski ES. DAI/ZBP1/DLM-1 complexes with RIP3 to mediate virus-induced programmed necrosis that is targeted by murine cytomegalovirus vIRA. Cell Host Microbe. 2012;11(3):290-297. 11. Newton K, Wickliffe KE, Maltzman A, et al. RIPK1 inhibits ZBP1-driven necroptosis during development. Nature. 2016; 540(7631):129-133. 12. Lin J, Kumari S, Kim C, et al. RIPK1 counteracts ZBP1-mediated necroptosis to inhibit inflammation. Nature. 2016; 540(7631):124-128. 13. Jiao H, Wachsmuth L, Kumari S, et al. Znucleic-acid sensing triggers ZBP1-dependent necroptosis and inflammation. Nature. 2020;580(7803):391-395. 14. Tanaka Y, Chen ZJ. STING specifies IRF3 phosphorylation by TBK1 in the cytosolic DNA signaling pathway. Sci Signal. 2012;5(214):ra20. 15. Ablasser A, Goldeck M, Cavlar T, et al. cGAS produces a 2′-5′-linked cyclic dinucleotide second messenger that activates STING. Nature. 2013;498(7454):380-384. 16. Mankan AK, Schmidt T, Chauhan D, et al. Cytosolic RNA:DNA hybrids activate the cGAS-STING axis. EMBO J. 2014; 33(24):2937-2946. 17. Takaoka A, Wang Z, Choi MK, et al. DAI

732

(DLM-1/ZBP1) is a cytosolic DNA sensor and an activator of innate immune response. Nature. 2007;448(7152):501-505. 18. Ishii KJ, Kawagoe T, Koyama S, et al. TANK-binding kinase-1 delineates innate and adaptive immune responses to DNA vaccines. Nature. 2008;451(7179):725-729. 19. Yang D, Liang Y, Zhao S, et al. ZBP1 mediates interferon-induced necroptosis. Cell Mol Immunol. 2020;17:356-368. 20. Kuriakose T, Man SM, Malireddi RKS, et al. ZBP1/DAI is an innate sensor of influenza virus triggering the NLRP3 inflammasome and programmed cell death pathways. Sci Immunol. 2016;1(2):aag2045. 21. Sarin V, Yu K, Ferguson ID, et al. Evaluating the efficacy of multiple myeloma cell lines as models for patient tumors via transcriptomic correlation analysis. Leukemia. 2020;34(10):2754–2765. 22. Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29(1):15-21. 23. Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017;14(4):417419. 24. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):550. 25. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43(7):e47. 26. Zhang Y, Liu T, Meyer CA, et al. Modelbased analysis of ChIP-Seq (MACS). Genome Biol. 2008;9(9):R137. 27. Ramírez F, Ryan DP, Grüning B, et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res. 2016;44(W1):W160-165. 28. Thorvaldsdóttir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinform. 2013;14(2):178-192. 29. Heinz S, Benner C, Spann N, et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell. 2010;38(4):576-589. 30. Wang S, Sun H, Ma J, et al. Target analysis by integration of transcriptome and ChIPseq data with BETA. Nat Protoc. 2013;8(12):2502-2515. 31. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545-15550. 32. Kuleshov MV, Jones MR, Rouillard AD, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res. 2016;44(W1): W90-97. 33. Cattoretti G, Shaknovich R, Smith PM, Jäck H-M, Murty VV, Alobeid B. Stages of germinal center transit are defined by B cell transcription factor coexpression and relative abundance. J Immunol. 2006; 177(10):6930-6939. 34. Ochiai K, Maienschein-Cline M, Simonetti G, et al. Transcriptional regulation of ger-

minal center B and plasma cell fates by dynamical control of IRF4. Immunity. 2013;38(5):918-929. 35. Buehler E, Chen Y-C, Martin S. C911: a bench-level control for sequence specific siRNA off-target effects. PLoS One. 2012;7(12):e51942. 36. Zuber J, McJunkin K, Fellmann C, et al. Toolkit for evaluating genes required for proliferation and survival using tetracycline-regulated RNAi. Nat Biotechnol. 2011;29(1):79-83. 37. Laganà A, Perumal D, Melnekoff D, et al. Integrative network analysis identifies novel drivers of pathogenesis and progression in newly diagnosed multiple myeloma. Leukemia. 2018;32(1):120-130. 38. Kuiper R, Broyl A, de Knegt Y, et al. A gene expression signature for high-risk multiple myeloma. Leukemia. 2012;26(11):24062413. 39. Muvaffak A, Pan Q, Yan H, et al. Evaluating TBK1 as a therapeutic target in cancers with activated IRF3. Mol Cancer Res. 2014;12(7):1055-1066. 40. Wang Z, Choi MK, Ban T, et al. Regulation of innate immune responses by DAI (DLM1/ZBP1) and other DNA-sensing molecules. Proc Natl Acad Sci U S A. 2008; 105(14):5477-5482. 41. Shaffer AL, Emre NCT, Lamy L, et al. IRF4 addiction in multiple myeloma. Nature. 2008;454(7201):226-231. 42. Zheng R, Wan C, Mei S, et al. Cistrome Data Browser: expanded datasets and new tools for gene regulatory analysis. Nucleic Acids Res. 2019;47(D1):D729-D735. 43. Jin Y, Chen K, De Paepe A, et al. Active enhancer and chromatin accessibility landscapes chart the regulatory network of primary multiple myeloma. Blood. 2018; 131(19):2138-2150. 44. Kesavardhana S, Malireddi RKS, Burton AR, et al. The Zα2 domain of ZBP1 is a molecular switch regulating influenzainduced PANoptosis and perinatal lethality during development. J Biol Chem. 2020;295(24): 8325-8330. 45. Yang D, Liang Y, Zhao S, et al. ZBP1 mediates interferon-induced necroptosis. Cell Mol Immunol. 2020;17(4):356-368. 46. Hose D, Rème 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):8795. 47. Rebsamen M, Heinz LX, Meylan E, et al. DAI/ZBP1 recruits RIP1 and RIP3 through RIP homotypic interaction motifs to activate NF-kappaB. EMBO Rep. 2009; 10(8):916-922. 48. Tian W-L, Jiang Z-X, Wang F, et al. IRF3 is involved in human acute myeloid leukemia through regulating the expression of miR155. Biochem Biophys Res Commun. 2016;478(3):1130-1135. 49. Pillai S, Nguyen J, Johnson J, Haura E, Coppola D, Chellappan S. Tank binding kinase 1 is a centrosome-associated kinase necessary for microtubule dynamics and mitosis. Nat Commun. 2015;6(1):1-14. 50. Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841-842.

haematologica | 2022; 107(3)

LETTERS TO THE EDITOR Early mortality and survival improvements for adolescents and young adults with acute promyelocytic leukemia in California: an updated analysis In the last two decades, survival from newly diagnosed acute promyelocytic leukemia (APL) has improved substantially after the introduction of all-trans retinoic acid (ATRA) and arsenic trioxide (ATO).1 Nevertheless, recent population-based studies revealed that mortality within 30 days of diagnosis remained high despite greater awareness of APL-specific coagulopathy and guidelines recommending prompt initiation of ATRA and aggressive supportive care as soon as APL is suspected.2,3 We previously showed that among patients aged ≤39 years diagnosed with APL in California, 30-day mortality decreased from 26% before ATRA (1988-1995) to 14% after ATRA

(2004-2011).4 However, 7-day mortality did not differ between the pre- and post-ATRA eras. A higher risk of 30-day mortality and inferior overall survival were observed among patients without health insurance and those of Hispanic race/ethnicity. In this update, we examined whether early mortality decreased in the most recent eras of treatment and also examined overall survival trends among adolescents and young adults (15-39 years) with APL. We used data from the California Cancer Registry linked to Medicaid enrollment files. Health insurance was classified based on Medicaid enrollment from 6 months prior to 6 months after APL diagnosis. The following categories were created: continuous Medicaid (enrollment 5 or 6 months prior to APL diagnosis), uninsurance or Medicaid at diagnosis (enrollment begins in the month prior to or within 2 months after diagnosis to account for reactive enrollment), discontinuous Medicaid

Table 1. Characteristics of adolescents and young adults (n=524) with acute promyelocytic leukemia by period of diagnosis, California, 20052017.

Early mortality 7-day mortality 30-day mortality Median age at diagnosis (IQR), years Sex Female Male Health insurance* Private/military& Continuous Medicaid¥ Discontinuous Medicaid¶ Medicaid at diagnosis/uninsuredµ Care facility NCI-CC Non-NCI-CC Race/ethnicity** Non-Hispanic White Non-Hispanic Black Hispanic Asian/Pacific Islander Neighborhood SES (tertiles) Highest Medium Lowest Total deceased

Pre-ACA† (2005–2010) N=216 n (%)

Period of diagnosis Early ACA‡ (2010–2013) N=144 n (%)

Full ACA& (2014–2017) N=164 n (%)

P-valuea

28 (13.0) 31 (14.4) 29 (22–35)

11 (7.6) 15 (10.4) 29 (23–35)

9.0 (5.5) 12 (7.3) 28 (23–33)

0.011 0.029 0.567

103 (47.7) 113 (52.3)

78 (54.2) 66 (45.8)

85 (52.0) 79 (48.2)

0.458

94 (43.5) 30 (13.9) 14 (6.5) 61 (28.2)

59 (41.0) 20 (13.9) 15 (10.4) 47 (32.6)

49 (29.9) 66 (40.2) 13 (7.9) 30 (18.3)

<0.001

77 (35.7) 139 (64.4)

54 (37.5) 90 (62.5)

68 (41.5) 96 (58.5)

0.507

62 (28.7) 12 (5.6) 113 (52.3) 25 (11.6)

45 (31.3) 6 (4.2) 76 (52.8) 16 (11.1)

35 (21.3) 14 (8.5) 88 (53.7) 24 (14.6)

0.472

54 (25.0) 72 (33.3) 90 (41.7) 49 (22.7)

32 (22.2) 62 (43.1) 50 (34.7) 19 (13.2)

44 (26.8) 52 (31.7) 68 (41.5) 12 (7.3)

0.271

<0.001

ACA: Affordable Care Act; SES: socioeconomic status; NCI-CC: National Cancer Institute-Designated Cancer Center; IQR: interquartile range. †Pre-ACA: March 2005 to September 2010; ‡Early ACA: October 2010 to December 2013; &Full ACA: 2014 to 2017. &Private insurance includes: Military, Health Maintenance Organization, Preferred Provider Organization, and Medicare. ¥Continuous Medicaid: enrollment 5 or 6 months prior to diagnosis. mMedicaid at diagnosis/uninsurance: coverage beginning in the month prior to or within 2 months after diagnosis to account for reactive enrollment. ¶Discontinuous Medicaid: enrollment that does not meet the definitions for continuous enrollees or Medicaid at diagnosis or Medicaid insurance recorded in the California Cancer Registry but without a match in the Medicaid enrollment files. *Other public (n=3, 0.6%) and unknown (n=23, 4.4%) insurances are not presented in the table. Other public insurance includes Indian/ Public Health Service, County Funded, not otherwise specified (NOS); Medicare without supplement; Medicare, NOS. **Other race/ethnicity (American Indian, n=4) and unknown (n=4) are not presented in the table. Data on Hispanics14 and Asian/Pacific Islanders15 are derived from a combination of variables based on the North American Association of Central Cancer Registries. ac2 P-values were used to compare frequency distributions of sociodemographic and clinical characteristics by treatment. Tests for trends were used to examine early mortality by treatment era.

haematologica | 2022; 107(3)

733

Letters to the Editor

Table 2. Relation of early mortality and overall survival with period of diagnosis, sociodemographic factors and care facility, California, 20052017.

Period of diagnosis Pre-ACA† Early ACA‡ Full ACA& Sex Female Male Age at diagnosis (years) 15–20 21–25 26–39 Race/ethnicity* Non-Hispanic White Non-Hispanic Black Hispanic Asian/Pacific Islander Neighborhood SES Highest Middle Lowest Care facility NCI-CC Non-NCI-CC Health insurance* Private/Military& Continuous Medicaid¥ Discontinuous Medicaid¶ Medicaid at diagnosis/uninsuredµ

N. of patients (%)

7-day mortality Adjusted OR§ (95% CI)

30-day mortality Adjusted OR§ (95% CI)

Overall survival Adjusted HR§ (95% CI)

216 (41.2) 144 (27.5) 164 (31.3)

Reference 0.57 (0.27–1.23) 0.42 (0.18–0.98)

Reference 0.72 (0.36–1.44) 0.51 (0.24–1.08)

Reference 0.60 (0.35–1.03) 0.39 (0.20–0.76)

266 (50.8) 258 (49.2)

Reference 1.21 (0.64–2.30)

Reference 0.99 (0.55–1.78)

Reference 1.25 (0.78–1.98)

95 (18.1) 100 (19.1) 329 (62.8)

Reference 1.07 (0.39–2.90) 0.87 (0.38–1.98)

Reference 0.72 (0.28–1.83) 0.77 (0.37–1.63)

Reference 0.94 (0.45–1.96) 0.93 (0.52–1.67)

142 (27.1) 32 (6.1) 277 (52.9) 65 (12.4)

Reference 0.95 (0.18–4.99) 1.97 (0.82–4.74) 1.45 (0.43–4.92)

Reference 1.19 (0.28–4.98) 2.16 (0.95–4.94) 1.25 (0.38–4.10)

Reference 1.27 (0.45–3.64) 1.56 (0.84–2.91) 1.40 (0.60–3.23)

130 (24.8) 186 (35.5) 208 (39.7)

Reference 1.44 (0.54–3.87) 2.09 (0.78–5.60)

Reference 1.38 (0.55–3.46) 2.27 (0.91–5.66)

Reference 1.04 (0.52–2.08) 1.69 (0.86–3.33)

199 (38.0) 325 (62.0)

Reference 4.85 (1.96–12.0)

Reference 5.24 (2.26–12.2)

Reference 3.71 (1.97–6.99)

202 (38.6) 116 (22.1) 42 (8.0) 138 (26.3)

Reference 0.77 (0.29–2.07) 1.12 (0.35–3.57) 1.00 (0.44–2.29)

Reference 0.71 (0.29–1.76) 1.28 (0.46–3.61) 0.98 (0.46–2.10)

Reference 0.86 (0.42–1.77) 1.52 (0.70–3.29) 1.04 (0.57–1.92)

OR: odds ratio; HR: hazard ratio; ACA: Affordable Care Act; SES: socioeconomic status; NCI-CC, National Cancer Institute-Designated Cancer Center. †Pre-ACA: March 2005 to September 2010, ‡Early ACA: October 2010 to December 2013; &Full ACA: 2014 to 2017 *Other/unknown race/ethnicity and other/unknown health insurance was omitted from the table due to the small number of events. Data on Hispanics14 and Asian/Pacific Islanders15 are derived from a combination of variables based on the North American Association of Central Cancer Registries. &Private insurance includes: Military, Health Maintenance Organization, Preferred Provider Organization, and Medicare. ¥ Continuous Medicaid: enrollment 5 or 6 months prior to the diagnosis of acute prolmyelocytic leukemia. µMedicaid at diagnosis/uninsurance: coverage beginning in the month prior to or within 2 months after diagnosis to account for reactive enrollment. ¶Discontinuous Medicaid: enrollment that does not meet the definitions for continuous enrollees or Medicaid at diagnosis or Medicaid insurance recorded in the California Cancer Registry but without a match in the Medicaid enrollment files. § Multivariable logistic regression (for early mortality) and Cox proportional regression models (for overall survival) were adjusted for all variables in the table. The proportional hazard assumption, assessed by examining log-log survival plots and confirmed using Schoenfeld residuals, was met for all variables in the multivariate Cox regression model.

(enrollment that does not meet the definitions for continuous enrollees, Medicaid at diagnosis or Medicaid insurance recorded in the California Cancer Registry but without a match in the Medicaid enrollment files), private, other public or unknown. We included three periods of diagnosis that reflect insurance policy changes in California under the Affordable Care Act (ACA): March 2005 to September 2010 (pre-ACA), October 2010 to December 2013 (early ACA: Dependent Care Expansion and early Medicaid expansion), and January 2014 to December 2017 (full ACA: Medicaid expansion and private health insurance marketplace). Neighborhood socioeconomic status is an aggregate measure at the census block level,5 which contains U.S. Census or American Community Survey information on education, occupation, unemployment, household income, poverty, rent and house values. 734

We examined the associations of 7- and 30-day mortality and overall survival with period of diagnosis, sociodemographic characteristics, and care facility using multivariable logistic regression and Cox proportional hazards regression, respectively. Results are presented as odds ratios (OR) or hazard ratios (HR) and corresponding 95% confidence intervals (95% CI). We also estimated 3-year overall survival using the Kaplan-Meier method. Patients were followed from the date of diagnosis to the date of death, loss to follow-up or end of the study (December 31, 2018), whichever occurred first. Five-hundred twenty-four adolescents and young adults diagnosed with APL between 2005 and 2017 were included in the analyses. Of these, 50.8% were female, 52.9% were of Hispanic race/ethnicity, 38.6% had private insurance, and 39.7% resided in the lowest socioeconomic neighborhoods. Fewer patients (38.0%) were haematologica | 2022; 107(3)

Letters to the Editor

treated at National Cancer Institute-Designated Cancer Centers (NCI-CC). Overall, 7- and 30-day mortality were 9.2% and 11.1%, respectively. From the pre-ACA to full ACA era, 7-day mortality decreased from 13.0% to 5.5% (P-value for trend=0.011) and 30-day mortality decreased from 14.4% to 7.3% (P-value for trend=0.029) (Table 1). With a median follow-up of 4.7 years (range, 0-14.1), 80 patients died during the study period: 48 (60%) died within 7 days and 58 (72.5%) died within 30 days of the diagnosis of APL. The 3-year survival increased from 79.8% before the ACA to 92.7% in the full ACA era (Pvalue for trend=0.0004). Across treatment eras, there was a decrease in adolescents and young adults who were uninsured or obtained Medicaid insurance at APL diagnosis, from 28.2% (preACA) to 18.3% (full ACA). Among those, a few adolescents and young adults (n=9, 1.7%) remained uninsured in the pre- and early-ACA eras, but none was uninsured in the full ACA era. In addition, there was a substantial increase in continuous Medicaid enrollment prior to diagnosis, from 13.9% (pre-ACA) to 40.2% (full ACA) (P-value<0.0001). Furthermore, we observed a trend towards increased initial care at NCI-CC, from 35.7% before the ACA to 41.5% in the full ACA era, although this difference was not statistically significant. In multivariable models, compared with the pre-ACA era, the odds of both 7- and 30-day mortality were about 50-60% lower in the full ACA era (OR=0.42, 95% CI: 0.18-0.98 and OR=0.51, 95% CI: 0.24-1.08, respectively) (Table 2). Remarkably, the odds of 7- and 30-day mortality were approximately 5-fold higher among adolescents and young adults who did not receive initial care at a NCI-CC (OR=4.85, 95% CI: 1.96-12.0 and OR=5.28, 95% CI: 2.26–12.2, respectively). Likewise, the hazard of death decreased in the full ACA era compared to that in the pre-ACA era (HR=0.39, 95% CI: 0.20–0.76), and was nearly 4 times higher for patients who did not receive initial care at a NCI-CC (HR=3.71, 95% CI: 1.97-6.99). Age at diagnosis, sex, race/ethnicity, socioeconomic status and health insurance were not associated with early mortality or overall survival. Our findings of reduced early mortality and improved overall survival during the study period are relevant. Most early deaths after APL result from severe hemorrhage (intracranial and, less often, pulmonary) and are the major cause of treatment failure.6 Another potentially fatal manifestation of APL is thrombosis (e.g., stroke, acute myocardial infarction, and pulmonary embolism). These manifestations can occur within a few hours or days of diagnosis or even prior to the recognition of APL, supporting the concept that APL should be considered as a medical emergency.7 Thus, timely access to optimal treatment is lifesaving. Several factors may have contributed to the better outcomes we observed among adolescents and young adults with APL in the more recent treatment eras. These include increased awareness of APL-related coagulopathy, improved physicians’ adherence to treatment guidelines, and early referral to hematology/oncology centers.6,8 In our previous study, Hispanic patients and those without health insurance experienced worse outcomes. In this update, we did not observe differences in early mortality or survival by race/ethnicity. This observation is encouraging and is likely the result of the increased insurance coverage we observed among Hispanics over time (uninsurance decreased from 46.3% before the ACA to 21.1% in the full ACA era). Historically, in the USA, adolescents and young adults have been the most highly uninsured or underinsured population.9 It is likely that haematologica | 2022; 107(3)

adolescents and young adults without health insurance delay seeking medical care until the APL symptoms become severe, whereas those with insurance may be encouraged to seek medical attention earlier, when the first manifestations of APL appear. In this current study, we found increases in continuous Medicaid and reduction in Medicaid enrollments at diagnosis or uninsurance among adolescents and young adults with APL. Before implementation of the ACA, a restricted number of adults with a cancer diagnosis would qualify for Medicaid coverage.10 In California, a Medicaid expansion state, the ACA increased health insurance coverage to low-income young adults through the Dependent Coverage Expansion and Low-Income Health Plan as early as 2010. At the beginning of 2014, the Low-Income Health Plan ended, and patients obtained coverage under the California regular Medicaid program (Medi-Cal) or acquired private health insurance through the establishment of marketplace coverage. Marketplace insurance plans are prohibited to deny coverage to insurers or set higher premiums for pre-existing conditions such as cancer.10 Although we did not find an association between insurance and APL outcomes in our cohort, when we separately analyzed patients who remained uninsured (n=9), we found that their 7-day and 30-day mortality were significantly higher. This suggests that uninsured patients are at risk of not having prompt access to optimal care during the initial diagnosis of APL. Importantly, we found that higher early mortality and worse survival were strongly associated with location of initial care, consistent with prior reports.2 This may be related, in part, to greater awareness and provider expertise gained by treating a larger volume of patients with acute leukemias at NCI-CC.11 In addition, these facilities generally have better access to blood products and broader specialist support to care for severely ill patients. Changes in therapy during the study period are risk-stratified approaches that include the combination of ATRA and ATO, cytotoxic chemotherapy for high-risk patients, pre-emptive therapy for differentiation syndrome, and more aggressive blood product support for coagulopathy.8,12,13 Adherence to more modern approaches is more likely at NCI-CC and may have contributed to the improvement in outcomes we observed. In California, there are eight NCI-CC in seven cities, so timely access to NCI-CC may not be possible for many adolescents and young adults with APL, reinforcing the need to improve care at non-NCI-CC. Our study limitations include a lack of treatment details, which prevented us from assessing adherence to treatment guidelines. In addition, we did not have clinical information for APL risk stratification. These data could contribute to our understanding of different factors that can influence early mortality in young patients with APL. Regardless of these limitations, we used populationbased data from the most populous and diverse U.S. state where cancer registration is mandatory, providing information on virtually all adolescents and young adults diagnosed with APL during the study period. We showed a considerable reduction in early mortality and increased survival over time. The higher 7-day and 30-day mortality and lower survival observed in adolescents and young adults cared for outside of NCI-CC underscore the need for earlier recognition and better implementation of treatment guidelines, particularly at non-NCI-CC. Renata Abrahão,1,2 Raul C. Ribeiro,3 Marcio H. Malogolowkin,4 Ted Wun1 and Theresa H.M. Keegan1 735

Letters to the Editor

1 Center for Oncology Hematology Outcomes Research and Training (COHORT), Division of Hematology and Oncology, University of California Comprehensive Cancer Center, Sacramento, CA; 2Center for Healthcare Policy and Research, University of California, Davis, School of Medicine, Sacramento, CA; 3Department of Oncology, Division of Hematology and Oncology, St. Jude Children’s Research Hospital, Memphis, TN and 4Department of Pediatrics, Division of Hematology and Oncology, University of California, Davis, School of Medicine, Sacramento, CA, USA Correspondence: RENATA ABRAHAO - rabrahao@ucdavis.edu doi:10.3324/haematol.2021.278851 Received: March 26, 2021. Accepted: May 4, 2021. Pre-published: July 29, 2021. Disclosures: no conflicts of interest to disclose. Contributions: RA and THMK conceived and designed the study; THMK and TW acquired data; RA analyzed data and drafted the manuscript. All authors reviewed and approved the final version of the manuscript. Funding: RA was supported by the Health Resources and Services Administration (HRSA) of the U.S. Department of Health and Human Services (HHS) under grant number T32HP30037 for Research in Primary Care. THMK was supported by the UC Davis Comprehensive Cancer Center (P30CA093373). TW was supported by UL1 0000860, National Center for Advancing Translational Science, National Institute of Health (NIH). RCR was partially funded by the National Cancer Institute (NCI) grant CA21765 and by the American Lebanese and Syrian Associated Charities. The collection of cancer incidence data used in this study was supported by the California Department of Public Health pursuant to California Health and Safety Code Section 103885; Centers for Disease Control and Prevention’s (CDC) National Program of Cancer Registries, under cooperative agreement 5NU58DP006344; the NCI’s Surveillance, Epidemiology and End Results (SEER) Program under contract HHSN261201800032I awarded to the University of California, San Francisco, contract HHSN261201800015I awarded to the University of Southern California, and contract HHSN261201800009I awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the authors and do not necessarily reflect the opinions of the State of California, Department of Public Health, the NCI, and the CDC or their Contractors and Subcontractors. Likewise, they should not be construed as the official position or policy of, nor should any endorsements be inferred by HRSA, HHS or the U.S. Government.

736

References 1. Sanz MA, Grimwade D, Tallman MS, et al. Management of acute promyelocytic leukemia: recommendations from an expert panel on behalf of the European LeukemiaNet. Blood. 2009;113(9):18751891. 2. Ho G, Li Q, Brunson A, et al. Complications and early mortality in patients with acute promyelocytic leukemia treated in California. Am J Hematol. 2018;93(11): E370-E372. 3. Lehmann S, Deneberg S, Antunovic P, et al. Early death rates remain high in high-risk APL: update from the Swedish Acute Leukemia Registry 1997-2013. Leukemia. 2017;31(6):1457-1459. 4. Abrahao R, Ribeiro RC, Medeiros BC, et al. Disparities in early death and survival in children, adolescents, and young adults with acute promyelocytic leukemia in California. Cancer. 2015; 121(22):3990-3997. 5. Yost K, Perkins C, Cohen R, et al. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control. 2001;12(8):703-711. 6. Abla O, Lo-Coco F, Sanz MA. Acute Promyelocytic Leukemia. First edition: Springer International Publishing (Chapter 6). 2018. 7. Lo-Coco F, Cicconi L, Voso MT. Progress and criticalities in the management of acute promyelocytic leukemia. Oncotarget. 2017; 8(59): 99221-99222. 8. Burnett AK, Russell NH, Hills RK, et al. Arsenic trioxide and all-trans retinoic acid treatment for acute promyelocytic leukaemia in all risk groups (AML17): results of a randomised, controlled, phase 3 trial. Lancet Oncol. 2015;16(13):1295-1305. 9. Kirchhoff AC, Lyles CR, Fluchel M, et al. Limitations in health care access and utilization among long-term survivors of adolescent and young adult cancer. Cancer. 2012;118(23):5964-5972. 10. Zhao J, Mao Z, Fedewa SA, et al. The Affordable Care Act and access to care across the cancer control continuum: a review at 10 years. CA Cancer J Clin. 2020;70(3):165-181. 11. Giri S, Pathak R, Aryal MR, et al. Impact of hospital volume on outcomes of patients undergoing chemotherapy for acute myeloid leukemia: a matched cohort study. Blood. 2015;125(21):3359-3360. 12. Lo-Coco F, Di Donato L, Schlenk RF. Targeted therapy alone for acute promyelocytic leukemia. N Engl J Med. 2016;374(12):11971198. 13. Cicconi L, Platzbecker U, Avvisati G, et al. Long-term results of alltrans retinoic acid and arsenic trioxide in non-high-risk acute promyelocytic leukemia: update of the APL0406 Italian-German randomized trial. Leukemia. 2020;34(3):914-918. 14. NAACCR Latino Research Work Group. NAACCR Guideline for Enhancing Hispanic-Latino Identification: Revised NAACCR Hispanic/Latino Identification Algorithm [NHIA v2]. Springfield (IL). North American Association of Central Cancer Registries, September 2005. 15. NAACCR Race and Ethnicity Work Group. NAACCR Asian Pacific Islander Identification Algorithm [NAPIIA v1.2.1]. Springfield (IL), August 2011.

haematologica | 2022; 107(3)

Letters to the Editor

Dual pyroptotic biomarkers predict erythroid response in lower-risk non-del(5q) myelodysplastic syndromes treated with lenalidomide and recombinant erythropoietin Symptomatic anemia is the most common manifestation in patients with lower-risk myelodysplastic syndrome (MDS) that requires treatment. Although both recombinant erythropoietin and lenalidomide are modestly effective as single agents, combined treatment yields a significantly higher and durable erythroid response.1,2 However, less than 50% of patients respond, therefore predictive biomarkers may assist in treatment selection for these patients. NLRP3 inflammasome-directed pyroptosis, irrespective of the functional classes of underlying somatic mutations, drives ineffective erythropoiesis, the common phenotype of lower-risk MDS.3 Although how MDS-related somatic mutations of disparate functional

classes (e.g., RNA splicing and epigenetic regulation) activate the NLRP3 inflammasome remains elusive, they are linked to genomic instability with excessive reactive oxygen species production.4,5 Furthermore, NLRP3 can also be activated by aberrantly elevated proinflammatory cytokines, such as S100A9, in the bone marrow microenvironment of MDS.4 Apoptosis-associated speck-like protein containing a CARD (ASC) is the adapter molecule that binds NLRP3 in response to activating signals, which then polymerizes to generate docking sites for the caspase-1 effector that instructs a number of cellular processes, including interleukin-1β production and lytic cell death, i.e., pyroptosis.6 ASC filaments cluster into a solitary signal complex referred to as a speck, which is released upon cytolysis and circulates for extended periods because of its inherent resistance to proteolytic degradation.7 Circulating ASC speck is a measurable biomarker of pyroptosis, and elevated levels have previously been found in patients with known NLRP3 inflammasome acti-

Table 1. Characteristics of patients.

Median (range) or N (%) Demographics Age (years) Male (%) Clinical features Hemoglobin (g/dL) ANC (x109/L) Platelet (x109/L) Serum EPO > 200 mU/mL (%) Bone marrow blast > 2% (%) Bone marrow erythroblast (%) RARS (%) Favorable karyotype (%) Splicing gene mutations SF3B1 (%) U2AF (%) SRSF2 (%) ZRSR2 (%) Epigenetic mutation TET2 (%) IDH (%) DNMT3A (%) ASXL1 (%) EZH2 (%) Number of mutations IPSS low risk (%) Biomarker ASC specks > median (%) Treatment Prior responder to recombinant erythropoietin (%) Lenalidomide (%) Lenalidomide plus epoetin (%)

All (n=69)

Responder (n=28)

Non-responder (n=41)

73 (46-86) 48 (69.6)

74 (46-86) 21 (75.0)

73 (54-85) 27 (65.9)

0.250 0.418

8.1 (6.3-10.0) 2.1 (0.15-10.06) (n=67) 235 (58-587) (n=67) 27 (48.2) (n=56) 27 (39.1) 28.9 (1.0-79.0) (n=67) 38 (55.1) 59 (86.8) (n=68)

8.1 (6.3-10.0) 1.56 (0.40-5.70) 241 (58-417) 6 (28.6) (n=21) 11 (39.3) 28.5 (4.4-79.0) (n=27) 12 (42.9) 24 (88.9) (n=27)

8.0 (6.8-9.7) 2.43 (0.15-10.06) (n=39) 231 (68-587) (n=39) 21 (60.0) (n=35) 16 (39.0) 30.5 (1.0-65.0) (n=40) 26 (63.4) 35 (85.4)

0.815 0.075 0.598 0.029 0.983 0.618 0.092 0.733

52 (75.4) 0 (0.0) 6 (8.7) 5 (7.2)

21 (75.0) 0 (0.0) 2 (7.1) 2 (7.1)

31 (75.6) 0 (0.0) 4 (9.8) 3 (7.3)

0.954 1.000 1.000 1.000

34 (49.3) 2 (2.9) 13 (18.8) 12 (17.4) 2 (2.9) 2 (0-6) 33 (48.5) (n=68)

17 (60.7) 0 (0.0) 8 (28.6) 3 (10.7) 2 (7.1) 3 (0-6) 10 (37.0) (n=27)

17 (41.5) 2 (4.9) 5 (12.2) 9 (22.0) 0 (0.0) 2 (0-6) 23 (56.1)

0.116 0.511 0.120 0.335 0.161 0.291 0.124

34 (49.3)

20 (71.4)

14 (34.1)

0.002

24 (35.8) (n=67)

11 (40.7) (n=27)

13 (32.5) (n=40)

0.490

31 (44.9) 38 (55.1)

7 (25.0) 21 (75.0)

24 (58.5) 17 (41.5)

0.006

*Continuous variables were compared by the Mann-Whitney test, and categorical variables were compared using the Fisher exact test or c2 test, as appropriate. ANC: absolute neutrophil count; EPO: erythropoietin; RARS: refractory anemia with ring sideroblasts; IPSS: International Prognostic Score System; ASC: apoptosis-associated speck-like protein containing a CARD.

haematologica | 2022; 107(3)

737

Letters to the Editor

vation, such as those with cryopyrin-associated periodic syndrome and Schnitzler syndrome.8 Capturing elevated circulating interleukin-1β, the typical downstream effector of inflammasome activation, is less practical given its very short half-life9 and it is virtually undetectable in plasma even in patients with active cryopyrin-associated periodic syndrome and Schnitzler syndrome.8 Peripheral blood plasma ASC speck levels are almost exclusively elevated in patients with lower-risk MDS, compared to other hematologic malignancies, and like the myeloid-related inflammatory protein S100A9, provide an index of the magnitude of medullary pyroptosis.4,10 In the current study, we explored whether peripheral blood plasma ASC speck level may serve as a biomarker of response in lower-risk MDS patients who received lenalidomide with or without epoetin β. Peripheral blood plasma samples collected at the time of enrolment were available from 69 of 99 patients with International Prognostic Scoring System (IPSS) low- or intermediate-1-risk (i.e., lower-risk), non-deletion (5q) MDS who were enrolled in the GFM-Len-Epo-08 trial and completed four cycles (16 weeks) of study treatment. This Groupe Francophone des Myelodysplasies trial, registered as NCT01718379, was a randomized multicenter phase III study in transfusion-dependent lower-risk MDS patients comparing treatment with lenalidomide to the combination of lenalidomide and epoetin β. All patients had failed to respond to or relapsed after initial treatment with recombinant erythropoietin. Hematologic improvement of erythroid lineage (HI-E), the primary endpoint of the study, was evaluated after four cycles of treatment according to International Working Group 2006 criteria. Further details of the trial were described previously.2 The mean percentage of ASC specks was measured in peripheral blood plasma using flow cytometry and its level was glucose-adjusted with log10-transformation to adjust for hyperglycemia-induced inflammasome activation.10,11 Covariates associated with HI-E were analyzed using binary logistic regression by including variables with clinical and statistical relevance in univariate analysis (P<0.05) into the multivariate model. Interactions between variables and treatment were assessed by the Gail-Simon test. SPSS 26 (Armonk, NY, USA) and R 4.0.3 (R Core Team) were used for statistical analyses. This study was approved by the institutional review board. The median age of the 69 patients at enrolment was 73 years (range, 46-86) and 69.6% were male (Online Supplementary Table S1). Based on the World Health Organization 2008 classification, 38 patients (55.1%) had refractory anemia with ring sideroblasts. Conventional karyotype was classified as favorable in 59 of 68 patients (86.8%) according to the classical IPSS. The median number of mutations in each patient was two (range, 0-6). Mutations in SF3B1 and TET2 were detected in 52 (75.4%) and 34 (49.3%) patients, respectively. Lenalidomide and lenalidomide plus epoetin β were administered to 31 and 38 patients, respectively. After completion of 16 weeks of study treatment, 7 of 31 (22.6%) and 21 of 38 (55.3%) patients achieved a HI-E (P=0.006). Comparison of response according to baseline clinical parameters showed a significantly higher frequency of elevated ASC specks (defined as above the median level, 71.4% vs. 34.1%, P=0.002) but less frequent elevation of serum erythropoietin level (defined as >200 mU/mL, 28.6% vs. 60.0%, P=0.029) in responders. Although limited by sample size, a higher frequency of elevated ASC specks was consistently found in responders compared to non-responders of each treatment arm (lenalidomide, 57.1% vs. 29.2%, P=0.210; lenalidomide 738

Figure 1. Erythroid response stratified by status of biomarkers. ASC: apoptosis-associated speck-like protein containing a CARD; sEPO, serum erythropoietin.

plus epoetin β, 76.2% vs. 41.2%, P=0.046). DNMT3A was more frequently mutated in responders, but the difference was not statistically significant (28.6% vs. 12.2%, P=0.120). Other baseline parameters, including bone marrow blast percentage, karyotype classification and numbers of mutations, were also not significantly different between responders and non-responders (Table 1). In the multivariate analysis including ASC specks and serum erythropoietin, and adjusted for DNMT3A mutational status1,12 and treatment arm, both elevated ASC specks (odds ratio [OR]=4.963, 95% confidence interval [95% CI]: 1.445-17.045, P=0.011) and serum erythropoietin (OR=0.119, 95% CI: 0.025-0.561, P=0.007) were independently associated with HI-E. As expected, combined treatment remained significant (OR=6.535, 95% CI: 1.622-26.334, P=0.008). Of note, there was no significant interaction between each of these two biomarkers and treatment arm. Of 56 patients with both biomarkers available at inclusion, elevation in ASC specks showed a sensitivity and specificity of 71.4% and 74.3%, respectively, to predict HI-E, and addition of serum erythropoietin could improve the specificity to 88.6% with a reduction of sensitivity to 42.9%. No response (0.0%) was observed in patients with low ASC specks plus elevated serum erythropoietin, while the response rate was 69.2% in those with elevated ASC specks plus low serum erythropoietin (P<0.001) (Figure 1). Ineffective erythropoiesis is a hallmark of lower-risk MDS, and NLRP3 inflammasome-directed pyroptosis serves as a common underlying driver licensed by both the extrinsic proinflammatory cytokine milieu and intrinsic somatic mutations.3,4 Compared to single-agent treatment, lenalidomide with recombinant erythropoietin produces a higher rate of erythroid response.1,2 Our study showed that elevation of ASC specks, a specific and measurable surrogate of pyroptosis, was independently predictive of this response. Moreover, the lower serum erythropoietin in responders may similarly reflect suppression of erythropoietin liberation in response to anemia by pyroptotic inflammatory proteins. Levels of the myeloid-related inflammatory protein S100A9 increase directly with ASC specks in lower-risk MDS, and similarly serve as an index of the extent of medullary pyroptosis. S100A9 suppresses erythropoietin release in vitro in HepG2 hepatoma cells and its levels in lower-risk MDS patients correlate inversely with serum erythropoietin concentration, supporting the notion that inflammatory cytokines also suppress erythropoietin production in vivo.13 These observations suggest that patients with greater medullary inflammasome haematologica | 2022; 107(3)

Letters to the Editor

activation, identified by the dual pyroptotic biomarkers (ASC specks and erythropoietin), may represent a distinct subset of patients among those with lower-risk MDS who could benefit more from treatment with lenalidomide and recombinant erythropoietin. Mechanistically, lenalidomide was shown to reduce the steady-state S100A9 production by peripheral blood mononuclear cells from patients with non-del(5q) MDS and diminished the suppression of erythropoietin elaboration by S100A9.13 It could also stabilize the erythropoietin receptor by inhibiting the E3 ubiquitin ligase RNF41, to strengthen erythropoietin receptor signaling.14 Practically, ASC speck is a relatively novel biomarker that is not yet available as a routine diagnostic test. Additional studies are required to identify a clinically applicable cutoff to specifically recognize lower-risk MDS with inflammasome activation. Patients with these characteristics may also be optimal candidates for NLRP3 or other inflammasome-targeted therapy. In particular, targeting interleukin-1β represents a promising approach given the successful experience of interleukin-1β blockade (e.g., by canakinumab, an interleukin-1β neutralizing monoclonal antibody) in conditions with NLRP3 inflammasome activation, such as cryopyrin-associated periodic syndrome and Schnitzler syndrome.15 We recently launched a phase Ib/II trial (NCT04798339) to evaluate canakinumab plus recombinant erythropoietin in lower-risk MDS patients in whom prior treatment with an erythropoiesis-stimulating agent failed. This trial may provide more information on inflammasome-targeted therapies in MDS patients in the near future. Chen Wang,1,2 Kathy L. McGraw,2 Amy F. McLemore,2 Rami Komrokji,2 Ashley A. Basiorka,2 Najla Al Ali,2 Jeffrey E. Lancet,2 Eric Padron,2 Olivier Kosmider,3 Michaela Fontenay,3 Pierre Fenaux,4 Alan F. List2# and David A. Sallman2# 1 Department of Internal Medicine, University of South Florida, Morsani College of Medicine, Tampa, FL, USA; 2Department of Malignant Hematology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, USA; 3Service d’Hématologie Biologique, Assistance Publique-Hôpitaux de Paris. Centre-Université de Paris, Hôpital Cochin, Paris, France and 4Groupe Francophone des Myélodysplasies, Hopital Saint Louis, Paris, France # AFL and DAS contributed equally as co-senior authors. Correspondence: DAVID A. SALLMAN - david.sallman@moffitt.org doi:10.3324/haematol.2021.278855 Received: March 29, 2021. Accepted: June 11, 2021. Pre-published: August 5, 2021. Disclosures: DAS has played a consulting or advisory role for Celyad, Agios, Abbvie, Aprea AB, Bristol-Myers Squibb, Gilead Sciences, Intellia Therapeutics, Kite Pharma, Magenta Therapeutics, Novartis, and Syndax; has sat on speakers' bureau for Agios, Incyte, and Bristol-Myers Squibb; has received research funding from Celgene and Jazz Pharmaceuticals; and is holder of an intellectual property patent for LB-100 in MDS. AFL has received honoraria from Celgene, Aileron Therapeutics and Cellular Biomedicine Group; has played a consulting or advisory role for Celgene, Cellular Biomedicine Group, Aileron Therapeutics, Acceleron Pharma, International Personalized Cancer Center, Precision Biosciences, CTI BioPharma Corp, and Prelude Therapeutics; has received research funding from Celgene; has received travel, accommodation or other expenses from Celgene and Cellular Biomedicine Group; and is the holder of a Thousand haematologica | 2022; 107(3)

Talents Award. RSK has stock and other ownership interests in Abbvie; has played a consulting or advisory role for Novartis, Incyte, Bristol-Myers Squibb, Jazz Pharmaceuticals, Abbvie, Geron, and Acceleron Pharma; has sat on speakers' bureau for Jazz Pharmaceuticals, Bristol-Myers Squibb, and Agios; and has received travel, accommodation, or other expenses from Incy6te, Jazz Pharmaceuticals, Bristol-Myers Squibb, and Agios. JFL holds stock and other ownership interests in Arvinas, has played consulting or advisory roles for Jazz Pharmaceuticals, Astellas Pharma, Abbvie,Agios, BerGenBio, Daiichi Sankyo, and ElevateBioResearch; and has received funding from Pfizer. PF has received honoraria and research funding from Celgene. EP has received honoraria from Stemline Therapeutics and Blueprint Medicines; has received research funding from Incyte, Bristol-Myers Squibb, and Kura Oncology and sat on speakers' bureau for Novartis and Taiho Pharmaceutical. KLM has received research funding from Genentech and Celgene. MF has received research funding from Geron. Contributions: CW conceived the study, analyzed the results, and wrote the manuscript; KLM, AFM and AAB managed the samples and performed the experiments; RK, NAA, JEL and EP involved in study design; OK, MF and PF were involved in study design and provided the samples; AFL and DAS conceived the study, analyzed the results and revised the manuscript.

References 1. List AF, Sun Z, Verma A, et al. Lenalidomide-epoetin alfa versus

lenalidomide monotherapy in myelodysplastic syndromes refractory to recombinant erythropoietin. J Clin Oncol. 2021;39(9):10011009. 2. Toma A, Kosmider O, Chevret S, et al. Lenalidomide with or without erythropoietin in transfusion-dependent erythropoiesis-stimulating agent-refractory lower-risk MDS without 5q deletion. Leukemia. 2016;30(4):897-905. 3. Sallman DA, List A. The central role of inflammatory signaling in the pathogenesis of myelodysplastic syndromes. Blood. 2019; 133(10):1039-1048. 4. Basiorka AA, McGraw KL, Eksioglu EA, et al. The NLRP3 inflammasome functions as a driver of the myelodysplastic syndrome phenotype. Blood. 2016;128(25):2960-2975. 5. Chen L, Chen JY, Huang YJ, et al. The augmented R-loop is a unifying mechanism for myelodysplastic syndromes induced by highrisk splicing factor mutations. Mol Cell. 2018;69(3):412-425. 6. Broz P, Dixit VM. Inflammasomes: mechanism of assembly, regulation and signalling. Nat Rev Immunol. 2016;16(7):407-420. 7. Lu A, Magupalli VG, Ruan J, et al. Unified polymerization mechanism for the assembly of ASC-dependent inflammasomes. Cell. 2014;156(6):1193-1206. 8. Rowczenio DM, Pathak S, Arostegui JI, et al. Molecular genetic investigation, clinical features, and response to treatment in 21 patients with Schnitzler syndrome. Blood. 2018;131(9):974-981. 9. Lachmann HJ, Lowe P, Felix SD, et al. In vivo regulation of interleukin 1beta in patients with cryopyrin-associated periodic syndromes. J Exp Med. 2009;206(5):1029-1036. 10. Basiorka AA, McGraw KL, Abbas-Aghababazadeh F, et al. Assessment of ASC specks as a putative biomarker of pyroptosis in myelodysplastic syndromes: an observational cohort study. Lancet Haematol. 2018;5(9):e393-e402. 11. Lee HM, Kim JJ, Kim HJ, Shong M, Ku BJ, Jo EK. Upregulated NLRP3 inflammasome activation in patients with type 2 diabetes. Diabetes. 2013;62(1):194-204. 12. Chesnais V, Renneville A, Toma A, et al. Effect of lenalidomide treatment on clonal architecture of myelodysplastic syndromes without 5q deletion. Blood. 2016;127(6):749-670. 13. Cluzeau T, McGraw KL, Irvine B, et al. Pro-inflammatory proteins S100A9 and tumor necrosis factor-α suppress erythropoietin elaboration in myelodysplastic syndromes. Haematologica. 2017; 102(12):2015-2020. 14. Basiorka AA, McGraw KL, De Ceuninck L, et al. Lenalidomide stabilizes the erythropoietin receptor by inhibiting the E3 ubiquitin ligase RNF41. Cancer Res. 2016;76(12):3531-3540. 15. Jesus AA, Goldbach-Mansky R. IL-1 blockade in autoinflammatory syndromes. Annu Rev Med. 2014;65:223-244.

739

Letters to the Editor

Survival in mantle cell lymphoma after frontline treatment with R-bendamustine, R-CHOP and the Nordic MCL2 regimen – a real world study on patients diagnosed in Sweden 2007-2017 The optimal frontline treatment of mantle cell lymphoma (MCL) with respect to long-term survival remains undefined. Intensified immunochemotherapy including rituximab (R), cytarabine and autologous hematopoetic cell transplant (HD-AHCT) upfront, such as the Nordic MCL2 protocol, has demonstrated improved disease control.1-3 Elderly patients may benefit from R-CHOP or R-bendamustine (BR), albeit the regimens have not been robustly evaluated in a randomized setting or in observational studies.1,4,5 Here, we used the Swedish Lymphoma Register (SLR), a nationwide register initiated in 2000 with a reported coverage of ∼95%, 6 in order to investigate overall and relative survival in a population-based cohort of patients diagnosed with MCL between Jan 2007 and Sept 2017. Particularly, we report outcome by given treatment with specific focus on the currently recommended treatment strategies upfront: BR, R-CHOP, the Nordic MCL2 protocol and curative radiotherapy to limited stage disease. Data on patient characteristics and frontline treatment administered was retrieved from the SLR. Survival status was obtained from the national population register. Cases fulfilling all of the following criteria were categorized as having been treated with curative radiotherapy: stage I-II disease; treatment with single radiotherapy; reported curative intent and radiation dose of 30-40 Gy. Chemotherapy treatment subgroups (MCL2, BR and RCHOP) were compared using t-test or chi2-test (continuous or categorical variables). Follow-up began 90 days after diagnosis to allow for treatment completion and ended on date of death (from any cause) or on 20 October 2018, whichever occurred first. Cox regression models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) comparing all-cause mortality rates between treatment groups. Both uni- and multivariable models were considered, allowing for adjustments for age at diagnosis, calendar year of diagnosis, sex, WHO performance status (PS), white blood cell count (WBC), lactate dehydrogenase (LDH) and the MCL International Prognostic Index (MIPI). Interaction models were fitted with age at diagnosis (dichotomized into <70/≥70 years) as an effect modifier. Non-parametric estimates of overall survival (OS) were calculated by age group using the Kaplan-Meier method. Marginal (standardized) OS was estimated using predictions from a flexible parametric survival model. 7 As a measure of net (cause-specific) survival, non-parametric relative survival (RS) estimates were calculated using the actuarial method and standardized estimates were predicted from a flexible parametric relative survival model corresponding to that in the OS analysis. All statistical analyses were performed using Stata (StatCorp, 2017. Stata Statistical Software: Release 16. College station, TX: StataCorp LLC). The study was approved by the Regional Board of the Ethical Committee in Lund, Sweden (2018/739). In total, 1277 patients were included in the study. The median age at diagnosis was 71 years, the majority were male (70%) and Ann Arbor stage III-IV (80%) (Table 1). Frontline systemic treatment was reported in 818 patients (63%). Among systemic treatments, MCL2 (n=268, 33%), BR (n=231, 29%) and R-CHOP (n=93, 12%) were the regimens most frequently applied. The 740

number of patients receiving maintenance rituximab was 10 (11%) after R-CHOP, 18 (8%) after BR and 14 (5%) after MCL2. In total, 218 patients received HDAHCT after MCL2 (82%) and one (<1%) after BR (Online Supplementary Table S1). Patients who received BR and R-CHOP were comparable in terms of mean age (75.5 and 73.5, P=0.39) at diagnosis, performance status (WHO PS 0-1 87% vs. 82%, P=0.32) and high risk MIPI (55% vs. 49%, P=0.36). Patients who were treated with MCL2 had a lower mean age (60 years, P<0.0001), lower PS (WHO PS 0-1 n=251; 94%, P=0.01 (vs. BR, P<0.0001; vs. R-CHOP, P<0.0001)) and were less likely to score a high-risk MIPI (26%, (vs. R-CHOP, P=0.02; vs. BR, P<0.0001)) compared to R-CHOP and BR, respectively. A total of 1182 patients were included in the survival analyses (95 were excluded due to lack of follow-up (FU) within 90 days after diagnosis). In patients receiving any systemic therapy, median OS was 4.9 years (IQR 1.5NR) at a median FU time of 3.5 years (IQR 1.4-5.8). Among patients of all ages, treatment with MCL2 was associated with a lower all-cause mortality than BR in univariable models (HR=0.49 (95% CI 0.37-0.66)) or when adjusting for MIPI (HR=0.66 (95% CI 0.48-0.91)) but not when adjusting for the individual prognostic factors included in MIPI (HR=1.06 (95% CI 0.71-1.56)). By age stratification (<70/≥70), no differences in all-cause mortality were observed in patients receiving MCL2 and BR. Patients treated with R-CHOP demonstrated higher all-cause mortality rates compared to BR-treated patients in univariable analysis, either including all patients or by age stratification. In multivariable models, no differences in all-cause mortality were observed. Survival proportions by age and treatment groups are presented in Figure 1 and Online supplementary Table S2. In patients <70 years, the unadjusted three-year OS was 0.80 (95% CI 0.75-0.85) for MCL2, 0.79 (95% CI 0.62-0.89) for BR, and 0.62 (95% CI 0.39-0.78) for R-CHOP. The corresponding marginal three-year OS was 0.78 (95% CI 0.74-0.82), 0.75 (95% CI 0.67-0.84) and 0.62 (95% CI 0.51-0.76). In patients aged ≥70 years, the unadjusted three-year OS was 0.68 (95% CI 0.60-0.74) for BR and 0.49 (95% CI 0.36-0.61) for R-CHOP and the corresponding marginal three-year OS was 0.55 (95% CI 0.450.67) and 0.40 (95% CI 0.22-0.71). Analysis of RS demonstrated very similar results as the analysis of OS (Online Supplemenetary Figure S1). In patients given curative radiotherapy (n=26 (2%), median age was 64 years; all had WHO PS 0-1) and five-year OS was 0.75 (95% CI 0.51-0.88). This study reports overall and relative survival in an unselected cohort of patients diagnosed with MCL in 2007-2017, receiving frontline treatment with R-CHOP, BR, or the Nordic MCL2 protocol. The results demonstrate no significant difference in overall or relative survival by intensified MCL2 protocol or R-CHOP compared to BR, by adjustment for prognostic factors or in age-stratified analysis. Furthermore, our analysis demonstrates long-term survival with curative radiotherapy in limited stage MCL. R-CHOP and BR patients were comparable whereas MCL2 patients were younger, had better performance status and less frequent MIPI high-risk score, thus the adjusted models were relevant. The lower all-cause mortality after MCL2 compared to BR in a univariable model or by adjustment for MIPI, but not when adjusting for individual prognostic factors or by age stratification, could be explained by the fact that MIPI is largely driven by age and WHO PS. Moreover, MCL2 is seldom considered in patients ≥70 years, thus the results from age stratification are expected to be haematologica | 2022; 107(3)

Letters to the Editor

more reliable. The three-year OS in BR and R-CHOP treated patients is similar to the population-based cohort reported by Villa et al.8 but lower than after R-CHOP in the European MCL Elderly trial, as may be expected from an unselected cohort.1 In the latter trial, maintenance rituximab after R-CHOP was associated with a benefit in OS, which was also demonstrated in younger patients. 1,9 Unfortunately, we were not able to confirm these data in the real-world setting due to a limited number of cases. As our study cannot prove that BR issuperior to R-CHOP, durther analysis on the effect of mainte-

nance rituximab would be valuable. HD-AHCT was established in MCL based on its association with improved OS in a randomized European pre-rituximab trial,10 although not confirmed after the addition of rituximab and cytarabine to CHOP-based induction and the use of maintenance with rituximab.3,9-11 Consequently, HD-AHCT is currently challenged in the ongoing phase III TRIANGLE (NCT02858258) trial. Reviewing the standardized estimates of OS in our analysis, survival after MCL2 and BR may be comparable during FU initial time, as represented by chemo-sensitive cases. The less steep

Table 1. Patient characteristics.

Variable Overall (row %) Year of diagnosis 2007-2012 2013-2017 Age at diagnosis Median (IQR) <50 50-59 60-69 70-79 ≥80 Sex Male Female Ann Arbor stage I II III IV Missing MIPI Median (IQR) Low (<5.7) Intermediate(5.7-6.1) High (≥6.1) Missing LDH Normal Elevated Missing WHO PS 0-1 2-4 Missing WBC Normal (<9x109/L) Elevated Missing

All patients

MCL2

R-CHOP

N (col %)

Other systemic* N (col %)

N (col %)

Curative Other/ radiotherapy** Missing*** N (col %) N (col%)

1,277 (100)

268 (21)

231 (18)

93 (7)

226 (18)

26 (2)

433 (34)

667 (52) 610 (48)

151 (56) 117 (44)

81 (35) 150 (65)

76 (82) 17 (18)

170 (75) 56 (25)

18 (69) 8 (31)

171 (39) 262 (61)

71 (64-79) 45 (4) 137 (11) 376 (29) 411 (32) 308 (24)

62 (56-66) 26 (10) 85 (32) 142 (53) 15 (6) 0 (0)

75 (71-80) 0 (0) 2 (1) 39 (17) 127 (55) 63 (27)

74 (69-79) 0 (0) 3 (3) 24 (26) 46 (49) 20 (22)

76 (70-83) 1 (0) 5 (2) 48 (21) 75 (33) 94 (42)

64 (59-73) 2 (8) 6 (23) 11 (42) 3 (12) 4 (15)

73 (66-81) 16 (4) 37 (8) 113 (26) 144 (33) 125 (29)

912 (71) 365 (29)

202 (75) 66 (25)

161 (70) 70 (30)

67 (72) 26 (28)

167 (74) 59 (26)

19 (73) 7 (27)

296 (68) 137 (32)

70 (5) 124 (10) 144 (11) 885 (69) 54 (4)

5 (2) 18 (7) 30 (11) 212 (79) 3 (1)

2 (1) 26 (11) 28 (12) 166 (72) 9 (4)

3 (3) 8 (9) 15 (16) 65 (70) 2 (2)

7 (3) 19 (8) 34 (15) 154 (68) 12 (5)

15 (58) 11 (42) (0) (0) 2 (8)

38 (9) 42 (9) 37 (9) 288 (62) 28 (6)

6.3 (5.9-6.8) 147 (12) 323 (25) 573 (45) 234 (18)

5.9(5.6-6.3) 70 (26) 84 (31) 69 (26) 45 (17)

6.4 (6.1-6.9) 7 (3) 64 (28) 128 (55) 32 (14)

6.5 (6.1-6.9) 2 (2) 20 (22) 44 (47) 27 (29)

6.5 (6.2-7.1) 7 (2) 36 (16) 126 (55) 57 (26)

5.9(5.4-6.2) 6 (23) 4 (15) 5 (19) 9 (35)

6.3 (5.9-6.9) 57 (13) 115 (26) 199 (49) 65 (14)

714 (56) 523 (41) 40 (3)

117 (44) 147 (55) 4 (1)

148 (64) 79 (34) 4 (2)

40 (43) 51 (55) 2 (2)

127 (56) 94 (42) 5 (2)

24 (9) 1 (4) 1 (4)

258 60) 151 (35) 24 (5)

1,070 (84) 186 (15) 21 (2)

251 (93) 15 (6) 2 (1)

200 (87) 26 (11) 5 (2)

76 (82) 15 (16) 2 (2)

169 (75) 54 (24) 3 (1)

26 (100) 0 (0) 0 (0)

348 (80) 76 (18) 7 (2)

580 (45) 503 (39) 194 (15)

131 (49) 99 (37) 38 (11)

103 (45) 103 (45) 25 (11)

35 (38) 33 (35) 25 (27)

91 (40) 79 (37) 51 (24)

13 (50) 4 (15) 9 (35)

207 (48) 185 (42) 46 (10)

* Includes treatment with BAC (bendamustine, cytarabine), chlorambucil, CHOP/cytarabine, cyclophosphamide, cytarabine, CVP (cyclophosphamide, vincristine, prednisone), CVIP (cyclophosphamide, etoposide, idarubicin, prednisone), FC (fludarabine, cytarabine), ibrutinib+bendamustine, idarubicin, ixoten, lenalidomide+bendamustine. Among these, the most frequently reported regimens were chlorambucil (n= 69) and CHOP/cytarabine (n=59).** Includes patients with AA stage I-II MCL, treated with single radiotherapy with curative intent, radiation dose 30-40 Gy.*** Other/missing includes no treatment given, treatment given but not specified, missing information on treatment and non-curative radiotherapy. MCL: mantle-cell lymphoma; R: Rituximab; CHOP: cyclophosphamide, doxorubicin, vincristine, prednisone; MIPI: MCL International Prognostic Index; N: number; col: column; WBC: white blood cell count; WHO PS: World Health Organization Performance Status. Due to rounding, not all percentages add up to 100.

haematologica | 2022; 107(3)

741

Letters to the Editor

Figure 1. Unadjusted (top panel) and standardized (bottom panel) overall survival among MCL patients diagnosed in Sweden between 2007 and 2017, by frontline treatment (BR, R-CHOP, MCL2) and age at diagnosis (<70 and ≥70 years). Standardization was performed over year of diagnosis, sex, age at diagnosis, and performance status (separately for age groups). In patients ≥70 years, MCL2 is not presented due to the limited number of cases. OS: overall survival; R: Rituximab; CHOP: cyclophosphamide, doxorubicin, vincristine, prednisone; MCL2/3: the Nordic MCL2 protocol (R-alternating maxi CHOP/cytarabine with consolidative high-dose chemotherapy with autologous hematopoetic cell transplant). Number at risk table denotes selected time points of estimates (0.3; 2; 4; 6; 8; 10 years).

Table 2. Hazard ratio (HRs) with 95% confidence intervals (CI) comparing all-cause mortality in relation to given frontline treatment among all patients (top panel) and by age (<70 and ≥70 years) (bottom panel). The analysis included univariable (a) and multivariable models (be). For categorical variables, female sex, WHO PS 0-1 and MIPI low risk were reference groups, respectively. Age and LDH were treated as continuous variables.

Treatment regimen All (n) 607 BR MCL2 R-CHOP Other systemic

< 70 yrs

BR MCL2 R-CHOP Other systemic

≥ 70 yrs

BR R-CHOP Other systemic

HRa (95% CI) 756

HRb (95% CI) 595

HRc (95% CI) 607

HRd (95% CI)

1.00 0.49 (0.37-0.66) 1.51 (1.15–2.06) 1.49 (1.15-1.91)

1.00 1.06 (0.71-1.56) 1.10 (0.73-1.66) 1.41 (1.02-1.92)

1.00 0.66 (0.48-0.91) 1.07 (0.73-1.58) 1.37 (1.01-1.86)

1.00 0.71 (0.51-0.98) 1.23 (0.84-1.80) 1.44 (1.06-1.95)

1.00 0.77 (0.44-1.33) 1.68 (0.82–3.43) 1.46 (0.78-2.75)

1.00 0.79 (0.42-1.46) 1.23 (0.50-3.03) 1.78 (0.86-3.70)

1.00 0.88 (0.49-1.60) 1.10 (0.47-2.58) 1.51 (0.74-3.08)

1.00 0.92 (0.5-1.67) 1.08 (0.46-2.55) 1.34 (0.65-2.74)

1.00 1.56 (1.10–2.19) 1.56 (1.19-2.06)

1.00 1.03 (0.65-1.64) 1.55 (1.10-2.18)

1.00 1.10 (0.72-1.69) 1.36 (0.97-1.90)

1.00 1.33 (0.87-2.02) 1.49 (1.07-2.07)

a) From Cox regression model adjusted for time since diagnosis (time scale). b) From multivariable Cox regression model adjusted for age, sex, WHO PS, WBC and LDH. c) From multivariable Cox regression model adjusted for sex and MIPI as a continuous variable. d) From multivariable Cox regression model adjusted for sex and MIPI as a categorical variable including high and intermediate risk versus low risk. WHO PS: World Health Organization performance status; WBC: white blood cell count; LDH: lactate dehydrogenase; MIPI: Mantle cell lymphoma International Prognostic Index; HR: high-risk; IMR: intermediate risk.

742

haematologica | 2022; 107(3)

Letters to the Editor

curve over time among MCL2 treated patients could possibly be related to a deeper remission and longer time to relapse, as previously demonstrated by the intensified strategy.3,11 In any case, the lack of a plateau in any of the curves is probably related to chemo-resistant disease, as previously demonstrated in biologic high-risk MCL.12,13 The favorable OS in limited-stage MCL patients receiving curative radiotherapy confirms the efficacy of this strategy. 14 However, the low number of patients studied, the uncertainty of defining a group based on criteria in a retrospective cohort and the potential influence of confounders, i.e., biologic good prognostic factors, should not be neglected. A major strength of the work presented is the population-based setting and the large size of the cohort, retrieved from a time period after rituximab was introduced in clinical routine. Moreover, the standardized models in both overall and relative survival models improve the reliability of the results. Limitations were the lack of data on molecular markers, comorbidity, second primary malignancies and other factors with potential impact on treatment choice and mortality15 as well as the low number of patients treated with maintenance rituximab. Furthermore, the delayed entry may potentially have excluded cases with the most treatment-resistant disease. To conclude, this study demonstrates that BR may be comparable to intensified treatment strategies in a proportion of patients with MCL. Awaiting the results from ongoing prospective trials on novel combinatory regimens, future studies should focus on a deeper evaluation of predictive markers in relation to established treatment concepts. Alexandra Albertsson-Lindblad,1 Thorgerdur Palsdottir,2,3 Karin E. Smedby,2,4 Caroline E. Weibull,2 Ingrid Glimelius,2,5 and Mats Jerkeman1 1 Division of Oncology, Skane University Hospital, Lund University, Lund; 2 Department of Medicine Solna, Clinical Epidemiology Division, Karolinska Institutet and Karolinska University Hospital, Stockholm; 3 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm; 4Department of Medicine Solna, Division of Hematology, Karolinska University Hospital, Stockholm and 5 Department of Immunology, Genetics and Pathology, Clinical and Experimental Oncology, Uppsala University and Uppsala Akademiska Hospital, Sweden Correspondence: ALEXANDRA ALBERTSSON LINDBLADAlexandra.albertsson_lindblad@med.lu.se doi:10.3324/haematol.2021.279037 Received: April 22, 2021. Accepted: November 5, 2021. Pre-published: November 18, 2021. Disclosures: AAL received grant support from Janssen Pharmaceutical NV as part of a collaboration between Karolinska Institutet and Janssen Pharmaceutical NV; CW and TP are part of a research collaboration between Karolinska Institutet and Janssen Pharmaceutical NV for which Karolinska Institutet has received grant support; IG participated in educational seminars run by Janssen Pharmaceutical NV; KES received a research grant from Janssen, honoraria from Takeda and Celgene; MJ received research support from Janssen Pharmaceutica NV, Celgene, Abbvie, Gilead, Roche and Astra Zeneca and honoraria from Roche, Gilead, BMS, Astra Zeneca, Janssen Genmab and Incyte. Contributions: AAL, TP, IG, CEW, KES, MJ designed the study; TP, CEW, IG, AAL prepared data; AAL, TP and CEW performed data analysis, tables and figures; AAL, TP, IG, CEW, KES, MJ haematologica | 2022; 107(3)

participated in the analysis and interpretation of results; AAL prepared the draft manuscript; AAL, TP, IG, CEW, KES, MJ critically reviewed the manuscript prior to submission. Funding: This study was financed partly through the Swedish Cancer Society and partly through a public-private real-world evidence collaboration between Karolinska Institutet and Janssen Pharmaceutical NV.

References 1. Kluin-Nelemans HC, Hoster E, Hermine O, et al. Treatment of older patients with mantle cell lymphoma (MCL): long-term follow-up of the Randomized European MCL Elderly Trial. J Clin Oncol. 2020;38(3):248-256. 2. Geisler CH, Kolstad A, Laurell A, et al. Long-term progressionfree survival of mantle cell lymphoma after intensive front-line immunochemotherapy with in vivo-purged stem cell rescue: a nonrandomized phase 2 multicenter study by the Nordic Lymphoma Group. Blood. 2008;112(7):2687-2693. 3. Hermine O, Hoster E, Walewski J, et al. Addition of high-dose cytarabine to immunochemotherapy before autologous stem-cell transplantation in patients aged 65 years or younger with mantle cell lymphoma (MCL Younger): a randomised, open-label, phase 3 trial of the European Mantle Cell Lymphoma Network. Lancet. 2016;388(10044):565-575. 4. Rummel MJ, Niederle N, Maschmeyer G, et al. Bendamustine plus rituximab versus CHOP plus rituximab as first-line treatment for patients with indolent and mantle-cell lymphomas: an openlabel, multicentre, randomised, phase 3 non-inferiority trial. Lancet 2013;381(9873):1203-1210. 5. Flinn IW, van der Jagt R, Kahl BS, et al. Randomized trial of bendamustine-rituximab or R-CHOP/R-CVP in first-line treatment of indolent NHL or MCL: the BRIGHT study. Blood. 2014;123(19):2944-2952. 6. Cancercentrum i samverkan R. Nationella kvalitetsregistret för lymfom - Årsrapport från Nationell kvalitetsregistret för lymfom, diagnosperioden 2000-2019. 2020. https://cancercentrum.se/globalassets/cancerdiagnoser/blod-lymfom-myelom/lymfom/rapporter/lymfomarsrapport2000-2019.pdf (accessed 16 February 2021). 7. Royston P, Lambert PC. Flexible parametric survival analysis using stata: beyond the Cox model. College Station, TX. Stata Press. StataCorp LP; 2011. 8. Villa D, Sehn LH, Savage KJ, et al. Bendamustine and rituximab as induction therapy in both transplant-eligible and -ineligible patients with mantle cell lymphoma. Blood Adv. 2020;4(15):34863494. 9. Le Gouill S, Thieblemont C, Oberic L, et al. Rituximab after autologous stem-cell transplantation in mantle-cell lymphoma. N Engl J Med. 2017;377(13):1250-1260. 10. Zoellner AK, Unterhalt M, Stilgenbauer S, et al. Long-term survival of patients with mantle cell lymphoma after autologous haematopoietic stem-cell transplantation in first remission: a post-hoc analysis of an open-label, multicentre, randomised, phase 3 trial. Lancet Haematol. 2021;8(9):e648-e657. 11. Gerson JN, Handorf E, Villa D, et al. Survival outcomes of younger patients with mantle cell lymphoma treated in the Rituximab era. J Clin Oncol. 2019;37(6):471-480. 12. Delfau-Larue M-H, Klapper W, Berger F, et al. High-dose cytarabine does not overcome the adverse prognostic value of CDKN2A and TP53 deletions in mantle cell lymphoma. Blood. 2015;126(5):604-611. 13. Eskelund CW, Dahl C, Hansen JW, et al. TP53 mutations identify younger mantle cell lymphoma patients who do not benefit from intensive chemoimmunotherapy. Blood. 2017;130(17):1903-1910. 14. Bernard M, Tsang RW, Le LW, et al. Limited-stage mantle cell lymphoma: treatment outcomes at the Princess Margaret Hospital. Leuk Lymphoma. 2013;54(2):261-267. 15. Glimelius I, Smedby KE, Albertsson-Lindblad A, et al. Unmarried or less-educated patients with mantle cell lymphoma are less likely to undergo a transplant, leading to lower survival. Blood Adv. 2021;5(6):1638-1647.

743

Letters to the Editor

Mechanical unloading aggravates bone destruction and tumor expansion in myeloma The importance of retaining physical functions has been increasingly emphasized to maintain the quality of life in patients with a variety of cancers, especially those with bone metastasis. Moreover, physical functions may impact prognosis of patients with cancers. Multiple myeloma (MM) has a unique propensity to develop and expand almost exclusively in the bone marrow and to generate destructive bone disease. Patients with MM in advanced stages often suffer from immobilization or are in a bed-ridden state with vertebral fracture and/or lower limb paralysis due to spinal cord compression by tumors expanding outside of bone. The skeleton and skeletal muscles are sensitive to their mechanical environment such as mechanical loading with exercise. Patients in a bed-ridden state or those with lower limb paralysis are exposed to mechanical unloading to decrease bone volume and strength along with muscle atrophy. However, the effect of mechanical unloading on the progression of MM tumor has not been studied. We hypothesized that immobilization or a paralytic state not only negatively affect bone health but also may aggravate tumor growth in patients with MM. In the present study, we therefore aimed to clarify the deleterious impact of paralytic immobilization and mechanical unloading on tumor expansion and bone destruction in MM. Unilateral hind legs of mice were immobilized to expose to mechanical unloading by sciatic denervation1 or casting with an adhesive bandage.2 These procedures reduced hind leg muscle volume as shown by the weight as well as the outer appearance of the anterior tibial and gastrocnemius muscles at 2 weeks (Online Supplementary Figure S1A and B). Atrophy was more marked in the muscles in the hind legs paralyzed with sciatic denervation than those immobilized by casting with adhesive bandage. Micro–computed tomography (mCT) revealed substantial reduction of the bone volume in the trabecular bone in the tibiae in the immobilized hind legs (Figure 1A). Bone morphometric analysis also showed the reduction of bone volume in the immobilized hind legs, as indicated by an increase of bone volume over total volume and trabecular numbers with reduced trabecular separation (Figure 1B). These mCT findings were consistent with the previous results in mice upon mechanical unloading with hind legs paralyzed by surgical denervation1 and tail suspension.3 Tartrate-resistant acid phosphatase (TRAP)-positive multinucleated osteoclasts increased in number on the surface of the trabecular bone in the immobilized hind legs (Figure 1C and D). TRAP-5b, a bone resorption marker, was increased in sera from the mice with hind legs paralyzed by sciatic denervation, while the levels of bone formation markers, bone alkaline phosphatase and osteocalcin, were not significantly changed in their sera (Figure 1E). These results demonstrate acute activation of bone resorption by osteoclasts and thereby trabecular bone reduction alone with muscle atrophy in immobilized hind legs. Osteocytes are embedded in the bone matrix, and major sensors of mechanical stress to regulate bone remodeling through interaction with bone marrow cells by their dendritic processes.4 Osteocytes produce critical molecules for bone metabolism, including receptor activator of nuclear factor-κB ligand (RANKL) and its inhibitor osteoprotegerin (OPG), and sclerostin (SOST). After flushing out bone cavities to remove bone marrow 744

cells, femurs were used for gene analysis in osteocytes embedded in bone. Consistent with a previous report, 1 the gene expression of Rankl but neither Opg nor Sost was upregulated in the femurs from the hind legs immobilized with the sciatic denervation or casting (Figure 1F). Serum levels of Rankl were significantly increased in the mechanical unloading with sciatic denervation (Figure 1F). These results suggest that the role of RANKL upregulated in osteocytes in osteoclastogenesis is enhanced in immobilized hind legs. We next looked at the effects of the immobilization or mechanical unloading of hind legs on MM tumor growth in bone. We inoculated luciferase-transfected mouse 5TGM1 MM cells into tibiae 2 weeks after sciatic denervation or sham operation, and compared tumor growth in the hind legs with or without mechanical loading. Tumor growth was more robust in the immobilized hind legs with the denervation than in sham-operated hind legs as shown in IVIS images (Figure 1G). Previous reports demonstrated that osteoclasts directly enhance MM cell proliferation5 and that RANKL-stimulated osteoclastogenesis triggers the proliferation of MM cells in vivo.6 Indeed, osteoclasts generated from mouse bone marrow cells directly enhanced the growth of 5TGM1 cells (Online Supplementary Figure S2A). As immobilization of legs acutely enhanced osteoclastogenesis (Figure 1C and D), osteoclasts induced in bone are suggested to play a causative role in MM cell expansion accelerated in mechanical unloading. Therefore, we next looked at the effects of the anti-bone resorbing agent zoledronic acid on MM tumor growth under hind leg immobilization. Treatment with zoledronic acid twice weekly after sciatic denervation resulted in the maintenance of bone volume (Figure 2A) along with the reduction of osteoclast numbers (Figure 2B and C). The treatment with zoledronic acid retarded MM tumor growth in the immobilized hind legs (Figure 2D), suggesting the role of osteoclasts in the acceleration of MM tumor growth in immobilized hind legs. In order to further confirm the acceleration of tumor growth in immobilized hind legs, we next simultaneously inoculated 5TGM1 MM cells into bilateral tibiae in immobilized (right) and intact (left) hind legs in the same mice, and compared tumor growth between the immobilized or intact hind legs. MM tumor growth was assessed with IVIS images, and was more accelerated in the immobilized legs with sciatic denervation (Figure 3A, top) or in those in a cast (Online Supplementary Figure S2B). We reported that proviral integrations of Moloney virus 2 kinase (PIM2) is constitutively overexpressed and further upregulated in MM cells through the interaction with cellular components in MM bone marrow microenvironment, including osteaolclasts (OC).7 We subsequently reported that TGF-β-activated kinase-1 (TAK1) is also overexpressed and phosphorylated to transcriptionally induce PIM2 expression in MM cells and OC.8 Consistently, CD138-positive MM cells and cathepsin Kpositive OC expressed both PIM2 and phosphorylated TAK1 in the tibiae with 5TGM1 MM cell inoculation (Online Supplementary Figure S3A). The TAK1 inhibitor LL-Z1640-2 as well as PIM inhibitor SMI16a are able to efficaciously reduce MM growth and osteoclastic bone destruction in in vivo MM models with 5TGM1 MM cells,8,9 suggesting the pivotal role of the TAK1-PIM2 pathway in MM cells in MM tumor growth and bone destruction. Interestingly, TAK1 phosphorylation and PIM2 protein levels are further upregulated in MM tumor lesions in immobilized hind legs by sciatic denervation compared to those in intact hind legs (Figure 3B). haematologica | 2022; 107(3)

Letters to the Editor

Figure 1. Enhancement of osteoclastogenesis and multiple myeloma tumor growth by immobilization. SCID mice were subject to sham-operation (control), sciatic denervation (DN) or casting with adhesive bandages (CAST) of the right hind legs. Two weeks later, the right tibiae were taken out, and histomorphometrically analyzed. (A) Micro-computed tomography (mCT) image of the representative tibiae resected from each treatment group. (B) Tibiae were resected from mice, carefully separated from surrounding tissues, and fixed overnight in 10 % formaldehyde solution. The dissected tibiae were then examined with a SkyScan 1176 unit (SkyScan 1176 scanner and analytical software; Buruker, Billerica, MA) using a 0.5 mm aluminum filter, rotation of 360°, rotation step of 0.5°, voltage of 50 kV, current of 200 mA and image size of 18 mm voxel size. The regions of interest for trabecular bones analyzed with a mCT were set on a 1.5 mm region of metaphyseal spongiosa in the proximal tibia located 0.5 mm above the growth plate. The threshold was set with 93 (lower) and 255 (upper), which was able to clearly indicate the trabecular bone. Bone volume over total volume (BV/TV), trabecular thickness (Tb.Th), trabecular numbers (Tb.N) and trabecular separation (Tb.Sp) were assessed in trabecular bones. Data are expressed as the mean ± standard deviation (SD) (n=6). (C) TRAP staining was performed in the sections of the resected tibiae using a TRAP/ALP stain kit (FUJIFILM Wako Chemicals USA, Richmond, VA, USA). TRAP-positive cells containing 3 or more nuclei on the bone surface were counted as osteoclasts (OC) under a light microscope (BZ-X800; Keyence, Osaka, Japan). Three fields were counted for each sample. Representative results were shown (magnification, ×100) (upper). Bars: 200 mm. Higher magnification (×400) of the boxed areas in the upper panels were shown in the lower panels. Bars: 50 mm. (D) Numbers of osteoclasts (N.Oc)/bone surface (BS) (/mm) were counted. Data are expressed as the mean ± SD (n=6). *P<0.05. (E) Serum levels of TRAP-5b (mg/mL), ALP (mU/mL) and Gla-osteocalcin (OCN) (ng/mL) were measured 2 weeks after the sciatic denervation (DN). In order to measure the serum levels of bone metabolic parameters, Mouse Osteocalcin EIA Kit (Biomedical Technologies Inc., MA, USA), Mouse TRAP-5b Assay (Immuno diagnostic system Ltd, UK), and alkaline phosphatase (ALP) test kit (Wako, Osaka, Japan) were used in accordance with the manufactures’ protocols. Data are expressed as the mean ± SD (n=6). *P<0.05. (F) Femurs were taken out at 2 weeks after the immobilization, and their bone marrow cavities were flushed out. Expression of RANKL, SOST, and OPG mRNA were analyzed by real-time reverse transcription polymerase chain reaction (RT-PCR) using the femurs. The primer sequences were as follows: mouse Rankl F: GTTCCTGTACTTTCGAGCGCAGAT, R: TGACTTTATGGGAACCCGATGGGA mouse Opg F: TTGCCCTGACCACCACTCTTATACGGA, R: CTTTTGCGTGGCTTCTCTCTGTTTCC mouse SOST F: TCCTCCTGAGAACAACCAGAC, R: TGTCAGGAAGCGGGTAGTC mouse Gapdh F: ATGTGTCCGTCGTGGATCTGA, R: TTGAAGTCGCAGGAGACAACCT. Serum level of RANKL (pg/mL) were measured by Mouse TRANCE/RANKL/TNFSF11 ELISA kit (R&D systems). Data are expressed as the mean ± SD (n=6). *P<0.05. (G) Luciferase-transfected 5TGM1 (5TGM1/luc) multiple myeloma (MM) cells were inoculated into tibiae 2 weeks after DN or sham operation. Two weeks after the immobilization, we injected 1x105 5TGM1/luc MM cells in 20 μL saline, and directly through the tibial plateau into the bone marrow cavity of the tibiae with a 27-gauge needle while flexing the knee. IVIS images were taken 2 and 4 weeks after the inoculation. Tumor areas with luminescence shown in green, yellow and red were measured. Px: pixel. Data are expressed as the mean ± SD (n=3). *P<0.05.

haematologica | 2022; 107(3)

745

Letters to the Editor

Figure 2. Effects of zoledronic acid on multiple myeloma tumor growth upon immobilization. Mice were subjected to sham-operation (control), sciatic denervation (DN) or casting with adhesive bandages (CAST) of the right hind legs. Zoledronic acid (ZOL, 120 mg/kg) or saline were subcutaneously injected twice a week. Two weeks later, the right tibiae were taken out, and histomorphometrically analyzed. (A) Bone volume over total volume (BV/TV), trabecular thickness (Tb.Th), trabecular numbers (Tb.N) and trabecular separation (Tb.Sp) were assessed in trabecular bones. Data are expressed as the mean ± standard deviation (SD) (n=6). (B) TRAP staining was performed in the resected tibiae. Representative results were shown (magnification, ×100) (upper). Bars: 200 mm. Higher magnification (×400) of the boxed areas in the upper panels were shown in the lower panels. Bars: 50 mm. (C) Numbers of osteoclasts (N.Oc)/bone surface (BS) (/mm) were counted. Data are expressed as the mean ± SD (n=6). **P<0.01. (D) 5TGM1/luc multiple myeloma (MM) cells were inoculated into the right tibiae 2 weeks after DN. The mice were subcutaneously injected with 120 mg/kg zoledronic acid (ZOL) or saline twice a week. IVIS images were taken 4 weeks after the inoculation. Control groups were given saline as a vehicle. For IVIS imaging, 100 mL of 15 mg/mL D-luciferin in phosphate-buffered saline was injected intraperitoneally to mice before taking images. Five minutes after D-luciferin injection, the mice were placed in the imager with the mice anesthetized with 2% isoflurane while IVIS images were taken. Tumor areas with luminescence shown in green, yellow and red were measured. Px: pixel. Data are expressed as the mean ± SD (n=5). **P<0.01.

746

haematologica | 2022; 107(3)

Letters to the Editor

TGF-β is released from the bone matrices through enhanced bone resorption, and activated by an acid and matrix metalloproteinases secreted from OC. Consistently, the phosphorylation of Smad2 was increased in MM tumor lesions in immobilized hind legs by sciatic denervation (Figure 3B), which may in part contribute to the further activation of TAK1 in MM lesions under mechanical unloading with enhanced osteoclastogenesis. Importantly, treatment with the

TAK1 inhibitor LL-Z1640-2 as well as the PIM inhibitor SMI16a was able to efficaciously reduce MM growth enhancement in the immobilized hind legs (Figure 3A), suggesting the therapeutic efficacy of these inhibitors under mechanical unloading. We are currently studying the therapeutic efficacy of LL-Z1640-2 and SMI16a in combination with proteasome inhibitors to maximize their anabolic as well as antitumor effects against MM. In addition, we noticed that multiple palpable solid

Figure 3. Legend on following page.

haematologica | 2022; 107(3)

747

Letters to the Editor

Figure 3. Multiple myeloma tumor growth and dissemination upon immobilization. (A) Right and left hind legs in the same mice were subjected to sciatic denervation (DN) and sham operation (control), respectively. Two weeks later, luciferase-transfected mouse 5TGM1 multiple myeloma (MM) cells were simultaneously inoculated into tibiae in both immobilized (right) and intact (left) hind legs in the same mice. The TAK1 inhibitor LL-Z1640-2 or the PIM inhibitor SMI16a were intraperitoneally injected at 20 mg/kg twice a week. Control groups were given saline as a vehicle. IVIS images taken at 4 weeks. Tumor areas with luminescence shown in green, yellow and red were measured. Px: pixel. Data are expressed as the mean ± standard deviation (SD) (n=5). **P<0.01. (B) 5TGM1 MM cells were inoculated into right tibiae 2 weeks after DN or sham operation. The right tibiae with tumor lesions were harvested at 4 weeks after the MM cell inoculation. Cell lysates were then collected from MM tumor lesions, and protein levels of the indicated factors were analyzed by western blotting analysis. βactin was used as a loading control. The following reagents were purchased from the indicated manufacturers: antibodies against phosphor-MAP3K7 (Thr187) from Cusabio (Cusabio Biotech, Wuhan, China); and antibodies against TAK1, PIM2, phosho-Smad2, Smad2, and β-actin from Cell Signaling Technology. (C) 5TGM1 MM cells transfected with green fluorescent protein (Gfp) or red fluorescent protein (Rfp) genes were inoculated into tibiae of right hind legs with DN and left sham-operated hind legs, respectively. The GFP-expressing 5TGM1/luc (5TGM1-GFP/Luc) cell line and the RFP-expressing 5TGM1/luc (5TGM1-RFP/Luc) cell line were generated by lentiviral transduction with the pLKO.1-puro-CMV-TurboGFP vector and pLKO.1-puro-CMV-TagRFP vector (Sigma-Aldrich, MO, USA), respectively. Four weeks later, IVIS images were taken. Blood was drawn from retro-orbital plexus, and tumors detected in the IVIS images were resected with surrounding tissues in the mice. Tumors emitting green or red fluorescence were visualized in resected samples with a fluorescence microscope (OLYMPUS SZX16). (D) Circulating 5TGM1-GFP and 5TGM1-RFP cells were analyzed in the blood samples from the mice at 4 weeks by flow cytometery.

tumorous lesions appeared in mice with sciatic denervation over time at around 4 weeks or later at sites distant from the tibiae where MM cells were inoculated. In order to better analyze the metastatic expansion of MM cells inoculated into the tibiae, we transfected 5TGM1 MM cells with either the green fluorescent protein (Gfp) or red fluorescent protein (Rfp) gene. The in vitro proliferation of the Gfp- and Rfp-transfected MM cells was the identical (Online Supplementary Figure S2C). The Gfp- and Rfp-transfected cells were inoculated into tibiae of the right hind legs with sciatic denervation and left shamoperated hind legs, respectively. Tumor lesions were detected at sites distant from the tibiae at 4 weeks in IVIS images (Figure 3C). Interestingly, all tumorous lesions metastasized outside of the tibiae were found to be composed of the MM cells labeled with GFP, indicating preferential extraosseous expansion of MM cells inoculated into tibiae in hind legs paralyzed with sciatic denervation. Furthermore, substantial numbers of GFPpositive cells but not RFP-positive cells were detected in sera drawn from the mice at 4 weeks (Figure 3D). These results demonstrate the acceleration of MM tumor growth with egression from the bone marrow into circulation and thereby extraosseous dissemination under immobilization or mechanical unloading. There were no significant changes in the serum levels of sclerostin (Online Supplementary Figure S3B) as well as Sost gene expression in mice with the sciatic denervation (Figure 1F). Robling et al. investigated the mechanoregulation of Sost mRNA and sclerostin under enhanced (ulnar loading) and reduced (hindlimb unloading) loading conditions.10 Sost transcripts and sclerostin protein levels were significantly reduced at 24 hours in the ulna fixed to the loading platens and actuator after 360 cycles of mechanical loading per day. In contrast, mice subjected to tail suspension (hindlimb unloading) for 3 days exhibited a significant increase in Sost mRNA expression in the tibia compared to those in ground control mice. Intriguingly, this upregulation subsided to be non-significant after 7 days of tail suspension. In our experiments, we analyzed Sost mRNA expression at 14 days after the immobilization, and found that Sost mRNA was not increased significantly. Sost mRNA induction by mechanical unloading may be temporal and should be studied in a timesequence manner. However, serum levels of sclerostin have been demonstrated to be increased in MM patients with active bone lesions11,12 and positively correlate with lumbar spinal bone mineral density in postmenopausal women.13 Consistent with the patients’ observation, serum levels of sclerostin were increased after MM cell inoculation in mice, and more in mice with mechanical unloading than in control mice (Online Supplementary Figure S3B), which may be due to the acceleration of MM tumor expansion resulting from mechanical unloading. 748

Bone is a unique microenvironment for MM cell growth and survival, which provides niches to foster clonogenic and dormant MM cells. The present study demonstrates that hind leg immobilization or mechanical unloading aggravates bone destruction and MM tumor expansion. In contrast, mechanical loading with repeated forced compression 14 and low intensity vibration 15 has been reported to suppress osteolysis and the growth of MM cells in bone. In order to keep bone mass in MM, repeated mechanical loading appears to play an important role. These observations warrant further study on the therapeutic merit of mechanical stress or loading in MM. Kotaro Tanimoto,1 Masahiro Hiasa,1 Hirofumi Tenshin,1 Jumpei Teramachi,2 Asuka Oda,3 Takeshi Harada,3 Yoshiki Higa,1 Kimiko Sogabe,3 Masahiro Oura,3 Ryohei Sumitani,3 Tomoyo Hara,3 Itsuro Endo,4 Toshio Matsumoto,5 Eiji Tanaka,1 and Masahiro Abe3 1 Department of Orthodontics and Dentofacial Orthopedics, Tokushima University Graduate School of Biomedical Sciences, Tokushima; 2Department of Oral Function and Anatomy, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Science, Okayama; 3Department of Hematology, Endocrinology and Metabolism, Tokushima University Graduate School of Biomedical Sciences, Tokushima; 4Department of Bioregulatory Sciences, Tokushima University Graduate School of Medical Sciences, Tokushima and 5Fujii Memorial Institute of Medical Sciences, Tokushima University, Tokushima, Japan Correspondence: MASAHIRO ABE - masabe@tokushima-u.ac.jp MASAHIRO HIASA - mhiasa@tokushima-u.ac.jp doi:10.3324/haematol.2021.278295 Received: January 12, 2021 Accepted: November 11, 2021. Pre-published: November 18, 2021. Disclosures: MA received research funding from Chugai Pharmaceutical, Sanofi K.K., Pfizer Seiyaku K.K., Kyowa Hakko Kirin, MSD K.K., Astellas Pharma, Takeda Pharmaceutical, Teijin Pharma and Ono Pharmaceutical, and honoraria from Daiichi Sankyo Company. All other authors declare no competing financial interests. Contributions:KT, MH and MA designed the research and conceived the project; KT, MH, HT, JT, AO, TH, YH, KS, MO and RS conducted animal experiments; TH, IE, TM, ET evaluated muscle and bone specimens; KT, MH, KS, MO and RS conducted in vitro cultures; KT, MH, JT, HT, AO, TH and YH performed immunoblotting and immunohistochemical analyses; KT, MH and AO performed flow cytometric analysis; and KT, MH, JT and HT performed transfection and PCR; KT, MH, IE, TM, ET and MA analyzed the data; KT, MH and MA wrote the manuscript. haematologica | 2022; 107(3)

Letters to the Editor

Funding: this work was supported in part by the JSPS KAKENHI grant numbers JP16K11504, JP17H05104, JP17KK0169, JP18K16118, JP18K08329, JP18H06294, JP19K21382 and 19K22719; and the Research Clusters program of Tokushima University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References 1. Miyazaki T, Zhao Z, Ichihara Y, et al. Mechanical regulation of bone homeostasis through p130Cas-mediated alleviation of NFkappaB activity. Sci Adv. 2019;5(9):eaau7802. 2. Friedman MA, Zhang Y, Wayne JS, Farber CR, Donahue HJ. Single limb immobilization model for bone loss from unloading. J Biomech. 2019;83:181-189. 3. Amblard D, Lafage-Proust MH, Laib A, et al. Tail suspension induces bone loss in skeletally mature mice in the C57BL/6J strain but not in the C3H/HeJ strain. J Bone Miner Res. 2003;18(3):561569. 4. Nakashima T, Hayashi M, Takayanagi H. New insights into osteoclastogenic signaling mechanisms. Trends Endocrinol Metab. 2012;23(11):582-590. 5. Abe M, Hiura K, Wilde J, et al. Osteoclasts enhance myeloma cell growth and survival via cell-cell contact: a vicious cycle between bone destruction and myeloma expansion. Blood. 2004;104(8):2484-2491. 6. Lawson MA, McDonald MM, Kovacic N, et al. Osteoclasts control reactivation of dormant myeloma cells by remodelling the endosteal niche. Nat Commun. 2015;6:8983.

haematologica | 2022; 107(3)

7. Asano J, Nakano A, Oda A, et al. The serine/threonine kinase Pim-2 is a novel anti-apoptotic mediator in myeloma cells. Leukemia. 2011;25(7):1182-1188. 8. Teramachi J, Tenshin H, Hiasa M, et al. TAK1 is a pivotal therapeutic target for tumor progression and bone destruction in myeloma. Haematologica. 2021;106(5):1401-1413. 9. Hiasa M, Teramachi J, Oda A, et al. Pim-2 kinase is an important target of treatment for tumor progression and bone loss in myeloma. Leukemia. 2015;29(1):207-217. 10. Robling AG, Niziolek PJ, Baldridge LA, et al. Mechanical stimulation of bone in vivo reduces osteocyte expression of Sost/sclerostin. J Biol Chem. 2008;283(9):5866-5875. 11. Terpos E, Christoulas D, Katodritou E, et al. Elevated circulating sclerostin correlates with advanced disease features and abnormal bone remodeling in symptomatic myeloma: reduction post-bortezomib monotherapy. Int J Cancer. 2012;131(6):1466-1471. 12. Terpos E, Berenson J, Raje N, Roodman GD. Management of bone disease in multiple myeloma. Expert Rev Hematol. 2014;7(1):113125. 13. Polyzos SA, Anastasilakis AD, Bratengeier C, Woloszczuk W, Papatheodorou A, Terpos E. Serum sclerostin levels positively correlate with lumbar spinal bone mineral density in postmenopausal women--the six-month effect of risedronate and teriparatide. Osteoporos Int. 2012;23(3):1171-1176. 14. Rummler M, Ziouti F, Bouchard AL, et al. Mechanical loading prevents bone destruction and exerts anti-tumor effects in the MOPC315.BM.Luc model of myeloma bone disease. Acta Biomater. 2021;119:247-258. 15. Pagnotti GM, Chan ME, Adler BJ, et al. Low intensity vibration mitigates tumor progression and protects bone quantity and quality in a murine model of myeloma. Bone. 2016;90:69-79.

749

Letters to the Editor

GNE-related thrombocytopenia: evidence for a mutational hotspot in the ADP/substrate domain of the GNE bifunctional enzyme The GNE gene encodes UDP-N-acetylglucosamine (UDP-GlcNAc) 2-epimerase/N-acetylmannosamine (ManNAc) kinase (GNE), a bifunctional enzyme catalyzing the synthesis of a sialic acid called 5-acetylneuraminic acid (Neu5Ac).1 Mutations of GNE are responsible for GNE myopathy (OMIM #605820), an autosomal recessive late-onset progressive muscle disorder1 and sialuria (OMIM #269921), an autosomal dominant disease characterized by a congenital impairment of sialic acid metabolism.1 In a small set of patients, biallelic mutations of GNE have only recently been associated with thrombocytopenia, either isolated or combined with muscle weakness.2-6 However, it is unclear why only a few of the almost 1,000 individuals carrying biallelic mutations of GNE show thrombocytopenia. We studied two families with severe thrombocytopenia using whole exome sequencing (WES). Family 1 proband (P1) was an 18 months-old boy born to consanguineous Egyptian parents with a platelet count of 5x109/L at birth. The proband (P2) of family 2 was a 4year-old boy, the third child of consanguineous Moroccan parents, who had scattered petechiae associated with severe thrombocytopenia (platelet count 4x109/L) in his first hours of life (Figure 1A). In both P1 and P2 allo- and auto-antibodies against platelet antigens were not found in the mother's serum. Parents and siblings were healthy with normal blood counts, and no family history of thrombocytopenia was reported in either family. Splenomegaly or dysmorphic features were not observed in either of the patients, and the neurological assessment showed normal psychomotor development. Creatine phosphokinase (CPK) level was average, and no sign of myopathy was detected. In P1, both karyotyping and comparative genomic hybridization (CGH) array did not reveal any chromosomal alteration. The WES analysis allowed us to identify two novel homozygous variants (c.1546_1547delinsAG and c.1724C>G) of the GNE gene, leading to the p.Val516Arg and p.Thr575Arg missense substitutions, respectively (Figure 1A and B). Extensive analysis of the exome data did not yield any other potential pathogenic variant, not even in the inherited thrombocytopeniacausing genes (IT-related genes; Online Supplementary Table S1). Moreover, we analyzed the runs of homozygosity (ROH) shared by P1 and P2 searching for potentially deleterious variants. Candidate genes were regarded as those whose mutations are associated with thrombocytopenia (n=56 from Online Supplementary Table S1) and those enlisted in the gene ontology (GO) term “Hemopoiesis” (n=788). Except for the mutations in GNE, all the other homozygous variants were excluded based on pathogenicity and splicing bioinformatic predictions (Online Supplementary Table S2). Substitutions p.Val516Arg and p.Thr575Arg are rare variants affecting well conserved residues during evolution (Figure 1C). Their potential deleterious effect was supported by segregation analysis, bioinformatics predictions (Online Supplementary Table S2), and significant reduction of the GNE protein expression, which was likely to maintain some residual activity due to the incompatibility of complete loss of the GNE function with life (Figure 1D).7 Consistent with alteration of the 750

GNE kinase activity, the transferrin serum glycoforms analysis revealed a higher level of the asialo, disialo and trisialo forms and a correspondent decrease in the tetrasialotransferrin form in both patients (Figure 2A). The hematological and clinical features of P1 and P2 are strikingly similar. Except for petechiae and minor post-traumatic bruises (grade 1/5) occurring when the platelet count was below 10-20x109/L, no clinically significant bleeding was reported regardless of treatment. Consistent with data in the literature,4 they had increased mean platelet volume (MPV) (MPV 11.9 fL in P1 and MPV 10.8 fL in P2) and high immature platelet fraction (IPF) (IPF 50-80% in P1 and IPF 39-89% in P2) (Figure 2B). In bone marrow aspirates of both the affected individuals the number of megakaryocytes was markedly increased, and several immature small-sized and hypolobulated megakaryocytes were observed (Figure 2C to E). Interestingly, this condition mimicked the pathophysiology of another inherited defect of platelet sialylation, namely SLC35A1 deficiency, which affects the same biochemical pathway,8 and is partly reminiscent of the peripheral platelet destruction in immune thrombocytopenia (ITP). High-dose intravenous immunoglobulin and steroid treatment resulted in no improvement of the patients’ platelet count, therefore both probands required regular platelet transfusions in the first year of life (Figure 2F). The response to platelet transfusion and the megakaryocytes features suggest that the patients’ platelets are more rapidly removed from circulation for intrinsic cellular defects, such as sialylation reduction, rather than a decreased platelet production.4 Both patients were treated with romiplostim to reduce the need for transfusions. P1 responded at low doses (4 µg/kg/week) and his platelet count was higher than 25x109/L during treatment, which is currently ongoing (Figure 2G), while P2 required a high dose of romiplostim (up to 10 µg/kg/week) to obtain a substantial, extremely unstable, response (Figure 2G), thus discouraging the continuation. Fluctuating response to romiplostim was also reported in patients with ITP further supporting the similarity with GNE-related thrombocytopenia.8 Currently, despite a low platelet count (510x109/L), patient P2 did not experience any bleeding up to the last follow-up. The absence of severe bleeding despite extremely low platelet counts was in line with other cases reported in the literature (Online Supplementary Table S3). Among these, only two (P6 and P17) experienced severe or lifethreatening hemorrhages. The mild bleeding diathesis might be attributed to the abundance of young enlarged platelets, which have a prothrombotic potential.9 These data suggested that prophylactic treatment might be needed only in specific conditions (e.g., neonatal period; severe hemorrhage; surgical procedures), and an approach aimed at treating only acute events with platelet transfusion might be considered in most patients. Several lines of evidence supported the hypothesis that GNE mutations can cause thrombocytopenia. At least to our knowledge, 20 patients from nine unrelated families of the nearly 1,000 individuals with alterations of this gene have been reported to have thrombocytopenia (Online Supplementary Table S3).2-4,10,11 Of the 15 different variants identified in these patients, including P1 and P2, seven have been previously reported in patients with GNE myopathy (Figure 3). Except for p.His188Tyr, which is in cis with a known mutation (p.Asn550Ser) associated with myopathy,4 the others (p.Asp444Tyr, haematologica | 2022; 107(3)

Letters to the Editor

p.Gly447Arg, p.Gly506Phe, p.Val516Arg, p.Leu517Pro, p.Thr575Arg, and p.Gly578Ser) were regarded as specific variants associated with isolated thrombocytopenia. Of note, these seven variants spread over a region of approximately 130 amino acids of the ManNac kinase domain (Figure 3).

We analyzed the structure of N-acetylmannosamine kinase in complex with N-acetylmannosamine and ADP (2yhy)12 to assess the potential effect of the mutations on the structure and function of the enzyme (Figure 3). Residues Asp444, Gly447 and Thr575 in the ADP pocket, and Gly506 and Gly578 close to the substrate domain

Figure 1. Identification of novel mutations of the GNE gene. (A) Pedigrees and segregation analysis in the two families F1 and F2. (B) Electropherograms of exon 9 showing the c.1546_1547delinsAG and c.1724C>G substitutions in probands P1 and P2, respectively. Sanger sequencing was performed using the following primers: 9F/5’-TTCTAGAAATCTTTAAGGTGCTATGG-3’ and 9R/5’-CCACCTGACCATGTTGAAGA-3’. (C) Protein multiple alignments, showing conservation through different species at residues (in red) affected by the p.Val516Arg and p.Thr575Arg mutations. H. sapiens (NP_001121699.1), P. troglodytes (XP_003312121.1), M. mulatta (XP_001082113.2), C. lupus (XP_003431623.1), B. taurus (NP_001178072.2), M. musculus (NP_056643.3), R. norvegicus (NP_446217.1), G. gallus (NP_001026603.2), D. rerio (NP_957177.1), and X. tropicalis (NP_001072728.1). (D) Western blot and of total lysates from lymphoblast cells of P1 and P2. Total protein lysates were prepared from these cells using M-PERTM Mammalian Protein Extraction Reagent (Thermo Fisher Scientific). Protein quantification shows only a partial expression (39% and 79%, respectively) of GNE expression compared to wild-type (CTRL) (***P<0,002). Actin was used as a loading control for protein quantification. The antibodies were used as follows: anti-GNE (Santa Cruz Biotechnology, sc-376057, 1:500) and anti-β-actin (Santa Cruz Biotechnology, sc-47778, 1:4,000) as primary antibodies, anti-mouse immunoglobulin conjugated with horseradish peroxidase (HRP) (Bethyl, A90-116P, 1:10,000) as a secondary antibody. Statistical analysis was performed using the t-test. Error bars represent the standard deviation of 4 independent experiments.

haematologica | 2022; 107(3)

751

Letters to the Editor

Figure 2. Blood, bone marrow features, and response to transfusion and treatment. (A) Sialotransferrin profile determined by ion-exchange chromatography using a commercial kit (CDT in Serum, Recipe München). The different isoforms were pointed out by UV detection at 460 nm and quantified by the “area percent method” (i.e., the relative abundance of each isoform is expressed as the percentage ratio of the peak area compared to the sum of the areas of all the peaks). (B) Peripheral blood smear of P2 showing enlarged platelets. (C to E) Bone marrow aspirates with an increased number of megakaryocytes at different stages of maturation. May-Grünwald-Giemsa staining; original magnification 100X (B), 10X (C), 20X (D), and 40X (E). (F) Response to platelet transfusion. Median platelet count before platelet transfusion and up to 7 days following transfusion are shown for both patients (P1 and P2’s specific values are indicated in brackets). (G) Time course of platelet count in response to treatments for P1 (left) and P2 (right) in response to romiplostim administration at different dosages. The dark grey bar indicates a period of complete transfusion dependency, with transfusion every 5-7 days. Values of platelets measured within 7 days after platelet transfusion are not shown.

may directly affect the enzyme active site, impairing ADP or substrate binding, respectively (Figure 3). Otherwise, p.Val516 and p.Leu517 are localized in the hydrophobic core and when mutated into an arginine or proline, respectively, may destabilize the fold and the conformation of the entire protein. Accordingly, in vitro mutagenesis of the highly conserved Asp444 residue in the ADP binding pocket resulted in the complete loss of the kinase function, though retaining the epimerase activity.13 Therefore we could hypothesize that megakaryocytes and platelets are more sensitive than other cells to defective kinetic activity or substrate-binding affinity, thus explaining the occurrence of thrombocytopenia. Nevertheless, whether GNE mutations are responsible for thrombocytopenia either isolated or in combination with muscle wasting remains to be elucidated. Indeed, considering that the GNE myopathy typically appears in 752

the third decade of life, we cannot exclude that patients with only thrombocytopenia will develop myopathy later in their life.2-4 Patients carrying the p.Asp444Tyr (F3), p.Gly447Arg (F5), p.Val516Arg (F1), p.Leu517Pro (F7), and p.Thr575Arg (F2) mutations were neonates or in their first/second decade of life. Instead, among the six individuals homozygous for p.Gly506Phe (F6) or p. Gly578Ser (F4) and all between 24-42 years of age, only two have subclinical features of myopathy, suggesting that, in addition to a low platelet count, this mutation could correlate with a mild form of muscle wasting of late onset. Trying to explain why only few patients with GNE mutations have a low platelet count, we cannot exclude that GNE variants cause thrombocytopenia only when cosegregating with other genetic factors. Whereas WES analysis did not provide any other plausible candidate in our families, the recessive transmission of variants in other haematologica | 2022; 107(3)

Letters to the Editor

Figure 3. Localization of the GNE mutations associated with thrombocytopenia and three-dimensional structure of the GNE enzyme. (A) Among the 15 mutations (Online Supplementary Table S3), the 7 associated with myopathy are indicated below the schematic representation of the protein structure; the other 8 are depicted above the protein (in red, the novel mutations reported in this paper). All the mutations identified in patients are in a homozygous state except those represented in matched colored boxes that are heterozygous biallelic GNE mutations found in a single patient. H188Y and N550S (light grey boxes) are homozygous mutations identified in the same patient. UDP-GlcNAc 2-epimerase: UDP-N-acetylglucosamine 2-epimerase; ManNAc kinase: N-acetylmannosamine; NH2: NH2-terminus; COOH: COOH-terminus; RUF: a region with unknown function; NES: putative nuclear export signal; AS: allosteric site. Nomenclature of mutations was referred to the NM_001128227.3 transcript. (B) The overall structure (left) and the zoom of the enzymatic pocket (right) of GNE. The side chains of the positions affected by the mutations discussed in this article are explicitly indicated. The structure (Protein Data Bank [PDB] entry 2yhy) corresponds to the GNE complex with ManNAc and ADP. The structure was analyzed by PyMOL and MOLMOL graphic support tools. The degree of exposure of the residues affected by mutations was established by DSSP (Define Secondary Structure of Proteins) analysis.

genes, such as ANKRD18A, FRMPD1, FLNB, PRKACG, have been reported in other cases.2,3,5,14 However, their potential impact was not further investigated, except for p.Ile74Met in PRKACG, whose functional studies demonstrated its effect in thrombocytopenia.14 In summary, although the role of GNE mutations is well-documented in GNE myopathy and sialuria, we identified two novel GNE variants, which together with a few other mutations reported in the literature could explain thrombocytopenia and extend the clinical phenotype of the GNE defects. In both patients, as well as in those of families F3 and F5 from a literature review, severe thrombocytopenia was reported since the first days of life, when the differential diagnosis of thrombocytopenia was broad and included thrombocytopenia secondary to sepsis and critical care, neonatal allo- and auto-immune thrombocytopenia, or ITP, such as congenital amegakaryocytic thrombocytopenia.15 Therefore, evaluating the sialotransferrin profile in patients with suspected inherited thrombocytopenia, large platelets and increased reticulated fraction might provide an important diagnostic clue. Roberta Bottega,1* Antonio Marzollo,2,3* Maddalena Marinoni,4 Emmanouil Athanasakis,1 Ilaria Persico,5,6 Anna Monica Bianco,1 Michela Faleschini,1 Erica Valencic,1 haematologica | 2022; 107(3)

Daniela Simoncini,4 Linda Rossini,2 Fabio Corsolini,7 Martina La Bianca,1 Giuseppe Robustelli,4 Maria Gabelli,2 Massimo Agosti,4 Alessandra Biffi,2 Paolo Grotto,8 Valeria Bozzi,9 Patrizia Noris,9,10 Alberto B. Burlina,11 Adamo Pio d'Adamo,1,5 Alberto Tommasini,1,5 Flavio Faletra,1 Annalisa Pastore12,13 and Anna Savoia1,5 1 Istitute for Maternal and Child Health – IRCCS Burlo Garofolo, Trieste, Italy; 2Pediatric Hematology, Oncology and Stem Cell Transplant Division, Padua University Hospital, Padua, Italy; 3 Fondazione Città della Speranza, Istituto di Ricerca Pediatrica, Padua, Italy; 4Maternal and Child Department, F. Del Ponte Hospital, Varese, Italy; 5Department of Medical Sciences, University of Trieste, Trieste, Italy; 6Department of Genetics and Microbiology, Universitat Autonoma de Barcelona, Barcelona, Spain; 7LABSIEM - Laboratory for the Study of Inborn Errors of Metabolism, Pediatric Clinic and Endocrinology, Istituto Giannina Gaslini, Genova, Italy; 8Pediatric Department, Hospital of Treviso - Oderzo, Treviso, Italy; 9 Biotechnology Research Laboratories, IRCCS Policlinico San Matteo Foundation, Pavia, Italy; 10Department of Internal Medicine and Medical Therapy, University of Pavia, Pavia, Italy; 11Division of Inherited Metabolic Diseases, Regional Center for Expanded Neonatal Screening Department of Women and Children’s Health, University Hospital of Padova, Padova, Italy; 12King’s College London, Department of Clinical Neuroscience, Denmark Hill Campus, London, UK and 13European Synchrotron Radiation Facility 71, Grenoble, France 753

Letters to the Editor

*RB and AM contributed equally as co-first authors. Correspondence: ANNA SAVOIA - anna.savoia@burlo.trieste.it doi:10.3324/haematol.2021.279689 Received: August 11, 2021. Accepted: November 11, 2021. Pre-published: November 18, 2021. Disclosures: no conflicts of interest to disclose. Contributions: RB, AM, MM and AS designed research, interpreted data, and wrote the manuscript; EA, AMB, MF, MLB and APDA performed NGS analysis; IP, EV, FC, ABB and AP performed experiments; DS, LR, GR, MG, MA, AB, PG, VB, PN, AT and FF collected hematological and other clinical data. All authors critically revised the paper and approved the final version. Acknowledgements: we thank Samuela Francescato and Giulia Bossi for their assistance in the production of the figures, and Dr. Martina Bradaschia for the language revision of the manuscript. The authors would like to acknowledge Prof. F. de Cothi and Prof. A. Dhele for the constructive discussions. Funding: this work was supported by IRCCS Burlo Garofolo (grant numbers RC 31/17 and RC 02/18) and IRCCS Policlinico San Matteo Foundation (Intramural research grant).We thank Fondazione Città della Speranza ONLUS (http://cittadellasperanza.org/) and Fondazione Giacomo Ascoli Onlus (https://www.fondazionegiacomoascoli.it/) for their support of our scientific work.

References 1. Hinderlich S, Weidemann W, Yardeni T, Horstkorte R, Huizing M. UDP-GlcNAc 2-epimerase/ManNAc kinase (GNE): a master regulator of sialic acid synthesis. Top Curr Chem. 2015;366:97-137. 2. Futterer J, Dalby A, Lowe GC, et al. Mutation in GNE is associated with severe congenital thrombocytopenia. Blood. 2018;132(17):1855-1858. 3. Johnson B, Lowe GC, Futterer J, et al. Whole exome sequencing

754

identifies genetic variants in inherited thrombocytopenia with secondary qualitative function defects. Haematologica. 2016;101(10):1170-1179. 4. Revel-Vilk S, Shai E, Turro E, et al. variants causing autosomal recessive macrothrombocytopenia without associated muscle wasting. Blood. 2018;132(17):1851-1854. 5. Izumi R, Niihori T, Suzuki N, et al. GNE myopathy associated with congenital thrombocytopenia: a report of two siblings. Neuromuscul Disord. 2014;24(12):1068-1072. 6. Zhen C, Guo F, Fang X, Liu Y, Wang X. A family with distal myopathy with rimmed vacuoles associated with thrombocytopenia. Neurol Sci. 2014;35(9):1479-1481. 7. Schwarzkopf M, Knobeloch KP, Rohde E, et al. Sialylation is essential for early development in mice. Proc Natl Acad Sci U S A. 2002;99(8):5267-5270. 8. Kauskot A, Pascreau T, Adam F, et al. A mutation in the gene coding for the sialic acid transporter SLC35A1 is required for platelet life span but not proplatelet formation. Haematologica. 2018;103(12):e613-e617. 9. Bongiovanni D, Santamaria G, Klug M, et al. Transcriptome analysis of reticulated platelets reveals a prothrombotic profile. Thromb Haemost. 2019;119(11):1795-1806. 10. Li X, Li Y, Lei M, et al. Congenital thrombocytopenia associated with GNE mutations in twin sisters: a case report and literature review. BMC Med Genet. 2020;21(1):224. 11. Mekchay P, Ittiwut C, Ittiwut R, et al. Whole exome sequencing for diagnosis of hereditary thrombocytopenia. Medicine (Baltimore). 2020;99(47):e23275. 12. Martinez J, Nguyen LD, Hinderlich S, et al. Crystal structures of N-acetylmannosamine kinase provide insights into enzyme activity and inhibition. J Biol Chem. 2012;287(17):13656-13665. 13. Effertz K, Hinderlich S, Reutter W. Selective loss of either the epimerase or kinase activity of UDP-N-acetylglucosamine 2epimerase/N-acetylmannosamine kinase due to site-directed mutagenesis based on sequence alignments. J Biol Chem. 1999;274(40):28771-28778. 14. Manchev VT, Hilpert M, Berrou E, et al. A new form of macrothrombocytopenia induced by a germ-line mutation in the PRKACG gene. Blood. 2014;124(16):2554-2563. 15. Sillers L, Van Slambrouck C, Lapping-Carr G. Neonatal thrombocytopenia: etiology and diagnosis. Pediatr Ann. 2015;44(7):e175180.

haematologica | 2022; 107(3)

Letters to the Editor

Dissociated humoral and cellular immune responses after a three-dose schema of BNT162b2 vaccine in patients receiving anti-CD20 monoclonal antibody maintenance treatment for B-cell lymphomas The anti-CD20 monoclonal antibodies rituximab (R) and obinutuzumab (O) are used as maintenance therapy every two months for two-three years in patients with follicular lymphoma (FL) and mantle cell lymphoma (MCL). This strategy improves event-free survival (EFS) and/or overall survival (OS) after immunochemotherapy in responding patients.1-3 Severe forms of COVID-19 with prolonged carriage of the virus in patients on R maintenance have been reported due to a significant alteration of humoral immunity in this context.4,5 Therefore, during the COVID-19 epidemic, clinicians are faced with the question of whether to discontinue maintenance therapy or not. The Pfizer/BioNTech RNA vaccine BNT162b2 (Comirnaty®) has shown efficacy and safety data in all age groups and against several variants. 6,7 The specificities of mRNA vaccines also suggest that they may induce a better T lymphocyte response.8 The aim of our study was to evaluate the post-vaccination humoral and T cell response based on the serological data and enumeration of interferon gamma (IFNγ)-producing T cells in response to SARS-COV-2 specific antigens (Elispot assay) in a group of patients with a longterm anti-CD20 antibody lymphoma treatment. Patients with lymphoma receiving or initiating their maintenance therapy by anti-CD20 antibodies (R or O) and treated in a single center - Centre Henri Becquerel, France - were selected. The study protocol was approved by the local and national ethics committees (Internal Review board N° 2106B and Comité de Protection des Personnes, Ile de France IV, registered as number NCT04918940 at clinical.trial.gov) and written informed consent was obtained. Serologies and Elispot assays were repeated between D21-D28 after the first vaccination (V1), one month after the second vaccination (V2) and 1 month after the third vaccination (V3). Twenty patients were enrolled in this study: 16 FL (80%); 3 MCL (15%); 1 Marginal Zone Lymphoma (5%) between 21 May 2021 and 1 July 2021. They received three doses of BNT162b2 vaccine after the initial phase of immunochemotherapy, in a steady state regarding lymphoma status (Table 1), during the maintenance phase. Ten patients were treated with R and ten with O. It should be noted that, in order to vaccinate patients as quickly as possible, the vaccine injections were carried out without any time interval rules with respect to the anti-CD20 antibody injections. The median number of R or O infusions received during the maintenance phase at the time of the first vaccination was four infusions (range 0-15, Table 1). The B-cell depletion was profound for all patients with no detectable CD19+/CD20+ cells. T-cell counts also indicated a low rate of both CD4+ and CD8+ T cells in 12/17 (70%) and 9/17 (53%) cases, respectively. Conversely, NK CD56+CD16+ cells were in the normal range in 16/17 (94%) cases (Table 2). With a median follow-up of 96 days (range 75-150), no COVID-19 infection was detected in the cohort. The anti–SARS-CoV-2 post-vaccine antibody response against the spike protein (RBD) was centrally assessed using the ARCHITECT SARS-Cov-2 IgG II Quant (Abbott) CMIA test, with titers >50 arbitrary units (AU) per milliliter considered positive (measurement interval: 6.8–80,000 AU/ml; positive agreement, 99.4%; negative agreement, 99.6%). At baseline, antibodies against the haematologica | 2022; 107(3)

SARS-CoV-2 virus spike protein were detected in 1/17 patients (5.9%, patient n°7, discussed hereafter). After one, two and three vaccine injections, all except this patient remained below the threshold of 50 AU/ml, indicating that the humoral response is deeply impaired in this cohort and does not seem improved by a triple-injection schedule. Therefore, the positivity rate one month after the third injection in this cohort remained unchanged at 5% (1/20 tested patients) with no additional responder. Of note, antibody titers only increase after vaccination in patient n°7 despite the fact that, as with other patients, no CD19+/CD20+ B-cells were detectable by cytometry. This male FL patient was treated initially by O-chlorambucil as a first-line treatment. He developed a COVID-19 disease after the first cycle (C1-D8) of this treatment. Other than oxygen therapy, no intensive care was required but the infection led to the postponement of immunochemotherapy for two months. The patient was thereafter treated by R-CHOP (six cycles). This led to the patient receiving one infusion of binutuzumab and six infusions of rituximab before his first vaccine injection. Contrasting with the deep alteration of the humoral response to SARS-COV-2 vaccination, we demonstrated Table 1. Population clinical features

Study population (n=20) Age (years) Median (q1-q3) 65.5 (58.5-74) Min-max 46-77 Sex Male 7 (35%) Female 13 (65%) Lymphoma subtype MCL 3 (15%) MZL 1 (5%) FL 16 (80%) Maintenance antibody Rituximab 10 (50%) Obinutuzumab 10 (50%) Time between first vaccination and last chemotherapy cycle (days) Median (q1-q3) 243.5 (131-430) Min-max 0-781 Number of cycles (R/O) received before first vaccination Median (q1-q3) 4.0 (2-8) Min-max 0-15 Disease status at the time of vaccination Complete remission 18 (90%) Partial response 2 (10%) Progressive disease 0 Chemotherapy performed before maintenance O-CHOP x 6 7 (35%) R-CHOP x 6 6 (30%) O-Bendamustine x 6 3 (15%) Lenalidomide /rituximab (R2) x 9 2 (10%) R-CHOP/R-HAD 2 (10%) COVID status before vaccination Proven COVID 1 (5%) No history of previous infection 19 (95%) Patients with active or uncontrolled lymphoma diseases were excluded. FL: follicular lymphoma; MCL: Mantle cell lymphoma; MZL: Marginal Zone lymphoma; O: obinutuzumab; R:rituximab; R-CHOP: Rituximab, Cyclophosphamide, Doxorubicin, Prednisolone and Vincristine; R-HAD: Rituximab, high-dose Ara-C and dexamethasone.

755

Letters to the Editor

the induction of an anti-spike T-cell response, as assessed by IFNγ Elispot assay, irrespective of the antibody response. Elispot assays were performed, as previously reported.9 At baseline, and after one, two and three injections, the median number of SFC/106 CD3+ T cells were 0 (0-20), 112.5 (0;339), 679.0 (202;1551) and 845.0 (243; 1305), respectively. Overall, one month after V2 and V3, 17/19 patients (89%) displayed IFNγ-producing T cells reactive to S (S1+S2) peptide pools. A significant increase in T-cell response, compared to baseline, was observed throughout the vaccination process until V2. The magnitude of T-cell response increased after the third vaccination in eight patients but appears globally unchanged for the entire cohort (Figure 1). T-cell response did not differ significantly after the complete vaccination schedule according to the type of anti-CD20 antibodies used (mean SFC/106 CD3+ T cells = 1045.0 (749; 1338) for O and 600.0 (160; 845) for R) or according to the number of anti-CD20 infusions received before vaccination (data not shown). In this cohort, no pre-existing immune memory was suggested by responses to other non-spike antigens in 19/20 patients. Conversely, as expected, patient n°7 displayed IFNγ-producing T cells reactive to N, M, and N7A SARS-CoV-2 proteins. Importantly, the level of T-cell response was at least equivalent to that observed in kidney-transplanted patients. We previously reported, using the same assay, a median rate of 212 SFCs /106 CD3+ T cells after two doses in responding patients (versus 679 SFCs /106 CD3+ T in the present cohort) and 330 SFC/106 CD3+ T cells after the third injection (versus 845 SFCs/106 CD3+T)(paper submitted, in revision).9 Similarly, the level of T-cell response was not below the that reported in healthy individuals, showing a median of 165 SFU/106 PBMCs 28 days after two doses (Angyal et al, Lancet 2021, in press). These comparisons suggest that T-cell response at least remains conserved in these highly immunocompromised patients. However, there is no clear demonstration as to whether T-cell activity is sufficient to protect vaccinated patients from COVID-19 infection.10 Robust T-cell responses to the SARS-CoV-2 virus occur in most individuals with COVID-19.11 Furthermore, SARS-CoV-2-specific T cells have been detectable in antibody-seronegative exposed family members and convalescent individuals with a history of asymptomatic and mild COVID-19, consistent with a non-redundant role of immune protection against COVID-19.11 Importantly, after the third vaccine dose, we still observed an increased interferon-γ response to the SARS-Cov-2 spike protein. This was particularly visible for patient n°7, characterized by a previous COVID-19 disease that occurred at the beginning of his lymphoma treatment, when only one dose of obinutuzumab had been delivered. To improve the rate of seroconversion or to maintain a humoral response in elderly individuals or immunocompromised patients, a third injection was proposed. However, the impact of such a strategy in such a case of very deep B-cell depletion is still uncertain. In the kidney transplant setting, one study has shown that a third dose of mRNA-1273 vaccine induced a serologic response in 49% of kidney transplant recipients who did not respond after two doses.12 Administration of a third dose of the BNT162b2 vaccine to solid organ transplant recipients, or to recipients of allogeneic HSCT, also significantly improved the immunogenicity of the vaccine.13,14 In our cohort characterized by complete B-cell depletion, we did not observe any improvement of the antibody response 756

after the third injection. The vaccination schedule could also be improved by increasing time-lapse between the second and third injection or by offering heterologous prime-boost strategies that demonstrated increased levels of neutralizing antibody titers in immunocompetent patients, as compared to homologous prime-boost strategies.15 Nevertheless, whether this optimization might result in an improved level of protection remains uncertain. REGN-COV2, a neutralizing antibody cocktail, may also be proposed as a prophylactic strategy in these immunosuppressed patients after or before SarsCov2 exposition. The question of delaying or ceasing maintenance therapy is clearly raised for patients with MCL and FL in the current pandemic context. The benefit in terms of OS in MCL, contrasting with the benefit only for PFS in FL, is an important point to consider when considering the benefit-risk balance. Other points such as social context, the ability to follow physical protective measures and access to passive immunization with anti-S monoclonal antibody therapy, should be considered for each individual patient in order to correctly evaluate their benefit-risk balance. Finally, given the lack of post-vaccinal humoral response observed in our cohort, vaccination may still provide a limited but significant protection by triggering Table 2. Population biological parameters at baseline

Biological parameter Lymphocytes (109/L) N median (q1;q3) min - max T-cell (109/L) N median (q1;q3) min - max CD19+/CD20+ cell (109/L) N median (q1;q3) min - max CD4+ cell (109/L) N median (q1;q3) min - max CD8+ cell (109/L) N median (q1;q3) min - max NK-cell (109/L) N median (q1;q3) min - max Neutrophil count (109/L) N median (q1;q3) min - max Albumin (G/L) N median (q1;q3) min - max Gammaglobulin (G/L) N median (q1;q3) min - max

Reference values 1.5-4

1-2.2

0.11-0.57

0.53-1.4

0.33-0.92

0.07-0.48

1.5-7

42-50

5-15

Study population n=20 17 0.90 (0.7; 1.2) 0.30 ; 2.10 17 0.74 (0.6; 1.0) 0.20 ; 1.95 17 0.00 (0.0; 0.0) 0.00 - 0.00 17 0.41 (0.2; 0.5) 0.11 - 1.10 17 0.32 (0.1; 0.4) 0.07 - 1.10 16 0.15 (0.1; 0.2) 0.00 - 0.41 17 3.40 (2.1; 4.7) 1.80 - 5.90 16 41.00 (39.8; 43.2) 30.00 - 69.00 16 7.00 (5.8; 8.0) 3.00 - 8.00

haematologica | 2022; 107(3)

Letters to the Editor

Figure 1. SARS-CoV-2–reactive IFNγ-producing T cells after vaccination. Elispot assays were performed, as previously reported (8). Briefly, PBMCs (in concentrations adjusted to 2x105 CD3+ T cells per well) were plated in anti-IFNγ–coated Elispot 96-well plates in the presence of overlapping 15-mer peptide pools spanning the sequence of SARS-CoV-2 structural and nonstructural proteins: S (pool S1 spanning the N-terminal part of the protein including the S1-subunit, and pool S2 spanning the C-terminal part), N, M, NS3A, NS7A (JPT, Strassberg, Germany). Spots were counted with an automated ELISPOT reader (AID, Strassberg, Germany). Results were expressed as spot forming cells (SFC) per 106 CD3+ T cells. For each assay, a specific response was considered positive if the SFC number was superior to 3 standard deviations of spot numbers observed in wells without antigens (ranging between 9 and 20 SFC/106 CD3+ T cells).

a T-cell response. This should encourage patients and physicians to maintain a proactive vaccine policy. Sophie Candon,1,2 Véronique Lémée,3 Emilie Lévêque,4 Pascaline Etancelin,5 Cédric Paquin,5 Marion Carette,1 Nathalie Contentin,6 Victor Bobée,7 Mustafa Alani,6 Nathalie Cardinaël,6 Stéphane Leprêtre,6 Vincent Camus,6 Florian Bouclet,6 Edwige Boulet,8 Anne-Lise Ménard,6 Hélène Lanic,6 Aspasia Stamatoullas,6,9 Emilie Lemasle,6 Louis-Ferdinand Pépin,4 Doriane Richard,4 Sydney Dubois,6 Hervé Tilly,6,9 Alain Dalleac,5 Jean-Christophe Plantier,3 Manuel Etienne10 and Fabrice Jardin.6,9 1 Laboratory of Immunology and Biotherapies, Rouen University Hospital, France; 2INSERM U1234, University of Rouen Normandy, Rouen, France; 3Normandie Univ, UNIROUEN, UNICAEN, GRAM 2.0, Rouen University Hospital, Department of Virology, Rouen, France; 4Unit of Clinical Research, Centre Henri Becquerel, Rouen, France; 5Department of Biopathology, Centre Henri Becquerel, Rouen, France; 6Department of Clinical Hematology, Centre Henri Becquerel, Rouen, France; 7Laboratory of Hematology, Rouen University Hospital, France; 8Cliniques universitaires UCL de Mont Godinne, Namur, Belgium; 9INSERM U1245, University of Rouen Normandy, Rouen, France and 10Normandie Univ, UNIROUEN, UNICAEN, GRAM 2.0, Rouen University Hospital, Department of Infectious Diseases, Rouen, France Correspondence: FABRICE JARDIN- : fabrice.jardin@chb.unicancer.fr doi:10.3324/haematol.2021.280139 Received: October 2, 2021. Accepted: November 22, 2021. Pre-published: December 2, 2021. Disclosures: no conflicts of interest to disclose. haematologica | 2022; 107(3)

Contributions: FJ, ME, JCP, SG, HT, LFP and DR designed the trial; VL performed serologic analysis; SG and MC performed Elispot analysis; VB performed cytometry analysis; PE, CP performed additional biological analysis and logistical tasks; EL performed statistical analysis and collected data; HL, SL, EL, NC, VC, ALM, EB, MA, NC enrolled patients in the trial; FJ, HT, ME, SD and SC wrote the manuscript. All authors approved the final manuscript. Acknowledgments: the authors thank for their help and technical support: Delphine Robbe, Julie Libraire, Nathalie Breda, Laure Gaillon, Justine Loret, Laure Sulpice and Julie Lamulle.

References 1. Le Gouill S, Thieblemont C, Oberic L, et al. Rituximab after autologous stem-cell transplantation in mantle-cell lymphoma. N Engl J Med. 2017;377(13):1250-1260. 2. 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 chemotherapy (PRIMA): a phase 3, randomised controlled trial. Lancet. 2011;377(9759):42-51. 3. Marcus R, Davies A, Ando K, et al. Obinutuzumab for the firstline treatment of follicular lymphoma. N Engl J Med. 2017;377(14):1331-1344. 4. Yasuda H, Tsukune Y, Watanabe N, et al. Persistent COVID-19 pneumonia and failure to develop anti-SARS-CoV-2 antibodies during rituximab maintenance therapy for follicular lymphoma. Clin Lymphoma Myeloma Leuk. 2020;20(11):774-776. 5. Dulery R, Lamure S, Delord M, et al. Prolonged in-hospital stay and higher mortality after Covid-19 among patients with nonHodgkin lymphoma treated with B-cell depleting immunotherapy. Am J Hematol. 2021;96(8):934-944. 6. Polack FP, Thomas SJ, Kitchin N, et al. Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine. N Engl J Med. 2020;383(27):2603-2615. 7. Dagan N, Barda N, Kepten E, et al. BNT162b2 mRNA Covid-19 Vaccine in a nationwide mass vaccination setting. N Engl J Med. 2021;384(15):1412-1423. 8. Sahin U, Muik A, Derhovanessian E, et al. COVID-19 vaccine

757

Letters to the Editor

BNT162b1 elicits human antibody and TH1 T cell responses. Nature. 2020;586(7830):594-599. 9. Candon S, Guerrot D, Drouot L, et al. T cell and antibody responses to SARS-CoV-2: Experience from a French transplantation and hemodialysis center during the COVID-19 pandemic. Am J Transplant. 2021;21(2):854-863. 10. Bonelli MM, Mrak D, Perkmann T, Haslacher H, Aletaha D. SARS-CoV-2 vaccination in rituximab-treated patients: evidence for impaired humoral but inducible cellular immune response. Ann Rheum Dis. 2021;80(10):1355-1356. 11. Sekine T, Perez-Potti A, Rivera-Ballesteros O, et al. Robust T Cell immunity in convalescent individuals with asymptomatic or mild COVID-19. Cell. 2020;183(1):158-168.e14.

758

12. Benotmane I, Gautier G, Perrin P, et al. Antibody response after a third dose of the mRNA-1273 SARS-CoV-2 vaccine in kidney transplant recipients with minimal serologic response to 2 doses. JAMA. 2021;326(11):1063-1065. 13. Kamar N, Abravanel F, Marion O, Couat C, Izopet J, Del Bello A. Three doses of an mRNA Covid-19 vaccine in solid-organ transplant recipients. N Engl J Med. 2021;385(7):661-662. 14. Redjoul R, Le Bouter A, Parinet V, Fourati S, Maury S. Antibody response after third BNT162b2 dose in recipients of allogeneic HSCT. Lancet Haematol. 2021;8(10):e681-e683. 15. Tenbusch M, Schumacher S, Vogel E, et al. Heterologous primeboost vaccination with ChAdOx1 nCoV-19 and BNT162b2. Lancet Infect Dis. 2021;21(9):1212-1213.

haematologica | 2022; 107(3)

Letters to the Editor

Guideline for management of non-Down syndrome neonates with a myeloproliferative disease on behalf of the I-BFM AML Study Group and EWOG-MDS^ In neonates with myeloid hyperproliferation, apart from benign causes, Down syndrome (DS) related transient abnormal myelopoiesis (TAM), acute myeloid leukemia (AML) and juvenile myelomonocytic leukemia (JMML) are considered.1-3 Besides TAM, rarely, non-DS related transient myeloproliferative diseases occur, making clinical decisions challenging.4 TAM, according to World Health Organization (WHO) classification, only applies to children with (mosaic) Down syndrome.5 In the past, different terminology has been used in non-DS patients, such as transient myeloproliferative disease (TMD) and transient leukemia. Since distinction from TAM is important, and it is challenging to determine whether this disease will be transient, the consensus group introduced the novel term ‘infantile myeloproliferative disease’ (IMD), in order to distinguish it from TAM. Both TAM and IMD can usually be managed with a ‘watch and wait’ strategy, while most fullblown AML or JMML cases require intensive treatment. We collected rare IMD cases from study groups collaborating in the International Berlin-Frankfurt-Münster AML Study Group (I-BFM AML SG). In addition, we reviewed the literature for neonatal cases of malignant myeloid hyperproliferation without DS. Based on these data, we developed, together with I-BFM AML SG and the European Working Group of Myelodysplastic syndromes in Childhood (EWOG-MDS) members, by consensus, clinical recommendations for the diagnostic approach and current adequate classification of malignant myeloid hyperproliferation in infancy. This is meant guiding clinicians in choosing the right strategy, i.e., whether to ‘watch and wait’ or start highly intensive treatment in individual cases. We centrally collected detailed information from databases of I-BFM AML SG collaborators to identify clinical and genetic characteristics of additional, not yet reported, cases with IMD. Children younger than one year, diagnosed between 1990 and 2020, were included. Ethical approval and informed consent were obtained by each study group individually. Registration and data forms involved clinical features, hematological data, morphology and immunology, treatment, outcome and follow-up data. Available written reports of cytogenetic findings were collected and centrally reviewed by Dr. A. Buijs (University Medical Center Utrecht) and Prof. Dr. S. Raimondi (St. Jude Children’s Hospital, Memphis). We identified 15 new cases of IMD with, in some cases, novel recurrent molecular aberrations (Table 1). No germline aberrations were identified; however, standardized diagnostics did not always include germline testing. Thirteen patients had somatic trisomy 21 (T21) with or without a GATA1 mutation, one patient had low mosaic somatic trisomy 8 and a SETD2 mutation and one patient was not tested for somatic aberrations. Notably, among the 15 newly-added cases, in four patients, evaluation for GATA1 mutations was not performed. The search for available literature and case reports of non-DS transient leukemia was performed in the PubMed database. Publications indexed until 1 January 2021 were included. Search terms included TMD, TAM and transient leukemia, used separately and combined with non-Down, non-Down syndrome, and without Down syndrome. A cross-reference check was performed in key articles. We included 23 articles that described one or multiple patients that met our search criteria (Table 2). Unfortunately, in these cases too, routine testing of somatic GATA1 and haematologica | 2022; 107(3)

potential germline mosaic T21 was not always performed. Congenital/infant leukemia accounts for <1% of all childhood leukemias.6 When the rare event occurs in which a neonate is suspected of myeloid leukemia, TAM or IMD, clinical decision making can be challenging. Here, representatives of the I-BFM AML SG, together with JMML experts from the EWOG MDS, provide a clinically-applicable consensus of diagnostic logistics for children younger than six weeks. This is based on literature and newly-added cases from our international survey, which may support clinical decision-making in individual cases (Figure 1). During two meetings with leading members from both the I-BFM AML SG and EWOG MDS, relevant literature was discussed and expert experience shared. We reached consensus on diagnostic strategies of neonates with myeloproliferation. The differential diagnosis of myeloproliferation in infants includes, apart from (congenital) infections and other stressors, JMML, AML, TAM and other types of IMD.4,6 More frequent benign underlying conditions should be seriously considered before diagnosing a neonate with leukemia and beginning intensive treatment (Figure 1). A medical history and physical examination are important to reveal initial clues regarding infectious causes, other factors inducing stress-hematopoiesis and genetic predisposition (presence of dysmorphic and congenital abnormalities). A physical examination will also reveal hepatosplenomegaly, fluid accumulation and/or skin infiltration. A total blood count and morphological assessment of the peripheral blood smear carried out by an experienced hematologist or morphologist in an expert laboratory are mandatory, and peripheral blood immunophenotyping is, as a minimum measure, advised.4 If a malignant condition is conceivable, the most important challenge is to discriminate a rare transient case, where a ‘watch and wait’ strategy may be justified, from an aggressive leukemia subtype that may require intensive treatment in a limited time span. First, a distinction between megakaryocytic and non-megakaryocytic leukemia is important, based on the morphology and immunophenotyping of the peripheral blood blasts. Megakaryocytic hyperproliferation (French-AmericanBritish - FAB - classification M7) can be recognized by moderately basophilic agranular cytoplasm with blebs on morphology, combined with expression of CD41, CD42 and/or CD61 on flow cytometry.5 In case of megakaryocytic hyperproliferation, germline T21 and GATA1 mutations may point towards TAM. TAM blasts can also present without megakaryocytic markers, FAB M0 (undifferentiated).7 In TAM, early onset and hepatosplenomegaly with monoclonal megakaryocytic hyperproliferation with T21 and a GATA1 mutation can be confirmed.8 The origin of TAM lies in the fetal liver which is why, in most cases, peripheral blood sampling is sufficient for a diagnosis and a bone marrow puncture is unnecessary.1 Without life-threatening disease, a ‘watch and wait’ policy with close monitoring, including regular physical examination and blood counts, is justified.8 Low-dose cytarabine treatment is advised in case of multiorgan failure, high WBC >100 x 109/l, hepatopathy (high bilirubin/transaminases, ascites), severe hepatosplenomegaly, hydrops fetalis, pleural or pericardial effusions, renal failure, or disseminated intravascular coagulation.8 This treatment does not prevent the development of ML-DS (myeloid leukemia related to Down syndrome), but substantially reduces mortality in symptomatic patients.9 After remission, follow-up is advised every three months until the age of four years, because of a 20% chance of ML-DS development during that life span.8 MLDS requires more intensive treatment, however this treat759

Letters to the Editor

Figure 1. Consensus on diagnostics in neonates with myeloblasts based on available literature and newly added cases 1In case of doubt always refer to a clinical geneticist. 2If these are not identified, deep sequencing techniques (SNP-array, RNA-seq, WGS) should be considered. Sporadic identified aberrations are listed in the text. 3Can be both transient and aggressive leukemia. 4Only if clinical presentation allows, with close monitoring of clinical symptoms and regular blood counts. 5In case of doubt, consider consulting international study groups (International Berlin-Frankfurt-Münster AML Study Group, European Working Groups of Myelodysplastic syndromes). References on individual IMD-related aberrations can be found in Table 2. 2-4, 10-12 AML: acute myeloid leukemia; BM: bone marrow; CT: chemotherapy; FISH: fluorescence in situ hybridization; HbF: fetal hemoglobin; HSCT: hematopoietic stem cell transplantation; IMD: infantile myeloproliferative disease (unrelated to Down syndrome); JMML: juvenile myelomonocytic leukemia; NS: Noonan syndrome; PB: peripheral blood; SNP: single nucleotide polymorphism; T21: trisomy 21; TAM: transient abnormal myelopoiesis related to Down syndrome; WGS: whole genome sequencing

ment has high success rates.8 In megakaryoblastic cases without germline (mosaic) T21 and a GATA1 mutation, a bone marrow puncture can be considered. Furthermore, additional mutational analyses for recurrent germline or somatic IMD-related aberrations (such as somatic T21), as well as analyses for recurrent infant AML translocations, are advised (Figure 1; discussed below). In neonatal non-M7/M0 hyperproliferation, first, discrimination between JMML and AML, and in rare cases, a non-M7 IMD, is important. Bone marrow investigation can be considered for immunophenotyping, karyotyping, fluorescence in situ hybridization (FISH) and targeted mutational analyses. Collection of germline material for sequencing discrimination purposes is advisable. In monocytic proliferation, JMML diagnostics are advised and morphology of the peripheral blood smear, which shows (meta)myelocytes and nucleated red cells combined with the clinical phenotype, is of utmost importance.3 It is important to identify dysmorphic features of RAS pathway related syndromes.3 Other JMML characteristics are splenomegaly, an elevated fetal hemoglobin value and a normal or moderately increased bone marrow blast 760

count.3 JMML is in 90% of the cases characterized by mutations in PTPN11, NRAS, KRAS, NF1 or CBL.3 Germline CBL, KRAS, NRAS, PTPN11 or RIT1 mutations indicate an RAS pathway driven JMML, in which spontaneous remission often occurs and a ‘watch and wait’ policy may be considered if clinically feasible.3 In contrast, patients with a somatic RAS driver mutation commonly have aggressive disease requiring allogeneic hematopoietic stem cell transplantation in most cases.3 When the clinical picture of a non-megakaryoblastic leukemia is not consistent with JMML, IMD and AML may be seriously considered. Such cases mainly consist of monoblastic AML (FAB M5; immunophenotype CD4+CD11b+CD64+), characteristically present with leukemia cutis, hepatosplenomegaly, hyperleukocytosis and KMT2A fusions, and require AML-directed chemotherapy.2,5,6,10 A diagnostic bone marrow puncture is advised for molecular blast cell characterization. Recurrent translocations, characteristic for infant AML, are t(1;22)(p13.3;q13.1)/RBM15-MKL1, 11q23.3/KMT2A translocation and t(8;16) (p11.2; p13.3)/KAT6A-CREBBP. Further, t(8;21)(q22;q22)/RUNX1-RUNX1T1, t(8;1)(p11;q22), t(5;6)(q31;q21), t(6;17)(q23;q11.2) and haematologica | 2022; 107(3)

Letters to the Editor

t(X;6)(p11.1;q23) have been identified.2,10-12 Most of these karyotypes are associated with aggressive AML, requiring intensive treatment.13-15 Interestingly, in rare myeloid leukemia cases, a ‘watch and wait’ policy can be considered, as illustrated by reports of incidental cases with successful ‘watch and wait’ strategies (Tables 1,2). These cases include monoclonal infant AML M4/M5-cases with somatic t(8;16); however, t(8;16) can also be present in full-blown AML.10 IMD associated with germline THPO mutations should be seriously considered in families with a positive history of essential throm-

bocytosis and myeloproliferative disease in the elderly (Table 2). Furthermore, we found increasing evidence on somatic T21, GATA1 mutations and del(8)(q23.2q24) in IMD (Table 1,2). SNP array analysis can aid in the identification of subclonal T21 with small clone sizes. Finally, some aberrations have only been described once, nevertheless, they might become recurrent, such as a del(5q), SETD2 or germline NSD1 mutation (Tables 1,2). In conclusion, this review and consensus-based diagnostic guideline may aid in clinical decision-making for the rare infant cases with myeloid hyperproliferation (Figure 1),

Table 1. IMD-cases without germline (mosaic) trisomy 21 from international database*

UPN

Study group

Age

Sex

Clinical FAB presentation1

Genetic tests

Germline

Somatic

Treatment

CR/event

Slovakia

Newborn

HSM

FISH, PCR

Normal

T212

N/A

Japan

1.5 months

HSM

Karyotype

Normal

T212

Japan

1 month

HM, CL, VSD Alagille syndrome)

N/A

Karyotype, FISH

Normal

T212

Czech

Newborn

None1

Normal

T21, GATA1

Sweden

6 days

None1

N/A

Normal

T21, GATA1

Austria

5 days

None1

Karyotype, FISH Karyotype, FISH, PCR Karyotype, FISH, PCR

Normal

T21, GATA1

Slovakia

Newborn

HM, CL

N/A

FISH, PCR

Normal

T21, GATA1

Slovakia

1 month

FISH, PCR

Normal

T21, GATA1

Slovakia

N/A3

HM, CL

N/A

Not tested at time of IMD3

Normal

Spain

Newborn

Few petechiae

Karyotype, FISH, CGH, NGS (117 genes)

Normal

SETD2, trisomy 8

Germany

6 weeks

None

N/A

Normal

T21, GATA1

Germany

Newborn

None

N/A

Karyotype, PCR Karyotype, PCR

13 Germany

Newborn

ASD II

N/A

Germany

3 weeks

N/A

HSM, VSD

N/A

Germany

Newborn

None

N/A

Karyotype, FISH, PCR Karyotype, FISH, PCR Karyotype, FISH, PCR

N/A

Normal

Alive (6.5 years) None CR Alive (5 years) Low-dose AML (at 5.5 months); Died AraC received AML-DS (at 18 months) treatment; progressive disease; respiratory failure None CR Alive (9 years) None CR Alive (3 years) None CR (1 month); AML M7 (at 15 months), same aberrations None CR Alive (9.5 years) None CR Alive (11 years) None AML (at 3 years) Alive with somatic T21 (6.5 years) and GATA1-mutation AML BFM 2004 protocol; CR at day 15; ASCT. None CR, developed Alive AML (at 4 months), (3 years) CR after first induction None CR Alive (3 years) None CR Alive

T21, (mosaic BM) GATA1 Normal Mosaic T21, Prednison4 (fibroblasts) GATA1 Normal T21, None GATA1 46,XX,idic(21) (p11)c [15]

T21, GATA1

Vital status (FU time)

None

Alive (1 year) CR (8 weeks), Died developed AML (2 days after (at 10 months) AML diagnosis) CR Alive (1 year)

*Inclusion criteria: historical non-TAM, non-JMML cases, cured with no/only symptomatic treatment, age <1 year at diagnosis, diagnosed from 1990-2020. Exclusion criteria: transient abnormal myelopoiesis (TAM) according to WHO definition. 1Questioned for hepatosplenomegaly (HSM), intravascular coagulation, cutaneous lesions (CL), central nervous system (CNS)-involvement or other extramedullary disease. 2GATA1 not tested in every case. 3IMD diagnosis not definite, was made in retrospect, based on blood counts.4Initial diagnosis acute lymphoblastic leukemia (ALL). AML: acute myeloid leukemia; araC = cytarabine; ASCT: allogenic stem cell transplantation; ASD: atrial septum defect; BM: bone marrow; CGH: comparative genomic hybridization; CR: complete remission; DS: Down syndrome F: female; FAB: French-American-British classification; FISH: fluorescence in situ hybridization; FU: follow-up; HM: hepatomegaly; IMD: infantile myeloproliferative disease (unrelated to Down syndrome); M: male; N/A: data not available; NGS: next generation sequencing; PCR: polymerase chain reaction; T21: trisomy 21; UPN: unique patient number; VSD: ventricular septum defect; WHO: World Health Organization.

haematologica | 2022; 107(3)

761

762

Age

Newborn

4 weeks

6 weeks

2 months Newborn

12 days

6 days

Newborn

Newborn Newborn

Newborn

5 days

Neonate

Newborn

UPNref

16a

17b

18c

19d 20e

21f

22g

23h

24i 25j

26k

27l

28m

29-36n

37o

F (n=4) M (n=3) F

M M

M F

Sex presentation1

HSM

N/A

HSM

None

None HSM

None

HSM

None

HSM None

HSM

None

Clinical

Myeloid

N/A

M0/M7

Myeloid

Myeloid, megakaryocytic Myeloid M7

Myeloid

Myelo-monocytic Myelo-monocytic

N/A

Immature

Not specific

FAB/IF

FISH, GATA1-analyses

Karyotype, FISH (also skin fibroblasts) Karyotype, FISH (also in CR) Karyotype, PCR, RT-PCR, FISH (also oral mucosa and skin fibroblasts) Karyotype, GATA1-analysis (also in CR) FISH (also buccal mucosal cells, urine epithelial cells and hair follicles) Karyotype (also skin fibroblasts), FISH, PCR Karyotype (also in CR), FISH (PB in CR, skin or buccal)

Karyotype, FISH, GATA1-analysis, SNP-array, WES (also in CR) Karyotype#, FISH (also in CR)

Karyotype, RT-PCR, FISH DNA sequencing

Karyotype (PB lymphocytes), GATA1 screening Karyotype (also fibroblasts), monoblasts N/A

Genetic tests

Table 2. Previously reported IMD-cases without germline (mosaic) trisomy 21 from literature

Normal

Normal Normal

THPO mutation unclear mutational analyses Familial thrombocytosis (THPO) NS (clinical diagnosis) PTPN11 mutation (NS) (also hair follicles) NSD1 mutation (Sotos syndrome)5 Chr.6 duplication within q25.3-q26 Yqs4

Mosaic trisomy 122

Germline

T21, GATA1 mutation

T21; At 3 months: del(13)(q13q31) T21, GATA1 mutation T21, GATA1 mutation

Del(8) (q23.2q24) & del(5)(q31.1q31.3) T213

None None

None

GATA1

Somatic

Low-dose AraC

None (n=3), Low-dose AraC (n=4)

None

None Low-dose AraC None

None

None None

Low-dose AraC None

None

Treatment

Vital status (FU time)

Alive (3 years)

N/A

Alive (7 months)

Table 2. Continued on following page.

CR; progression to AML (n=2)

CR; AML ,(7 months) Alive (6 years) CT; ML-DS protocol CR Alive (2 years)

CR; leukemia N/A (20 months), CT CR Alive (2.5 years) CR N/A

N/A

Alive

Alive (3.8 years) N/A

Alive (5 years)

Alive (3 years)

Alive (3 months)

CR; AML (11 months), CT CR

CR CR

CR/event

Letters to the Editor

haematologica | 2022; 107(3)

Newborn

haematologica | 2022; 107(3)

Newborn

1-30 days

Newborn

43r

44-51s

52t

53u

54v

55w

N/A

HSM, CL

Myeloid, AML M2

Megakaryocytic

Myeloid

FAB/IF

Megakaryocytic F

Megakaryocytic

M4 (myeloid sarcoma)

M4/M5

Immature myeloid

Blasts in cerebral spinal fluid

None

CL (n=6)

HSM

None

Clinical

ISH, WES, whole transcriptome sequencing

Karyotype, FISH (also in CR), molecular testing Karyotype, FISH, chromosomal microarray, GATA1-analysis on UCB Karyotype, FISH, BAC-array, SNP-array

Karyotype (repeated in CR) Karyotype, FISH or RT-PCR

Karyotype (also in skin fibroblast)

FISH, karyotype, NGS (35 myeloid genes) Karyotype (also in CR), PCR GATA1

Genetic tests

Normal

13q12.11 deletion (300 kb; 3 genes: GJB6, MIR4499, CRYL1)

None found

Normal

Germline

Del(3)(q21.2q23), del(7)(q22.1q31.1), del(7)(q31.1q31.2), del(7)(q36.1) & del(8)(q23.2q24) GATA1, JAK1, SPIRE2 & FN1 mutation

None found

Cryptic t(8;16)(p11.2; p13.3) insertional translocation

t(8;16)(p11.2;p13.3)

T21

T21, +223

T21, GATA1 mutation T21, GATA1 mutation

Somatic

None

Treatment

N/A

Alive (3 years)

Alive (2 years)

CR; recurrence Alive (n=6; <48 months (n=4) variable FU) CT, SCT (n=1) Died (n=1) CR Alive (23 months)

leukemia (17 months) Alive CT CR (16 months) CR; AML (7 months,) Alive t(1;10), (5 years) +16, +21, +22, CT CR Alive (5 years)

N/A

Alive (1.5 years)

CR; MDS (14 months),

CR/event Vital status (FU time)

1 Checked for hepatosplenomegaly (HSM), intravascular coagulation, cutaneous lesions (CL), CNS - central nervous system involvement or other extramedullary disease. 2Uncertain whether this was germline mosaic. 3GATA1 not tested. 4Satellited Y chromosome. 5This case was previously described, at that time Sotos diagnosis was not known yet (WES was performed after). AML: acute myeloid leukemia; araC = cytarabine; BAC: bacterial artificial chromosome; CR: complete remission; CT: chemotherapy; F: female; FAB: French-American-British classification; FISH: fluorescence in situ hybridization; FU: follow-up; HM: hepatomegaly; IF: immunophenotype markers; IMD: infantile myeloproliferative disease (unrelated to Down syndrome); M: male; MDS: myelodysplastic syndrome; ML-DS: myeloid leukemia related to Down syndrome; N/A: data not available; NGS: next generation sequencing; NS: Noonan syndrome; PB: peripheral blood; PCR: polymerase chain reaction; RT-PCR: reverse transcription PCR; SCT: stem cell transplantation; SNP: single nucleotide polymorphism; T21: trisomy 21; UCB: umbilical cord blood; UPN: unique patient number; WES: whole exome sequencing. a Basu B et al. Pediatr. Hematol. Oncol. 2010; bHouwing ME et al. Int. J. Hematol. 2015; cVan Dijken et al. Acta Paediatr. 1996; dSilvio F et al. J Pediatr Hematol Oncol. 2002; eMalone A et al. Br. J. Haematol. 2017; fBertrums EJM et al. Pediatr. Blood Cancer. 2017; g Richards M et al. Arch. Dis. Child. Fetal Neonatal Ed. 1998; hPolski JM et al. J Pediatr Hematol Oncol. 2002; iRozen L et al. Eur. J. Pediatr. 2014; jOhkawa T et al. Pediatr Int. 2015; kOno R et al. Eur. J. Pediatr. 2015; lCarruthers V et al. J. Paediatr. Child Health. 2017; m Salvatori G et al. Oncol. Lett. 2017; nYuzawa K et al. Pediatr. Blood Cancer. 2020; oDosedla E et al. Actual Gyn. 2019; pRoseman AS et al. Cancer Genet. 2020; qTsai MH et al. Indian J Pediatr. 2011; rApollonsky N et al. J Pediatr Hematol Oncol. 2008; sCoenen EA et al. Blood 2013; tBarrett R et al. Pediatr. Blood Cancer. 2017; uNakashima et al. Pediatr. Blood Cancer. 2015; vSchifferli A et al. Eur. J. Haematol. 2015; wLukes J et al. Leukemia. 2020.

Newborn

42r

40-41q

Newborn (twins) Newborn

38-39p

Sex presentation1

Age

UPNref

Letters to the Editor

763

Letters to the Editor

especially if a ‘watch and wait’ policy is considered and clinically feasible. Despite our extensive research, we were only able to include a limited number of patients; this underlines the rarity of the disease and makes general conclusions challenging. To identify these individual cases, an extensive and ongoing (international) collaboration of pediatric oncologists, cytogeneticists, immunologists, molecular biologists and clinical geneticists is mandatory for clinical decision-making and the development of diagnostics tools and treatment. Genomic sequencing can identify novel aberrations that could be recurrent. We here present a consensus for the preferred diagnostic logistics, based on a broad international consortium with clinicians and investigators from the I-BFM AML SG and EWOG MDS. This consensus may support decision-making in these rare infants presenting with myeloproliferative disease. Eline J.M. Bertrums,1,2,3 C. Michel Zwaan,1,2 Daisuke Hasegawa,4 Valerie de Haas,1,5 Dirk N. Reinhardt,6 Franco Locatelli,7 Barbara de Moerloose,8 Michael Dworzak,9 Arjan Buijs,10 Petr Smisek,11 Alexandra Kolenova,12 Cornelis Jan Pronk,13 Jan-Henning Klusmann,14 Ana Carboné,15 Alina Ferster,16 Evangelia Antoniou,6 Soheil Meshinchi,17 Susana C. Raimondi,18 Charlotte M. Niemeyer,19 Henrik Hasle,20 Marry M. van den Heuvel-Eibrink1,21 and Bianca F. Goemans1,5 1 Princess Máxima Center for Pediatric Oncology, Utrecht, the Netherlands; 2Department of Pediatric Oncology, Erasmus MC, Rotterdam, the Netherlands; 3Oncode Institute, Utrecht, the Netherlands; 4Department of Pediatrics, St. Luke’s International Hospital, Tokyo, Japan; 5Dutch Childhood Oncology Group (DCOG), Utrecht, the Netherlands; 6Department of Pediatric Oncology, University of Duisburg-Essen, Essen, Germany; 7Sapienza, University of Rome Department of Pediatric Hematology-Oncology, IRCCS Bambino Gesù Children’s Hospital, Rome, Italy; 8Department of Pediatric Hematology-Oncology and Stem Cell Transplantation, Ghent University Hospital, Ghent, Belgium; 9Children's Cancer Research Institute and St. Anna Kinderspital, Department of Pediatrics, Medical University of Vienna, Vienna, Austria; 10 Department of Genetics, University Medical Center, Utrecht, the Netherlands; 11Department of Pediatric Hematology and Oncology, 2nd Faculty of Medicine, Charles University and University Hospital Motol, Prague, Czech Republic; 12Department of Pediatric Hematology and Oncology, Comenius University Children's Hospital, Bratislava, Slovakia; 13Childhood Cancer Center, Skåne University Hospital, Lund, Sweden; 14Pediatric Hematology, Oncology and Hemostaseology, Hospital for Children and Adolescents, University Hospital of Frankfurt/Main, Goethe-University Frankfurt/Main, Frankfurt, Germany; 15Department of Pediatric Onco-Hematology, University Hospital Miguel Servet, Zaragoza, Spain; 16Department of Hemato-Oncology, Immunology and Transplantation, University Pediatric Hospital Reine Fabiola (ULB), Brussels, Belgium; 17Fred Hutchinson Cancer Research Center, Clinical Research Division, Seattle, WA, USA; 18Department of Pathology, St. Jude Children’s Hospital, Memphis, TN, United States; 19Division of Pediatric Hematology and Oncology, Department of Pediatrics and Adolescent Medicine, Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany; 20Department of Pediatrics, Aarhus University Hospital, Aarhus, Denmark and 21University of Utrecht, Utrecht, the Netherlands. ^Members of I-BFM AML SG and EWOG-MDS are stated in the appendix. Correspondence: ELINE J.M. BERTRUMS - e.j.m.bertrums@prinsesmaximacentrum.nl

764

doi:10.3324/haematol.2021.279507 Received: June 24, 2021. Accepted: November 25, 2021. Pre-published: December 2, 2021. Disclosures: no conflicts of interest to disclose. Contributions: EJMB, BFG and MMHE designed the study and wrote the manuscript; EJMB performed literature review and included the data. BFG and MMHE supervised the study. AK, DJMH, CJP, MD, AC, EA and DNR included patients. AB and SCR reviewed the cytogenetic analyses. MMHE, CJP, DNR, BM, HH, SM, EJMB, CMN, JHK and DH participated in consensus meetings. All authors reviewed the manuscript and provided feedback. ^Appendix: International Berlin-Frankfurt-Münster AML Study Group (I-BFM AML SG) members are: C.M. Zwaan, D. Hasegawa, D.N. Reinhardt, F. Locatelli, B. de Moerloose, M., Dworzak, P. Smisek, A. Kolenova, C.J. Pronk, J.H. Klusmann, A. Carboné, E. Antoniou, H. Hasle, M.M. van den Heuvel-Eibrink and B.F. Goemans. European Working Group of MDS in childhood (EWOG-MDS) members are: F. Locatelli, B. de Moerloose, M. Dworzak, C.M. Niemeyer, H. Hasle and M.M. van den HeuvelEibrink

References 1. Roberts I, Izraeli S. Haematopoietic development and leukaemia in Down syndrome. Br J Haematol. 2014;167(5):587-599. 2. Roberts I, Fordham NJ, Rao A, Bain BJ. Neonatal leukaemia. Br J Haematol. 2018;182(2):170-184. 3. Niemeyer CM, Flotho C. Juvenile myelomonocytic leukemia: who's the driver at the wheel? Blood. 2019;133(10):1060-1070. 4. van der Linden MH, Creemers S, Pieters R. Diagnosis and management of neonatal leukaemia. Semin Fetal Neonatal Med. 2012;17(4):192-195. 5. Bain BJ, Bene MC. Morphological and immunophenotypic clues to the WHO categories of acute myeloid leukaemia. Acta Haematol. 2019;141(4):232-244. 6. Bresters D, Reus AC, Veerman AJ, et al. Congenital leukaemia: the Dutch experience and review of the literature. Br J Haematol. 2002;117(3):513-524. 7. Boztug H, Schumich A, Pötschger U, et al. Blast cell deficiency of CD11a as a marker of acute megakaryoblastic leukemia and transient myeloproliferative disease in children with and without Down syndrome. Cytometry B Clin Cytom. 2013;84(6):370-378. 8. Tunstall O, Bhatnagar N, James B, et al. Guidelines for the investigation and management of transient leukaemia of Down syndrome. Br J Haematol. 2018;182(2):200-211. 9. Flasinski M, Scheibke K, Zimmermann M, et al. Low-dose cytarabine to prevent myeloid leukemia in children with Down syndrome: TMD Prevention 2007 study. Blood Adv. 2018;2(13):1532-1540. 10. Coenen EA, Zwaan CM, Reinhardt D, et al. Pediatric acute myeloid leukemia with t(8;16)(p11;p13), a distinct clinical and biological entity: a collaborative study by the International-Berlin-FrankfurtMunster AML-study group. Blood. 2013;122(15):2704-2713. 11. van Dongen JC, Dalinghaus M, Kroon AA, de Vries AC, van den Heuvel-Eibrink MM. Successful treatment of congenital acute myeloid leukemia (AML-M6) in a premature infant. J Pediatr Hematol Oncol. 2009;31(11):853-854. 12. Dastugue N, Duchayne E, Kuhlein E, et al. Acute basophilic leukaemia and translocation t(X;6)(p11;q23). Br J Haematol. 1997;98(1):170-176. 13. Gruber TA, Downing JR. The biology of pediatric acute megakaryoblastic leukemia. Blood. 2015;126(8):943-949. 14. de Rooij JD, Branstetter C, Ma J, et al. Pediatric non-Down syndrome acute megakaryoblastic leukemia is characterized by distinct genomic subsets with varying outcomes. Nat Genet. 2017;49(3):451-456. 15. Noort S, Zimmermann M, Reinhardt D, et al. Prognostic impact of t(16;21)(p11;q22) and t(16;21)(q24;q22) in pediatric AML: a retrospective study by the I-BFM Study Group. Blood. 2018;132(15):15841592.

haematologica | 2022; 107(3)

Letters to the Editor

Biallelic CXCR2 loss-of-function mutations define a distinct congenital neutropenia entity Neutrophil homeostasis results from a balance between neutrophil production, release from the bone marrow and clearance from the circulation, where chemokines and their receptors play central roles.1,2 Studies on mice demonstrated that CXCR4 and CXCR2 receptors antagonistically regulate bone marrow neutrophil release.2 While CXCR4 and its chemokine CXCL12, which is constitutively expressed in the bone marrow, provide key signals for neutrophil retention, CXCR2 activation by the CXCL8 subfamily of chemokines promotes their release from the bone marrow.1,2 Those events were shown in patients carrying heterozygous CXCR4 gain-of-function mutations causing the rare autosomal dominant WHIM syndrome, characterized by human papillomavirus-induced warts, hypogammaglobulinemia, recurrent bacterial infections

and myelokathexis reflecting an accumulation of senescent neutrophils in the bone marrow.3 Profound neutropenia associated with myelokathexis was previously reported in two siblings carrying a homozygous truncating CXCR2 loss-of-function mutation, supporting the importance of CXCR2 signaling in neutrophil mobilization.4 Myelokathexis and recurrent severe infections5 in that single pedigree led to it being included in the large series of WHIM syndrome and WHIM syndrome-like cases,6 and it remains the only published example of CXCR2 deficiency. Herein, we report biallelic CXCR2 mutations, including one complete gene deletion, in four patients with chronic neutropenia, harboring a wild-type (WT) CXCR4 gene. Patients were diagnosed during childhood with profound neutropenia in the context of recurrent gingivitis and oral ulcerations (Table 1). Bone marrow smears showed no major granulocytic maturation defect. Myelokathexis was present in only patient 1 (P1) and affected 35% of myeloid cells. Values of the other hema-

Table 1. Clinical profile of the four patients with biallelic CXCR2 loss-of-function mutations.

Characteristic P1 (8364) Clinical profile Age at diagnosis (years) Oral lesions Severe infectionsa (age) Prophylactic treatment G-CSF therapy (dose, period) Age at last follow-up (years) Hematologic values at diagnosis Neutrophils (x109/L) Monocytes (x109/L) Lymphocytes (x109/L) Hemoglobin level (g/dL) Platelets (x109/L) Hematologic values during follow-up Blood counts (n) Neutrophils (x109/L) Monocytes (x109/L) Lymphocytes (x109/L) Hemoglobin (g/dL) Platelets (x109/L) Differential bone-marrow count Myeloblasts Promyelocytes & myelocytes Metamyelocytes & mature neutrophils Myelokathexis Immunoglobulin levels (g/L) IgG IgM IgA Lymphocyte subsets Subset determinations (n) CD3+CD4+ T cells (x109/L) CD3+CD8+ T cells (x109/L) CD19+ B cells (x109/L) CD3–CD16+CD56+ NK cells (x109/L)

2.9 Yes No

Patient (Registry ID) P2 P3 (6487) (6902)

Reference range

1.2 Yes No

22.5

1.8 Yes 1 pneumonitis (22 months) Cotrimoxazole Yes (5 mg/kg, 1 month) 10

0.56 0.31 2.0 12.4 359

0.6 0.46 6.3 12.2 371

0.18 0.52 6.2 11.8 368

0.6 0.46 4.6 10 225

1.5-8.0 0.1-1.0 1.5-6.5 11.5-15.5 150-400

18 0.57 (0.28-1.8) 0.29 (0.02-0.51) 1.66 (1-2.7) 12.6 (11.5 -13.3) 357 (257-439)

24 0.45 (0.1-0.85) 0.43 (0.16-0.68) 2.03 (1-6.3) 11.5 (9.9-13.2) 277 (215-572)

23 0.54 (0.1-10.8b) 0.53 (0.19-1.3) 4.0 (0.97-10.8) 12.2 (10.4-13.1) 324 (167-544)

7 0.35 (0.3-1) 0.52 (0.39-0.7) 4.58 (1.9-4.8) 10.7 (9.6-12.7) 523 (225-670)

1.5-8.0 0.1-1.0 1.5-6.5 11.5-15.5 150-400

4% 15% 43% 35%

0% 5% 29% 0%

1% 13% 38% 0%

2% 2% 37% 0%

0.3-4% 12-25% 33-48%

15.1 (14.2-20.5) 0.91 (0.81-1.12) 2.98 (1.73-3.06)

16.7 (14.3-17.3) 1.77 (1.51-2.0) 3.75 (3.14-3.96)

10.9 (10.2-11.8) 2.41 (2.12-2.53) 1.75 (1.69-2.56)

14.8 1.63 3.35

5.98-11.1 0.56-1.59 0.49-1.53

5 0.544 (0.333-0.700) 0.288 (0.170-0.429) 0.157 (0.121-0.177) 0.088 (0.078-0.129)

10 0.677 (0.611-0.853) 0.501 (0.432-0.629) 0.295 (0.240-0.365) 0.171 (0.134-0.209

1c 1.456 1.295 0.769 0.778

0.53-1.3 0.33-0.92 0.11-0.57 0.07-0.48

No Yes (2 mg/kg, 2 years) 36

1.9 Yes 1 cellulitis (13 months) Cotrimoxazole No

P4 (8497)

3 0.920 (0.766-1.075) 0.611 (0.498-0.725) 0.468 (0.344-0.591) 0.153 (0.134-0.172)

No No 6.5

Results are expressed as medians (range), unless stated otherwise. Bold type indicates lower values and bold italics indicate higher values compared to the reference range. a Intravenous antibiotic-treated infections. bAt the time of a fever of unknown origin (likely viral), the absolute neutrophil count increased spontaneously. cEvaluated at age of 4 years. G-CSF: granulocyte colony-stimulating factor; NK natural killer.

haematologica | 2022; 107(3)

765

Letters to the Editor

tologic lineages, including lymphocyte subsets, were within their normal ranges. These four patients had high levels of circulating IgG and/or IgA at diagnosis which persisted throughout follow-up. We investigated a possible genetic etiology using targeted sequencing of genes known to be involved in inherited neutropenia and exome-sequencing. We excluded CXCR4

mutations and identified a homozygous CXCR2-gene deletion in P1, homozygous CXCR2 missense mutations in P2 and P3, and compound heterozygous CXCR2 mutations in P4 (Figure 1A, Online Supplementary Figure S1A, B). The CXCR2 deletion was further confirmed by single nucleotide polymorphism-array analysis (data not shown) that revealed a homozygous 13.4-kb deletion in 2q35

Figure 1. Characterization of germline biallelic CXCR2 mutations identified in four patients with chronic neutropenia. (A) Family pedigrees with identified homozygous (patients P1, P2, and P3) or compound heterozygous (P4) CXCR2 mutations. Healthy parents were heterozygous carriers for the identified mutations. (B) Cell-surface CXCR2 immunostaining on neutrophils from P1, P2, and P3, one heterozygous carrier, and healthy donors. (C) Dose-dependent CXCL8induced chemotaxis of neutrophils without or with SB265610 (SB), its specific CXCR2 inhibitor. Chemotaxis assays were run in duplicate, with whole blood samples (diluted 1:4 in RPMI with 1% human serum) using 12 mm diameter transwell devises with 5 mm pores. For each assay including patient, parent and control, blood samples were collected concomitantly and treated equally. Samples were added in the upper chamber, CXCL8 in the lower chamber and SB in both chambers. Control wells without chambers were also added to determine the number and phenotype of total seeded cells. After incubation for 1 hour, cells recovered in the lower chambers (responding cells) were counted and identified by flow cytometry. Results are expressed as percentage of responding neutrophils, calculated as [(Number of neutrophils recovered in the lower chamber with CXCL8) - (Number of neutrophils recovered in the lower chamber without CXCL8)] / (Number of total seeded neutrophils) x 100. WT: wild-type; na: not available.

766

haematologica | 2022; 107(3)

Letters to the Editor

Table 2. Comparison of the clinical characteristics of 14 patients with CXCR4 gain-of-function mutations and four patients with CXCR2 lossof-function mutations enrolled in the French Severe Congenital Neutropenia Registry.

Characteristic Number of patients Age at diagnosis (years) Age at last visit (years) Oral lesions Severe infections Warts Hematologic values (all available CBC) Neutrophils (x109/L) Monocytes (x109/L) Lymphocytes (x109/L) Hemoglobin (g/dL) Platelets (x109/L) Myelokathexis Bone-marrow myeloid/erythroid ratio G-CSF treatment Immunoglobulin levels (g/L) IgG IgM IgA Lymphocyte subsets CD3+CD4+ T cells (x109/L) CD3+CD8+ T cells (x109/L) CD19+ B cells (x109/L) CD3–CD16+CD56+ NK cells (x109/L) Solid tumors Tetralogy of Fallot Deaths

Patients with CXCR4 mutations

Patients with CXCR2 mutations

Relevant differences

14 4.9 (0.1-33) 31.9 (8.9-77) 3/14 10/14 8/14

4 1.8 (1.2-2.9) 16.3 (7.2-36.5) 4/4 2/4 0/4

0.221 (0.13-1.4) 0.156 (0.06-0.44) 0.577 (0.16-1.9) 12.1 (8.6-13.8) 220 (169-479) 14/14 3.5 3/14, poor responses

0.496 (0.18-0.57) 0.477 (0.29-0.54) 3.2 (1.6-4.5) 11.8 (10.7-12.6) 338 (277-523) 1/4 3 2/4, good responses

8 (4.2-15) 0.64 (0.24-1.7) 0.86 (0.1-2.5)

16.2 (10.2-20.5) 1.63 (0.81-2.53) 3.33 (1.69-3.96)

** ** **

0.37 (0.17-0.51) 0.09 (0.04-0.10) 0.02 (0.01-0.05) 0.12 (0.06-0.16) 8/14 5/14 3a

0.80 (0.54-1.46) 0.50 (0.29-1.29) 0.38 (0.12-0.77) 0.15 (0.08-0.78) 0/4 0/4 0/4

** ** ** **

** ** * *

Results are expressed as medians (range), unless stated otherwise. Bold type indicates lower values and bold italics indicate higher values compared to the reference range. a Two deaths occurred between 30 and 40 years of age from vulvar cancer or atypical mycobacteria with liver failure, and one 77-year-old died of pneumonitis. Owing to the very low number of patients to be compared, the most relevant differences are indicated as *P<0.01 or **P<0.001. CBC: complete blood count; G-CSF: granulocyte colony-stimulating factor; NK: natural killer.

(218,988,774_219,002,220) encompassing only CXCR2. To exclude other causal variants in P2, P3 and P4, who harbor missense CXCR2 mutations, DNA from the probands and their parents were subjected to wholeexome sequencing. The mean depth of exome coverage was 74X with 96% covered at least 20X. The CXCR2 mutations were confirmed and no other potentially causative candidate variants were identified. The homozygous CXCR2 genotypes of P1, P2, and P3 were consistent with the reported consanguinity of these pedigrees. Parents were heterozygous carriers and their bloodcell counts were within normal ranges. The three CXCR2 missense mutations (p.Arg144Cys, p.Arg212Trp and p.Arg289Cys) had been entered into the Genome Aggregation Database (gnomAD) with an allele frequency <5x10–5 but never as being homozygous. The mutation in P2 affects Arg144 which constitutes the critical DRY motif for G-protein activation.7 The mutations affect Arg184 in P3, which is highly conserved between CXCR2 and CXCR1, and Arg212 and Arg289 in P4, which belong to domains cooperating with the CXCR2 N-terminal for the efficient docking of the CXCL8-chemokine ligand (Online Supplementary Figure S1C).8 We then examined cell-surface CXCR2 expression in neutrophils (Figure 1B), monocytes (Online Supplementary Figure S2A) and natural killer cells (data not shown) from P1, P2 and P3, their parents, and healthy control blood donors. As expected, CXCR2 was not expressed in the haematologica | 2022; 107(3)

different cell populations derived from patient P1, who has a homozygous CXCR2-gene deletion. Her mother, who carries a heterozygous CXCR2 deletion, had intermediate CXCR2 expression between P1 and control values. That mutant-dosage effect was also observed in carriers of CXCR2 missense mutations, e.g., all pedigree-P2 blood cell populations noted above and pedigree-P3 monocytes (Figure 1B, Online Supplementary Figure S2A). Whether the underlying mechanisms implicate altered turnover of the Arg144Cys mutant and, in a more cellrestricted fashion, of the Arg212Trp mutant, remains to be investigated. As expected based on the patients’ WT CXCR4 genotypes, cell-surface CXCR4 expression was within the normal range for all tested blood cell populations as illustrated for P1 and P3 (Online Supplementary Figure S2B). We evaluated the potential impact of CXCR2 mutations on the CXCL8-driven chemotactic response of blood neutrophils derived from P1 and P3 pedigrees (Figure 1C). In transwell migration assays, healthy donors’ neutrophils responded to CXCL8, yielding a typical bell-shaped, dose-dependent, chemotaxis-response curve. Blockade with the specific CXCR2 inhibitor SB265610 confirmed the involvement of CXCR2 in the observed chemotaxis. Neutrophils from parents migrated similarly to controls despite lower cell-surface CXCR2 expression, supporting the reported dissociation between the expression level of chemokines-receptors and their 767

Letters to the Editor

functions.9 In contrast, efficacy of the CXCL8-induced chemotaxis for P1-derived neutrophils was drastically reduced (up to 86%) for all tested CXCL8 concentrations. For P3-derived neutrophils, this response was more weakly lowered (up to 59%) indicating that the Arg212Trp CXCR2 mutation only partially abrogates CXCR2 function. This was further confirmed by the SB265610-mediated inhibition of the remaining Arg212Trp CXCR2-driven chemotaxis (Figure 1C). P3derived neutrophils expressed similar levels of CXCR2 than control neutrophils (Figure 1B) and their remaining chemotactic responses toward CXCL8 were out of the range of the ones provided by control neutrophils (Online Supplementary Figure S3A), further supporting the CXCR2 loss-of-function phenotype. We extrapolated that this loss-of-function phenotype would be similarly conferred by P2’s and P4’s CXCR2 missense mutations, affecting the protein’s critical DRY domain7 or N-terminal domain,8 respectively. CXCR1 could account for the remaining migration of P1’s neutrophils, which were not affected by the inhibitor SB265610.10 Indeed, although CXCR1 and CXCR2 have closely linked actions, they differ notably in their signaling properties and chemokineligand spectra, with CXCR1 being engaged by CXCL5 and CXCL6 and having high affinity for CXCL8, while CXCR2 promiscuously binds to all seven CXCL8-family chemokines.11 CXCR1 expression levels on P1 and control neutrophils were within the same range (Online Supplementary Figure S3B), thereby substantiating that hypothesis. The patients described herein did not experience severe recurrent bacterial infections, suggesting that although CXCR2 actively participated in neutrophil recruitment into inflammatory tissues, this function was largely counterbalanced. Indeed, patients’ neutrophils remained responsive to N-formylmethionine-leucylphenylalanine (fMLP) (Online Supplementary Figure S3C), indicating that they might be efficiently guided to inflammatory sites by chemoattractant signals, such as fMLP and possibly others including the C5a complement factor, both abundantly generated in foci of bacterial infection.12 Likewise, CXCL12-driven migration was equivalent for CD3+CD4+ cells (Online Supplementary Figure S3D) and the other lymphocyte subpopulations (data not shown) from P1 and P3, their parents and controls. These findings support the postulate of normal CXCR4 function in patients harboring CXCR2 mutations acting as drivers of congenital neutropenias although it remains to be experimentally demonstrated. Different clinical manifestations distinguish these four patients with CXCR2 mutations from the clinical spectrum of the 14 WHIM syndrome cases enrolled in the French Severe Chronic Neutropenia Registry, as summarized in Table 2. Myelokathexis, a pathognomonic feature of WHIM syndrome,6 was solely detected in P1, harboring the CXCR2 gene deletion, thereby extending the description of the two previously reported cases with CXCR2 loss-of-function mutations.5 Its absence in the clinical pictures of P2, P3 and P4, together with the partial CXCR2-chemotaxis response retained by Arg212Trp, further suggests that their chronic neutropenia is not the only consequence of a CXCR2-dependent mobilization defect; neutrophil homeostasis also seems to be affected. That hypothesis is supported by the reported association of rare heterozygous CXCR2 missense variants, including the one carried by P4, with low white blood-cell counts.4 Elucidating the mechanisms underlying the relationship between the biallelic CXCR2 mutations identified herein and neutropenia will require the development of 768

relevant experimental models. Alternative models to mice should be considered, in light of the lack of a murine CXCL8 homologue and the neutrophilia of mice lacking Cxcr2.13,14 However, targeted Cxcr2 invalidation in mouse neutrophils led to their retention in bone marrow, reproducing a myelokathexis phenotype,4 thereby suggesting a role for Cxcr2 in the regulation of neutrophil biology and, intrinsically, in neutrophil trafficking. In contrast to patients with WHIM syndrome, who suffer from chronic lymphopenia, often associated with hypogammaglobulinemia,6,15 patients with CXCR2 mutations experienced only transient episodes of lymphopenia and had elevated levels of immunoglobulins, mostly IgG and IgA (Table 2). B-lymphocyte counts were normal, unlike those in mice with invalidated Cxcr2, which exhibited B-cell expansion,13 highlighting the limitation of mice to model CXCR2 deficiency. No papilloma virus-induced warts, neoplasia or syndromic features, such as tetralogy of Fallot, observed in WHIM syndrome15 were noted during the follow-up of the patients. However, we could not exclude incomplete penetrance of these phenotypes, as reported in WHIM syndrome.6,15 In conclusion, CXCR2 deficiency seems to be a distinct molecular entity associated with congenital neutropenia with clinical severity and pathogenic mechanisms distinct from those of WHIM syndrome, thereby highlighting the importance of determining CXCR2 mutational status in patients with chronic neutropenia. Viviana Marin-Esteban,1 Jenny Youn,2 Blandine Beaupain,3,4 Agnieszka Jaracz-Ros,1 Vincent Barlogis,5 Odile Fenneteau,6 Thierry Leblanc,7 Florence Bellanger,8 Philippe Pellet,8 Julien Buratti,8 Hélène Lapillonne,9 Françoise Bachelerie,1 Jean Donadieu2,3,4 and Christine Bellanné-Chantelot4,8,10 1 Université Paris-Saclay, Inserm UMR996, Inflammation, Microbiome and Immunosurveillance, Clamart; 2Sorbonne Université, Service d’Hémato-oncologie Pédiatrique, Assistance Publique–Hopitaux de Paris (AP-HP), Hôpital Trousseau, Paris; 3Registre Français des Neutropénies Congénitales, Hôpital Trousseau, Paris; 4Centre de Référence des Neutropénies Chroniques, AP-HP, Hôpital Trousseau, Paris; 5CHU Marseille, Hôpital La Timone, Service d’Hémato-oncologie Pédiatrique, Assistance Publique–Hôpitaux de Marseille, Marseille; 6 AP-HP, Laboratoire d’Hématologie, Hôpital Robert-Debré, Paris; 7APHP, Hôpital Robert-Debré, Service d’Hématologie Pédiatrique, Paris; 8 Sorbonne Université, Département de Génétique Médicale, AP-HP, Hôpital Pitié–Salpêtrière, Paris; 9Sorbonne Université, CRSA–Unité INSERM, AP-HP, Hôpital Trousseau, Paris and 10Inserm U1287, Villejuif, France Correspondence: CHRISTINE BELLANNÉ-CHANTELOT christine.bellanne-chantelot@aphp.fr VIVIANA MARIN-ESTEBAN viviana.marin-esteban@universite-paris-saclay.fr doi:10.3324/haematol.2021.279254 Received: May 20, 2021. Accepted: November 25, 2021. Pre-published: December 2, 2021. Disclosures: no conflicts of interest to disclose. Contributions: VM-E, FB, JD and CB-C designed the study. VME collected and interpreted functional data. JY analyzed clinical data. BB collected biological and clinical data. AJ-R performed chemotaxis assays. VB and TL provided samples and clinical data. OF performed and reviewed bone marrow examinations. FB and PP performed molecular experiments and exome sequencing. JB performed exome annotation. HL performed cytological analysis. JD analyzed clinical data and performed the statistical analysis. CB-C analyzed exome haematologica | 2022; 107(3)

Letters to the Editor

sequencing and performed variant interpretation; VM-E, FB, JD and CB-C analyzed the data and wrote the manuscript which was reviewed and edited by all authors. Acknowledgments: the authors would like to thank the families involved in the study. They thank the IPSIT “Ingénierie et Plateformes au Service de l’Innovation Thérapeutique” Laboratory and, especially, Mrs M.-L. Aknin for her support with flow-cytometry analysis (PLAIMMO platform). Funding: the French Severe Chronic Neutropenia Registry is supported by grants from X4 Pharma, Prolong Pharma and Chugai SA (B. Beaupain, J. Donadieu). This work was also funded by the Association Laurette Fugain and the CEREDIH. Data-sharing statement: technical information is available on request in order to assist other laboratories with characterization of CXCR2 variants.

References 1. Martin C, Burdon PC, Bridger G, Gutierrez-Ramos JC, Williams TJ, Rankin SM. Chemokines acting via CXCR2 and CXCR4 control the release of neutrophils from the bone marrow and their return following senescence. Immunity. 2003;19(4):583-593. 2. Eash KJ, Greenbaum AM, Gopalan PK, Link DC. CXCR2 and CXCR4 antagonistically regulate neutrophil trafficking from murine bone marrow. J Clin Invest. 2010;120(7):2423-2431. 3. Hernandez PA, Gorlin RJ, Lukens JN, et al. Mutations in the chemokine receptor gene CXCR4 are associated with WHIM syndrome, a combined immunodeficiency disease. Nat Genet. 2003;34(1):70-74. 4. Auer PL, Teumer A, Schick U, et al. Rare and low-frequency coding variants in CXCR2 and other genes are associated with hematologi-

haematologica | 2022; 107(3)

cal traits. Nat Genet. 2014;46(6):629-634. 5. Bohinjec J. Myelokathexis: chronic neutropenia with hyperplastic bone marrow and hypersegmented neutrophils in two siblings. Blut. 1981;42(3):191-196. 6. Heusinkveld LE, Majumdar S, Gao J-L, McDermott DH, Murphy PM. WHIM syndrome: from pathogenesis towards personalized medicine and cure. J Clin Immunol. 2019;39(6):532-556. 7. Rovati GE, Capra V, Neubig RR. The highly conserved DRY motif of class A G protein-coupled receptors: beyond the ground state. Mol Pharmacol. 2007;71(4):959-964. 8. Liu K, Wu L, Yuan S, et al. Structural basis of CXC chemokine receptor 2 activation and signalling. Nature. 2020;585(7823):135-140. 9. Honczarenko M, Douglas RS, Mathias C, Lee B, Ratajczak MZ, Silberstein LE. SDF-1 responsiveness does not correlate with CXCR4 expression levels of developing human bone marrow B cells. Blood. 1999;94(9):2990-2998. 10. Richardson RM, Pridgen BC, Haribabu B, Ali H, Snyderman R. Differential cross-regulation of the human chemokine receptors CXCR1 and CXCR2. Evidence for time-dependent signal generation. J Biol Chem. 1998;273(37):23830-23836. 11. Nasser MW, Raghuwanshi SK, Grant DJ, Jala VR, Rajarathnam K, Richardson RM. Differential activation and regulation of CXCR1 and CXCR2 by CXCL8 monomer and dimer. J Immunol. 2009;183(5):3425-3432. 12. Petri B, Sanz MJ. Neutrophil chemotaxis. Cell Tissue Res. 2018;371:425-436. 13. Cacalano G, Lee J, Kikly K, et al. Neutrophil and B cell expansion in mice that lack the murine IL-8 receptor homolog. Science. 1994;265(5172):682-684. 14. Broxmeyer HE, Cooper S, Cacalano G, Hague NL, Bailish E, Moore MW. Involvement of interleukin (IL) 8 receptor in negative regulation of myeloid progenitor cells in vivo: evidence from mice lacking the murine IL-8 receptor homologue. J Exp Med. 1996;184(5):1825-1832. 15. Beaussant Cohen S, Fenneteau O, Plouvier E, et al. Description and outcome of a cohort of 8 patients with WHIM syndrome from the French Severe Chronic Neutropenia Registry. Orphanet J Rare Dis. 2012;7:71.

769

Letters to the Editor

A phase I study of the fully human, fragment crystallizable-engineered, anti-CD-33 monoclonal antibody BI 836858 in patients with previously-treated acute myeloid leukemia In recent years, several research programs in acute myeloid leukemia (AML) have investigated the use of therapeutic monoclonal antibodies, which primarily elicit their effects through direct cell killing (apoptosis), via antibody-dependent cellular cytotoxicity (ADCC) or antibody-dependent cellular phagocytosis (ADCP).1,2 Attention has been particularly focused on the myeloid differentiation antigen CD33,2 which is expressed on the surface of leukemic blast cells of almost all AML patients.1 While the activity of unconjugated anti-CD33 antibodies such as lintuzumab has been generally disappointing to date,3-6 clinical experience with gemtuzumab ozogamicin, a humanized anti-CD33 antibody-drug conjugate, provides proof-of-principle for targeting CD33 in patients with AML.7,8 BI 836858 is a fully humanized IgG1 unconjugated anti-CD33 monoclonal antibody.1,9 Unlike lintuzumab, BI 836858 was fragment crystallizable (Fc)-engineered for increased binding to FcγRIIIa (CD16),9 a receptor found on the surface of natural killer (NK) cells and known to be involved in ADCC signaling.1,10 Accordingly, BI 836858 demonstrated superior ADCC to lintuzumab in AML cells in the laboratory.9 Here, we report the findings of a phase I dose escalation study of BI 836858 in patients with relapsed or refractory (R/R) AML, according to World Health Organization 2016 criteria (clinicaltrials gov. Identifier: NCT01690624). Details of the study are available on request. Briefly, BI 836858 was administered as an intravenous (i.v.) infusion every 7 days in a 14-day treatment cycle (days 1 and 8). Premedication was obligatory prior to the first three administrations (acetaminophen/paracetamol; antihistamine; glucocorticoid). The starting dose of BI 836858 was 10 mg, and in the absence of dose-limiting toxicities (DLT), dose escalation up to 320 mg was planned. Patients could receive up to eight repeated administrations of BI 836858 and in the case of clinical benefit and acceptable tolerability were allowed to continue treatment beyond that until disease progression. In the first seven patients, infusions were administered over a 3-hour period; however, due to the occurrence of a high number of infusion-related reactions (IRR), administration was adapted to a stepwise rate-controlled infusion with an increased premedication glucocorticoid dose (100 mg prednisolone or equivalent). If no IRR were apparent after first administration, the dose was reduced to 50 mg for adminstrations two to four, 25 mg for administration 5 and zero thereafter. Also, the protocol was adapted according to tumor load; patients with >5,000 leukocytes/mL peripheral blood were excluded. The primary endpoints for the study were the maximum tolerated dose (MTD) and number of patients with DLT during the MTD evaluation period (the first 2 treatment cycles). Secondary efficacy endpoints included best overall response according to International Working Group Criteria and progression-free survival (PFS), defined as the time from first treatment with BI 836858 until disease progression, relapse or death. Fifty-five patients were screened and 27 were treated with BI 836858 (10-40 mg; Table 1). The median duration of treatment was 21 days (range, 1-99 days), with a median of three infusions given (range, 1-14 infusions). The DLT evaluation period was not completed by 13 patients 770

for reasons other than DLT (progressive disease, n=7; fatal intracranial hemorrhage, n=1; other adverse event [AE], n=2; patient refusal, n=1; not evaluable per protocol, n=1; persistent disease/lack of efficacy, n=1), therefore the MTD analysis set comprised 14 patients. Two patients in the 10 mg cohort had DLT. One patient had drug-related grade 3 elevated alanine aminotransferase (ALT) and aspartate transaminase (AST) 2 days after the first dose of BI 836858, which resolved within 8 days and 5 days, respectively. No change was made to study treatment for this patient. A second patient had a treatmentrelated grade 3 liver function test increased 3 days after the first dose of BI 836858, which resolved within 6 days. BI 836858 was discontinued in this patient. Following protocol amendment, no further DLT were reported in the 10, 20 or 40 mg dose cohorts. The MTD was not reached because the study was prematurely terminated by the sponsor, based on interim pharmacodynamic evaluations. The most common AE were febrile neutropenia (44%), nausea (44%), IRR (41%) and anemia (37%; Table 2). Febrile neutropenia (41%) was the most frequent grade 3 AE. Grade 4 AE were reported in six patients (22%). Seventeen (63%) patients had AE considered related to BI 836858 (9 in the 10 mg cohort, 5 in the 20 mg cohort, and 3 in the 40 mg cohort). The most common were IRR (11 patients [41%]). The rate of IRR was higher prior to adaptation of the infusion protocol (57% vs. 35% after adaptation) with all IRR occurring in patients with a white blood cell (WBC) count of >10x103/mL. The majority of IRR occurred during the first six infusions. All cases of IRR were manageable with established supportive care measures. No patients experienced AE that led to dose reductions. Six (22%) patients discontinued treatment due to AE, including one with grade 3 febrile neutropenia (20 mg dosing cohort), one with grade 2 leukocytosis (10 mg), one grade 3 cardiac failure (40 mg), one grade 3 IRR (20 mg), one grade 3 liver function test increased (10 mg) and one patient with a grade 4 neutrophil count decreased (10 mg). Twentythree patients (85%) had serious AE (SAE); five SAE were considered to be related to the study drug and were recorded in four patients (grade 3 IRR; grade 3 IRR and grade 3 liver function test increased; grade 3 non-cardiac chest pain; grade 3 elevated ALT/AST). Sixteen patients died during the study (8 during the on-treatment period; 4 of which were due to disease progression and 4 were due to unknown reasons). None of the deaths were related to study treatment. Special search categories for AE (i.e., user-defined search categories) were used for this study to adequately monitor the frequency and severity of anticipated AE (e.g., potential class effects). The most frequent user-defined AE were neutropenia (48%), nausea (44%), IRR (41%) and drug-related hepatic disorders (33%). Laboratory evaluation demonstrated that 24 (89%) patients had low WBC counts (grade 3: 33%; grade 4: 52%) BI 836858 exhibited two-compartmental pharmacokinetic behavior, with maximum plasma concentrations generally achieved at the end of the infusion (Table 3). From 10 to 40 mg BI 836858, maximum plasma concentrations for patients with R/R AML patients increased in a dose-proportional manner (Online Supplementary Figure S1). BI 836858 exposure and apparent terminal half-life (t ) increased with increasing doses (Table 3). Total plasma clearance was low and decreased in a dose-dependent manner. The volume of distribution was small, at approximately 6-7 L. No accumulation of BI 836858 was observed in plasma after weekly doses of up to 20 mg. 1/2

haematologica | 2022; 107(3)

Letters to the Editor

Slight accumulation was observed in patients receiving 40 mg. Anti-drug antibodies were detected in one of 25 patients with R/R AML after the first infusion in cycle 1; however, this patient had tested positive prior to treatment commencement. Baseline absolute counts of CD33+ blasts in peripheral

blood and bone marrow were highly variable, ranging from 15% to 95% and 28% to 95% of total blasts, respectively. Overall, total blood blasts remained unchanged following treatment with BI 836858. However, pharmacodynamic analyses showed decreased levels of CD33+ blasts in blood after adminis-

Table 1. Baseline demographics and characteristics of patients with relapsed/refractory acute myeloid leukemia treated with BI 836858.

Characteristics

10 mg N=12

BI 836858 dose 20 mg N=8

40 mg N=7

All patients N=27

Male Race White Black/African American Multiple* Age, years Median (range) <65 ≥65 ECOG PS 0 1 2 Type of AML De novo Secondary Blasts in bone marrow Median, % (range) Bone marrow blasts category <30% ≥30% Missing Previous systemic anti-leukemia therapies Median, n (range) ≥1 line of iHD ≥1 line of pLD ≥1 line of ASCT ≥1 line of other

6 (50)

6 (75)

3 (43)

15 (56)

12 (100) 0 0

6 (75) 2 (25) 0

6 (86) 0 1 (14)

24 (89) 2 (7) 1 (4)

63.5 (36-77) 6 (50) 6 (50)

62.0 (45-67) 4 (50) 4 (50)

70.0 (63-81) 1 (14) 6 (86)

67.0 (36-81) 11 (41) 16 (59)

4 (33) 8 (67) 0

1 (13) 5 (63) 2 (25)

1 (14) 3 (43) 3 (43)

6 (22) 16 (59) 5 (19)

7 (58) 5 (42)

3 (38) 5 (63)

5 (71) 2 (29)

15 (56) 12 (44)

18.0 (4.0-96.0)

56.0 (0-90.0)

50.0 (1.0-92.0)

41.5 (0-96.0)

7 (58) 4 (33) 1 (8)

2 (25) 6 (75) 0

2 (29) 5 (71) 0

11 (41) 15 (56) 1 (4)

2.0 (1-7) 10 (83) 3 (25) 4 (33) 2 (17)

3.5 (1-9) 8 (100) 3 (38) 2 (25) 1 (13)

2.0 (1-5) 7 (100) 3 (43) 1 (14) 0

3.0 (1-9) 25 (93) 9 (33) 7 (26) 3 (11)

*American Indian/Alaskan native and White. Data are n (%), unless otherwise stated. iHD: intensive high dose; pLD: palliative low dose; ECOG PS: Eastern Cooperative Oncology Group performance status; R/R AML: relapsed or refractory acute myeloid leukemia; ASCT: allogeneic stem cell transplantation.

Table 2. All-cause adverse events by MedDRA preferred terms and highest CTCAE grade in patients with relapsed or refractory acute myeloid leukemia treated with BI 836858 during the on-treatment period.

AE, n (%) Total with AE Febrile neutropenia Nausea Infusion-related reactions Anemia Dyspnea Platelet count decreased Constipation White blood cell count decreased Hypokalemia Fatigue Sepsis Atrial fibrillation

All grades

Grade 1/2

Grade 3

Grade 4

Grade 5

27 (100) 12 (44) 12 (44) 11 (41) 10 (37) 9 (33) 8 (30) 7 (26) 7 (26) 7 (26) 6 (22) 4 (15) 3 (11)

2 (7) 1 (4) 12 (44) 8 (30) 1 (4) 8 (30) 0 7 (26) 1 (4) 6 (22) 5 (19) 0 0

11 (41) 11 (41) 0 3 (11) 8 (30) 1 (4) 0 0 1 (4) 1 (4) 1 (4) 1 (4) 3 (11)

6 (22) 0 0 0 1 (4) 0 8 (30) 0 5 (19) 0 0 2 (7) 0

8 (30) 0 0 0 0 0 0 0 0 0 0 1 (4) 0

Adverse events (AE) shown are those occurring in >20% of patients for all grades and grade 1/2, >10% for grade 3 and all grade 4. CTCAE: common terminology criteria for adverse events; R/R AML: relapsed or refractory acute myeloid leukemia. Medical dictionary for drug regulatory activities (MedDRA) version used for reporting: 21.0.

haematologica | 2022; 107(3)

771

Letters to the Editor

Table 3. Summary of pharmacokinetic parameters following a single infusion of BI 836858 in patients with relapsed or refractory acute myeloid leukemia (n=25, infusion duration 3-6 hours in different dose groups).

10 mg† (n=12) 20 mg‡ (n=6) 40 mg‡ (n=7)

AUC0-∞ ng.h/mL

AUC0-∞, norm (ng.h/mL)/mg

23,400 (137) 97,900 (114) 296,000 (256)

2,340 (137) 4,890 (114) 7,400 (256)

Pharmacokinetic parameters* Cmax Cmax, norm tmax ng/mL (ng/mL)/mg hours 873 (207) 3,270 (35.4) 6,100 (82.1)

87.3 (207) 164 (35.4) 152 (82.1)

4.43 (3.00-7.00) 6.10 (5.00-8.80) 5.87 (5.00-8.45)

t½ hours

CL mL/min

Vss L

11.3 (64.3) 21.4 (91.7) 34.5 (107)

7.11 (137) 3.40 (114) 2.25 (256)

7.01 (53.0) 6.30 (29.4) 6.86 (69.5)

*Shown are geometric mean (%gCV) non-compartmental pharmacokinetic parameters of BI 836858 except for tmax which is shown as median (range); †infusion range from 3-6 hours; ‡5-hour infusion. AUC0-∞: area under the plasma concentration-time curve over the time interval from zero extrapolated to infinity; CL: clearance; Cmax: maximum concentration; gCV: geometric coefficient of variance; NA: not assessable; NC: not calculated; norm: normalized to administered dose; R/R AML: relapsed or refractory acute myeloid leukemia; tmax: time point at which maximum concentration is reached; t : apparent half-life; Vss: volume of distribution at steady state ½

tration of 40 mg BI 836858, indicating target engagement (Online Supplementary Figure S2). In the 40 mg dose cohort, CD33+ blasts were undetectable in the bone marrow of five of seven patients on cycle 1 day 8, suggesting target saturation. Conversely, in the 10 mg cohort, the majority of patients had detectable CD33+ blasts at this time point. Most patients had low NK counts (CD3-negative, CD16-positive, CD56-positive) in the blood and bone marrow at screening (Online Supplementary Figure S3). Peripheral blood NK cells were below the lower limit of normal (<30 cells/mL), within the lower end of normal range (30-150 cells/mL), normal (>150 cells/mL) and missing in 30%, 56%, 11% and 4% of patients, respectively. Of note, there was a transient decline in NK cell counts at cycle 1 day 4 compared to the screening values in patients across all treatment cohorts. NK cells in the bone marrow ranged from very few to 7%. In both blood and bone marrow, there were no significant changes in the numbers of activated NK cells (expressing CD69 or CD158b) during treatment (data not shown). Monocyte counts were also below, or at the very end of the lower limit of the normal range in the majority (59%) of patients. No responses to treatment were detected; 23 patients (85%) were removed from the study due to lack of response with persistent AML, two (7%) patients were not evaluable, and two (7%) did not have a post-baseline assessment. Median PFS was 29 days (95% confidence interval: 27-50). There were no notable changes in the percentage of myeloid blasts in the bone marrow during the treatment period. To conclude, this study provides valuable information about the safety, efficacy, pharmacokinetics and pharmacodynamics of BI 836858 in patients with R/R AML which may have important implications for potential future development of immunotherapies in this setting. BI 836858 had predictable and manageable tolerability with no unexpected AE. However, although there was evidence of target engagement in the blood and to some extent in the bone marrow, no responses to BI 836858 were observed. The primary objective of MTD was not met due to premature termination. We hypothesize that the low levels of baseline effector cells observed in this study were relevant to the lack of efficacy of BI 836858. Other studies indicate that baseline NK cell phenotype and function is defective in patients with AML.11 Interestingly, phenotypic and functional abnormalities of NK cells appear to be partially restored in AML patients achieving a complete remission (CR).11 As BI 836858 relies on ADCC for functionality, the low levels of effector cells detected in the patient population, and the lack of pharmacodynamic effects (see Online Supplementary Appendix), underpinned the decision to terminate the study based on a lack of perceived benefit over risk. 772

In contrast to the current study, lintuzumab conferred objective responses including CR in a phase I dose escalation trial in R/R AML.12 Given that BI 836858 was a similar design to lintuzumab (both were fully humanized IgG1 monoclonal antibodies, though BI 836858 was Fc engineered to increase ADCC) and was considered superior to lintuzumab in preclinical experiments,9 it is not clear why it did not show clinical activity in this phase I study. Lintuzumab is dependent on engagement of several immune effector cells including macrophages/monocytes (that facilitate ADCP)13 as well as NK cells (that facilitate ADCC).13 The impact of BI 836858 on immune effector cell function in patients requires more evaluation. It is possible that differences in internalization kinetics of CD33 following engagement with lintuzumab or BI 836858 may influence immunogenicity. While internalization is slower with BI 836858 and this correlates with superior ADCC in cell-based assays,9 it is uncertain how this may influence other effector functions. Also, lack of efficacy could potentially relate to pre-administration of glucocorticoids. Despite steps to reduce glucocorticoid dose as soon as possible, administration may interfere with NK cell function, as indicated in previous preclinical studies.14,15 However, unpublished observations suggest that BI 836858 may activate NK cells when combined with decitabine, despite premedication with glucocorticoids. Further clinical studies have been designed to assess combination regimens that may potentiate the activity of BI 836858 by increasing effector immune function. For example, based on preclinical evidence that decitabine increased BI 836858 activity via upregulation of the NK group 2D ligand,9 a phase II study is exploring the efficacy and safety of this combination (clinicaltrials gov. Identifier: NCT02632721). Other BI 836858-based combinations are also being assessed in a multi-sub-study phase Ib/II trial (clinicaltrials gov. Identifier: NCT03013998). These studies will include post-remission therapy when immune effectors might be more numerous and active. Sumithira Vasu,1 Jessica K. Altman,2 Geoffrey L. Uy,3 Martin S. Tallman,4 Ivana Gojo,5 Gerard Lozanski,6 Ute Burkard,7 Annika Osswald,7 Pamela James,8 Björn Rüter8 and William Blum9 1 Division of Hematology, Department of Internal Medicine, The Ohio State University Comprehensive Cancer Center, Columbus, OH, USA; 2Division of Hematology/Oncology, Northwestern University, Chicago, IL, USA; 3Washington University School of Medicine, Siteman Cancer Center, St Louis, MO, USA; 4Leukemia Service, Memorial Sloan Kettering Cancer Center, Weill Cornell Medical College, New York, NY, USA; 5The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD, USA; 6Department of Pathology, The Ohio State University, Columbus, OH, USA; haematologica | 2022; 107(3)

Letters to the Editor

Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach an der Riß, Germany; 8Boehringer Ingelheim Pharmaceuticals, Inc, Ridgefield, CT, USA and 9Department of Hematology and Medical Oncology, Emory University School of Medicine, Winship Cancer Institute, Atlanta, GA, USA Correspondence: WILLIAM BLUM - william.g.blum@emory.edu doi:10.3324/haematol.2020.274118 Received: October 22, 2020. Accepted: November 25, 2021. Pre-published: December 2, 2021. Disclosures: SV reports receiving funding from Boehringer Ingelheim for Advisory Board participation. JKA reports being part of an advisory council or committee for Astellas, Novartis, Cancer Expert Now, Agios, Glycomimetics, Theradex, AbbVie and Daiichi Sankyo, and being a speaker for PeerView, prIME Oncology, France Foundation. GLU reports receiving honoraria from Astellas and consulting fees from Genentech and Jazz. MST reports being part of an advisory council or committee for AbbVie, Daiichi-Sankyo, Orsenix, KAHR, Rigel, Delta Fly Pharma, Tetraphase, Oncolyze, Jazz Pharma, Roche, Biosight and Novartis, receiving grants or funds from AbbVie, Orsenix, ADC Therapeutics, Biosight, Glycomimetics, Rafael and Amgen, and other potential financial relationship with UpToDate. IG reports receiving grants or funds from Merck, Amgen, Amphivena, Celgene and Genentech. GL reports receiving funding of flow cytometry analysis of samples for this study from Boehringer Ingelheim. UB and AO report employment with Boehringer Ingelheim Pharma GmbH & Co. KG. PJ reports no conflict of interests. BR reports employment and other potential financial relationship with Boehringer Ingelheim Pharmaceuticals Inc. WB reports receiving consulting fees from Amerisource Bergen and grants or funds from Leukemia and Lymphoma Society, Xencor, Forma, Celyad and Novartis. Contributions: JKA, GLU, MST, BR and WB designed the study; SV, WB, MT, JKA, MST, GLU and GL conducted the study, analyzed data, wrote and edited the manuscript. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work, which includes ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Acknowledgements: the authors would like to thank the patients and their families, the research staff and Anne Esler for statistical analysis. Medical writing assistance during the preparation of this manuscript was provided by Lynn Pritchard, Ashfield MedComms, an Ashfield Health Company, which was funded by Boehringer Ingelheim. Funding: this work was supported by Boehringer Ingelheim. Data sharing statement: the clinical study report (including appendices, but without line listings) and other clinical documents related to this study may be accessed on request. Prior to providing access, the documents and data will be examined, and, if necessary, redacted and de-identified to protect the personal data of study participants and personnel, and to respect the boundaries of the informed consent of the

haematologica | 2022; 107(3)

study participants. See https://trials.boehringeringelheim.com/data_sharing/sharing.html#accordion-1-2 for further details. Bona fide, qualified scientific and medical researchers may request access de-identified, analyzable patient-level study data, together with documentation describing the structure and content of the datasets. Researchers should use https://vivli.org/ to request access to raw data from this study.

References 1. Walter RB. Investigational CD33-targeted therapeutics for acute

myeloid leukemia. Expert Opin Investig Drugs. 2018;27(4):339-348. 2. Williams BA, Law A, Hunyadkurti J, Desilets S, Leyton JV, Keating A. Antibody therapies for acute myeloid leukemia: unconjugated, toxin-conjugated, radio-conjugated and multivalent formats. J Clin Med. 2019;8(8):1261. 3. Feldman EJ, Brandwein J, Stone R, et al. Phase III randomized multicenter study of a humanized anti-CD33 monoclonal antibody, lintuzumab, in combination with chemotherapy, versus chemotherapy alone in patients with refractory or first-relapsed acute myeloid leukemia. J Clin Oncol. 2005;23(18):4110-4116. 4. Borthakur G, Rosenblum MG, Talpaz M, et al. Phase 1 study of an anti-CD33 immunotoxin, humanized monoclonal antibody M195 conjugated to recombinant gelonin (HUM-195/rGEL), in patients with advanced myeloid malignancies. Haematologica. 2013;98 (2):217-221. 5. Sekeres MA, Lancet JE, Wood BL, et al. Randomized phase IIb study of low-dose cytarabine and lintuzumab versus low-dose cytarabine and placebo in older adults with untreated acute myeloid leukemia. Haematologica. 2013;98(1):119-128. 6. Feldman E, Kalaycio M, Weiner G, et al. Treatment of relapsed or refractory acute myeloid leukemia with humanized anti-CD33 monoclonal antibody HuM195. Leukemia. 2003;17(2):314-318. 7. Castaigne S, Pautas C, Terre C, et al. Effect of gemtuzumab ozogamicin on survival of adult patients with de-novo acute myeloid leukaemia (ALFA-0701): a randomised, open-label, phase 3 study. Lancet. 2012;379(9825):1508-1516. 8. Hills RK, Castaigne S, Appelbaum FR, et al. Addition of gemtuzumab ozogamicin to induction chemotherapy in adult patients with acute myeloid leukaemia: a meta-analysis of individual patient data from randomised controlled trials. Lancet Oncol. 2014;15(9):986-996. 9. Vasu S, He S, Cheney C, et al. Decitabine enhances anti-CD33 monoclonal antibody BI 836858-mediated natural killer ADCC against AML blasts. Blood. 2016;127(23):2879-2889. 10. Romain G, Senyukov V, Rey-Villamizar N, et al. Antibody Fc engineering improves frequency and promotes kinetic boosting of serial killing mediated by NK cells. Blood. 2014;124(22):3241-3249. 11. Stringaris K, Sekine T, Khoder A, et al. Leukemia-induced phenotypic and functional defects in natural killer cells predict failure to achieve remission in acute myeloid leukemia. Haematologica. 2014;99(5): 836-847. 12. Raza A, Jurcic JG, Roboz GJ, et al. Complete remissions observed in acute myeloid leukemia following prolonged exposure to lintuzumab: a phase 1 trial. Leuk Lymphoma. 2009;50(8):1336-1344. 13. Sutherland MK, Yu C, Lewis TS, et al. Anti-leukemic activity of lintuzumab (SGN-33) in preclinical models of acute myeloid leukemia. MAbs. 2009;1(5):481-490. 14. Kumai T, Oikawa K, Aoki N, Kimura S, Harabuchi Y, Kobayashi H. Assessment of the change in cetuximab-induced antibody-dependent cellular cytotoxicity activity of natural killer cells by steroid. Head Neck. 2016;38(3):410-416. 15. Eddy JL, Krukowski K, Janusek L, Mathews HL. Glucocorticoids regulate natural killer cell function epigenetically. Cell Immunol. 2014;290(1):120-130.

773

ERRATA CORRIGE Isatuximab plus pomalidomide and dexamethasone in elderly patients with relapsed/refractory multiple myeloma: ICARIA-MM subgroup analysis Fredrik Schjesvold,1 Paul G. Richardson,2 Thierry Facon,3 Adrián Alegre,4 Andrew Spencer,5 Artur Jurczyszyn,6 Kazutaka Sunami,7 Laurent Frenzel,8 Chang-Ki Min,9 Sophie Guillonneau,10 Peggy L. Lin,11 Solenn Le-Guennec,12 Frank Campana,13 Helgi van de Velde,13 Samira Bensfia11 and Sara Bringhen14 1 Oslo Myeloma Center, Oslo University Hospital and KG Jebsen Center for B Cell Malignancies, University of Oslo, Oslo, Norway; 2Dana-Farber Cancer Institute, Boston, MA, USA; 3Lille University Hospital, Lille, France; 4Hospital Universitario La Princesa & Hospital Quironsalud, Madrid, Spain; 5Department of Clinical Hematology, Alfred Health-Monash University, Melbourne, Australia; 6Department of Hematology, Jagiellonian University Medical College, Krakow, Poland; 7Department of Hematology, National Hospital Organization Okayama Medical Center, Okayama, Japan; 8Hôpital Necker-Enfants Malades, Paris, France; 9Department of Hematology, Catholic Hematology Hospital and Leukemia Research Institute, Seoul St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of South Korea; 10Sanofi CMO, Chilly-Mazarin, France; 11Sanofi Global Oncology, Cambridge, MA, USA; 12Sanofi R&D, Vitry-sur-Seine, France; 13Sanofi R&D, Cambridge, MA, USA and 14Myeloma Unit, Division of Hematology, University of Torino, Azienda-Ospedaliero Universitaria Città della Salute e della Scienza di Torino, Torino, Italy Correspondence: SARA BRINGHEN - sarabringhen@yahoo.com Published in Haematologica 2021;106(4):1182-1187

doi:10.3324/haematol.2021.279160

In the article by Schjesvold et al., entitled “Isatuximab plus pomalidomide and dexamethasone in elderly patients with relapsed/refractory multiple myeloma: ICARIA-MM subgroup analysis”, which appeared in the April 2021 issue of Haematologica (volume 106, pages 1182-1187), the values entered for “Refractory status” at the bottom of Table 1 were incorrect. The authors have prepared a new, corrected version of Table 1, which is reported here. The authors apologize to the Editor and readers for their mistake. They want to underscore that the results and conclusions of the paper are unaffected by the error.

774

haematologica | 2022; 107(3)

Errata Corrige

Table 1. Patients’ baseline characteristics at study entry by age group in the intent-to-treat population.

≥75 years (n=61)

Age (years) Mean (SD) Median (range) MM subtype, n (%) IgG IgA IgM Kappa light chain only Lambda light chain only ISS stage*, n (%) Stage I Stage II Stage III Unknown ECOG Performance Status, n (%) 0 1 2 Cytogenetic risk†, n (%) High-risk CA Standard-risk CA Unknown or missing N. of patients with a medical history of Asthma or COPD, n (%) N. of patients with renal impairment‡, n (%) eGFR, n (%) ≥60-<90 mL/min/1.73 m² (mild impairment) ≥45-<60 mL/min/1.73 m² ≥30-<45 mL/min/1.73 m² ≥15-<30 mL/min/1.73 m² (severe impairment) N. of prior lines of therapy, Median (range) Prior therapy, n (%) Alkylating agent Proteasome inhibitor Lenalidomide Refractory status, n (%) Lenalidomide refractory PI refractory Lenalidomide and PI refractory

65–74 years (n=122) Isa-Pd Pd (n=68) (n=54)

Isa-Pd (n=32)

Pd (n=29)

77.9 (2.0) 77 (75-83)

78.3 (3.2) 78 (75-86)

69.4 (2.9) 69 (65-74)

21 (65.6) 9 (28.1) 0 1 (3.1) 1 (3.1)

22 (75.9) 4 (13.8) 0 2 (6.9) 1 (3.4)

7 (21.9) 12 (37.5) 13 (40.6) 0

<65 years (n=124) Isa-Pd (n=54)

Pd (n=70)

69.0 (2.5) 69 (65-74)

56.5 (5.9) 57.5 (36-64)

57.0 (6.1) 58 (41-64)

45 (66.2) 17 (25.0) 1 (1.5) 2 (2.9) 3 (4.4)

32 (59.3) 19 (35.2) 0 1 (1.9) 2 (3.7)

38 (70.4) 7 (13.0) 1 (1.9) 5 (9.3) 3 (5.6)

47 (67.1) 18 (25.7) 0 4 (5.7) 1 (1.4)

4 (13.8) 12 (41.4) 12 (41.4) 1 (3.4)

31 (45.6) 22 (32.4) 14 (20.6) 1 (1.5)

18 (33.3) 23 (42.6) 13 (24.1) 0

26 (48.1) 19 (35.2) 7 (13.0) 2 (3.7)

29 (41 .4) 21 (30.0) 18 (25.7) 2 (2.9)

9 (28.1) 18 (56.3) 5 (15.6)

14 (48.3) 8 (27.6) 7 (24.1)

24 (35.3) 36 (52.9) 8 (11.8)

18 (33.3) 31 (57.4) 5 (9.3)

22 (40.7) 29 (53.7) 3 (5.6)

37 (52.9) 29 (41.4) 4 (5.7)

7 (21.9) 20 (62.5) 5 (15.6)

11 (37.9) 9 (31.0) 9 (31.0)

9 (13.2) 47 (69.1) 12 (17.6)

6 (11.1) 32 (59.3) 16 (29.6)

8 (14.8) 36 (66.7) 10 (18.5)

19 (27.1) 37 (52.9) 14 (20.0)

5 (15.6) 30 (93.8)

5 (17.2) 27 (93.1)

7 (10.3) 63 (92.6)

8 (14.8) 51 (94.4)

4 (7.4) 49 (90.7)

4 (5.7) 67 (95.7)

10 (33.3)

11 (40.7)

31 (49.2)

25 (49.0)

20 (40.8)

33 (49.3)

13 (43.3) 6 (20.0) 0

9 (33.3) 5 (18.5) 1 (3.7)

14 (22.2) 7 (11.1) 0

12 (23.5) 4 (7.8) 0

8 (16.3) 6 (12.2) 1 (2.0)

11 (16.4) 7 (10.4) 0

3 (2–11)

3 (2–10)

3 (2–8)

3 (2–6)

3 (2–10)

3 (2–7)

27 (84.4) 32 (100) 32 (100)

29 (100) 29 (100) 29 (100)

60 (88.2) 68 (100) 68 (100)

51 (94.4) 54 (100) 54 (100)

52 (96.3) 54 (100) 54 (100)

68 (97.1) 70 (100) 70 (100)

30 (93.8) 25 (78.1) 24 (75.0)

28 (96.6) 21 (72.4) 20 (69.0)

63 (92.6) 54 (79.4) 50 (73.5)

47 (87.0) 41 (75.9) 37 (68.5)

51 (94.4) 39 (72.2) 37 (68.5)

65 (92.9) 53 (75.7) 50 (71.4)

*International Staging System staging was derived based on the combination of serum β2-microglobulin and albumin concentrations. †High risk chromosomal abnormalities were defined as the presence of del(17p), and/or t(4;14), and/or t(14;16) by fluorescence in situ hybridization. Cytogenetics was performed by a central laboratory with a cut-off of analyzed plasma cells of 50% for del(17p), and of 30% for t(4;14) and t(14;16). ‡Renal impairment was defined as an estimated glomerular filtration rate <60 mL/min/1.73 m² as determined using the Modification of Diet in Renal Disease (MDRD) equation. Isa: isatuximab: Pd: pomalidine and dexamethasone; SD: standard deviation; MM: multiple myeloma; Ig: immunoglobulin; ISS: International Staging System; ECOG: Eastern Cooperative Oncology Group; CA: chromosomal abnormalities; COPD: chronic obstructive pulmonary disease; eGFR: estimated glomerular filtration rate; PI: proteasome inhibitor.

haematologica | 2022; 107(3)

775

haematologica Journal of the Ferrata Storti Foundation

List of the 1117 reviewers who in 2021 made an essential contribution to the high scientific quality of the journal Maria Abbattista Miguel Abboud Omar Abdel-Wahab Renata Abrahao Charles Abrams Ranjana Advani Walter Ageno Nadine Aguilera Anat Aharon Haifa Kathrin Al-Ali Lorenzo Alberio Sarah Alexander Slimane Allali Carl Allen Hanny Al-Samkari Sandro Altamura Pascal Amireault Claudio Anasetti Tessa Andermann Immacolata Andolfo Charalambos Andreadis Petya Apostolova Jane Apperley R. Ariëns Philippe Armand Paul Armstrong Joseph Aslan Sarit Assouline Ehab Atallah Hervé Avet-Loiseau Irit Avivi Irit Avivi Slim Azouzi Christian Babbs Daria Babushok Ulrike Bacher Emmanuel Bachy Andrea Bacigalupo Peter Bader Tamam Bakchoul Carlo Balduini Alessandra Balduini Enrico Balleari Katrin Bankov Joao Barata Silvia Barbieri Luciano Baronciani Giovanni Barosi 776

Virginie Barraud-Lange Sally Barrington Marije Bartels Matthew Barth Nancy Bartlett Pablo Bartolucci André Baruchel Renato Bassan Connie Batlevi Lisa Baumann Kreuziger Susanne Baumeister Ali Bazarbachi Cecilia Becattini Joan Beckman Christine Bellanne Chantelot Marie Bene Zsuzsanna Bereczky James Berenson Sigbjorn Berentsen Wolfgang Bergmeier Ellin Berman Olivier Bernard Sonja Berndt Francesco Bertoni Julien Bertrand Caroline Besson Ravi Bhatia Shruti Bhatt Paola Bianchi Alexander Bick Marc Bierings Andrea Biondi James Blachly Lionel Blanc Olga Blau Igor Blau Zeev Blumenfeld Dominique Bluteau Steffen Boettcher Nicolas Boissel Niccolò Bolli Katherine Borden Arndt Borkhardt Martin Bornhauser Beat Bornhauser Stefania Bortoluzzi Floris Bosch Jean Bourquin

Teresa Bowman Susan Branford Marc Braunstein Massimo Breccia Elizabeth Brem Benjamin Brenner Emery Bresnick Benoit Brethon John Brewin Pauline Brice Sara Bringhen Annamaria Brioli Eolia Brissot Paul Bröckelmann Andrew Brooks Catherine Broome Patrick Brown Jennifer Brown Brian Brown Lesley Bruce Carlo Brugnara Benedetto Bruno Heiko Bruns Rena Buckstein Lars Bullinger Geert Bultynck Michael Burke James Bussel John Byrd Minji Byun Megan Bywater Jo Caers Sophie Caillat-Zucman Mitchell Cairo Andrea Cairo George Calin Katherine Calvo Clara Camaschella Elias Campo Laura Cannella Alessio Cantore Marcus Carden Gunnar Cario Alessandra Carobbio Marc Carrier Guillaume Cartron Olivier Casasnovas Alberto Cascon

James Casella Ryan Cassaday Bruno Cassinat Jorge Castillo Carla Casulo Marco Cattaneo Michele Cavo Giovanni Cazzaniga Claudio Cerchione Ronjon Chakraverty Subarna Chakravorty John Chan John Chan Steven Chan Melissa Chan Hong Chang Nicolas Chapuis Dharminder Chauhan Chan Yoon Cheah Deepak Chellapandian Roy Chemaly Jichun Chen Luke Chen Weina Chen Edwin Chen Raffaella Chiaramonte Shigeru Chiba Sung Choi Stella Chou Douglas Cines Stefan Ciurea Emmanuelle Clappier Fina Climent Esteller Jacqueline Cloos Thomas Cluzeau Fabrice Cognasse Jonathon Cohen Gisele Colleoni Matthew Collin Raffaella Colombatti Michela Colombo Raymond Comenzo Shannon Conneely Nicola Conran Stefan Constantinescu Valentino Conter Catherine Coombs Mhairi Copland

haematologica | 2022; 107(3)

Paul Coppo Raul Cordoba Catherine Cordonnier Seth Corey Paolo Corradini Javier Corral Irene Corrales Insa Jorge Cortes Benilde Cosmi Luciano Costa Charles Craddock Nick Cross Merlin Crossley Adam Cuker Aubrey Cunnington Corey Cutler Angelo D'Alessandro Frederik Damm Sophia Danhof Debbie Daniels Alexey Danilov Giovanni D'Arena Yesim Dargaud Debabrata Dash Matthew Davids Richard Davis Kara Davis Sara De Biasi Raimondo De Cristofaro Lucia De Franceschi Daphne de Jong Laurence de Leval Luigi De Marco Barbara De Moerloose Mariane de Montalembert Adam de Smith Valerio De Stefano Huguesde The Huguesde Thé Silvia Deaglio Joachim Deeg Zachariah DeFilipp Jan Delabie Michel Delforge Matteo Della Porta Selami Demirci Anne Demulder Enrico Derenzini Karl Desch Ani Deshpande Katrien Devreese Sascha Dietrich Richard Dillon Melios Dimopoulos Wei Ding Angela Dispenzieri Ahmet Dogan Inderjeet Dokal haematologica | 2022; 107(3)

Jean Donadieu Jeanette Doorduijn Adrienne Dorrance Sinisa Dovat Peter Dreger Christoph Driessen Yigal Dror Xiaoping Du Daniel Duerschmid Carlo Dufour Matt Dun Kieron Dunleavy Benjamin Durham Michael Dworzak Martin Dyer Mary Eapen Ashley Eckel Dimitar Efremov Stephane Egee Dennis Eichenauer Sabine Eichinger Anna Eiring Ann Eisfeld Paul Ekert Sara El Hoss Wassim El Nemer Shokrollah Elahi Eric Eldering Shannon Elf Fredrik Ellin Steven Ellis Jean-François Emile Monika Engelhardt Cristina Epalza Ibarrondo Daniel Erdelyi Aurélie Eude-Caye William Evans Andrew Evens Solene Evrard Toby Eyre Marco Falasca Lorenzo Falchi Anahita Fathi Bruno Fattizzo Emmanuel Favaloro Andrew Feldman Falko Fend Christian Fernandez Andres Ferreri Thierry Fest Jean Feuillard Michaela Feuring-Buske Adele Fielding Claire Fieschi Marie-Dominique Filippi Juergen Finke Jude Fitzgibbon Angela Fleischman

Mark Fleming Magali Fontaine Michaela Fontenay Jan Frayne Henrik Frederiksen Sylvie Freeman Michela Frenquelli Kai Fu Richard Furman Gosta Gahrton Dave Gailani Robert Gale Andrea Gallamini R Gallo Karthik Ganapathi Shreyans Gandhi Arnold Ganser Guillermo Garcia-Manero Jean Gaschignard Moshe Gatt Philippe Gaulard Pascale Gaussem Jordan Gauthier Eleni Gavriilaki Meinrad Gawaz Francesca Gay Lindsey George Fabiana Geraci Alina Gerrie Waleed Ghanima Cedric Ghevaert Paolo Ghia Asish Ghosh Nilajan Ghosh Arianna Giacomini Jon Gibbins John Gibson FJ Giles Jennifer Gillette Sergio Giralt Stephane Giraudier Domenico Girelli Andreas Glenthøj Robert Gniadecki Ronald Go Nicola Goekbuget Christian Goetz Josee Golay Hartmut Goldschmidt Victor Gordeuk Leo Gordon Peter Gordon Gaurav Goyal Solomon Graf Martin Gramatzki Dita Gratzinger Giulia Graziani Michael Green

Peter Greenberg Christine Greil Andreas Greinacher Hildegard Greinix Paolo Gresele William Grey Martin Griesshammer Christian Grommes Greta Guarda Romain Guieze François Guilhot Jay Gunawardana Kalpna Gupta Arjun Gupta Ronit Gurion Alejandro Gutierrez Fredrik H.Schjesvold Anoosha Habibi Claudia Haferlach Rana Zeeshan Haider Corinne Haioun Morgan Hakki Christina Halsey Nada Hamad Mehdi Hamadani Paul Hampel Maher Hanoun Hinrich Hansen Yasmine Hardy Julien Haroche Matthew Harper Christine Harrison Claire Harrison Simon Harrison Sylvia Hartmann Tanja Hartmann Daigo Hashimoto Hans Hasselbalch Patrick Hayden Mark Heaney Jennifer Heaney Robert Hebbel Johan Heemskerk Ute Hegenbart Vignir Helgason Kristoffer Hellstrand Kari Hemminki Andrea Henden Jeanne Hendrickson Charles Herbaux Georg Herget Sébastien Héritier Marco Herling Angel Hernandez-Hernandez Daniel Herranz Alex Herrera Chaim Hershko Mark Hertzberg 777

Georg Heß Michael Heuser Mats Heymann Joshua Hill Shinsuke Hirabayashi Caleb Ho Theodore Ho Andreas Hochhaus Daniel Hodson Martin Hoglund Steven Holland Ernst Holler dengli Hong Netanel Horowitz Steven Horwitz Benoît Ho-Tin-Noé Eugenio Hottz Lewis Hsu Shimin Hu Timothy Hughes Susanta Hui Gerwin Huls Michael Hummel Vera Ignjatovic Gabriele Ihorst Julie Irving Deena Iskander Prioty Islam Raphael Itzykson Vincent Jachiet Caron Jacobson Elaine Jaffe Pudur Jagadeeswaran Preetesh Jain Michael Jain Andrzej Jakubowiak Chloé James Fabrice Jardin Robert Jenq Madlen Jentzsch Peng Ji Erel Joffe William Johnson Courtney Jones Gunnar Juliusson Hassan Jumaa Johannes Jung Przemyslaw Juszczynski Lars Kaestner Walter Kahr Martin Kaiser Manali Kamdar Rashmi Kanagal-Shamanna Theoni Kanellopoulou Jun Kang Hagop Kantarjian Alesia Kaplan Reuben Kapur 778

Kennosuke Karube Ana Kasirer-Friede Efstathios Kastritis Arnon Kater Jakob Kather Gregory Kato Lev Kats Antonis Kattamis Jonathan Kaufman Léon Kautz Jun-ichi Kawada Sabine Kayser Colm Keane K Kemparaju Marie Kersten Rachel Kesse-Adu Nigel Key Michael Kharas Hossein Khiabanian Won Seog Kim Fumihiko Kimura Fenella Kirkham Adam Kittai Ulf Klein Catherine Klersy Justin Kline Jan-Henning Klusmann Steven Knapper Stefan Knop Berengere Koehl Thomas Koehnke Scott Kogan Natalie Köhler Seiji Kojima Anders Kolb Krishna Komanduri John Koreth Steffen Koschmieder Maja Krajinovic Daniel Krappmann Christian Kratz Daniela Krause Stefan Krause Robert Kreitman Robert Kridel Sigurdur Kristinsson Nicolaus Kröger Jan Krönke Kazuko Kudo Austin Kulasekararaj Andreas Kulozik Shaji Kumar Anita Kumar Andrea Kündgen Ralf Küppers Mineo Kurokawa John Kuruvilla John Kuruvilla

Shigeru Kusumoto Paul Kyrle Ann LaCasce Laurence Lamant Jatinder Lamba Bernhard Lammle Daniel Landsburg Steven Lane Anton Langerak Petra Langerbeins Claudia Lanvers Riitta Lassila Camille Laurent Michelle Lavin Steven Le Gouill Caroline Le Van Kim Eric Lechman Lydia Lee Alexander Leeksma Sylvain Lefort Sören Lehmann Roberto Lemoli François Lemonnier Peter Lenting John Leonard Sirpa Leppa Stephane Lepretre Will Lester Remi Letestu M Levi Mark Levis Robert Levy Ludovic Lhermitte Linheng Li Zongjin Li Jian Li Nailin Li Jane Liesveld Sean Lim Richard Lin Coleman Lindsley Brian Link Daniel Lipka Ton Lisman Jane Little Feng Liu Zhiqiang Liu Junling Liu Per Ljungman Franco Locatelli Frederick Locke Mignon Loh Thomas Loughran Philip Low Siew-Kee Low Maria Lozano Michael Lübbert Heinz Ludwig

Stefano Luminari Martina Lundberg Slingsby Turid Helen Lunde Leo Luznik Nigel Mackman Kylee Maclachlan Isabelle Mahè Reiner Mailer Alexandros Makis Michael Makris Florent Malard Sebastien Malinge Maksim Mamonkin Ankit Mangla Pier Mannuccio Mannucci Deepa Manwani Ambroise Marçais Francesco Marchesi Rossella Marcucci Xavier Mariette Reinhard Marks Rolf Marschalek Alberto Martelli Joost Martens Rebecca Martin Ida Martinelli JoseA Martinez-Climent Jose Martin-Subero Anna Martner Nabih Maslah Maria-Victoria Mateos Stephen Mathew Anthony Mato Masao Matsuoka Jonas Mattsson Matthew Maurer Luca Mazzarella Kenneth McClain Patrick McGann Pamela McKay Steven McKenzie Mary Frances McMullin Lisa McReynolds Adam Mead Robert Means L. Jeffrey Medeiros Hind Medyouf Neha Mehta-Shah Michel Meignan Jules Meijerink Ester Mejstrikova Christopher Melani Jan Melenhorst David Melzer Fanjie Meng Stephan Menzel Reid Merryman Sachith Mettananda haematologica | 2022; 107(3)

Lüder Meyer HJ Meyerson Sylvain Meylan Hira Mian Gerard Michel Cornelius Miething Aleksandar Mijovic Malgorzata Mikulska Thomas Milne Michael Milsom Roberto Mina Valentina Minciacchi Moshe Mittelman Stefano Molica Thierry Molina Santiago Montes-Moreno Marco Montillo Anthony Moorman Jonathan Moraes Pierre-Emmanuel Morange Philippe Moreau Richard Moriggl Ryan Morin Claudia Morris LM Morton Alison Moskowitz Guillaume Moulis Martina Muckenthaler Antonia Mueller Ann Mullally Ann Mullaly Stephen Mulligan Javier Munoz Andrew Muntean Jagan Muppidi Philip Murphy Satu Mustjoki Pellegrino Musto Elie Nader Shigekazu Nagata Arnon Nagler Michael Nagler Prajeeda Nair Amina Nardo-Marino Mohandas Narla Loretta Nastoupil Yaso Natkunam Yulia Nefedova Elizabeta Nemeth Antonino Neri Clara Nervi Jakob Neubauer Ellis Neufeld Heyu Ni Leo Nicolai Charlotte Niemeyer Tom Nightingale Satoshi Nishizuka haematologica | 2022; 107(3)

Contance Tom Noguchi Josep Nomdedeu Ajay Nooka Ruggiero Norfo Marina Noris Enrico Novelli Ariela Noy Panagiotis Ntziachristos Erfan Nur AT Nurden Paquita Nurden Esther Obeng Niels Odum Vivian Oehler Massimo Offidani Yishai Ofran Stephen Oh Robert Ohgami Shinichiro Okamoto Jessica Okosun Eric Oksenhendler Timothy Olson Adam Olszewski Choon Kiat Ong Stephen Opat Joseph Opferman Etan Orgel Maha Othman German Ott Carolyn Owen Ingrid Pabinger Jerome Paggetti Maria Teresa Pagliari James Palis Maria Palomba Weijun Pan Qishen Pang Milena Pantic Samir Parekh Christopher Park Steven Park Charles Parker Laura Pasqualucci Francesco Passamonti Diana Passaro Bhumika Patel Mrinal Patnaik Paola Patrignani Robert Paulson Annalisa Paviglianiti Steven Pavletic Sue Pavord Charlotte Pawlyn Alessandro Pecci Francesco Pegoraro Jonathan Peled Olavi Pelkonen Sandra Pellegrini

Daniel Peltier Jun Peng Victor Peperzak Melanie Percy Jose Perez-Simon Andrew Perkins Florian Perner Christina Peters Kseniya Petrova Sjaak Philipsen Tycel Phillips Caroline Philpott Pier-Paolo Piccaluga Eric Pietras Pasquale Pignatelli Yana Pikman Stefano Pileri Asha Pillai Vinodh Pillai John Pimanda Miguel Piris Cynthia Pise-Masison Sean Platton Uwe Platzbecker Torben Plesner Christopher Pleyer Isabelle Plo Klaus Podar Bernd Poetzsch Catherine Poirot Pierluigi Porcu Christiane Pott James Poulter Paolo Prandoni Josef Prchal Anuja Premawardena Claude Preudhomme Miles Prince Barbara Pro Kamil Przyborowski Bethan Psaila Xose Puente Fabio Pulcinelli Louise Purton Zhijian Qian Charles Quinn Marc Raab David Rabbolini Jerry Radich Elizabeth Raetz Emmanuel Raffoux Margaret Ragni Alessandro Rambaldi Raajit Rampal Margaret Rand Anna Randi Vijay Rao V. Rao

Leo Rasche Michael Rassner Rachel Rau Michael Rauh Farhad Ravandi Julie Rayes Patrick Reagan Michele Redell Andreas Reiter Mary Relling Raffaele Renella Thomas Renné Linda Resar Roberto Ria Jean-Antoine Ribeil Josep-Maria Ribera Paul Richardson Anita Rijneveld Lisa Rimsza Stefano Rivella Jose Rivera Irene Roberts Kathryn Roberts Aldo Roccaro Jacques Rochette Claire Roddie Francesco Rodeghiero Alba Rodriguez Meira Casper Roed Lindsey Roeker Kerry Rogers Pierre Rohrlich Antonella Ronchi Mark Roschewski Richard Rosenquist Mikhail Roshal Davide Rossi Lisa Roth Alexander Röth Kevin Rouault-Pierre Milad Rouhimoghadam Lubka Roumenina Philippe Rousselot Noemi Roy Jeffrey Rubnitz Ulrich Sachs Laia Sadeghi Brooke Sadler Giuseppe Saglio Shoji Saito Amandeep Salhotra Helmut Salih Gilles Salles David Sallman Takaomi Sanda Vijay Sankaran Miguel Sanz Timothy Satchwell 779

Craig Sauter Sharon Savage Anna Savoia Lydia Scarfo Johannes Schetelig Denis Schewe Charles Schiffer Gary Schiller Alfred Schinkel Uwe Schlegel Richard Schlenk Alvin Schmaier Christoph Schmid Rebekka Schneider Johanna Schneider Dominik Schnerch Heiko Schoder Martin Schreder Andre Schuh Julia Schüler Martin Schumacher Jan Schuringa David Scott Marie Scully Omid Seidizadeh Martina Seiffert John Semple Pierre Sesques John Seymour Mazyar Shadman NIrav Shah Mala Shanmugam Claire Sharpe Bronwen Shaw Vivien Sheehan Ekaterina Shelysheva Sethi Shenon Patricia Shi Qizhen Shi Avichai Shimoni William Shomali Khalid Shoumariyeh David Sibon Jorge Sierra Heinz Sill Sergio Siragusa Susan Slager Soni Smith Melody Smith Kah Teong Soh Eric Solary Antonio Solimando Kevin Song Joo Song Cynthia So-Osman Emmanouil Spanoudakis Silvia Spena Wolfgang Sperr 780

Karsten Spiekermann Luca Spiezia Alessandro Squizzato Fabio Stagno Martin Stanulla Simon Stanworth Jan Stary Jan Stary Anastasios Stathis Deborah Stephens William Stevenson Brian Storrie Catherine Strassel Paolo Strati David Straus Sabine Strehl John Strouboulis Sara Stuart-Smith Marion Subklewe Nathan Subramaniam Daisuke Sugiyama Pierre Sujobert Maria Luisa Sulis Prithu Sundd Katsue Suzuki-Inoue Jakub Svoboda Steven Swerdlow Stephen Sykes Jeff Szer Yu-Tzu Tai Minoru Takata Hideto Tamura Liang Tang Jianguo Tao Samir Taoudi Sarah Tasian Pierfrancesco Tassone Isao Tawara Malcolm Taylor Paul Telfer Hugo ten Cate Arina ten Cate-Hoek Evangelos Terpos Evangelos Terpos Siok Tey Astha Thakkar Kim Theilgaard-Mønch Swee Lay Thein Thomas Thiele Felicitas Thol Christopher Thom Daniel Thomas Margot Thome André Tichelli Anastasia Tikhonova Emanuela Tolosano Daisuke Tomizawa Ciprian Tomuleasa

Mauro Torti Valeria Tosello Ivo Touw Ashley Toye Cindy Toze Lisa Traeger Thai Hoa Tran Alexandra Traverse-Glehen Judith Trotman Eric Tse Leon Tshilolo Henna Tyynismaa Dimitrios Tzachanis Athanasios Tzioufas Hannah Uckelmann Daniela Ungureanu Saad Usmani Anne Uyttebroeck Angelo Vacca Augusto Vaglio Ben Valdez Peter Valent Niels van de Donk Arjan van de Loosdrecht Maarten van den Buuse Vincent Van der Velden Frank van Leeuwen Frits van Rhee Pieter Van Vlierberghe Richard Van Wijk Santosha Vardhana Antiopi Varelias George Vassiliou Adriano Venditti Girish Venkataraman Gwenny Verstappen Yoann Vial Mark Vickers Corinne Vigouroux Diego Villa Martin Villalba Francesca Vinchi Aaron Viny Francesco Violi Conrad-Amadeus Voltin Bastian von Tresckow Maria Teresa Voso Hideo Wada Spencer Waddle Johannes Waldschmidt Carl Walkley Edmund Waller Sa Wang Michael Wang Russell Ware T Warkentin Ralph Wäsch Tove Wästerlid

John Waye Ashutosh Wechalekar Andrew Wei Oliver Weigert Niels Weinhold David Weinstock Guenter Weiss John Welch Robert Welner Clemens Wendtner Jorg Westermann Jason Westin Deborah White Adrian Wiestner Tanya Wildes Brian Wilhelm Bettina Willie Wyndham Wilson Naomi Winick Marcin Wlodarski Sandra Woerner Wilhelm Woessmann Ofir Wolach Daniel Wolff John Wood Bas Wouters Yi Wu Shmuel Yaccoby Hamideh Yadegari Pearlly Yan Zhi-Zhang Yang Yang Yang Feng-Chun Yang Hajime Yasuda Karina Yazdanbakhsh David Yeung Kwee Yong Ken Young Loic Ysebaert Weiping Yuan Marketa Zaliova Sasan Zandi Joanna Zdziarska Clive Zent Thorsten Zenz Yi Zhang X. Long Zheng Chengyun Zheng Jianfeng Zhou Jun Zhou Hong-Hu Zhu Martin Zimmermann James Zimring Gina Zini CM Zwaan

haematologica | 2022; 107(3)

haematologica — Vol. 107 n. 3 — March 2022 —565-780