Haematologica, Volume 106, Issue 12 by Haematologica

haematologica Journal of the Ferrata Storti Foundation

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

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

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

Statistical Consultant Catherine Klersy (Pavia)

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

Managing Director Antonio Majocchi (Pavia)

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

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

haematologica Journal of the Ferrata Storti Foundation

Brief information on Haematologica Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org). Haematologica publishes Editorials, Original articles, Review articles, Perspective articles, Editorials, Guideline articles, Letters to the Editor, Case reports & Case series and Comments. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org). Manuscripts should be submitted online at http://www.haematologica.org/. Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail at www.haematologica.org/content/policies. Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties. Subscription. Detailed information about subscriptions is available at www.haematologica.org. Haematologica is an open access journal and access to the online journal is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org). Rates of the printed edition for the year 2021 are as following: Institutional: Euro 700 Personal: Euro 170 Advertisements. Contact the Advertising Manager, Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, e-mail: marketing@haematologica.org). Disclaimer. Whilst every effort is made by the publishers and the editorial board to see that no inaccurate or misleading data, opinion or statement appears in this journal, they wish to make it clear that the data and opinions appearing in the articles or advertisements herein are the responsibility of the contributor or advisor concerned. Accordingly, the publisher, the editorial board and their respective employees, officers and agents accept no liability whatsoever for the consequences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage, and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published literature.

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

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

haematologica Journal of the Ferrata Storti Foundation

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

haematologica Journal of the Ferrata Storti Foundation

Table of Contents Volume 106, Issue 12: December 2021 About the Cover 3029

Images from the Haematologica Atlas of Hematologic Cytology: hemophagocytic syndrome Rosangela Invernizzi

https://doi.org/10.3324/haematol.2021.279797

Editorials 3030

Pulling the Pin on NPMc+ acute myeloid leukemia Lev M. Kats

https://doi.org/10.3324/haematol.2021.279070

3031

Arsenic and all-trans retinoic acid for acute promyelocytic leukemia: yes, it really is as good as it seems Mark Levis

https://doi.org/10.3324/haematol.2021.278984

3032

Daratumumab: new indications revolving around "off-targets" Yishai Ofran

https://doi.org/10.3324/haematol.2021.279487

Review Article 3034

Interactions of adenoviruses with platelets and coagulation and the vaccine-induced immune thrombotic thrombocytopenia syndrome Paolo Gresele et al.

https://doi.org/10.3324/haematol.2021.279289

Articles Acute Lymphoblastic Leukemia 3046 Multiclonal complexity of pediatric acute lymphoblastic leukemia and the prognostic relevance of subclonal mutations Željko Anticć et al.

https://doi.org/10.3324/haematol.2020.259226

Acute Myeloid Leukemia 3056 Plasmacytoid dendritic cells proliferation associated with acute myeloid leukemia: phenotype profile and mutation landscape Loria Zalmaï et al.

https://doi.org/10.3324/haematol.2020.253740

3067

The IL1-IL1RAP axis plays an important role in the inflammatory leukemic niche that favors acute myeloid leukemia proliferation over normal hematopoiesis Bauke de Boer et al.

https://doi.org/10.3324/haematol.2020.254987

3079

Networking for advanced molecular diagnosis in acute myeloid leukemia patients is possible: the PETHEMA NGS-AML project Claudia Sargas et al.

https://doi.org/10.3324/haematol.2020.263806

3090

A Pin1/PML/P53 axis activated by retinoic acid in NPM-1c acute myeloid leukemia Rita Hleihel et al.

https://doi.org/10.3324/haematol.2020.274878

Acute Promyelocytic Leukemia 3100 Characteristics and outcome of patients with low-/intermediate-risk acute promyelocytic leukemia treated with arsenic trioxide: an international collaborative study Sabine Kayser et al.

https://doi.org/10.3324/haematol.2021.278722

Haematologica 2021; vol. 106 no. 12 - December 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Bone Marrow Transplantation 3107 Use of the HLA-B leader to optimize cord blood transplantation Effie W. Petersdorf et al.

https://doi.org/10.3324/haematol.2020.264424

Cell Therapy & Immunotherapy 3115 PVRIG is a novel natural killer cell immune checkpoint receptor in acute myeloid leukemia Jessica Li et al.

https://doi.org/10.3324/haematol.2020.258574

Chronic Lymphocytic Leukemia 3125 SF3B1-mutated chronic lymphocytic leukemia shows evidence of NOTCH1 pathway activation including CD20 downregulation Federico Pozzo et al.

https://doi.org/10.3324/haematol.2020.261891

Hematopoiesis 3136 Inhibition of the anti-apoptotic protein MCL-1 severely suppresses human hematopoiesis Sheila Bohler et al.

https://doi.org/10.3324/haematol.2020.252130

Iron Metabolism & its Disorders 3149 Cell-specific expression of Hfe determines the outcome of Salmonella enterica serovar Typhimurium infection in mice Manfred Nairz et al.

https://doi.org/10.3324/haematol.2019.241745

Myeloproliferative Disorders

3162

Bone marrow megakaryocytic activation predicts fibrotic evolution of Philadelphia-negative myeloproliferative neoplasms Mattia Schino et al.

https://doi.org/10.3324/haematol.2020.264143

Non-Hodgkin Lymphoma 3170 Phase I study of selinexor in combination with dexamethasone, ifosfamide, carboplatin, etoposide chemotherapy in patients with relapsed or refractory peripheral T-cell or natural-killer/T-cell lymphoma Tiffany Tang et al.

https://doi.org/10.3324/haematol.2020.251454

Plasma Cell Disorders 3176 Expression of the chemokine receptor CCR1 promotes the dissemination of multiple myeloma plasma cells in vivo Mara N. Zeissig et al.

https://doi.org/10.3324/haematol.2020.253526

Red Cell Biology & its Disorders 3188 Danicopan: an oral complement factor D inhibitor for paroxysmal nocturnal hemoglobinuria Antonio M. Risitano et al.

https://doi.org/10.3324/haematol.2020.261826

Letters to the Editor 3198

Daratumumab, an original approach for treating multi-refractory autoimmune cytopenia Etienne Crickx et al.

https://doi.org/10.3324/haematol.2021.279232

3202

Biallelic mutations in the SARS2 gene presenting as congenital sideroblastic anemia Elia Colin et al.

https://doi.org/10.3324/haematol.2021.279138

Haematologica 2021; vol. 106 no. 12 - December 2021 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation 3206

Combined transcriptome and proteome profiling of SRC kinase activity in healthy and E527K defective megakaryocytes Lore De Kock et al.

https://doi.org/10.3324/haematol.2021.279248

3211

Melphalan dose intensity for autologous stem cell transplantation in multiple myeloma Samer A. Srour et al.

https://doi.org/10.3324/haematol.2021.279179

3215

Predicting risk of progression in relapsed multiple myeloma using traditional risk models, focal lesion assessment with PET-CT and minimal residual disease status David Baker, et al.

https://doi.org/10.3324/haematol.2021.278779

3219

Allogeneic hematopoietic cell transplantation outcomes in patients with Richter’s transformation Haesook T. Kim et al.

https://doi.org/10.3324/haematol.2021.279033

3223

BET inhibitors enhance embryonic and fetal globin expression in erythroleukemia cell lines John Z. Cao et al.

https://doi.org/10.3324/haematol.2021.278791

3228

Platelet activation and modulation in thrombosis with thrombocytopenia syndrome associated with ChAdOx1 nCov-19 vaccine Mariangela Scavone et al.

https://doi.org/10.3324/haematol.2021.279345

3232

3236

Lower respiratory tract infection with Staphylococcus aureus in sickle-cell adult patients with severe acute chest syndrome - the STAPHACS Study Alexandre Elabbadi et al.

https://doi.org/10.3324/haematol.2021.278827

3240

https://doi.org/10.3324/haematol.2021.279174

Case Reports 3245

Dominance of an UBA1 mutant clone over a CALR mutant clone: from essential thrombocytemia to VEXAS Mehdi Hage-Sleiman et al.

https://doi.org/10.3324/haematol.2021.279418

3249

https://doi.org/10.3324/haematol.2021.279509

Haematologica 2021; vol. 106 no. 12 - December 2021 http://www.haematologica.org/

ABOUT THE COVER Images from the Haematologica Atlas of Hematologic Cytology: hemophagocytic syndrome Rosangela Invernizzi University of Pavia, Pavia, Italy E-mail: ROSANGELA INVERNIZZI - rosangela.invernizzi@unipv.it doi:10.3324/haematol.2021.279797

emophagocytic syndromes include primary (familial) and secondary conditions characterized by uncontrolled T-cell and macrophage activation and phagocytosis of cells of various hematopoietic lineages by the histiocytes of bone marrow, spleen or lymph nodes. Secondary forms can be reactive to an underlying infection, particularly in immunocompromised hosts. They have also been reported in association with malignancies such as cancer and lymphomas. In this case of Epstein-Barr virus (EBV)-associated hemophagocytic syndrome, the bone marrow aspirate is slightly hypocellular and shows active hematopoiesis (A and B). Many histiocytes are scattered among the hematopoietic cells. They have a low nuclear:cytoplasmic ratio, an oval nucleus with a reticular chromatin pattern, inconspicuous nucleoli and abundant, often vacuolated cytoplasm. Some of them have ingested numerous red cells, erythroblasts, platelets (A-E) and leukocytes (F). These morphological features have diagnostic power.1 Disclosures No conflicts of interest to disclose.

Reference 1. Invernizzi R. Hemophagocytic syndrome. Haematologica. 2020;105(Suppl 1):40-43.

haematologica | 2021; 106(12)

3029

EDITORIALS Pulling the Pin on NPMc+ acute myeloid leukemia Lev M. Kats Peter MacCallum Cancer Centre and Sir Peter MacCallum Department of Oncology, The University of Melbourne, Melbourne, Victoria, Australia. E-mail: LEV M. KATS - lev.kats@petermac.org doi:10.3324/haematol.2021.279070

n this issue of Haematologica, Hleihel et al.1 identify the propyl isomerase Pin1 as a key target of retinoic acid (RA, also known as all-trans retinoic acid) in NPMc+ acute myeloid leukemia (AML). AML is an aggressive cancer with few effective treatment options and extremely poor outcomes in the majority of cases. Approximately one third of AML patients carry mutations in the NPM1 gene that encodes the multifunctional protein nucleophosmin. The mutations, collectively termed NPMc+, cluster at the 3’ end of the NPM1 open reading frame and introduce a nuclear export signal that causes relocalization of nucleophosmin from the nucleolus to the cytoplasm.2 Evidence from clinical trials had suggested that RA treatment may enhance the efficacy of intensive chemotherapy in a subset of NPMc+ patients.3 Excitingly, two concurrent studies in 2015 implicated RA as a degrader of mutant nucleophosmin,4,5 but did not elucidate the molecular target of RA responsible for this effect. In their article, Hleihel and colleagues provide novel mechanistic insights using AML cell lines and patients’ samples. They further demonstrate synergy between RA and chemotherapy or arsenic trioxide (ATO) in NPMc+ AML, with this synergy being dependent on expression of the protein PML. RA is a hormone that, at physiological concentrations, regulates a wide array of biological processes by activating gene expression via retinoic acid receptor (RAR) transcription factors. Seminal studies in the 1980s, initially in vitro and subsequently in clinical trials, identified the potent efficacy of RA against acute promyelocytic leukemia, a subtype of AML most often characterized by the oncogenic fusion protein PML-RARα. Although the molecular details remain debated, at pharmacological concentrations RA promotes both the transcriptional activation of PML-RARα target genes and the degradation of the fusion protein itself, driving differentiation of promyelocytic blasts to mature neutrophils.6 Notably, a study by Wei and colleagues in 2014 uncovered that as well as its effects on RAR signaling, RA is also a potent inhibitor of Pin1,7 a unique enzyme that binds to phosphorylated Ser/ThrPro motifs within target proteins and catalyzes their cis/trans conformation thereby altering their stability or activity. Known Pin1 targets include RARα, PML and PML-RARα, CyclinD1 and NF-kB.8 In the present study, Hleihel et al.1 began by expanding on earlier observations that RA treatment of NPMc+ AML cells leads to NPMc+ proteolysis, P53 activation, differentiation and apoptosis.4,5 They initially tested whether PML, the essential protein component of PML nuclear bodies, is required for RA activity in NPMc+ AML. PML nuclear bodies are small nuclear matrix-associated structures that provide a molecular docking station for a wide array of interacting proteins. Although seemingly dispensable for life (Pml knockout mice are viable), PML nuclear bodies are detectable in most cell types, are regulated by cellular stress and are associated

3030

with numerous biological processes and disease states.9 PML nuclear body formation is dysregulated in NPMc+ AML cells compared with NPM1 wild-type cells, and the authors found that PML knockout in the NPMc+ AML cell line OCI-AML3 abrogates its sensitivity to RA. Analyzing the kinetics of the response to RA by OCI-AML3 cells and NPMc+ primary AML blasts, they unexpectedly found that P53 activation can be untangled from NPMc+ degradation, with the former evident within 2 h of treatment and the latter occurring only after 2448 h. These observations in turn prompted the authors to investigate the role of Pin1 in the RA response. Both RA and a structurally distinct Pin1 inhibitor AG17724 triggered stabilization of PML and P53 proteins, solely in NPMc+ cells. These effects could be abrogated by shRNA-mediated Pin1 knockdown, although it is perhaps surprising that Pin1 knockdown itself is tolerated in the NPM1 mutant context. Importantly, the team found that OCI-AML3 cells and NPMc+ primary AML blasts have increased expression of Pin1 compared with NPM1 wild-type controls, providing a potential explanation for the selective effects of RA on mutant cells. They went on to validate their findings in vivo using an OCI-AML3 xenograft model. As was previously demonstrated in vitro,4,5 RA synergized with both ATO and DNA-damaging chemotherapy, with therapeutic efficacy and NPMc+ degradation dependent on PML expression. Excitingly, two NPMc+ AML patients treated with an RA/ATO combination on a compassionate basis demonstrated a significant albeit incomplete response. Together, the findings reported in this issue by Hleihel et al., as well as earlier work from their group and others demonstrate the potential for expanding the clinical use of RA beyond acute promyelocytic leukemia. The data support a model whereby RA induces multiple anti-leukemic effects in NPMc+ AML cells, most of which are initiated by and dependent on re-assembly of PML nuclear bodies triggered by inhibition of Pin1. A number of important questions do, however, remain. The mechanism of NPMc+ degradation and the significance of this phenomenon for the therapeutic response beyond OCI-AML3 cells are still unclear. Likewise, the role of RAR signaling in potentiating (or opposing) the Pin1/PML/P53 axis, or indeed AML differentiation in the NPMc+ context, has not been explored. Unbiased methodologies such as pooled CRISPR screening could identify essential nodes of the various aspects of the RA response such as PML stabilization, NPMc+ proteolysis, differentiation and cell death. Further validation using genetically engineered mouse models of NPMc+ AML as well as patient-derived xenografts will also be important for building confidence in the strategy. Collectively, these studies would bring us closer to extending the application of a safe existing drug to an area of unmet need.

haematologica | 2021; 106(12)

Editorials

Disclosures LMK has received research funding and consultancy payments from Agios Pharmaceuticals and Celgene Corporation.

References 1. Hleihel R, El Hajj H, Wu H-C, et al. A Pin1/PML/P53 axis activated by retinoic acid in NPM-1c acute myeloid leukemia. Haematologica. 2021;106(12):3090-3099 2. Patel SS, Kluk MJ, Weinberg OK. NPM1 biology in myeloid neoplasia. Curr Hematol Malig Rep. 2020;15(4):350-359. 3. Schlenk RF, Döhner K, Kneba M, et al. Gene mutations and response to treatment with all-trans retinoic acid in elderly patients with acute myeloid leukemia. Results from the AMLSG Trial AML HD98B.

Haematologica. 2009;94(1):54-60. 4. Martelli MP, Gionfriddo I, Mezzasoma F, et al. Arsenic trioxide and all-trans retinoic acid target NPM1 mutant oncoprotein levels and induce apoptosis in NPM1-mutated AML cells. Blood. 2015;125(22):3455-3465. 5. El Hajj H, Dassouki Z, Berthier C, et al. Retinoic acid and arsenic trioxide trigger degradation of mutated NPM1, resulting in apoptosis of AML cells. Blood. 2015;125(22):3447-3454. 6. de Thé H. Differentiation therapy revisited. Nat Rev Cancer. 2018;18(2):117-127. 7. Wei S, Kozono S, Kats L, et al. Active Pin1 is a key target of all-trans retinoic acid in acute promyelocytic leukemia and breast cancer. Nat Med. 2015;21(5):457-466. 8. Zhou XZ, Lu KP. The isomerase PIN1 controls numerous cancer-driving pathways and is a unique drug target. Nat Rev Cancer. 2016;16(7):463-478. 9. Lallemand-Breitenbach V, de Thé H. PML nuclear bodies: from architecture to function. Curr Opin Cell Biol. 2018;52:154-161.

Arsenic and all-trans retinoic acid for acute promyelocytic leukemia: yes, it really is as good as it seems Mark Levis Department of Oncology, Johns Hopkins University, Baltimore, MD, USA E-mail: MARK LEVIS - levisma@jhmi.edu doi:10.3324/haematol.2021.278984

n this issue of Haematologica, Kayser and colleagues report the results of an analysis of outcomes from the National Acute Promyelocytic Leukemia (APL) Observational study (NAPOLEON-Registry; NCT02192619), including 152 non-high-risk APL patients in Germany and France.1 In their study, which they claim represents a reflection of “real-life” outcomes, these authors specifically focused on APL patients treated upfront with all-trans retinoic acid (ATRA) and arsenic, according to the study led by the late Francesco Lo Coco.2 As with that original protocol, this present study excluded high-risk APL patients. When Lo-Coco’s study was published in 2013, the results seemed almost too good to be true.2 The eventfree survival rate of patients treated with ATRA and arsenic was 97%. In their study of the registry patients, Kayser and colleagues found an almost identical result (event-free survival of 95%, with no patient relapsing after achieving remission. The remarkable efficacy of this regimen seems to be every bit as high even outside of the context of a clinical trial. Two out of 152 patients died during induction, and both were older (64 and 70 years) than typical APL patients. Interestingly, differentiation syndrome was only reported in seven patients (6%), in contrast to the 19% reported in Lo-Coco’s study. One wonders whether this is more a reflection of clinicians’ comfort in managing and even preventing this condition as they grow more familiar with this regimen over time. Where to next with APL? Certainly, an oral version of arsenic would expand the use of this combination to many parts of the world lacking access to intravenous medication. It would also represent a major improvement in the quality of life of APL patients, who must trudge through months of daily arsenic infusions. Oral preparations are under investigation,3 but formulation challenges have thus far been an obstacle to their widespread use.

haematologica | 2021; 106(12)

High-risk APL patients were excluded from these studies, and of course they represent a significant challenge for physicians treating them. In one of the original pilot studies exploring the combination of ATRA and arsenic, gemtuzumab ozogamycin (GO) was used as a cytoreductive agent in the high-risk patients.4 This highly effective agent is not approved for such use worldwide, but studies to compare its efficacy against anthracyclines are warranted. Another way to potentially optimize this therapy is to determine how much arsenic is really needed to achieve these high-quality outcomes. The selection of four cycles of consolidation with arsenic was somewhat arbitrary, and no one should lose sight of the fact that arsenic is a group 1 human carcinogen with neurotoxic and hepatotoxic effects.5 Identifying the minimum necessary number of cycles would be a worthwhile endeavor for the field. Finally, lest we be too self-congratulatory about how well we are doing with this dreadful malignancy, let us not forget how many patients die of APL before their disease is recognized and treated.6 At present, in areas of the world that have complete access to standard-of-care leukemia treatment, most APL patients die because their care providers are unknowingly observing the natural history of untreated APL. The failure to recognize APL rapidly is a problem without an immediate solution. However, perhaps in this digital age, there is a ray of hope for this problem. The use of artificial intelligence algorithms combined with digital scanning technology may offer an automated way of identifying an APL patient,7 leading to the same sort of electronic red flag that occurs when a patient with a low electrolyte or platelet count is evaluated by an emergency room physician. We are probably not far off from that future. Disclosures No conflicts of interest to disclose.

3031

Editorials

References 1. Kayser S, Schlenk R, Lebon D, et al. Characteristics and outcome of patients with low-/intermediate-risk acute promyelocytic leukemia treated with arsenic trioxide - an international collaborative study. Haematologica. 2021;106(12):3100-3106. 2. Lo-Coco F, Avvisati G, Vignetti M, et al. Retinoic acid and arsenic trioxide for acute promyelocytic leukemia. N Engl J Med. 2013;369(2):111-121. 3. Zhu HH, Wu DP, Du X, et al. Oral arsenic plus retinoic acid versus intravenous arsenic plus retinoic acid for non-high-risk acute promyelocytic leukaemia: a non-inferiority, randomised phase 3 trial. Lancet Oncol. 2018;19(7):871-879.

4. Estey E, Garcia-Manero G, Ferrajoli A, et al. Use of all-trans retinoic acid plus arsenic trioxide as an alternative to chemotherapy in untreated acute promyelocytic leukemia. Blood. 2006;107(9):3469-3473. 5. Zhou Q, Xi S. A review on arsenic carcinogenesis: epidemiology, metabolism, genotoxicity and epigenetic changes. Regul Toxicol Pharmacol. 2018;99:78-88. 6. Lehmann S, Ravn A, Carlsson L, et al. Continuing high early death rate in acute promyelocytic leukemia: a population-based report from the Swedish Adult Acute Leukemia Registry. Leukemia. 2011;25(7):1128-1134. 7. Sidhom JW, Siddarthan IJ, Lai BS, et al. Deep learning for diagnosis of acute promyelocytic leukemia via recognition of genomically imprinted morphologic features. NPJ Precis Oncol. 2021;5(1):38.

Daratumumab: new indications revolving around "off-targets" Yishai Ofran Department of Hematology, Shaare Zedek Medical Center, Faculty of Medicine, Hebrew University of Jerusalem, Israel E-mail: YISHAI OFRAN - yofran@szmc.org.il doi:10.3324/haematol.2021.279487

he IgG1k monoclonal anti-CD38 antibody daratumumab, approved only 5 years ago, has become a principal agent in the treatment of patients with multiple myeloma. Evidence from multiple trials and realworld experience proved its safety and effectiveness.1 Daratumumab targets CD38, a membrane glycoprotein with various functions. Binding of CD38 to its ligand (CD31) enables plasma cells to interact with surrounding immune and endothelial cells. Additionally, CD38 serves as a dual active enzyme involved in generating and hydrolyzing ADP-ribosyl cyclase and, therefore, affects intracellular calcium signaling and energy metabolism.2 Normal human plasma cells, as well as myeloma cells, express CD38 highly. The anti-myeloma effect of daratumumab is mediated through the elimination of CD38expressing plasma cells. During anti-myeloma therapy, binding of daratumumab to CD38+ natural killer (NK), T and B cells, and erythrocytes leads to "off-target" effects as well as to some common side effects, such as interference with blood product cross-matching or potential immunemodulation through regulatory T-cell elimination. However, the clinical significance of "off-target" effects of daratumumab on CD38+ non-plasma cells is not yet fully characterized. In this issue of Haematologica, Crickx and colleagues3 reported the outcome of eight patients treated with daratumumab for refractory immune thrombocytopenia (ITP) or warm autoimmune hemolytic anemia (AIHA). Patients were struggling with long-lasting diseases, with a median duration of 84.5 months (range, 18–174), refractory to multiple lines of standard therapies. The protocol of daratumumab administration was weekly infusions of 16 mg/kg combined with oral dexamethasone for at least four doses. Three out of five ITP patients and one of two patients with warm AIHA responded. A decrease in gammaglobulin levels was reported, but the autoimmune suppressive effect of daratumumab in these patients most probably went beyond its effect on patients’ normal plasma cells. Notably, in addition to plasma and mature B cells, CD38 is also expressed by T and NK cells, and can also be induced by

3032

interferon and other cytokines. These cells are considered "off-targets" and in multiple myeloma patients treated with daratumumab, a reduction in regulatory T-cell count and expansion of CD4+ and CD8+ T cells were reported.4 Therefore, caution is required with co-administration of daratumumab and checkpoint inhibitors or other immune therapies. Studies are ongoing to confirm the safety of such combinations. Given its multiple targets, predicting which patients with autoimmune diseases will benefit from daratumumab is a challenge. The patients treated by Crickx and colleagues had longlasting ITP or warm AIHA resistant to various lines of therapy. ITP and warm AIHA are antibody-mediated diseases and therefore one can speculate that daratumumab targeted mature B or plasma cells which survived previous lines of therapy. Such a mechanism may apply in other antibody-mediated refractory diseases. Indeed, reports of successful treatments in similar situations are accumulating (Table 1), including daratumumab as a therapeutic option in ABO mismatch-derived post-allogeneic stem cell transplantation hemolysis/cytopenia or pure red cell aplasia, antibody-mediated rejection of transplanted kidney, and even in a refractory case of antiphospholipid syndrome. An attempt to investigate the immune pathophysiology of ITP was made through a comprehensive pathology evaluation of patients’ spleens. CD38 was identified as a prominent marker specifically present in clinically severe cases.5 However, despite broad-range staining for multiple markers, the authors could not definitely confirm that the CD38+ cells were of B-cell or plasma-cell phenotype. The potential activity of daratumumab in targeting T cells or early lymphoid precursors was demonstrated in a preclinical study in mice injected with T-cell acute lymphoblastic leukemia.6 Next came reports of successful treatment of patients with resistant cases of acute lymphoblastic leukemia with daratumumab,7-9 with best and lasting responses achieved in patients treated for minimal residual disease eradication. Interestingly, daratumumab was recently reported to be active in diseases in which the pathological immune response was complicated and

haematologica | 2021; 106(12)

Editorials

Table 1. List of conditions in which daratumumab has been reported to be clinically beneficial.

Condition

Reference

Post-allogeneic stem cell transplantation hemolysis/ cytopenia

Blood. 2016;128:4819 Blood Advances 2018;2(19):2550-2553 British Journal of Haematology 2019;187(2): e48-e51 Pediatric Blood & Cancer 2021;67(1):e28010. Blood Advances 2020;4(5): 815. Molecular and Cellular Pediatrics 2021;8(1):1-7. Frontiers in Immunology 2021;12:444.. New England Journal of Medicine 2018;379(19):1846-1850 American Journal of Hematology 2019;94(8):E216-E219. Bone Marrow Transplantation 2020;55(6):1191-1193. European Journal of Haematology 2020;104(2):145-147 Acta Haematologica 2021;Apr 22;1-5 [Online ahead of print] Transfusion 2019;59:3801-3802 American Journal of Hematology 2020 Jul 11 Annals of Hematology 2021;100(5);1351-1353 Case Reports in Nephrology and Dialysis 2019;9(3): 149-157 Frontiers in Immunology 2021;12:1133 New England Journal of Medicine 2020;383(12):1149-1155. Journal of the American Society of Nephrology 2021;32 (5): 1163-1173

Post-allogeneic stem cell transplantation pure red cell aplasia

Autoimmune hemolysis

Antibody-mediated rejection of transplanted kidney Antiphospholipid syndrome Systemic lupus erythematosus Proliferative glomerulonephritis

involved multiple coordinating cells such as systemic lupus erythematous, and proliferative glomerulonephritis (Table 1). The multiple aberrant immune mechanisms potentially involved make it difficult to identify the exact mechanisms of action of daratumumab in such conditions. Notably, a recent alarming report described that patients with COVID-19 can produce autoantibodies targeting CD38 which lead to exacerbation of immune responses resulting in autoimmune thyroiditis, insulin-dependent diabetes and even exacerbating the cytokine storm and other deleterious responses in COVID-19.10 Daratumumab is an effective anti-myeloma agent with a low toxicity profile. Its prominent effect is elimination of CD38-bearing cells, and in myeloma patients it targets mostly malignant plasma cells. The current report by Crickx et al. suggests that it should be considered as a therapeutic option in refractory cases of ITP and warm AIHA. A proposed mechanism of action is similar to that in myeloma, i.e., elimination of antibody-producing cells, but since CD38 is presented by many other immune cells, potential ‘off-target’ effects cannot be ruled out. Daratumumab's potential effectiveness against T-cell acute lymphoblastic leukemia is to be investigated in a future, planned, prospective study. The work by Crickx et al. is a step forwards in recognizing the potential role of daratumumab in autoimmune conditions. However, this treatment should be used with caution because its effect on multiple arms of the immune system may lead to paradoxical responses.

haematologica | 2021; 106(12)

Disclosures No conflicts of interest to disclose.

References 1. Kobayashi H, Tsushima T, Terao T, et al. Evaluation of the safety and efficacy of daratumumab outside of clinical trials. Int J Hematol. 2019;109(6):665-672. 2. van de Donk NW, Janmaat ML, Mutis T, et al. Monoclonal antibodies targeting CD38 in hematological malignancies and beyond. Immunol Rev. 2016;270(1):95-112. 3. Crickx E, Audia S, Robbins S, et al. Daratumumab, an original approach for treating multi-refractory autoimmune cytopenia. Haematologica. 2021;106(1):3198-3102. 4. Krejcik J, Casneuf T, Nijhof IS, et al. Daratumumab depletes CD38+ immune regulatory cells, promotes T-cell expansion, and skews Tcell repertoire in multiple myeloma. Blood. 2016;128(3):384-394. 5. Furudoï A, Rivière É, Lazaro E, Furudoï E, Viallard JF, Parrens M. Adult primary immune thrombocytopenia. Am J Surg Pathol. 2018;42(3):401-412. 6. Bride KL, Vincent TL, Im SY, et al. Preclinical efficacy of daratumumab in T-cell acute lymphoblastic leukemia. Blood 2018:131(9): 995-999. 7. Ofran Y, Ringelstein-Harlev S, Slouzkey I, et al. Daratumumab for eradication of minimal residual disease in high-risk advanced relapse of T-cell/CD19/CD22-negative acute lymphoblastic leukemia. Leukemia. 2020:34(1):293-295. 8. Ganzel C, Kharit M, Duksin C, et al. Daratumumab for relapsed/refractory Philadelphia-positive acute lymphoblastic leukemia. Haematologica. 2018;103(10):e489-e490. 9. Bonda A, Punatar S, Gokarn A, et al. Daratumumab at the frontiers of post-transplant refractory T-acute lymphoblastic leukemia - a worthwhile strategy? Bone Marrow Transplant. 2018;53(11):14871489. 10. Wang EY, Mao T, Klein J, et al. Diverse functional autoantibodies in patients with COVID-19. Nature. 2021;595(7866):283-288.

3033

REVIEW ARTICLE Ferrata Storti Foundation

Interactions of adenoviruses with platelets and coagulation and the vaccine-induced immune thrombotic thrombocytopenia syndrome Paolo Gresele,1 Stefania Momi,1 Rossella Marcucci,2 Francesco Ramundo,3 Valerio De Stefano3 and Armando Tripodi4 Department of Medicine and Surgery, Section of Internal and Cardiovascular Medicine, University of Perugia, Perugia; 2Department of Experimental and Clinical Medicine, University of Florence; Atherothrombosis Center, AOU Careggi, Florence; 3Section of Hematology, Department of Radiological and Hematological Sciences, Catholic University, Fondazione Policlinico A. Gemelli IRCCS, Rome and 4Fondazione IRCCS Ca' Granda, Ospedale Maggiore Policlinico, Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, Milan, Italy 1

Haematologica 2021 Volume 106(12):3034-3045

ABSTRACT

Correspondence: PAOLO GRESELE paolo.gresele@unipg.it Received: May 24, 2021 Accepted: August 4, 2021. Pre-published: August 19, 2021. https://doi.org/10.3324/haematol.2021.279289

he COVID-19 pandemic has had a heavy impact on global health and economy and vaccination remains the primary way of controlling the infection. During the ongoing vaccination campaign some unexpected thrombotic events have emerged in subjects who had recently received the AstraZeneca (Vaxzevria) vaccine or the Johnson&Johnson (Janssen) vaccine, two adenovirus vector-based vaccines. Epidemiological studies confirm that the observed/expected ratio of these unusual thromboses is abnormally increased, especially in women in fertile age. The characteristics of this complication, with venous thromboses at unusual sites, most frequently in the cerebral vein sinuses but also in splanchnic vessels, often with multiple associated thromboses, thrombocytopenia, and sometimes disseminated intravascular coagulation, are unique and the time course and tumultuous evolution are suggestive of an acute immunological reaction. Indeed, plateletactivating anti-PF4 antibodies have been detected in a large proportion of the affected patients. Several data suggest that adenoviruses may interact with platelets, the endothelium and the blood coagulation system. Here we review interactions between adenoviral vectors and the hemostatic system that are of possible relevance in vaccine-associated thrombotic thrombocytopenia syndrome. We systematically analyze the clinical data on the reported thrombotic complications of adenovirus-based therapeutics and discuss all the current hypotheses on the mechanisms triggering this novel syndrome. Although, considering current evidence, the benefit of vaccination clearly outweighs the potential risks, it is of paramount importance to fully unravel the mechanisms leading to vaccineassociated thrombotic thrombocytopenia syndrome and to identify prognostic factors through further research.

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

3034

Introduction The coronavirus disease 19 (COVID-19) pandemic has prompted an unprecedented effort to develop highly effective vaccines to prevent further spreading of the infection, the associated mortality and the enormous strain on healthcare systems. Indeed, in a previously unimaginable short time, many vaccines have been developed. Several of them underwent controlled randomized phase III clinical trials and, as of 22 June, 2021, 13 have been licensed globally for clinical use. By July 18, 2021 they had been administered to more than 1.9 billion subjects worldwide (923 million of whom are fully vaccinated; 3.66 billion doses have been administered globally; 26.3% of the world’s population has received at least one dose of a COVID-19 vaccine). This represents the most massive vaccination campaign ever undertaken

haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

(https://www.who.int/emergencies/diseases/novel-coronavirus2019/covid-19-vaccines; https://ourworldindata.org/covid-vaccinations). Although careful scrutiny of vaccine safety in controlled randomized phase III clinical trials did not highlight significant thrombotic risks, exceedingly rare events may have been missed and indeed during the vaccination campaign several cases of thrombosis, in particular thrombotic events at unusual sites associated with thrombocytopenia, were reported. Most events occurred in subjects who had received the ChAdOx1 (Vaxzevria) vaccine in the preceding weeks, but more recently several cases have also been reported following the Ad26-CoV2S Johnson&Johnson (Janssen) vaccine.1-9 Not only was the observed/expected ratio of these thromboses abnormally high in subjects receiving the Vaxzevria vaccine, but the clinical characteristics of the events were unique, associating unusual site venous thromboses, mainly cerebral vein sinus thrombosis (CVST), with thrombocytopenia and sometimes disseminated intravascular coagulation (DIC). In contrast, no thromboses were reported in about 90 million subjects who had received the messenger RNA (mRNA)-based Pfizer BioNTech and only very few in those who had received the Moderna vaccine (Spikevax), although the latter had characteristics apparently dissimilar from those observed in Vaxzevria recipients, with one exception.10 These findings suggest that the reported thrombotic complications, which have variously been called vaccineinduced prothrombotic immune thrombocytopenia (VIPIT), vaccine-induced immune thrombotic thrombocytopenia (VITT), thrombotic thrombocytopenia syndrome (TTS) and vaccine-associated thrombotic thrombocytopenia syndrome (VATTS),2,11-13 are peculiar to adenoviral (Ad) vector-based vaccines and have led to limitations and/or temporary suspensions of the use of such vaccines in several countries. From the most recently available UK pharmacovigilance data (July 7, 2021), CVST and other major thromboembolic events with concurrent thrombocytopenia had been reported in 147 (average age, 54 years) and 258 subjects (average age, 54 years), respectively, among an estimated 24.6 million recipients of a first dose and an estimated 22.3 million recipients of a second dose of the Vaxzevria vaccine. Thus, the overall incidence after first or unknown doses was 14.8 cases per million doses in the UK (https://www.gov.uk/government/publications/coronaviruscovid-19-vaccine-adverse-reactions/coronavirus-vaccine-summary-of-yellow-card-reporting). Concerning Europe, as of June 27, 2021, there were spontaneous reports to EudraVigilance of 479 suspected cases, 100 of which had had a fatal outcome, among recipients of about 51.4 million doses of Vaxzevria, i.e. 19.3 cases per million doses (https://www.ema.europa.eu/en/documents/covid-19-vaccinesafety-update/covid-19-vaccine-safety-update-vaxzevria-previously-covid-19-vaccine-astrazeneca-14-july-2021_en.pdf), and 21 cases of suspected TTS associated with the Janssen COVID-19 vaccine, four of which were fatal, among recipients of about 7 million doses of this vaccine, i.e. 3 cases per million doses (https://www.ema.europa.eu/en/documents/covid-19-vaccine-safety-update/covid-19-vaccine-safetyupdate-covid-19-vaccine-janssen-14-july-2021_en.pdf). This review aims to discuss the interactions between Ad vectors and Ad-based vaccines and the hemostatic system and the hypotheses on the mechanisms triggering VITT. haematologica | 2021; 106(12)

Adenoviruses, platelets and the blood coagulation system Based on available data and given that VITT has been associated with Ad-vector-based vaccines, hypotheses on a direct role of the interaction between Ad and blood components can be made. Ad are non-enveloped DNA viruses with a nucleoprotein core encapsulated by an icosahedral protein capsid from which proteinaceous fibers protrude. The C-terminal knob domain at the distal end of these fibers is responsible for virus binding to its primary cellular receptor, a 46-kDa transmembrane protein14-16 which also functions as a receptor for Coxsackie B virus and is, therefore, called coxsackie and Ad receptor (CAR).15-17 The high affinity binding of Ad to CAR starts receptor-mediated endocytosis.18 Moreover, Ad have evolved other mechanisms to facilitate cell entry via recognition of the arginine-glycine-aspartate (RGD) sequence on cell surface integrins. Molecules expressed on host cell surfaces involved in cell infection include the vitronectin-binding integrins α b and α b ,19 the fibronectin-binding integrin α b 20 and others, such as α b ,21 all characterized by a common RGD peptide sequence which is recognized by the RGD ligand in the HI fiber knob loop of the Ad penton base protein. Although the CAR is expressed in almost all tissues, including the adult nervous system and cerebral vasculature,22,23 muscle,24 heart25 and the hematopoietic system,26 its presence in platelets is debated. Othman et al. identified CAR (by flow cytometry) and its mRNA (by reverse transcriptase polymerase chain reaction) in human platelets27 while Shimony et al. did not confirm the presence of the receptor and proposed that binding of Ad to platelets is mediated by an interaction between RGD-binding motifs of Ad and platelet α b 28 (Figure 1). Indeed, human megakaryocytes either do not express mRNA for CAR or express it at extremely low levels (J. Rowley and A.S. Weyrich, University of Utah, personal communication). After intravenous inoculation in mice, Ad rapidly bind circulating platelets causing their activation and subsequent entrapment in liver sinusoids where virus-platelet aggregates are taken up by Küpffer cells and degraded. Platelet activation is followed by activation of blood coagulation, leading to DIC.29 Activated platelets also release cytokines promoting endothelial cell activation with secretion of von Willebrand factor, binding of platelets to endothelial cells and the formation of platelet/leukocyte aggregates, eventually triggering the development of microthrombi in liver sinusoids.21,29 There is also a complex interplay between Ad and the coagulation system. In fact, the distribution and activity of Ad in blood is affected by interactions with plasma proteins, including complement and vitamin K-dependent coagulation factors, which act as opsonizing agents. Our knowledge of these interactions derives mainly from in vitro observations and it is unknown whether the interplay of Ad with coagulation proteins affects the activity of the latter. Vitamin K-dependent coagulation factors, including the anticoagulant protein C, interact with Ad-5, the most widely used Ad vector. Activated protein C is generated on endothelial cells via the interaction of protein C with the thrombin-thrombomodulin complex and the endothelial protein C receptor (EPCR). Activated protein C requires protein S to express anticoagulant activity.30 Protein S circulates either free or associated with C4BP, a v

3035

P. Gresele et al.

Figure 1. Hypothesized interactions between platelets and adenoviruses. Adenoviruses (Ad) induce platelet activation either by binding to platelet coxsackie and adenovirus receptor (CAR) or to plateletsurface integrins, such as α b or α b . Moreover, circulating Ad-elicited IgG or immune complexes may directly activate platelets through FcgRIIa. Gas6 exposed by cerebral vein endothelial cells may bind Ad and activate platelets acting on Tyro3, Axl and Mer (TAM) receptors. Ad may also bind CAR expressed by cerebral vein vessels in this way activating endothelial cells which in turn may elicit platelet activation. v

regulatory protein of the complement system.31 C4BP binds to activated platelets through mechanisms involving chondroitin sulfate expressed on activated platelets32 and to membrane-associated protein S on platelets.33 Interestingly, the protein C anticoagulant pathway plays a peculiar pathophysiological role in CVST.34 Small but measurable amounts of EPCR are also found in plasma. Soluble EPCR binds both protein C and activated protein C with an affinity similar to that of membrane-bound EPCR35 but, in contrast to the latter, it inhibits activated protein C anticoagulant activity thus limiting its ability to inactivate activated factor V, and also binds protein C impeding its activation by thrombinthrombomodulin complexes.32,36 An increase in soluble EPCR was observed in CVST, possibly leading to a procoagulant condition and enhanced risk of thrombosis.37 Finally, Gas6 (encoded by the growth arrest-specific 6 gene), a vitamin K-dependent protein with 44% sequence homology with protein S but devoid of anticoagulant activity, is widely expressed in the cerebral nervous system where it is found on resting endothelial cells. Gas6 potentiates platelet activation acting on Tyro3, Axl and Mer (TAM) receptors leading to thrombus formation38 and in vitro studies have shown that Gas6 binds to Ad enhancing their gene expression.39 The affinity of different Ad for coagulation factors is variable, with a considerable number of Ad types unable to bind them. Ad-5, Ad-2 and Ad-16 bind strongly to factor X.40 Moreover, the ability of Ad to bind coagulation factors is species-specific, e.g., Ad-5 binds human and mouse factor X with similar affinity, but Ad-2 binds human factor X with 10-fold lower affinity than mouse factor X.41 Vitamin K-dependent coagulation factors VII, IX, X, and protein C mediate the binding of Ad to hepatocytes.42-44 For instance, for Ad-5 hepatotropism is critically dependent on the ability of the Ad-5 hexon to bind factor 3036

X. In contrast, non-factor X-binding Ad, such as Ad-48 and Ad-26, do not show hepatocyte tropism.45 The primary reason why factor X is required for Ad-5 transduction to the liver is that it protects Ad-5 from attack by complement.46 It has been previously ascertained that components of intramuscularly-injected vaccines, including the Ad vector, are disseminated in the circulation47 and it is thus conceivable that some of the above described activating interactions between Ad and platelets, endothelium and the blood clotting system can occur in recipients of Advector-based vaccines. However, so far no experimental evidence that this may have a role in VITT is available and actually it seems unlikely that sufficiently high circulating levels of a non-replicating Ad vector may be reached to trigger platelet activation or blood coagulation changes. In fact, it should be considered that around 2,500 billion virions/kg are required to trigger this reaction in mice and non-human primates,29,48 and even if all the Vaxzevria viral content were to spill-over into the blood after intramuscular administration, a concentration of 0.7 billion/kg Ad viral vectors would be reached, which is probably insufficient to activate platelets/coagulation.49

Antibody-dependent enhancement and vaccine-associated adverse events Antibody-dependent enhancement (ADE) is an immunological form of a more general phenomenon called enhanced respiratory disease, leading to the clinical worsening of respiratory viral infections. ADE can occur either through an antibody-mediated increase of virus uptake by Fcg receptor IIa (FcgRIIa)-expressing phagocytic cells, thus facilitating viral infection and replication, or by boosting immune activation through excessive Fc-mediated immunological cell effector functions or immune haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

complex formation with consequent increase of inflammation and immunopathology.50 Both ADE pathways can occur when non-neutralizing antibodies or antibodies at sub-neutralizing levels bind to viral antigens without blocking or clearing the infection. ADE has been reported for vaccines against both severe acute respiratory syndrome corona virus (SARS-CoV) and Middle East respiratory syndrome corona virus (MERS-CoV) in vitro and in animal models.50 The cytoplasmic tail of FcgRIIa activates the protein-tyrosine kinases Src 51,52 and Syk.53-55 Srcdependent signaling has been shown to be crucial for ADE triggered by Ebola virus, enhancing viral uptake into cells and thus worsening the infection.56 Circulating antibodies activating platelet IgG FcgRIIa may be key determinants of a host response leading to uncontrolled platelet aggregation and thrombosis. Studies in transgenic mice expressing human FcgRIIa on platelets showed that the administration of anti-CD9 antibodies caused thrombosis accompanied by platelet consumption, a response that was absent in mice lacking the receptor.57 The clinical relevance of this pathway for thrombotic disorders in humans is confirmed by the observation that FcgRIIa expression is higher in patients with stroke58 and that relatively common FcgRIIa polymorphisms are associated with increased risk of thrombosis in patients with heparin-induced thrombocytopenia (HIT).59 Immuno-complex formation, complement deposition and local immune activation are likely mechanisms triggered by SARS-CoV-2 vaccine. Furthermore, preexisting antibodies to coronavirus strains endemic in humans could mediate ADE by facilitating cross-reactive recognition of SARS-CoV-2 in the absence of viral neutralization.60 Interestingly, compared to Ad-5 and Ad-6, chimpanzee adenoviruses (ChAd) are much less frequently neutralized by pre-existing antibodies present in humans. The prevalence of vector-neutralizing antibodies against Y25, now renamed ChAdOx1, the vector of the Vaxzevria vaccine, in human sera from British and Gambian adults was found to be 0% (n=100) and 9% (n=57), respectively.61 The presence of these antibodies in rare patients in Europe might theoretically represent one potential mechanism triggering ADE, and possibly VITT, in vaccine recipients but no data on this are available yet. Despite the above hypotheses, preliminary in vitro evidence suggests that serum from convalescent COVID-19 patients does not induce either enhancement of SARSCoV-2 infection or innate immunity responses in human macrophages, indicating that ADE may not be involved in the immune-pathological processes associated with COVID-19 infection or immunization.62

Use of adenovirus vectors and thrombotic events Adenovirus vectors for gene therapy Ad vectors have been used therapeutically for their ability to transduce and deliver transgenes to different cell types. However, for these indications the clinical use of Ad vectors has been limited to a few tens of patients and the main concerns have been the development of humoral and cellular immunity occurring upon repeated administration and/or the possible neutralization of the vector by pre-existing immunity against the virus, while haematologica | 2021; 106(12)

little attention had been paid to the possible interactions of Ad vectors with platelets and the blood clotting system. The first use of Ad vectors for gene therapy of inherited disorders or to treat neoplasia dates back to the 1990s. An analysis of the risks associated with the use of Ad-vectored gene therapies among 90 individuals who received 140 administrations for various diseases (cystic fibrosis, metastatic colorectal cancer, cardiovascular disease), showed that 13 deaths were recorded. The authors concluded that none was linked to the Ad vector.63 The reported hematologic abnormalities were decreased hemoglobin, leukocytosis, thrombocytopenia, and prolongation of the activated partial thromboplastin time (aPTT), with no cases of DIC.63 It is, however, puzzling that a recently European Medicines Agency-licensed Ad-vectored gene therapy for spinal muscular atrophy received a warning about the possible risk of thrombotic microangiopathy based on the reporting of five cases in treated infants (https://www.ema.europa.eu/en/medicines/human/EPAR/zolgensma).

Adenovirus-vectored vaccines Beside SARS-CoV-2, Ad vectors have been used for the preparation of other vaccines, including the ChAdOx1vectored vaccines for MERS-CoV and Chikungunya; with only a few hundred volunteers having received these vaccines up to June 2020,64 no excess of thrombotic events had been noted.65 Even for the Ebola vaccination campaign, the largest previous example of large-scale vaccination using an Ad vector, a maximum of around 200,000 volunteers were treated, with only one vena cava thrombosis reported (Table 1). However, it may be extremely difficult to prove that adverse events following immunization are caused by the vaccine itself when their occurrence is extremely rare (https://www.nature.com/articles/d41586-021-00880-9. Accessed on April 9, 2021). Except for common mild/moderate reactome reactions, the most frequently recorded adverse events in clinical trials were hematologic (e.g., mild hemoglobin decrease, thrombocytopenia, leukopenia) the majority of which recovered a few days or weeks after vaccination. The extent and rate of hematologic adverse events associated with Ad-vectored vaccines are summarized in Table 1. Occasional abnormalities of coagulation were reported, with prolongation of the aPTT, possibly due to the development of transient antiphospholipid antibodies. Thrombotic events were rare both for human and nonhuman Ad-vectored vaccines. One case of phlebitis was observed among 114 volunteers who received a recombinant, replication-defective Ad-5-vectored vaccine expressing human immunodeficiency virus (HIV)-1 antigenic proteins.66 Another case of deep vein thrombosis was observed among 58 volunteers after administration of a recombinant, replication-defective Ad-35-vectored vaccine expressing HIV-1 antigens.67 Both events were considered unrelated to the vaccine. A systematic review identified 200 clinical studies on active immunization against SARS-CoV-2. The second most used vaccine platform, after mRNA-based vaccines, was represented by Ad vectors (24%).68 Concerning chimpanzee Ad-vectored vaccines (ChAdOx1 nCoV-19), neutropenia was the most common hematologic abnormality (Table 1). Across all studies, vaccines had a good 3037

P. Gresele et al.

Table 1. Studies with adenovirus-vectored vaccines reporting hematologic adverse effects.

Adenoviral vector

Pathogens

Chimpanzee Adenovirus Influenza A

Study (type of)

N. of Thrombocytopenia Venous Coagulation participants (n) thromboembolism disorders (n)

Phase Ia, dose-escalation (S1) MERS Phase I, dose-escalation, ChAdOx1 non-randomized, uncontrolled (64) SARS-CoV-2 Phase I/II, single-blind, randomized, controlled (70) Ebola Phase I, dose-escalation, open-label (S2) Phase I/IIa, double-blind, placebo-controlled, dose-finding (S3) Phase I, dose-escalation, open-label (S4) ChAd3 Phase II, randomized, observer-blind, placebo-controlled (S5) Phase II, randomized, observer-blind, placebo-controlled (S6) RSV Phase I, open-label, single-site, dose-escalation (S7) Human Adenovirus HIV Phase I, double-blind, randomized, placebo-controlled (66) Phase I, double-blinded, placebo-controlled (S8) Phase IIb, double-blind, randomized, controlled (S9) Ad5 Ebola Phase I, randomized, double-blind, placebo-controlled (S10) Phase I, single-site, double-blind, randomized, placebo-controlled, dose-escalation (S11) SARS-Cov2 Phase I, single-centre, dose-escalation, double-blind, non-randomized (S12) SARS-CoV-2 Phase III, randomized, double-blind, placebo-controlled (71) Ad26 Ebola Phase II, randomized, double-blind, placebo-controlled (S13) HIV Phase I, double-blind, randomized, placebo-controlled (67) Ad35 Plasmodium Phase Ib, randomized, Falciparum controlled, double-blind, dosage-escalation (S14) Phase I, randomized, placebo-controlled, dose-escalation (S15) Ad26-Ad5 SARS-CoV-2 Phase III, non-randomized, single-center (72)

Other hematologic complications

Other systemic adverse events

Leukopenia

Anemia, neutropenia, lymphopenia

Fatigue, malaise, headache Fatigue, headache, myalgia

1077

Neutropenia

Fatigue, headache

Leukopenia

Fever

120

aPTT prolongation (15%) aPTT prolongation (n=1)

Fatigue, malaise, headache

3030

Anemia, lymphopenia, neutropenia Leukopenia, eosinophilia Anemia

Fatigue, headache, myalgia Fever, headache

600

Anemia

Fever, headache

Anemia

Fatigue, headache, myalgia, nausea

114

Phlebitis (n=1)

Leukopenia, anemia

Fatigue, malaise, headache

Neutropenia

801

120

Neutropenia, anemia Anemia, leukopenia

Fever, malaise, myalgia, chills Headache, malaise, myalgia Fever

aPTT prolongation (n=2)

Malaise, myalgia, headache, chills

108

44325

DVT (n=6), PE (n=4), CSVT (n=1)

Fever, fatigue, headache, muscle pain Fatigue, headache, myalgia, nausea

423

Anemia, neutropenia

Fatigue, headache, myalgia, chills

DVT (n=1)

Malaise, myalgia, headache, chills

WBC count abnormalities

Neutropenia, eosinophilia

21977

DVT (n=1)

Myalgia, chills, headache, fever, vomiting Fever, headache, malaise, myalgia, nausea Fever, headache, fatigue, myalgia

aPTT prolongation (n=4) Vena cava thrombosis NR (n=1)

The reference numbers preceded by an S refer to references listed in the Online Supplementary Material. The adverse events reported are related only to the vaccinated groups and not to the placebo groups. MERS: Middle East respiratory syndrome virus; SARS-CoV-2: severe acute respiratory syndrome coronavirus 2; RSV: respiratory syncytial virus; HIV: human immunodeficiency virus; NR: not reported; DVT: deep vein thrombosis; PE: pulmonary embolism; CSVT: cerebral sinus venous thrombosis; aPTT: activated partial thromboplastin time; WBC: white blood cell.

3038

haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

safety profile with no difference in severe reactions between study arms.69,70 In a phase III trial with a recombinant, replication-incompetent human Ad-26 vector encoding the SARS-CoV-2 spike protein, with 43,783 participants, 11 venous thromboembolic events were observed in the vaccine group compared to three in the placebo group (Table 1); however, most subjects had underlying medical conditions that might have contributed to these events. In the vaccine group there were six cases of lower leg deep venous thrombosis and four cases of pulmonary embolism. Interestingly, however, a CVST, with cerebral hemorrhage and thrombocytopenia, occurred 21 days after vaccination in a 25-year-old male who had multiple predisposing factors, including preexisting cerebral sigmoid sinus stenosis and infection from an unknown pathogen. Subsequent testing identified anti-PF4 antibodies at the time of the event. The patient recovered.71 In a phase III controlled randomized clinical trial with a recombinant Ad-26-vectored and a recombinant Ad-5vectored vaccine (Sputnik V) among 16,501 participants, ten vascular events (0.061%) were observed including: one deep vein thrombosis (0.006%), one transient ischemic attack (0.006%), one cerebral circulation failure (0.006%), one vascular encephalopathy (0.00659%) and two acute myocardial infarctions (0.012%) (4 additional events were non thrombotic)72 (Table 1).

The vaccine-induced immune thrombotic thrombocytopenia syndrome When the anti-SARS-CoV-2 vaccination campaign was well underway worldwide a few cases of spontaneous, severe thromboembolic events in otherwise healthy subjects began to be reported, leading to a pause in the administration of the Vaxzevria vaccine in several European countries (https://www.ema.europa.eu/en/news/emas-safetycommittee-continues-investigation-covid-19-vaccine-astrazenecathromboembolic-events). Soon after several case reports were published, mainly concerning young females, with new ones continuing to accrue, although many were not subject to rigorous central review and with anti-PF4 antibodies measured using disparate methods, not allowing to conclude that all were typical VITT cases Up to July 17, 2021, 105 such cases with two Ad-vectored vaccines had been published (Table 2) with some common clinical features characterizing a new syndrome, including thrombocytopenia, often severe, venous thrombosis at unusual sites, in particular of the cerebral sinuses but also of the splanchnic veins, frequently associated with thromboses in multiple sites, both venous and arterial, and sometimes DIC combined with hemorrhage. A comparative evaluation of the clinical characteristics of the published ChAdOx1 or Ad26.CoV2.S VITT cases suggests that while clinical symptoms are comparable, Ad26.CoV2.S-associated cases show more thrombosis and intracerebral hemorrhage, lower D-dimer and less altered aPTT, but a similar mortality.9 In a recent, large nationwide healthcare register-based study in Denmark and Norway involving 281,264 ChAdOx1-S-vaccinated subjects aged 18-65 and as controls the entire agematched populations of the two countries studied in the period 2016-2019, the standardized morbidity ratio for CVST was 20.25 (8.14-41.7), with an excess of 2.5 events haematologica | 2021; 106(12)

per 100,000 vaccinations, particularly evident in women 18-44 years old,73 confirming the crucial relationship between Vaxzevria administration and occurrence of VITT. The catastrophic syndrome, burdened by a 20-50% mortality rate, has the time course and tumultuous evolution of an acute immunological reaction and indeed three groups of investigators identified, in several of their patients, circulating antibodies to PF4/heparin complexes using an enzyme-linked immunosorbent assay (ELISA) and a heparin-induced platelet activation assay,2-4 and thus proposed that this disorder is a peculiar form of autoimmune HIT.

The autoimmune heparin-induced thrombocytopenia hypothesis HIT is a rare immune-mediated adverse drug reaction that may occur after exposure to heparin. Circulating heparin binds to PF4, a positively charged platelet protein released in plasma upon activation. PF4 normally binds to negatively charged glycosaminoglycans on the endothelium, displacing antithrombin and thus activating coagulation. However, PF4 binds with greater affinity to heparin, forming heparin/PF4 complexes which become neoantigens inducing the formation of autoantibodies. HeparinPF4-IgG immune complexes in turn bind to platelet FcgRIIA receptors causing activation, aggregation, and additional release of PF4, with ignition of a positive feedback loop leading to further platelet activation and consumption. Moreover, these complexes also activate monocytes, which release tissue factor, thus promoting concomitant activation of coagulation. HIT is a potentially fatal condition, associated with the development of arterial or venous thrombosis.74 Thrombocytopenia occurs in more than 85% of HIT patients and is usually of moderate severity, with median platelet counts of approximately 50-60×109/L, although values <20×109/L can be found in approximately 10% of cases. Typically, the platelet count starts to decrease 5-10 days after initiation of heparin, but early-onset thrombocytopenia (rapid-onset), within 24 h of exposure, can develop in 25-30% of cases if patients have been treated with heparin in the preceding 3 months.75 Thromboembolic complications occur in 35-75% of HIT patients and are usually severe. They can be venous (i.e., deep vein thrombosis and pulmonary embolism, but rarely also CVST or splanchnic thrombosis), arterial (ischemic stroke, myocardial infarction, acute occlusion of limb arteries) or microvascular (digital infarction).76 Recently, another clinical picture not triggered by exposure to heparin has been recognized and defined as autoimmune HIT.77 The main characteristic of this condition is the presence of circulating antibodies able to activate platelets also in the absence of heparin. Polyanion molecules potentially involved in the development of autoimmune HIT are typically bacteria and virus components, hypersulfated chondroitin sulfate, DNA and RNA and polyphosphates.78 Patients with this syndrome show slightly different clinical features from those with classical HIT, including severe thrombocytopenia (<20x109/L), sometimes in combination with DIC, microvascular thrombosis and CVST in up to 40% of cases.78 From a therapeutic standpoint, besides the indication for an alternative anticoagulant, valid also for HIT, the intravenous 3039

P. Gresele et al.

Table 2. Cases of vaccine-induced immune thrombotic thrombocytopenia reported in the literature as of July 17, 2021.

Vaccine

VITT cases, n

ChAdOx-1 ChAdOx-1 11 ChAdOx-1 5 ChAdOx-1 23 ChAdOx-1 2 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 3 ChAdOx-1 1 ChAdOx-1 5 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 2 ChAdOx-1 7d ChAdOx-1 2 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 4f ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 1 ChAdOx-1 2 ChAdOx-1 3 Ad26.COV2.s Ad26.COV2.s 1 Ad26.COV2.s 12 Ad26.COV2.s 1 TOTAL 105

Female, n (%)

Age range, years

CVST, n/total

Platelet count nadir-range x109/L

Negative anti PF4, n (total)

Heparin-treated with success, n/treated

Time form References vaccination range, days

9(82%) 4(80%) 14(61%) 0(0%) 1(100%) 1(100%) 1(100%) 3(100%) 1(100%) 5(100%) 0(0%) 1(100%) 0(0%) 1(50%) 4(57%) 0(0%) 1(100%) 1(100%) 1(100%) 3(75%) 1(100%) 1(100%) 0/1(0%) 0/1(0%) 0/1(0%) 1(100%) 0(0%) 1(100%) 0(0%) 0(0%) 0(0%) 0(100%) 2(100%) 2(66.6)%

22-49 32-54 21-77 25-32 55 51 60 22-46 54 41-67 63 69 50 37-50 24-53 25-32 62 30 30 29-50 41 36 51 38 50 35 54 36 44 27 63 30 24-39 53-61

9/11 4/5 13/22a 2/2 0/1c 0/1 0/1 3/3 1/1 1/5 0/1 1/1 1/1 1/2 4/7 2/2 0/1e 1/1 0/1 1/4 1/1 1/1 1/1 1/1 1/1 1/1 1/1 0/1 0/1 1/1 0/1 1/1 2/2 2/3

8-107 10/70 7/113 7/17 30 37 5 60-92 NR 12-105 36 18 15 7-9 8/71 19-30 19 56 72 24-110 36 94 NRh 14 15 50 34 133 6 68 36 37 29-36 21-25

0/9 0/5 1/23 0/1 1/1 0/0 0/1 0/3 0/0 0/5 0/1 0/1 0/0 0/2 0/7 1/1 0/1 0/1 0/1 0/4 0/1 0/0 1/1 0/1 0/1 0/1 0/1 0/1 0/1 0/1 0/1 0/1 0/2 0/3

1/5 2(3)/5 NRb 0/1 0/1 1/1 0/1 0/3 3/3 0/1 (1) 0/1 0/0 0/2 NR 0/1 0/0 0/1 1/1 0/1g 0/0 0/1 1/1 0/0 0/0 0/0 0/0 0/1g 0/0 0/0 0/0 0/0 0/2 0/0g

5-16 5-30 6-24 6-9 10 11 7 4-17 12 5-11 20 11 11 10 6-20 6-9 8 8 8 7-20 7 14 6 8 7 14 21 17 8 2 20 7 8-12 10-16

2 4 3 S1 6 5 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 S27 S28 S29

1(100%) 12(100%) 1(100%) 73(69.5%)

48 18-<60l 40 18-77

0/1 7/12 1/1 65/105 (62%)

13 9-127 20 5-133

1/1 0/11 0/1 5/92

1/1 NR/6 0/0 10(4)/41

14 6-15 5 2-30

1 12 S30 -

The reference numbers preceded by an S refer to references listed in the Online Supplementary Material. aOne patient had only thrombocytopenia. bThe patient was reported to be thrombocytopenic. cBilateral superior ophthalmic vein thrombosis. dAn eighth case included in this series was reported in reference S10. eNo thromboses were detected (possibly not a VITT). fOne of the cases did not have thrombosis. gOne case was initially treated with full dose low molecular weight heparin, with no apparent worsening, and was then changed to a non-heparin anticoagulant upon identification of positive anti PF4 antibodies. hThe platelet count was stated to be normal. iAges are reported as ranges, not individual. NR: not reported; PF4: platelet factor 4; VITT: vaccine-induced immune thrombotic thrombocytopenia.

administration of high doses of IgG in combination with steroids has been proposed for autoimmune HIT.77 Very early after the first reports of unusual types of thrombosis associated with the Vaxzevria vaccine, a German group suggested a tentative pathogenic mechanism underlying these rare events, which they named VIPIT, based on findings in nine cases of previously vac3040

cinated subjects.11 Indeed, in sera from four of these subjects the investigators detected antibodies to PF4/heparin complexes using an enzyme-immunoassay; these antibodies were inhibited by the addition of high concentrations of heparin (i.e., 100 U/mL), and the sera were able to activate washed control platelets when either PF4 or the Vaxzevria vaccine was added in vitro to the samples.11 haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

Figure 2. The autoimmune heparin-induced thrombocytopenia hypothesis. Vaccine components leaking into the bloodstream from the vaccination site (facilitated by ethylenediaminetetraacetic acid present in the vaccine) activate platelets to release platelet factor 4 (PF4). Vaccine constituents, likely polyanions or viral DNA, form complexes with positively-charged PF4 which are recognized as neoantigens by B cells that then produce antibodies against these complexes. The resulting immune complexes activate platelets through FcgRIIa, triggering the release of additional PF4 and polyphosphates thereby initiating a positive feedback loop that leads to further platelet activation and consumption. Extracellular DNA in neutrophil extracellular traps binds PF4 and the resulting DNA/PF4 complexes further recruit anti-PF4 antibodies inducing massive Fcg receptor-dependent activation of neutrophils, platelets, monocytes and endothelial cells leading to massive activation of coagulation and thrombosis. EDTA: ethylenediaminetetraacetic acid.

The investigators showed that platelet activation was triggered by FcgRIIA stimulation, because an FcgRIIAblocking antibody prevented this phenomenon. Given the absence of previous exposure to heparin, the authors suggested a condition resembling autoimmune HIT. More recently, in a preliminary report published in a non-peerreviewed internet repository, the German group went on to suggest that the Ad vector and/or some protein components of the Vaxzevria vaccine activate platelets to release PF4 which then forms complexes with virus proteins and other anionic constituents of the vaccine, generating neoantigens against which antibodies develop and induce strong platelet activation via FcgRIIa stimulating granulocyte activation with NETosis and ultimately catastrophic thrombosis.47 The presence of EDTA in the vaccine would favor vascular leakage at the inoculation site, facilitating dissemination of the vaccine components in blood (Figure 2). Very recently a study using alanine scanning mutagenesis explored the binding sites on PF4 of antibodies isolated from patients with VITT or with classical HIT. While the binding of VITT anti-PF4 antibodies was restricted to eight surface amino acids, all located within the heparinbinding site of PF4, HIT anti-PF4 antibodies bound amino acids corresponding to two different sites on PF4; moreover, VITT antibodies had a stronger binding response than HIT antibodies. The authors concluded that VITT antibodies mimic the effect of heparin by binding to a similar site on PF4, allowing PF4 tetramers to cluster with the formation of immuno-complexes which, in turn, cause FcγRIIa-dependent platelet activation.79 These peculiar characteristics explain why the identifihaematologica | 2021; 106(12)

cation of VITT requires different tests from those needed for the identification of classical HIT.80 In suspected VITT anti-PF4 antibodies can be identified by an ELISA, but not by other rapid immunological assays typically positive in HIT, such as the STic® Expert HIT kit, latex immunoassays and chemiluminescence-based assays,4,80 and it is better characterized by a heparin-induced platelet aggregation (HIPA) or PF4-induced platelet activation (PIPA) test.2,4,81 It should be noted that no single ELISA method detected all possible/probable VITT cases.82 While the autoimmune HIT hypothesis is important and provides the basis for understanding this novel catastrophic autoimmune thrombotic syndrome, several aspects do not fit completely with the clinical presentation of VITT and various issues remain unclarified. First, among the reported VITT cases in which anti-PF4 antibodies were measured, in a few they were negative (Table 2).4,5 Second, it is expected that treatment with heparin would worsen the clinical evolution of these patients, and indeed it is generally cautiously recommended not to use this drug.2,4,13 However, in around one fourth of the heparin-treated cases, the use of this anticoagulant was successful (Table 2). Third, while VITT resembles autoimmune HIT in several respects, the latter is not as frequently associated with CVST and rarely with DIC, and the above summarized mechanistic hypothesis does not explain the preferential localization of the venous thrombotic events in the cerebral and splanchnic circulations. Other unclear aspects are its relatively precocious onset, as early as 4 days after vaccination, which seems too soon to generate high-titer, class-switched, high-affin3041

P. Gresele et al.

Figure 3. A suggested clinical surveillance and diagnostic approach to suspected vaccine-induced immune thrombotic thrombocytopenia. Subjects receiving the Vaxzevria and the Janssen vaccines who develop new onset headache, especially if severe or with unprecedented characteristics, and/or associated with other clinical manifestations (blurred vision, gait disturbance, focal neurological symptoms and/or abdominal pain, vomiting, bloody stool, shortness of breath, petechiae or ecchymoses) should be referred for immediate laboratory evaluation (platelet count and D-dimer measurement). If thrombocytopenia is detected they should undergo anti-PF4 antibody testing, imaging and evaluation for cerebral vein sinus thrombosis, splanchnic vein thrombosis or pulmonary embolism. If confirmed, therapy for vaccine-induced immune thrombocytopenic thrombosis (VITT) should be immediately started according to the statement from the Italian Society for the Study of Haemostasis and Thrombosis (SISET)13 Although almost one quarter of the reported patients with VITT in whom unfractionated or low-molecular weight heparin was used apparently responded well to treatment (Table 2), subjects positive for anti-PF4 antibodies, as determined by a heparin-induced platelet aggregation (HIPA) test and/or enzyme-linked immunosorbent assay, or who have not been tested should, for prudence, be treated with alternative anticoagulants, until new information becomes available. In subjects in whom these anti-PF4 antibodies do not cross-react with heparin, as shown by a HIPA test in the presence of a low concentration of heparin, the use of heparin as anticoagulant may be allowed. ELISA: enzyme-linked immunosorbent assay; MRI: magnetic resonance imaging; CT: computed tomography; US: ultrasound.

ity anti-PF4 antibodies. Furthermore, there is no evidence that the anti-PF4 antibodies isolated from patients with VITT cause thrombosis and thrombocytopenia in animal models.49,83 Finally, recent observations show that 1.2% to 8.0% of subjects receiving a first dose of Vaxzevria develop circulating anti-PF4 antibodies while the prevalence of VITT ranges from 0.0006% to 0.00125%.84,85

Other possible pathogenic mechanisms of vaccine-induced immune thrombotic thrombocytopenia Very recently a preliminary report, published in a nonpeer-reviewed repository, provided an interesting alternative potential pathogenic mechanism of VITT.86 COVID19 is caused by SARS-CoV-2 which is a single-strand RNA virus that is translated and replicates only in the cytosol of infected cells in the absence of processes which are necessary when nuclear-encoded genes are transcribed, and in particular of mRNA splicing. Nuclear 3042

encoded genes have intronic sequences, thus their transcripts require splice reactions at consensus RNA sequences to eliminate them. When an Ad-vectored viral RNA sequence is administered the vector infects host cells, adenoviral DNA enters the nucleus and is then transcribed by the host transcription machinery. However, the viral piece of DNA deriving from the SARS-CoV-2 virus is not optimized to be transcribed into the nucleus and its open reading frame may thus be disrupted by arbitrary splice events. These splice events would produce shorter spike protein variants, including forms missing the C-terminal membrane anchor, thus leading to soluble circulating spike protein molecules. The soluble spike protein may cause a strong activation of endothelial cells expressing ACE2.87 Moreover, when the host immune system starts to produce antibodies against the spike protein, endothelial cells binding soluble spike would also be decorated by these antibodies, triggering a strong inflammatory reaction through antibody-dependent or complement-dependent cytotoxicity, thus eliciting VITT. With this hypothesis, the preferential involvement of cerebral haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

veins could be explained by the non-unidirectional blood flow in these vessels due to the lack of venous valves, with prolonged residence time of the soluble spike protein in this district depending on body posture or when sleeping. The immunological part of this hypothesis is also in agreement with the apparent higher prevalence of VITT in young women, because they have stronger immune reactions than men and older people. To explain the rarity of VITT the authors hypothesized that only some individuals, due to specific major histocompatibility complex combinations, are not able to produce neutralizing anti-spike antibodies which would instead prevent the binding of soluble spike to endothelial ACE2 and its ominous consequences in most vaccine recipients. This hypothesis was partly validated by the identification through in silico analysis of potential splice sites in the AstraZeneca and Johnson&Johnson codon-optimized spike opening frames and by in vitro studies with HeLa cells showing that vaccine-transduced cells generate transcripts smaller than the full spike protein. It would also explain why VITT has not been reported with mRNA vaccines, which release their cargo mRNA directly into the host cells’ cytosol where it is translated into spike protein without undergoing splicing reactions. Finally, it would account for why the incidence of VITT seems to be lower with the Johnson&Johnson vaccine than with the AstraZeneca one, given that the latter carries more splice donor sequences than the former.86 Additional hypotheses on the mechanisms triggering VITT include a genetically determined enhanced expression of FcgRIIa in susceptible subjects, an altered glycosylation state of IgG produced in response to vaccination in some individuals making these antibodies more reactive to platelet FcgRIIa,49 the leakage of the Ad vector into the circulation and/or the prior presence of cross-reactive antibodies to other coronaviruses forming immune complexes activating platelets.88 However, the hypothesis that VITT develops in subjects with previous, not apparent SARS-CoV-2 infection with prior circulating IgG antibodies against the spike protein able to activate platelet FcgRIIa89 should be excluded by the observation that most VITT subjects tested for previous or recent COVID-19 infections were negative. Excessive transcription of the spike protein, which would then activate platelets binding to ACE2,90 and vaccine-induced expression of the spike by megakaryocytes and platelets, leading to a thrombo-inflammatory storm,49,91 have also been proposed. Another hypothesis starts from the observation that both the ChAdOx1 and the Ad26.CoV2.S vaccines use polysorbate 80 as an excipient. Polysorbate 80 is a non-anionic surfactant that crosses the blood-brain barrier and enhances microparticle uptake by endothelial cells. Therefore leakage of Ad vector and polysorbate into the circulation and the spike protein produced by vaccination could preferentially localize in brain vessels triggering endothelial activation.92 However, considering that VITT usually develops at least 1 week after vaccination, it is very unlikely that circulating Ad vector or vaccine excipients would still be present in blood, making the alterna-

haematologica | 2021; 106(12)

tive explanations, and in particular an immunological reaction, more likely.

Conclusive remarks At least two Ad vector-based vaccines against SARSCoV-2 have been associated with an excess rate of a special form of catastrophic thrombotic syndrome associated with thrombocytopenia of likely autoimmune origin, not observed so far with mRNA-based vaccines, suggesting that the vectors may play a role in eliciting it. Several elements of the Ad vectors and/or of vaccine composition may theoretically interact with platelets, the endothelium and the blood clotting system precipitating this rare complication. However, the exact sequence of events leading to the development of this syndrome and, most importantly, the reason why it evolves in only very few subjects without apparent predisposing factors remain to be clarified. It is clear that our understanding of the pathogenesis of VITT is far from complete and that more mechanistic studies are required to clarify it and, it is hoped, to identify risk factors predictive of its development. What is quite likely is that the Ad vector-based vaccine triggers an immunological reaction which, for unknown reasons, in some rare subjects involves especially blood platelets, and possibly some peculiar vascular endothelial districts such as those of the cerebral and splanchnic veins, precipitating the catastrophic vaccine-induced autoimmune thrombocytopenia thrombosis syndrome. Awareness of this condition and the prompt identification/evaluation of affected patients may lead to successful treatment and recovery (Figure 3).13,84 COVID-19 continues to be a serious global health problem and vaccination against SARS-CoV-2 is the most effective way of limiting illness and death due to the pandemic. Based on the current available information, and in light of the relative rarity of VITT, the benefits of vaccination clearly outweigh the potential risks (https://www.ema.europa.eu/en/documents/dhpc/direct-healthcare-professional-communication-dhpc-zolgensma-onasemnogene-abeparvovec-risk-thrombotic_en.pdf). However, once the global pandemic begins to retreat, the relative importance of even small risks will increase,65 making it critically important to understand the mechanisms leading to this ominous thrombotic syndrome, to identify the prognostic factors for its development and to define the best management strategies.13 Disclosures No conflicts of interest to disclose. Contributions PG, VDS, AT, RM, SM and FR wrote the review article. Acknowledgments This work was supported in part by grants from Fondazione Cassa di Risparmio di Perugia (#19663 (2020.0508) and FISR 2020 (# 1049) to PG.

3043

P. Gresele et al.

References 1. Muir KL, Kallam A, Koepsell SA, Gundabolu K. Thrombotic thrombocytopenia after Ad26.COV2.S vaccination. N Engl J Med. 2021;384(20):1964-1965. 2. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle PA, Eichinger S. Thrombotic thrombocytopenia after ChAdOx1 nCov-19 vaccination. N Engl J Med. 2021;384(22):2092-2101. 3. Schultz NH, Sørvoll IH, Michelsen AE, et al. Thrombosis and thrombocytopenia after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;384(22):2124-2130. 4. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021,384(23):2202-2211. 5. Bayas A, Menacher M, Christ M, Behrens L, Rank A, Naumann M. Bilateral superior ophthalmic vein thrombosis, ischaemic stroke, and immune thrombocytopenia after ChAdOx1 nCoV-19 vaccination. Lancet. 2021;397(10285):e11. 6. Muster V, Gary T, Raggam RB, WÖlfler A, Brodmann M. Pulmonary embolism and thrombocytopenia following ChAdOx1 vaccination. Lancet. 2021;397(10287):1842. 7. Shay DK, Gee J, Su JR, et al. Safety monitoring of the Janssen (Johnson & Johnson) COVID-19 vaccine - United States, MarchApril 2021. MMWR Morb Mortal Wkly Rep. 2021;70(18):680-684. 8. Rodeghiero F, Balduini CL. A new enemy is emerging in the fight against the SARSCoV-2 pandemic. Haematologica. 2021;106 (8):2040-2041. 9. Hwang J, Lee SB, Lee SW. Comparison of vaccine-induced thrombotic events between ChAdOx1 nCoV-19 and Ad26.COV.2.S vaccines. J Autoimmun. 2021;122:102681. 10. Sangli S, Virani A, Cheronis N. Thrombosis with thrombocytopenia after the messenger RNA-1273 vaccine. Ann Intern Med. 2021;L21-0244. 11. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle P, Eichinger S. A prothrombotic thrombocytopenic disorder resembling heparin-induced thrombocytopenia following coronavirus vaccination. Blood Transfus. 2021; 19(4): 281-283 12. See I, Su JR, Lale A, Woo EJ, et al. US case reports of cerebral venous sinus thrombosis with thrombocytopenia after Ad26.COV2.S vaccination, March 2 to April 21, 2021. JAMA. 2021;325(24):24482456. 13. Gresele P, Marietta M, Ageno W, et al. Management of cerebral and splanchnic vein thrombosis associated with thrombocytopenia in subjects previously vaccinated with Vaxzevria (AstraZeneca): a position statement from the Italian Society for the Study of Haemostasis and Thrombosis (SISET). Blood Transfus. 2021;19(4):281283. 14. Hsu KH, Lonberg-Holm K, Alstein B, Crowell RL. A monoclonal antibody specific for the cellular receptor for the group B coxsackieviruses. J Virol. 1988;62(5):16471652. 15. Bergelson JM, Cunningham JA, Droguett G, et al. Isolation of a common receptor for coxsackie B viruses and adenoviruses 2 and 5. Science. 1997;275(5304):1320-1323. 16. Bewley MC, Springer K, Zhang YB, Freimuth P, Flanagan JM. Structural analysis of the mechanism of adenovirus binding to its human cellular receptor CAR.

3044

Science. 1999;286(5444):1579-1583. 17. Roelvink PW, Mi Lee G, Einfeld DA, Kovesdi I, Wicham TJ. Identification of a conserved receptor-binding site on the fiber proteins of CAR-recognizing Adenoviridae. Science. 1999;286(5444):1568-1571. 18. Greber U, Willets M, Webster P, Helenius A. Stepwise dismantling of adenovirus 2 during entry into cells. Cell. 1993;75(3):477486. 19. Wickham TJ, Mathias P, Cheresh DA, Nemerow GR. Integrins alpha v beta 3 and alpha v beta 5 promote adenovirus internalization but not virus attachment. Cell. 1993;73(2):309-319. 20. Davison E, Diaz RM, Hart IR, Santis G, Marshall JF. Integrin a5b1-mediated adenovirus infection is enhanced by the integrin-activating antibody TS2/16. J Virol. 1997;71(8):6204-6207. 21. Zhang Y, Bergelson JM. AD receptors. J Virol. 2005;79(19):12125-12131. 22. Honda T, Saitoh H, Masuko M, et al. The coxsackievirus-adenovirus receptor protein as a cell adhesion molecule in the developing mouse brain. Brain Res Mol Brain Res. 2000;77(1):19-28. 23. Patzke C, Max KEA, Behlke J, et al. The coxsackievirus-adenovirus receptor reveals complex homophilic and heterophilic interactions on neural cells. J Neurosci. 2010;30(8):2897-2910. 24. Nalbantoglu J, Pari G, Karpati G, et al. Expression of the primary coxsackie and adenovirus receptor is downregulated during skeletal muscle maturation and limits the efficacy of adenovirus-mediated gene delivery to muscle cells. Hum Gene Ther. 1999;10(6):1009-1019. 25. Noutsias M, Fechner H, de Jonge H, et al. Human coxsackie-adenovirus receptor is colocalized with integrins alpha(v)beta(3) and alpha(v)beta(5) on the cardiomyocyte sarcolemma and upregulated in dilated cardiomyopathy: implications for cardiotropic viral infections. Circulation. 2001;104(3): 275-280. 26. Rebel V, Hartnett S, Denham J, et al. Maturation and lineage-specific expression of the coxsackie and adenovirus receptor in hematopoietic cells. Stem Cells. 2000;18 (13):176-182. 27. Othman M, Labelle A, Mazzetti I, Elbatarny HS, Lillicrap D. Adenovirusinduced thrombocytopenia: the role of von Willebrand factor and P-selectin in mediating accelerated platelet clearance. Blood. 2007;109(7):2832-2839. 28. Shimony N, Elkin G, Kolodkin-Gal D, Krasny L, Urieli-Shoval S, Haviv YS. Analysis of adenoviral attachment to human platelets. Virol J. 2009;6:25. 29. Stone D, Liu Y, Shayakhmetov D, Li ZY, Ni S, Lieber A. Adenovirus-platelet interaction in blood causes virus sequestration to the reticuloendothelial system of the liver. J Virol. 2007;81(9):4866-4871. 30. Griffin JH, Zlokovic BV, Mosnier LO. Activated protein C: biased for translation. Blood. 2015;125(19):2898-2907. 31. Esmon CT. Protein C. Prog Hemost Thromb. 1984;7:25-54. 32. Hamad OA, Nilsson PH, Lasaosa M, et al. Contribution of chondroitin sulfate A to the binding of complement proteins to activated platelets. PLoS One. 2010;5(9): e12889. 33. Dahlbäck B, Wiedmer T, Sims PJ. Binding of anticoagulant vitamin K-dependent protein S to platelet-derived microparticles. Biochemistry. 1992;31(51):12769-12777.

34. Ferro JM, Canhão P, Stam J, Bousser MG, Barinagarrementeria F; ISCVT Investigators. Prognosis of cerebral vein and dural sinus thrombosis: results of the International Study on Cerebral Vein and Dural Sinus Thrombosis (ISCVT). Stroke. 2004;35(3):664-670. 35. Kurosawa S, Stearns-Kurosawa DJ, Hidar N, Esmon CT. Identification of functional endothelial protein C receptor in human plasma. J Clin Invest. 1997;100(2):411-418. 36. Gandrille S. Endothelial cell protein C receptor and the risk of venous thrombosis. Haematologica. 2008;93(6):812-816. 37. Javanmard SH, Shahsavarzadeh T, Saadatnia M. Soluble thrombomodulin and endothelial cell protein C receptor levels in patients with cerebral venous and sinus thrombosis. Eur Neurol. 2013;70(3-4):156158. 38. Gould WR, Baxi SM, Schroeder R, et al. Gas6 receptors Axl, Sky and Mer enhance platelet activation and regulate thrombotic responses. J Thromb Haemost. 2005;3(4):733-741. 39. Nidetz NF, Gallagher TM, Wiethoff CM. Inhibition of type I interferon responses by adenovirus serotype-dependent Gas6 binding. Virology. 2018;515:150-157. 40. Allen RJ, Byrnes AP. Interaction of adenovirus with antibodies, complement, and coagulation factors. FEBS Letters. 2019;593 (24):3449-3460. 41. Lenman A, Muller S, Nygren MI, Frangsmyr L, Stehle T, Arnberg N. Coagulation factor IX mediates serotypespecific binding of species A adenoviruses to host cells. J Virol. 2011;85(24):1342013431. 42. Parker AL, McVey JH, Doctor JH, et al. Influence of coagulation factor zymogens on the infectivity of adenoviruses pseudotyped with fibers from subgroup D. J Virol. 2007;81(7):3627-3631. 43. Parker AL, Waddington SN, Nicol CG, et al. Multiple vitamin K-dependent coagulation zymogens promote adenovirus-mediated gene delivery to hepatocytes. Blood. 2006;108(8):2554-2561. 44. Shayakhmetov DM, Gaggar A, Ni S, Li, Lieber A. Adenovirus binding to blood factors results in liver cell infection and hepatotoxicity. J Virol. 2005;79(12):7478-7491. 45. Kalyuzhniy O, Di Paolo NC, Silvestry M, et al. Adenovirus serotype 5 hexon is critical for virus infection of hepatocytes in vivo. Proc Natl Acad Sci U S A. 2008;105(14): 5483-5488. 46. Duffy MR, Doszpoly A, Turner G, Nicklin SA, Baker AH. The relevance of coagulation factor X protection of adenoviruses in human sera. Gene Ther. 2016;23(7):592596. 47. Greinacher A, Selleng K, Wesche J, et al. Towards understanding ChAdOx1 nCov19 vaccine-induced immune thrombotic thrombocytopenia (VITT). Research Square. 2021 Apr 20. [Epub ahead of print] 48. Brunetti-Pierri N, Palmer DJ, Beaudet AL, Carey KD, Finegold M, Ng P. Acute toxicity after high-dose systemic injection of helper-dependent adenoviral vectors into nonhuman primates. Hum Gene Ther. 2004;15(1):35-46. 49. Kadkhoda K. Post-adenoviral-based COVID-19 vaccines thrombosis: a proposed mechanism. J Thromb Haemost. 2021;19(7):1831-1832. 50. Lee WS, Wheatley AK, Kent SJ, DeKosky BJ. Antibody-dependent enhancement and SARS-CoV-2 vaccines and therapies. Nat

haematologica | 2021; 106(12)

Adenovirus vaccines and thrombosis

Microbiol. 2020;5(10):1185-1191. 51. Hamada F, Aoki M, Akiyama T, Toyoshima K. Association of immunoglobulin G Fc receptor II with Src like protein-tyrosine kinase Fgr in neutrophils. Proc Natl Acad Sci U S A. 1993;90(13):6305-6309. 52. Ghazizadeh S, Bolen JB, Fleit HB. Physical and functional association of Src-related protein tyrosine kinases with Fc gamma RII in monocytic THP-1 cells. J Biol Chem. 1994;269(12):8878-8884. 53. Kiener PA, Rankin BM, Burkhardt AL, et al. Cross-linking of Fc gamma receptor I (Fc gamma RI) and receptor II (Fc gamma RII) on monocytic cells activates a signal transduction pathway common to both Fc receptors that involves the stimulation of p72 Syk protein tyrosine kinase. J Biol Chem. 1993;268(32):24442-24448. 54. Nimmerjahn F, Ravetch JV. Fcγ receptors as regulators of immune responses. Nat Rev Immunol. 2008;8(1):34-47. 55. Joshi T, Butchar JP, Tridandapani S. Fcg receptor signaling in phagocytes. Int J Hematol. 2006;84(3):210-216. 56. Furuyama W, Marzi A, Carmody AB, et al. Fcg-receptor IIa-mediated Src signaling pathway is essential for the antibodydependent enhancement of Ebola virus infection. PLoS Pathog. 2016;12(12): e1006139. 57. Taylor SM, Reilly MP, Schreiber AD, Chien P, Tuckosh JR, McKenzie SE. Thrombosis and shock induced by activating antiplatelet antibodies in human Fc gamma RIIA transgenic mice: the interplay among antibody, spleen, and Fc receptor. Blood. 2000;96(13):4254-4260. 58. Calverley DC, Brass E, Hacker MR, et al. Potential role of platelet FcgammaRIIA in collagen-mediated platelet activation associated with atherothrombosis. Atherosclerosis. 2002;164(2):261-267. 59. Pamela S, Anna Maria L, Elena D et al. Heparin-induced thrombocytopenia: the role of platelets genetic polymorphisms. Platelets. 2013;24(5):362-368. 60. Tetro JA. Is COVID-19 receiving ADE from other coronaviruses? Microbes Infect. 2020;22(2):72-73. 61. Collocca S, Barnes E, Folgori A, et al. Vaccine vectors derived from a large collection of simian adenoviruses induce potent cellular immunity across multiple species. Sci Transl Med. 2012;4(115):115ra2. 62. Obdulio GN, V'kovski V, Zettl F, Zimmer G, Thiel V, Summerfield A. No evidence for human monocyte-derived macrophage infection and antibody-mediated enhancement of SARS-CoV-2 Infection. Front Cell Infect Microbiol. 2021;11:644574. 63. Crystal RG, Harvey BG, Wisnivesky JP, et al. Analysis of risk factors for local delivery of low- and intermediate-dose adenovirus gene transfer vectors to individuals with a spectrum of comorbid conditions. Hum Gene Ther. 2002;13(1):65-100. 64. Folegatti PM, Bittaye M, Flaxman A, et al. Safety and immunogenicity of a candidate Middle East respiratory syndrome coronavirus viral-vectored vaccine: a dose-escalation, open-label, non-randomised, uncon-

haematologica | 2021; 106(12)

trolled, phase 1 trial. Lancet Infect Dis. 2020;20(7):816-826. 65. Kupferschmidt K, Vogel G. What’s the future of vaccines linked to rare clotting disorders? Science breaks down the latest. 2021 May 3. [Epub ahead of print] 66. Jaoko W, Karita E, Kayitenkore K, et al. Safety and immunogenicity study of multiclade HIV-1 Adenoviral vector vaccine alone or as boost following a multiclade HIV-1 DNA vaccine in Africa. PLoS One. 2010;5(9):e12873. 67. Keefer MC, Gilmour J, Hayes P, Gill D, et al. A phase I double blind, placebo-controlled, randomized study of a multigenic HIV-1 adenovirus subtype 35 vector vaccine in healthy uninfected adults. PLoS One. 2012;7(8):e41936. 68. Rego GNA, Nucci MP, Alves AH, et al. Current clinical trials protocols and the global effort for immunization against SARS-CoV-2. Vaccines (Basel). 2020;8(3): 474. 69. Van Doremalen N, Lambe T, Spencer A, et al. ChAdOx1 nCoV-19 vaccination prevents SARS-CoV-2 pneumonia in rhesus macaques. Nature. 2020;586(7830):578-582. 70. Folegatti PM, Ewer KJ, Aley PK, et al. Oxford COVID Vaccine Trial Group. Safety and immunogenicity of the ChAdOx1 nCoV-19 vaccine against SARSCoV-2: a preliminary report of a phase 1/2, single-blind, randomised controlled trial. Lancet. 2020;396(10249):467-478. 71. Sandoff J, Gray G, Vandebosch A, et al. Safety and efficacy of single-dose Ad26.COV2.S vaccine against Covid-19. N Engl J Med. 2021;384(23):2187-2201. 72. Logunov DY, Dolzhikova IV, Shcheblyakov DV, et al. Safety and efficacy of an rAd26 and rAd5 vector-based heterologous primeboost COVID-19 vaccine: an interim analysis of a randomized controlled phase 3 trial in Russia. Lancet. 2021;397(10275):671681. 73. Pottegård A, Lund LC, Karlstad Ø, et al. Arterial events, venous thromboembolism, thrombocytopenia, and bleeding after vaccination with Oxford-AstraZeneca ChAdOx1-S in Denmark and Norway: population based cohort study. BMJ. 2021;May 5;373:n1114. 74. Marcucci R, Berteotti M, Gori AM, et al. Heparin induced thrombocytopenia: position paper from the Italian Society on Thrombosis and Haemostasis (SISET). Blood Transfus. 2021;19(1):14-23. 75. Cuker A, Arepally GM, Chong BH, et al. American Society of Hematology 2018 guidelines for management of venous thromboembolism: heparin-induced thrombocytopenia. Blood Adv. 2018;2(22): 3360-3392. 76. Greinacher A, Farner B, Kroll H, et al. Clinical features of heparin-induced thrombocytopenia including risk factors for thrombosis. A retrospective of 408 patients. Thromb Haemost. 2005;94(1): 132-135. 77. Greinacher A, Selleng K, Warkentin TE. Autoimmune heparin-induced thrombocytopenia. J Thromb Haemost. 2017;15(11):

2099-2114. 78. Warkentin TE, Greinacher A. Spontaneous HIT syndrome: knee replacement, infection, and parallels with vaccine-induced immune thrombotic thrombocytopenia. Thromb Res. 2021;204:40-51. 79. Huynh A, Kelton JG, Arnold DM, Daka M, Nazy I. Antibody epitopes in vaccineinduced immune thrombotic thrombocytopenia. Nature. 2021;596(7873):565-569. 80. Favaloro EJ. Laboratory testing for suspected COVID-19 vaccine-induced (immune) thrombotic thrombocytopenia. Int J Lab Hematol. 2021;43(4): 559-570 81. Vayne C, Guery EA, Kizlik-Masson C, et al. Beneficial effect of exogenous platelet factor 4 for detecting pathogenic heparininduced thrombocytopenia antibodies. Br J Haematol. 2017;179(5):811-819. 82. Platton S, Bartlett A, MacCallum P, et al. Evaluation of laboratory assays for antiplatelet factor 4 antibodies after ChAdOx1 nCOV-19 vaccination. J Thromb Haemost. 2021;19(8):2007-2013. 83. Cines DB, Bussel JB. SARS-CoV-2 Vaccineinduced immune thrombotic thrombocytopenia. N Engl J Med. 2021;384(23):22542256. 84. Sørvoll IH, Horvei KD, Ernstsen SL, et al. An observational study to identify the prevalence of thrombocytopenia and antiPF4/polyanion antibodies in Norwegian health care workers after COVID-19 vaccination. J Thromb Haemost. 2021;19(7): 1813-1818. 85. Thiele T, Ulm L, Holtfreter S, et al. Frequency of positive anti-PF4/polyanion antibody tests after COVID-19 vaccination with ChAdOx1 nCoV-19 and BNT162b2. Blood. 2021;138(4):299-303. 86. Kowarz E, Krutzke L, Reis J, Bracharz S, Kochanek S, Marschalek R. Vaccineinduced Covid-19 spike open reading frame results in spike protein variant that may cause thromboembolic events in patients immunized with vector-based vaccine. Research Square. 2021 May 26. [Epub ahead of print] 87. Lei Y, Zhang J, Schiavon CR, et al. SARSCoV-2 spike protein impairs endothelial function via downregulation of ACE 2. Circ Res. 2021;128(9):1323-1326. 88. Chakraborty S, Gonzalez J, Edwards K. Proinflammatory IgG Fc structures in patients with severe COVID-19. Nat Immunol. 2021;22(1):67-73. 89. Douxfils J, Favresse J, Dogné JM. Hypotheses behind the very rare cases of thrombosis with thrombocytopenia syndrome after SARS-CoV-2 vaccination. Thromb Res. 2021;203:163-171. 90. Shen S, Zhang J, Fang Y, et al. SARS-CoV-2 interacts with platelets and megakaryocytes via ACE2-independent mechanism. J Hematol Oncol. 2021;29;14(1):72. 91. Millington-Burgess SL, Harper MT. A double-edged sword: antibody-mediated procoagulant platelets in COVID-19. Platelets. 2021; 32(5):579-581. 92. Choi PHI. Thrombotic thrombocytopenia after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;385(3):e11.

3045

ARTICLE Ferrata Storti Foundation

Acute Lymphoblastic Leukemia

Multiclonal complexity of pediatric acute lymphoblastic leukemia and the prognostic relevance of subclonal mutations Željko Antić,1* Jiangyan Yu,1,2* Simon V. van Reijmersdal,1,2 Anke van Dijk,2 Linde Dekker,1 Wouter H. Segerink,1 Edwin Sonneveld,1,3 Marta Fiocco,1,4,5 Rob Pieters,1,3 Peter M. Hoogerbrugge,1,3 Frank N. van Leeuwen,1 Ad Geurts van Kessel,2 Esmé Waanders1,6 and Roland P. Kuiper1,6

Haematologica 2021 Volume 106(12):3046-3055

Princess Máxima Center for Pediatric Oncology, Utrecht; 2Department of Human Genetics, Radboud Institute for Molecular Life Sciences, Radboud University Medical Center, Nijmegen; 3Dutch Childhood Oncology Group, Utrecht; 4Medical Statistics, Department of Biomedical Data Science, Leiden University Medical Center, Leiden; 5 Mathematical Institute, Leiden University and 6Department of Genetics, University Medical Center Utrecht, Utrecht, the Netherlands *ŽA and JY contributed equally as co-first authors.

ABSTRACT

Correspondence: ROLAND P. KUIPER r.kuiper@prinsesmaximacentrum.nl Received: May 13, 2020. Accepted: October 23, 2020. Pre-published: November 5, 2020. https://doi.org/10.3324/haematol.2020.259226

3046

enomic studies of pediatric acute lymphoblastic leukemia (ALL) have shown remarkable heterogeneity in initial diagnosis, with multiple (sub)clones harboring lesions in relapse-associated genes. However, the clinical relevance of these subclonal alterations remains unclear. We assessed the clinical relevance and prognostic value of subclonal alterations in the relapse-associated genes IKZF1, CREBBP, KRAS, NRAS, PTPN11, TP53, NT5C2, and WHSC1 in 503 ALL cases. Using molecular inversion probe sequencing and breakpoint-spanning polymerase chain reaction analysis we reliably detected alterations with an allele frequency below 1%. We identified 660 genomic alterations in 285 diagnostic samples of which 495 (75%) were subclonal. RAS pathway mutations were common, particularly in minor subclones, and comparisons between RAS hotspot mutations revealed differences in their capacity to drive clonal expansion in ALL. We did not find an association of subclonal alterations with unfavorable outcome. Particularly for IKZF1, an established prognostic marker in ALL, all clonal but none of the subclonal alterations were preserved at relapse. We conclude that, for the genes tested, there is no basis to consider subclonal alterations detected at diagnosis for risk group stratification of ALL treatment.

Introduction Improvements in the treatment of pediatric acute lymphoblastic leukemia (ALL) have resulted in high overall survival rates, now approaching 90%.1,2 Nevertheless, relapse still remains the most common cause of treatment failure and death in children with ALL, and better recognition of individuals at risk of developing relapse will likely aid further improvements in outcome. Recent studies describing the genomic landscape of relapsed ALL have shown that relapse often originates from a minor (sub)clone at diagnosis, at a cellular fraction often undetectable by routine diagnostic methods.3-5 These minor (sub)clones harbor genomic alterations acquired later during leukemia development, which could potentially contribute to clonal drift, but are unlikely to be essential for initiation of the primary disease. However, selective pressure of the upfront treatment may provide a competitive advantage to subclones that harbor alterations in cancer genes, enabling their selective survival, eventually leading to treatment failure. Both the number and clonal burden of the alterations in these genes are expected to be increased at the time of relapse, compared to initial diagnosis. Indeed, mutations in relapse-associated genes, such as those in the histone acetyltransferase (HAT) domain of the histone methyltransferase CREBBP, can often be traced back to minor subclones in the diagnostic sample.4,6,7

haematologica | 2021; 106(12)

Prognostic relevance of subclonal mutations in ALL

Genomic characterization of relapsed pediatric ALL has revealed multiple alterations that are enriched compared to diagnosis, including activating mutations in RAS pathway genes, HAT domain mutations in CREBBP and deletions or mutations in the B-cell transcription factor IKZF1.6-13 The presence of these aberrations at the time of diagnosis can be of potential prognostic relevance, as has been demonstrated extensively for IKZF1 in many different treatment protocols12,14-19 and can even lead to adjustments in stratification and treatment.14,20 However, it remains unclear whether mutations in relapse-associated genes when present in a minor subclone at initial diagnosis are also clinically relevant. Subclonal mutations can be identified using deep targeted, next-generation sequencing techniques.21,22 Despite the sensitivity of these techniques, both amplification and sequencing can easily lead to errors that hamper the reliable detection of low-level mosaic mutations. We previously demonstrated that single molecule molecular inversion probes (smMIP), which use unique molecular identifiers to barcode each DNA copy, can correct for sequencing and amplification artefacts, resulting in a reliable detection of low-level mosaic mutations, down to a variant allele frequency of 0.4%.23 In this study we used the smMIP-based sequencing approach to perform deep targeted sequencing of seven relapse-associated genes in a cohort of 503 pediatric ALL samples taken at initial diagnosis, resulting in the detection of 141 clonal and 469 subclonal mutations. In addition, we performed real time quantitative polymerase chain reaction (PCR) to sensitively detect subclonal IKZF1 exon 4-7 deletions (del 4-7), which were found at a similar frequency as full-clonal deletions. Subsequently, we estimated their potential as drivers of clonal expansion and prognostic markers for relapse development.

enrollment in the study, and the DCOG institutional review board approved the use of excess diagnostic material for this study (OC2017-024). In order to accurately detect subclonal alterations in diagnostic samples, 166 smMIP were designed in CREBBP, PTPN11, NT5C2, WHSC1, TP53, KRAS and NRAS, seven genes that are frequently mutated in relapsed ALL (Online Supplementary Table S4, Online Supplementary Materials and Methods). IKZF1 and ERG deletion status was assessed using the multiplex ligation-dependent probe amplification assay (MLPA) SALSA P335 ALL-IKZF1 and P327 iAMP-ERG kits, respectively (MRC-Holland, the Netherlands), according to the manufacturer’s instructions and as described before.12,24 Additionally, IKZF1 4-7 deletions were assessed with Sanger sequencing and real-time quantitative PCR, using an IQ SYBR Green supermix (Biorad, USA). For detailed descriptions of the smMIP-based sequencing, IKZF1 deletion detection and data analysis, see the Online Supplementary Materials and Methods (Online Supplementary Figures S1 and S2, Online Supplementary Tables S4-S6). To test continuous and categorical variables, the nonparametric Wilcoxon signed rank and Fisher exact tests were used, respectively (R packages ggpubr version 0.2 and stats version 3.5.1). Cumulative incidence of relapse (CIR) was estimated by employing a competing-risk model with death as a competing event.27 To assess the statistical difference between CIR, the Gray test28 was applied. To investigate the effect of prognostic factors on relapse, univariate and multivariate Cox proportional hazard regression models were estimated. Competing risk analysis was performed with the R packages cmprsk (version 2.2-7) and survminer (version 0.4.3). Univariate and multivariate Cox models were estimated using R package survival (version 3.1-12). Data were visualized using the R package ggplot2 (version 3.2.1) and cBioPortal MutationMapper.29,30

Results Methods In this study we analyzed two cohorts of diagnostic samples from B-cell precursor ALL patients treated according to the Dutch Childhood Oncology Group (DCOG) protocols DCOG-ALL9 (n=131)12,24 and DCOG-ALL10 (n=245) (Online Supplementary Table S1). Both cohorts were representative selections of the total studies12,24 (Online Supplementary Table S2). The median age at diagnosis of the patients in these cohorts was 4 and 5 years, and the median follow-up time, estimated with a reverse Kaplan-Meier method, was 138 and 104 months, respectively.25 Relapse occurred in 18% (24/131) and 11% (27/245) of the patients, while 0.7% (1/131) and 2.8% (7/245) died during the follow-up. DNA was isolated from mononuclear cells obtained from bone marrow or peripheral blood. The median blast percentage of the samples was 92% (Online Supplementary Table S3). To increase the number of patients for the comparisons between relapsed and non-relapsed cases, we used an extended cohort of diagnostic samples from 127 additional ALL patients treated according to the DCOG-ALL9 (n=76) or DCOG-ALL10 (n=51) protocols; this cohort was enriched for patients who had a relapse and also contained 55 patients with T-cell ALL. This latter cohort was not included in the survival analyses. In order to detect mutations preserved in major clones at relapse, we performed Sanger sequencing (73/171) or used previously published Ampliseq-based deep-sequencing data (98/171) to verify alterations observed at diagnosis.26 In accordance with the Declaration of Helsinki, written informed consent was obtained from all patients and/or their legal guardians before

haematologica | 2021; 106(12)

A total of 503 diagnostic samples from children with ALL (Online Supplementary Table S1) was subjected to targeted deep sequencing of the relapse-associated genes TP53, CREBBP (HAT domain), KRAS, NRAS, PTPN11, NT5C2 and WHSC1 using smMIP, which contain random molecular tags to accurately detect low-level mosaic variants.23 Each targeted region was covered with an average of 308 unique capture-based consensus reads (Figure 1, Online Supplementary Figure S1A, B), enabling the reliable detection of alterations with allele frequencies even below 1%. A total of 7,836 quality-filtered variants was detected, of which 610 were absent in public and private variant databases and were predicted as pathogenic. The allele frequency of these mutations ranged from 0.03-100% (Figure 2A, Online Supplementary Table S3). The majority of the mutations (473/610; 78%) was found in one of the three RAS pathway genes (KRAS, NRAS, PTPN11), of which 418 (88%) were known hotspot mutations. In addition to sequencing the seven relapse-associated genes, we performed sensitive screening for IKZF1 deletions, which are strongly associated with the occurrence of relapse. We chose to focus on exon 4-7 deletions, which represent 25% of all IKZF1 deletions, have a similar unfavorable outcome as other IKZF1 deletions,31 and show the strongest clustering of deletion breakpoints, thus enabling their sensitive upfront detection by breakpoint-spanning semi-quantitative PCR.32 Applying this strategy to the 503 diagnostic samples revealed all 22 IKZF1 exon 4-7 dele3047

Ž. Antić et al.

Figure 1. Schematic representation of the study design. Single-molecule molecular inversion probe-based sequencing approach and real-time quantitative polymerase chain reaction were used in order to detect alterations in known relapse-associated genes in a large cohort of diagnostic samples from patients with acute lymphocytic leukemia. Detected alterations were correlated with outcome and Sanger sequencing was performed on available relapse samples in order to confirm that exactly the same alteration was present in the major clone in relapse. smMIP: single-molecule molecular inversion probe; MLPA: multiplex ligation-dependent probe amplification assay; PCR: polymerase chain reaction; qPCR: real-time quantitative polymerase chain reaction.

tions previously identified using a standard MLPA method, as well as 28 additional cases carrying deletions that were missed with the MLPA technique. All breakpoints were sequenced to determine their unique breakpoint-spanning sequences (Online Supplementary Table S6). Using a dilution series of a control sample with a full-clonal IKZF1 exon 4-7 deletion, we determined the level of clonality of the deletions, which ranged from 100% down to 0.32% (Figures 1 and 2, Online Supplementary Figure S1C). All but one of the subclonal IKZF1 exon 4-7 deletions had allele frequencies below 10% (Online Supplementary Table S7).

Subclonal alterations in relapse-associated genes are common at diagnosis Combining sequence mutations and IKZF1 exon 4-7 deletions, we detected 660 genomic alterations in 285 diagnostic samples, of which 165 (25%) were present in the major fraction of cells (allele frequency ≥25%), which were referred to as high-clonal. The remaining 495 mutations (75%), most of which had an allele frequency <10% were referred to as subclonal (Online Supplementary Figure S2, Online Supplementary Table S7). A total of 147/285 patients carried at least one alteration in a major clone, while 138/285 (48%) patients carried exclusively subclonal alterations. NRAS and KRAS were the most frequently affected genes, showing major clone mutations in 6% and 8% of the cases and subclonal mutations in 20% and 15% of the cases, respectively (Figure 2A, B). The proportion of subclonal alterations, relative to major clone alterations, was variable among different genes, ranging from 59% for IKZF1 exon 4-7 deletions to 86% for PTPN11 mutations (Figure 2B). Only one thus far unknown (subclonal) NT5C2 mutation (p.Arg507Trp) was identified in a leukemia sample from a patient who did not relapse (Figure 2A, B). Subclonal mutations were relatively common in hyperdiploid ALL (184 cases), particularly for mutations in RAS pathway genes (190/256; 74%), WHSC1 (22/26; 85%) and CREBBP (13/27; 48%) (Online Supplementary Table S8). Major clone WHSC1 mutations 3048

were mostly identified in ETV6-RUNX1-positive cases (4/10, 40%).

Potency of RAS pathway genes as drivers of clonal expansion We identified 473 RAS pathway mutations in 225/503 (45%) cases, of which 78% were subclonal (median allele frequency = 3.5%). Over half of the RAS-affected cases were hyperdiploid (>47 chromosomes), in line with previous studies indicating that RAS mutations are associated with hyperdiploidy at diagnosis.10,33 The abundance of these mutations in major and minor clones suggests that these mutations drive clonal expansion during the development of leukemia. Major clone RAS pathway mutations (n=102; all being known hotspots) were found to be mutually exclusive, and 52/102 (51%) of these RAS-mutated cases had at least one additional subclonal mutation in one of the three RAS pathway genes. The mutations mostly affected codons 12 and 13 of KRAS and NRAS (Figure 3AC), and considerable variability in the level of clonality was observed between the different RAS hotspot mutations at the time of diagnosis. For example, NRAS G12A (10 cases), NRAS G12V (7 cases), and PTPN11 E76K (7 cases) were never found to be present in a major clone, whereas 55% (n=11) of the KRAS G13D and 27% (n=9) of the KRAS G12D mutations were found in major clones. With these high numbers of RAS mutations, the variability in clonal burden between hotspot mutations may provide an opportunity to compare the capacity of different hotspots to drive clonal expansion of ALL. In order to test this hypothesis we compared allele frequencies and performed statistical analyses. We found that KRAS hotspot mutations had a significantly higher allele frequency compared to both NRAS and PTPN11 mutations (Wilcoxon signed-rank test, P<0.01) (Figure 3D). When comparing the different hotspot mutations within KRAS, A146V showed the lowest allele frequency, indicating a weaker potential of this hotspot to drive clonal expansion compared to the other KRAS hotspots. Furthermore, the allele frequency of KRAS G13D was significantly higher than haematologica | 2021; 106(12)

Prognostic relevance of subclonal mutations in ALL

Figure 2. Prevalence and distribution of alterations in eight relapse-associated genes. (A) Violin plot showing the variability in mutation allele frequency at diagnosis in the genes studied. The color of the dots indicates whether the mutation was detected in a case without relapse (blue) or with relapse (red). (B) Bar plot showing the frequencies of major clone (high-clonal) and subclonal alterations per case in the genes studied. RAS pathway genes (NRAS, KRAS, PTPN11) were the most frequently mutated. Subclonal mutations (yellow bar) were highly prevalent in all genes tested. A subset of cases had both clonal and subclonal alterations in the same gene (gray bar).

haematologica | 2021; 106(12)

3049

Ž. Antić et al.

Table 1. Cox regression analysis in the combined representative ALL9 and ALL10 cohort (n=376). IKZF1

MRD

Age at diagnosis, years

Gender

Status

Number of patients

Univariate Cox regression1,2

Multivariate Cox regression1,2

High-clonal3 Subclonal Other high-clonal4 Low Medium High 0-4 5-9 10-14 15-18 Male Female

18 22 13 111 227 22 173 133 47 23 215 161

P<0.01; HR=7.22 [3.27-15.95] P=0.39; HR=1.69 [0.51-5.57] P<0.01; HR=19.92 [9.76-40.66] 1 (Ref) P=0.1; HR=2 [0.88-4.61] P<0.01; HR=10.85 [4.16-28.28] 1 (Ref) P=0.34; HR =1.34 [0.73-2.46] P=0.83; HR=0.9 [0.34-2.38] P=0.3; HR=1.76 [0.6-5.12] 1 (Ref) P=0.06; HR=0.56 [0.31-1.03]

P<0.01; HR=3.6 [1.38-9.55] P= 0.34; HR=1.8 [0.53-6.24] P<0.01; HR=13 [5.68-29.79] 1 (Ref) P=0.28; HR=1.6 [0.67-3.78] P<0.01; HR=5 [1.64-15.23] 1 (Ref) P=0.6; HR=1.2 [0.64-2.25] P=0.6; HR=0.8 [0.27-2.12] P=0.3; HR=0.34 [0.04-2.66] 1 (Ref) P=0.08; HR=0.53 [0.26-1.1]

The hazard ratio is given with a 95% confidence interval. Multivariate Cox regression analysis included gender, age at diagnosis and minimal residual disease status as covariates; Analysis was done on the combined cohort stratified on treatment protocol. Representative ALL9 and ALL10 cohorts are outlined in Online Supplementary Table S1. 3Clonality status based on detection at initial diagnosis; high-clonal: allele frequency ≥25%, subclonal: allele frequency <25%. 4Detectable using a multiplex ligation-dependent probe amplification assay. A full overview of all comparisons is given in Online Supplementary Table S11. MRD: minimal residual disease. 2

that of the NRAS hotspot mutations G13D, G12D and KRAS A146V (Figure 3E). This finding indicates that some RAS hotspot mutations (e.g., KRAS G12D, G13D, A146T) may result in a stronger expansion potential compared to others (e.g., KRAS A146V, NRAS G12D, G13D), and further illustrates the complex heterogeneity of RAS hotspot mutations in their potential to drive clonal expansion.

Relevance of gene alterations to relapse development The high number of alterations in these relapse-associated genes at the time of diagnosis triggers the hypothesis that these could be used as prognostic biomarkers for relapse development, even when present at subclonal levels. To test this hypothesis, we first explored whether alterations in each of the eight genes were enriched in diagnostic samples from patients who subsequently relapsed compared to diagnostic samples from patients who did not relapse. In general, subclonal alterations were very common at primary diagnosis in patients who relapsed (60/82; 73%) as well as in patients who did not (165/203; 81%). For high-clonal alterations, we only observed a higher percentage of relapse development in cases with IKZF1 deletions compared to wild-type cases, whereas an association with relapse development was not observed for diagnostic samples with subclonal alterations in any of the genes, including IKZF1 (Figure 4). Furthermore, patients with high-clonal IKZF1 4-7 deletions were more often classified as having high minimal residual disease (MRD; >5×10−4 at day 79 or 84 after start of the treatment) in both representative ALL9 and ALL10 cohorts (Fisher exact test, P<0.01 and P<0.05, respectively), compared to patients without an IKZF1 deletion (Online Supplementary Table S10). The CIR at 5 years was 41.7% (SE 0.04%) and 42.9% (SE 0.03%) in patients with high-clonal IKZF1 4-7 deletions treated according to the ALL9 and ALL10 protocols, respectively (Figure 5). The cause-specific hazard ratio (HRCS) in the two representative cohorts (n=376), estimated with a univariate Cox proportional hazards regression model, revealed an association of high-clonal IKZF1 exon 4-7 deletions with relapse (HR=7.22; 95% CI: 3.27-15.95; P<0.01). In the multivariate Cox model, in which age at diagnosis, gender and MRD status were included, the adjusted HRCS was 3.6 (95% CI: 1.38-9.55; P<0.01) (Table 1, Online Supplementary 3050

Table S11). These data are in line with those from earlier studies on these cohorts in which all IKZF1 deletions were included.12,24 However, when we assessed the clinical relevance of subclonal alterations for relapse development in IKZF1, or any of the other genes, Cox regression analysis revealed no significant associations in the combined ALL9 and ALL10 cohorts compared to wild-type cases (Table 1, Online Supplementary Table S11), and the CIR was similar in the two groups (Figure 5). Furthermore, patients with subclonal IKZF1 4-7 deletions did not have significantly different levels of MRD compared to IKZF1 wild-type patients (Online Supplementary Table S10). Since previous studies have shown a lack of association of IKZF1 deletion with relapse in patients who carry a deletion in ERG,34,35 we used MLPA to test whether there was an enrichment of ERG deletions in cases with subclonal IKZF1 exon 4-7 deletions compared to those with clonal IKZF1 exon 4-7, but these deletions were infrequent in both groups (Online Supplementary Table S12).

Tracing of major and minor clone mutations at the time of relapse To obtain further insight into the clinical relevance of the identified alterations in relapse development, we investigated whether these were preserved in the cases that relapsed. For this analysis, we used all 146 cases that later developed a relapse, of which 82 carried alterations in a major or minor clone in one or more of the genes (Online Supplementary Tables S13 and S14). Overall, we found that for most genes at the time of diagnosis the frequency of subclonal alterations was similar or slightly higher compared to that of the alterations detected in a major clone (Online Supplementary Figure S3A, Online Supplementary Tables S13 and S14). We collected 73 relapse samples from patients who carried these major or minor clone alterations at the time of diagnosis (89%), which enabled us to trace 171 of the 185 sequence mutations, and 25 of the IKZF1 exon 4-7 deletions. We did not assess whether mutations detected at diagnosis were still preserved in minor clones at relapse, since these clones were unlikely to be true relapse drivers. Overall, 56% (22/39) of the tested major clone mutations were found to be preserved in the major clone at relapse, whereas the value for the subclonal mutations was 7% haematologica | 2021; 106(12)

Prognostic relevance of subclonal mutations in ALL

Figure 3. Potency of RAS pathway mutations to drive clonal expansion. (A-C) Schematic representation of KRAS (A), NRAS (B) and PTPN11 (C) indicating the prevalence of common hotspot mutations. (D) Violin plot showing allele frequency in hotspot mutations of three investigated RAS pathway genes. The median allele frequency was significantly higher in KRAS, indicating a high potential of KRAS hotspot mutations to drive clonal expansion. (E) Violin plot showing allele frequencies in frequent KRAS and NRAS hotspot mutations. The median allele frequency was significantly higher in KRAS G13D, suggesting their higher potential to drive clonal expansion compared to other RAS hotspot mutations.

haematologica | 2021; 106(12)

3051

Ž. Antić et al.

Figure 4. Prevalence of relapse associated genomic alterations at diagnosis. Bar plot showing the percentage of relapses in cases with high-clonal (blue) or subclonal (yellow) mutations in seven relapse-associated genes, and in cases that were wild-type (black) for these genes. Only cases with high-clonal IKZF1 4-7 deletions showed a significantly higher percentage of relapse development compared to wild-type cases (Fisher exact test, P<0.01) (Online Supplementary Table S9).

Figure 5. Cumulative incidence of relapse for high-clonal and subclonal IKZF1 deletions. The cumulative incidence of relapse (CIR) was estimated using a competing-risk model with death as a competing event. CIR plots are presented for the representative ALL9 (left) and ALL10 (right) cohorts. Lines represent the IKZF1 deletion status and include wild-type (black line), subclonal exon 4-7 deletion (yellow), other high-clonal deletion (purple), and high-clonal exon 4-7 deletion (blue). Straight lines depict relapses and dotted lines death. P-values shown are obtained by employing the Gray test to compare CIR curves. The 5-year CIR was higher in cases with high-clonal IKZF1 deletions, compared to wild-type cases in both representative cohorts.

(9/132) (Fisher exact test, P<0.01) (Figure 6, Online Supplementary Figure S3B, Online Supplementary Table S7). For IKZF1 exon 4-7 deletions, the difference was even more striking. Here, the presence of deletions was studied in 19 available relapse samples using breakpoint-spanning PCR, followed by Sanger sequencing to confirm that the breakpoint sequences were identical at diagnosis and relapse (Figure 6, Online Supplementary Figure S3B, Online Supplementary Tables S6 and S7). All major clone IKZF1 exon 4-7 deletions were found to be preserved in the major clone at the time of relapse (n=12), which is in agreement with earlier findings and illustrates their relevance to relapse development in these treatment protocols.12,24 In contrast, none of the subclonal exon 4-7 deletions in IKZF1 (n=13) was preserved in either the major or a minor clone at relapse. Collectively, the data from the 3052

present study indicate that these deletions, when present at initial diagnosis at a subclonal level, do not drive relapse in pediatric ALL.

Discussion ALL is a heterogeneous disease in which specific genomic alterations show strong associations with relapse risk and outcome. In this study, we assessed the clinical relevance and prognostic value of subclonal alterations in eight genes frequently mutated in relapsed B-cell precursor ALL in a cohort of 503 diagnostic samples. Our data demonstrate that subclonal alterations in these genes are very common at the time of diagnosis, but that these mutations do not provide a basis for risk stratification in haematologica | 2021; 106(12)

Prognostic relevance of subclonal mutations in ALL

Figure 6. Preservation of clonal and subclonal mutations at the time of relapse. Tracing of major clone (top) and subclonal (bottom) alterations detected in initial diagnosis samples from relapsed patients in the matched relapse samples. The pie charts depict the fractions of preserved (blue) and lost (orange) alterations at the time of relapse.

pediatric ALL. This finding is particularly relevant for IKZF1 alterations, which are currently used or implemented for treatment stratification in multiple upfront treatment protocols.14,20 The selection of these genes was made based on enriched mutation frequencies in relapse found in previous studies. Of all alterations identified in this study, 75% were subclonal at diagnosis, suggesting that these relapseassociated gene mutations accumulate during progression of the leukemia before the initial diagnosis, thereby increasing the clonal complexity. Whereas seven of the genes selected in our study showed this high mutational burden at diagnosis, both in terms of numbers and level of clonality, we identified only a single, not previously reported, subclonal NT5C2 mutation in a non-relapsed case (follow-up time 9.5 years). NT5C2 encodes the cytosolic nucleotidase, which is responsible for inactivating cytotoxic thiopurine monophosphate nucleotides, and activating mutations in this gene are recurrently found in relapsed ALL, mainly T-cell ALL.4,9,36-38 One explanation for the low number of activating NT5C2 mutations at diagnosis is that these mutations decrease cell fitness, and only obtain their selective advantage during treatment with thiopurine.36 If already present at the time of initial diagnosis, these mutations are usually detectable in only a very small subset of cells, far below the detection level of our smMIP analysis.36 Hotspot RAS pathway mutations have been detected in nearly half of the cases, often of the hyperdiploid subtype, and their frequency and clonal burden varied between the different mutations. In our study, we used this variability to compare the potential of different hotspot mutations to drive clonal expansion under physiological conditions. Compared to diagnosis, we observed a less diverse spectrum of KRAS and NRAS hotspot mutations in relapse, with G12D, G12V and G13D together accounting for two-thirds of KRAS and NRAS hotspot mutations found in relapse-fated clones. Studies in other cancers have demonstrated that the prevalence of different RAS pathway mutations varies depending on the type of cancer and tissue of origin, with KRAS mutations G12D, G12V, G13D and G12C being among the most common ones.39,40 Comparison of oncogenic capacities of different RAS haematologica | 2021; 106(12)

hotspots has also been performed using in vitro and in vivo modeling studies, focusing primarily on KRAS. These studies identified KRAS mutations G12D, G12V and G13D as having higher proliferative and transforming potential compared to other common hotspots in various tumors of epithelial origin.39,41,42 Our data indicate that in competition of multiple RAS hotspot mutations, some of these not only confer a proliferative advantage but can also more effectively sustain a treatment-induced selective sweep.4,10 The presence of IKZF1 deletions has been shown to be associated with relapse and survival in multiple clinical ALL studies,12,14-19 and these deletions have been described to play a role in resistance to tyrosine kinase inhibitors and glucocorticoids.43-46 Therefore, with the advance of more sensitive detection techniques, the question of whether subclonal alterations are also associated with relapse is very relevant, both from biological and from clinical perspectives. We here demonstrate that, in contrast to major clone IKZF1 exon 4-7 deletions, cases that carry this deletion only in a subset of the cells do not show an association with relapse. Moreover, whereas all major clone exon 4-7 deletions were preserved in cases that relapsed, none of the relapses from cases with subclonal exon 4-7 deletions at diagnosis carried this deletion. Importantly, the majority of subclonal deletions had allele frequencies below 10% (Online Supplementary Table S7). Therefore, since a threshold to distinguish subclonal from major clonal deletions is difficult, deletions closer to our threshold of 25% should be evaluated with caution. Nevertheless, the difference between major and minor clone IKZF1 4-7 deletions is striking, and the reason behind this remains unclear. Possibly, the functional impact of full-clonal IKZF1 deletions, which arise early during leukemia development, is different from that of deletions that occur in later stages when the leukemia has already expanded. Other deletions in IKZF1 show much less clustering in their breakpoints and, therefore, screening for these subclonal deletions in diagnostic samples is much less efficient. We did not, therefore, directly assess the stability and potential prognostic importance of whole gene and rare intragenic IKZF1 deletions. However, a previous study showed that other IKZF1 deletion subtypes 3053

Ž. Antić et al.

have similar prognostic relevance as exon 4-7 deletions,31 suggesting that subclonal alterations in these other IKZF1 deletions may show the same lack of association. In summary, we show that subclonal alterations in the relapse-associated genes IKZF1, CREBBP, KRAS, NRAS, PTPN11, TP53, and WHSC1 in pediatric ALL are frequently present at initial diagnosis, often at a subclonal level. At relapse, however, most of these subclonal mutations are lost, suggesting that their selective advantages over wild-type clones during treatment is limited. This finding has direct implications for clinical practice, particularly in the case of IKZF1, where deletion status is used for routine risk stratification. We conclude that, at least for the investigated set of genes, there is no basis for the use of subclonal alterations at initial diagnosis as a prognostic marker. Disclosures No conflicts of interest to disclose.

References 1. Hunger SP, Lu X, Devidas M, et al. Improved survival for children and adolescents with acute lymphoblastic leukemia between 1990 and 2005: a report from the Children's Oncology Group. J Clin Oncol. 2012;30(14):1663-1669. 2. Pieters R, de Groot-Kruseman Hd, Velden V, et al. Successful therapy reduction and intensification for childhood acute lymphoblastic leukemia based on minimal residual disease monitoring: study ALL10 from the Dutch Childhood Oncology Group. J Clin Oncol. 2016;34(22):2591-2601. 3. Anderson K, Lutz C, van Delft FW, et al. Genetic variegation of clonal architecture and propagating cells in leukaemia. Nature. 2011;469(7330):356-361. 4. Ma X, Edmonson M, Yergeau D, et al. Rise and fall of subclones from diagnosis to relapse in pediatric B-acute lymphoblastic leukaemia. Nat Commun. 2015;6:6604. 5. Waanders E, Gu Z, Dobson SM, et al. Mutational landscape and patterns of clonal evolution in relapsed pediatric acute lymphoblastic leukemia. Blood Cancer Discov. 2020;1(1):96-111. 6. Malinowska-Ozdowy K, Frech C, Schönegger A, et al. KRAS and CREBBP mutations: a relapse-linked malicious liaison in childhood high hyperdiploid acute lymphoblastic leukemia. Leukemia. 2015;29(8): 1656-1667. 7. Mullighan CG, Zhang J, Kasper LH, et al. CREBBP mutations in relapsed acute lymphoblastic leukaemia. Nature. 2011; 471(7337):235-239. 8. Irving J, Matheson E, Minto L, et al. Ras pathway mutations are prevalent in relapsed childhood acute lymphoblastic leukemia and confer sensitivity to MEK inhibition. Blood. 2014;124(23):3420-3430. 9. Tzoneva G, Perez-Garcia A, Carpenter Z, et al. Activating mutations in the NT5C2 nucleotidase gene drive chemotherapy resistance in relapsed ALL. Nat Med. 2013;19(3):368-371. 10. Jerchel IS, Hoogkamer AQ, Ariës IM, et al. RAS pathway mutations as a predictive biomarker for treatment adaptation in pediatric B-cell precursor acute lymphoblastic leukemia. Leukemia. 2018; 32(4):931-940.

3054

Contributions RPK, EW, FvL, and AGvK conceived the study. RPK, EW, ŽA and JY designed the study. PMH and RP are responsible for the clinical outcome data of patients included in the study. ŽA, JY, SVvR, AvD, LD and WHS performed experiments and analyzed the data. ŽA and MF performed statistical analyses. ŽA, JY and RPK wrote the manuscript and created the figures and tables. ES provided samples and clinical information. All authors critically reviewed the manuscript and approved the final submitted manuscript. Acknowledgments We thank Radboud University Medical Center, Department of Human Genetics for bioinformatic support in data analysis. Funding This work was supported by grants from Stichting Kinderen Kankervrij (KIKA 150 to RPK), Stichting Bergh in het Zadel (to RPK), and the China Scholarship Council (CSC201304910347 to JY).

11. Mullighan CG, Su X, Zhang J, et al. Deletion of IKZF1 and prognosis in acute lymphoblastic leukemia. N Engl J Med. 2009;360(5):470-480. 12. Kuiper RP, Waanders E, van der Velden VH, et al. IKZF1 deletions predict relapse in uniformly treated pediatric precursor B-ALL. Leukemia. 2010;24(7):1258-1264. 13. Jaffe JD, Wang Y, Chan HM, Z, et al. Global chromatin profiling reveals NSD2 mutations in pediatric acute lymphoblastic leukemia. Nat Genet. 2013;45(11):1386-1391. 14. van der Veer A, Waanders E, Pieters R, et al. Independent prognostic value of BCR-ABL1like signature and IKZF1 deletion, but not high CRLF2 expression, in children with Bcell precursor ALL. Blood. 2013;122(15): 2622-2629. 15. Hinze L, Moricke A, Zimmermann M, et al. Prognostic impact of IKZF1 deletions in association with vincristine-dexamethasone pulses during maintenance treatment of childhood acute lymphoblastic leukemia on trial ALL-BFM 95. Leukemia. 2017;31(8): 1840-1842. 16. Clappier E, Grardel N, Bakkus M, et al. IKZF1 deletion is an independent prognostic marker in childhood B-cell precursor acute lymphoblastic leukemia, and distinguishes patients benefiting from pulses during maintenance therapy: results of the EORTC Children's Leukemia Group study 58951. Leukemia. 2015;29(11):2154-2161. 17. Ofverholm I, Tran AN, Heyman M, et al. Impact of IKZF1 deletions and PAX5 amplifications in pediatric B-cell precursor ALL treated according to NOPHO protocols. Leukemia. 2013;27(9):1936-1939. 18. Dörge P, Meissner B, Zimmermann M, et al. IKZF1 deletion is an independent predictor of outcome in pediatric acute lymphoblastic leukemia treated according to the ALL-BFM 2000 protocol. Haematologica. 2013;98(3):428-432. 19. Asai D, Imamura T, Suenobu S, et al. IKZF1 deletion is associated with a poor outcome in pediatric B-cell precursor acute lymphoblastic leukemia in Japan. Cancer Med. 2013;2(3):412-419. 20. Yeoh AEJ, Lu Y, Chin WHN, et al. Intensifying treatment of childhood B-lymphoblastic leukemia with IKZF1 deletion reduces relapse and improves overall sur-

vival: results of Malaysia-Singapore ALL 2010 study. J Clin Oncol. 2018;36(26):27262735. 21. Absalan F, Ronaghi M. Molecular inversion probe assay. Methods Mol Biol. 2007;396: 315-330. 22. Berglund EC, Lindqvist CM, Hayat S, et al. Accurate detection of subclonal single nucleotide variants in whole genome amplified and pooled cancer samples using HaloPlex target enrichment. BMC Genomics. 2013;14(1):856. 23. Yu J, Antić Ž, van Reijmersdal SV, et al. Accurate detection of low-level mosaic mutations in pediatric acute lymphoblastic leukemia using single molecule tagging and deep-sequencing. Leuk Lymphoma. 2018; 59(7):1690-1699. 24. Waanders E, van der Velden VH, van der Schoot CE, et al. Integrated use of minimal residual disease classification and IKZF1 alteration status accurately predicts 79% of relapses in pediatric acute lymphoblastic leukemia. Leukemia. 2011;25(2):254-258. 25. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343-346. 26. Yu J, Waanders E, van Reijmersdal SV, et al. Upfront treatment influences the composition of genetic alterations in relapsed pediatric B-cell precursor acute lymphoblastic leukemia. Hemasphere. 2020;4(1):e318. 27. Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics: competing risks and multi-state models. Stat Med. 2007;26(11):2389-2430. 28. Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat. 1988;16(3):1141-1154. 29. Gao J, Aksoy BA, Dogrusoz U, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013;6(269):pl1. 30. Cerami E, Gao J, Dogrusoz U, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012; 2(5):401-404. 31. Boer JM, van der Veer A, Rizopoulos D, et al. Prognostic value of rare IKZF1 deletion in childhood B-cell precursor acute lymphoblastic leukemia: an international collaborative study. Leukemia. 2016;30(1):32-38. 32. Caye A, Beldjord K, Mass-Malo K, et al.

haematologica | 2021; 106(12)

Prognostic relevance of subclonal mutations in ALL

Breakpoint-specific multiplex polymerase chain reaction allows the detection of IKZF1 intragenic deletions and minimal residual disease monitoring in B-cell precursor acute lymphoblastic leukemia. Haematologica. 2013;98(4):597-601. 33. Paulsson K, Lilljebjörn H, Biloglav A, et al. The genomic landscape of high hyperdiploid childhood acute lymphoblastic leukemia. Nat Genet. 2015;47(6):672-676. 34. Clappier E, Auclerc MF, Rapion J, et al. An intragenic ERG deletion is a marker of an oncogenic subtype of B-cell precursor acute lymphoblastic leukemia with a favorable outcome despite frequent IKZF1 deletions. Leukemia. 2014;28(1):70-77. 35. Zaliova M, Potuckova E, Hovorkova L, et al. ERG deletions in childhood acute lymphoblastic leukemia with DUX4 rearrangements are mostly polyclonal, prognostically relevant and their detection rate strongly depends on screening method sensitivity. Haematologica. 2019;104(7):1407-1416. 36. Tzoneva G, Dieck CL, Oshima K, et al.

haematologica | 2021; 106(12)

Clonal evolution mechanisms in NT5C2 mutant relapsed acute lymphoblastic leukemia. Nature. 2018;553(7689):511-514. 37. Meyer JA, Wang J, Hogan LE, et al. Relapsespecific mutations in NT5C2 in childhood acute lymphoblastic leukemia. Nat Genet. 2013;45(3):290-294. 38. Pieters R, Huismans DR, Loonen AH, et al. Relation of 5'-nucleotidase and phosphatase activities with immunophenotype, drug resistance and clinical prognosis in childhood leukemia. Leuk Res. 1992; 16(9):873880. 39. Miller MS, Miller LD. RAS mutations and oncogenesis: not all RAS mutations are created equally. Front Genet. 2012;2:100. 40. Prior IA, Lewis PD, Mattos C. A comprehensive survey of Ras mutations in cancer. Cancer Res. 2012;72(10):2457-2467. 41. Winters IP, Chiou S-H, Paulk NK, et al. Multiplexed in vivo homology-directed repair and tumor barcoding enables parallel quantification of Kras variant oncogenicity. Nat Commun. 2017;8(1):2053.

42. Stolze B, Reinhart S, Bulllinger L, Fröhling S, Scholl C. Comparative analysis of KRAS codon 12, 13, 18, 61, and 117 mutations using human MCF10A isogenic cell lines. Sci Rep. 2015;5:8535. 43. Marke R, Havinga J, Cloos J, et al. Tumor suppressor IKZF1 mediates glucocorticoid resistance in B-cell precursor acute lymphoblastic leukemia. Leukemia. 2016; 30(7):1599-1603. 44. Imamura T, Yano M, Asai D, et al. IKZF1 deletion is enriched in pediatric B-cell precursor acute lymphoblastic leukemia patients showing prednisolone resistance. Leukemia. 2016;30(8):1801-1803. 45. Churchman ML, Low J, Qu C, et al. Efficacy of retinoids in IKZF1-mutated BCR-ABL1 acute lymphoblastic leukemia. Cancer Cell. 2015;28(3):343-356. 46. van der Veer A, Zaliova M, Mottadelli F, et al. IKZF1 status as a prognostic feature in BCR-ABL1-positive childhood ALL. Blood. 2014;123(11):1691-1698.

3055

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(12):3056-3066

Acute Myeloid Leukemia

Plasmacytoid dendritic cells proliferation associated with acute myeloid leukemia: phenotype profile and mutation landscape Loria Zalmaï,1 Pierre-Julien Viailly,2 Sabeha Biichle,3 Meyling Cheok,4 Lou Soret,3 Fanny Angelot-Delettre,3 Tony Petrella,5 Marie-Agnès CollongeRame,6 Estelle Seilles,3 Sandrine Geffroy,4,7 Eric Deconinck,8 Etienne Daguindau,8 Sabrina Bouyer,9 Elodie Dindinaud,9 Victor Baunin,10 Magali Le Garff-Tavernier,11 Damien Roos-Weil,12 Orianne Wagner-Ballon,13 Véronique Salaun,14 Jean Feuillard,15 Sophie Brun,16 Bernard Drenou,17 Caroline Mayeur-Rousse,18 Patricia Okamba,19 Véronique Dorvaux,20 Michel Tichionni,21 Johann Rose,22 Marie Thérèse Rubio,23 Marie Christine Jacob,24 Victoria Raggueneau,25 Claude Preudhomme,4,7 Philippe Saas,3 Christophe Ferrand,3 Olivier Adotevi,3 Christophe Roumier,4,7 Fabrice Jardin,2 Francine Garnache-Ottou3 and Florian Renosi3 1

Correspondence: FRANCINE GARNACHE-OTTOU francine.garnache@efs.sante.fr Received: March 26, 2020. Accepted: September 24, 2020. Pre-published: October 13, 2020. https://doi.org/10.3324/haematol.2020.253740

3056

Service d’Hématologie Biologique, Hôpital Cochin, Assistance Publique-Hôpitaux de Paris (AP-HP), Paris, France; 2INSERM U1245, Centre Henri Becquerel, Rouen, France; 3 Université Bourgogne Franche-Comté, INSERM, EFS BFC, UMR1098, Interactions HôteGreffon-Tumeur/Ingénierie Cellulaire et Génique, Besançon, France; 4INSERM U837, CHRU Lille, IRCL Laboratoire d'Hématologie, Centre de Biologie Pathologie, Lille, France; 5 Department of Pathology, University of Montréal, Hôpital Maisonneuve-Rosemont, Montréal, Quebec, Canada; 6Laboratoire de Génétique Biologie, CHU Besançon, Besançon, France; 7Laboratoire d'Hématologie A, Centre de Biologie Pathologie, Boulevard du Pr Leclercq, Lille, France; 8Service Hématologie, CHU Besançon, Besançon, France; 9Service d’Hématologie Biologique, CHU La Milétrie, Poitiers, France; 10 Laboratoire du Groupe Hospitalier de La Rochelle-Ré-Aunis, CH de La Rochelle, La Rochelle, France; 11Laboratoire d’Hématologie, Hôpital Pitié-Salpêtrière, Assistance Publique-Hôpitaux de Paris (AP-HP), Paris, France; 12Service d’Hématologie Clinique, Sorbonne Université, Hôpital Pitié-Salpêtrière, Assistance Publique-Hôpitaux de Paris (AP-HP), Paris, France; 13Département d’Hématologie biologique, Hôpitaux Universitaires Henri Mondor, Assistance Publique-Hôpitaux de Paris (AP-HP), Créteil, France; 14 Laboratoire d’Hématologie, CHU de Caen Normandie, Normandie Université, UNICAEN, Caen, France; 15Laboratoire d'Hématologie, CHU Dupuytren, Limoges, France; 16 Laboratoire d'Hématologie et Consultations d'Hématologie Biologique, Hôpital Universitaire Carémeau, Nîmes, France; 17Service d’Hématologie, Groupe Hospitalier de la région Mulhouse Sud Alsace, Mulhouse, France; 18Laboratoire d’Hématologie, CHRU Strasbourg, Hôpital de Hautepierre, Strasbourg, France; 19Laboratoire d’Hématologie et Auto-immunité, Hôpital de Mercy, CHR de Metz-Thionville, France; 20Service d’Hématologie de l’Hôpital de Mercy, CHR de Metz-Thionville, France; 21Laboratoire d’Hématologie de l’Hôpital Pasteur, Nice, France; 22Laboratoire d’hématologie, CH du Mans, Le Mans, France; 23Service Hématologie, CNRS UMR7365, Biopôle Université de Lorraine, CHRU Nancy, Vandœuvre-lès-Nancy, France; 24Laboratoire d’Immunologie, CHU Grenoble, La Tronche, France and 25Service de Biologie Médicale, Centre Hospitalier de Versailles A. Mignot, Le Chesnay, France

ABSTRACT

eoplasms involving plasmacytoid dendritic cells (pDC) include blastic pDC neoplasms (BPDCN) and other pDC proliferations, where pDC are associated with myeloid malignancies: most frequently chronic myelomonocytic leukemia (CMML) but also acute myeloid leukemia (AML), hereafter named pDC-AML. We aimed to determine the reactive or neoplastic origin of pDC in pDC-AML, and their link with the CD34+ blasts, monocytes or conventional DC (cDC) associated in the same sample, by phenotypic and molecular analyses (targeted next-generation sequencing, 70 genes). We compared 15 pDCAML at diagnosis with 21 BPDCN and 11 normal pDC from healthy donors. CD45low CD34+ blasts were found in all cases (10-80% of medullar cells), associated with pDC (4-36%), monocytes in 14 cases (1-10%) and cDC (two cases, 4.8-19%). pDC in pDC-AML harbor a clearly different phenotype from BPDCN: CD4+ CD56– in 100% of cases, most frequently CD303+, CD304+ and CD34+; lower expression of haematologica | 2021; 106(12)

Frequent RUNX1 mutations in acute leukemia + pDC

cTCL1 and CD123 with isolated lymphoid markers (CD22/CD7/CD5) in some cases, suggesting a prepDC stage. In all cases, pDC, monocytes and cDC are neoplastic since they harbor the same mutations as CD34+ blasts. RUNX1 is the most commonly mutated gene: detected in all AML with minimal differentiation (M0-AML) but not in the other cases. Despite the low number of cases, the systematic association between M0-AML, RUNX1 mutations and an excess of pDC is puzzling. Further evaluation in a larger cohort is required to confirm RUNX1 mutations in pDC-AML with minimal differentiation and to investigate whether it represents a proliferation of blasts with macrophage and DC progenitor potential.

Introduction Plasmacytoid dendritic cells (pDC) are hematopoietic cells mainly developed from a myeloid branch including the macrophage DC progenitor (MDP) with monocyte, conventional DC (cDC) and pDC differentiation potential.1–4 Two types of neoplastic counterparts for pDC have been identified: the first is the well-known blastic pDC neoplasm (BPDCN), initially described as CD4+ CD56+ neoplasm;5-10 and the second is defined as mature pDC proliferation (MPDCP) associated with a myeloid neoplasm, frequently chronic myelomonocytic leukemia (CMML), but also myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML), especially with monocytic differentiation.11-17 MPDCP is not formally referenced in the World Health Organization 2017 classification, but mentioned as a differential diagnosis of BPDCN.7,8 As for BPDCN, MPDCP occur predominantly in male patients (75%) with a median age of 69 years18 and frequent lymph nodes or skin lesions. The mature pDC denomination refers to the morphologically mature and CD56– phenotype (as with normal pDC), in the absence of the blastic morphology of BPDCN.8 Flow cytometry or immunohistochemistry are mandatory for BPDCN diagnosis and relatively well defined with CD4+, CD56+, CD123+high, CD303+/-, CD304+/- cells7,19 expressing TCL1 at high levels.20 Conversely, only few cases of MPDCP phenotype have been described.7,14,15 In the same way, the genomic profile of MPDCP is still poorly understood, but a clonal relationship between pDC and the associated neoplasm has been demonstrated with pDC exhibiting leukemic abnormalities such as monosomy 7, trisomy 8, del(5q), CBFB-MYH11 or internal tandem duplication of FLT3 (FLT3-ITD) of blasts in AML16,17,21–23 and the mutational profile of monocytes in CMML.24 We collected AML harboring a heterogeneous phenotypic presentation with a population of immature blasts associated with an excess of pDC, but also monocytes and sometimes cDC. These cases are hereafter referred to as AML with pDC (pDC-AML). The purpose of this study was to better characterize pDC-AML by analyzing the phenotype of immature blasts and pDC and the mutational profiles of each cell population (blasts, pDC, monocytes, cDC) after cell sorting, in order to determine whether they share the same profiles.

CD34+ blasts of myeloid origin and pDC (Table 1). All cases were analyzed at diagnosis, except for N35 (day 81 post-induction). The analysis was performed on BM aspiration (n=12) or PB (n=3). BM biopsies were rarely available, preventing from anatomopathology analysis. Fifteen normal BM aspiration performed for peripheral thrombopenia/research of metastatic infiltration, ten PB from healthy donors and 21 previously described cases of BPDCN20 were used as controls after written informed consent. This study was approved by the Besançon Ethics Committee (CPP-Est II, Besançon, France).20

Immunophenotype Flow cytometry was performed using a FACSCanto II cytometer (BD Biosciences, San Jose, CA, USA) with DIVA 6.2 software (BD Biosciences) after cell labeling with monoclonal antibodies (Online Supplementary Table S1). The mean fluorescence intensity ratio (MFIR) of cTCL1 was obtained by dividing the mean fluorescence intensity (MFI) for cTCL1 by that of the isotype control monoclonal antibody (mAb). Cells were considered positive for cTCL1 expression when MFIR was greater than 2. Isotype control was not used for MFIR of CD123 because of its high expression on pDC; thus MFIR was calculated by dividing the MFI of CD123 on pDC by that on lymphocytes.

Cell sorting One to 10 million cells were sorted using an ARIA III FACS (Becton Dickinson Biosciences) after cell labeling with mAb (Online Supplementary Table S1) in order to select the populations of interest: T-cells (CD45+high, CD3+), immature blasts (CD34+, CD303–), pDC (CD123+high, CD303+), monocytes (CD14+ or CD64+, CD123+low, CD303–) and cDC (CD34-, CD1c+, CD303–).

Molecular biology Whole genome amplification was carried out using the REPLI-g® Single Cell kit (Qiagen Hilden, Germany), as recommended by the manufacturer. Next-generation sequencing (NGS) was performed, from a HaloPlexHS Target Enrichment System (Agilent Technologies Inc., Santa Clara, CA, USA) targeting 70 genes (Online Supplementary Table S2), in paired-end, 2x150 cycles on a MiSeq platform (Illumina Inc., San Diego, CA, USA).

Bioinformatics analysis

Methods

Raw data were analyzed using in-house bioinformatics pipelines (Online Supplementary Methods) with annotation of the variants via GenerateReportsTM25 yielding variant call files. Finally, several filters were applied to eliminate intronic regions, synonymous mutations and polymorphisms.

Patient samples

Statistical analyses

Primary cells are routinely referred to our center for BPDCN suspicion with cytologic and phenotypic arguments in peripheral blood (PB) or bone marrow (BM) (collection DC 2016-27 91). Some cases do not meet the criteria for a BPDCN diagnosis, based on the World Health Organization classification 2017.5,8 Especially, 15 cases were collected based on the association of

Statistical analyses were performed using Prism 6.0 software (GraphPad, San Diego, CA, USA). The distribution of MFIR was studied using the D’Agostino-Pearson normality test. Quantitative data were compared using ANOVA. Patient groups were compared using the Mann-Whitney non-parametric test for quantitative variables with a non-Gaussian distribution, the Student t-test

haematologica | 2021; 106(12)

3057

L. Zalmai et al.

Table 1. Clinical and biological features of the cohort. Patient Age Sex Material WHO FAB Secondary number (y) classification classification (yes/no)

Prior Extramedullar Anatomopathology therapy lesions of extramedullar lesions

Karyotype

FISH KMT2A (11q23)

pDC contingent

no anomaly detected ND

CD34–

M0 AML

46,XX[20]

M0 AML

AML with M0 AML mutated RUNX1 BM AML with M0 AML mutated RUNX1 BM AML with M0 AML mutated RUNX1 BM AML with M0 AML KMT2A(MLL) rearrangement and mutated RUNX1 PB AML with M0 AML mutated RUNX1

yes/cutaneous

AML

46,XY,del(7) (q36)[8]/46,XY[12] 46,XY

45,X,-Y[20]

46,XX[20]

yes/ cutaneous

AML

46,XY[20]

47,XX,+13[13] /46,XX[7]

pDCs

46,X,-Y,+13 [11]/46,XY[23]

pDCs

N13

AML with mutated RUNX1 AML-MRC

N16*

N20

N19

N9$

AML-MRC

N11

AML with mutated RUNX1

N12

N36

N14

N34¤ N35

73 65

M M

PB BM

AML with mutated RUNX1 BM AML with mutated RUNX1 BM AML without maturation AML-MRC

M0 AML Neutropenia Therapeutic yes/ and abstention, cutaneous thrombocytopenia, monitoring granulocytic dysplasia£ M0 AML no yes/ cutaneous

no anomaly detected no anomaly detected no anomaly detected rearranged

CD34– CD34–

CD34–

CD34+

CD34–

no anomaly detected ND

CD34+

47,XY,+13 [21]/46,XY[11]

no anomaly detected

CD34+

M0 AML

yes/ lymph nodes

46,XY

CD34–

M0 AML

46,XY[30]

CD34+

M1 AML

46,XY[28]

M4 AML

CMML

46,XY,-7,+mar [18]/46,XY[4]

no anomaly detected no anomaly detected ND

no no

ND ND

46,XY[20] 46,XY[20]

ND rearranged

CD34– CD34–

AML-MRC M4 AML AML with M5 AML mutated NPM1 and KMT2A(MLL) rearrangement

CMML MDS/MPN

Therapeutic abstention, monitoring Hydroxyurea Therapeutic abstention, monitoring

CD34+

BM: bone marrow; PB: peripheral blood; AML-MRC: AML with myelodysplasia-related changes; WHO classification: World Health Organization classification; FAB: FrenchAmerican-British classification; FISH: fluorescent in situ hybridization; ND: not done; F: female; M: male. $pDC nodules were described on BM biopsy. £Diagnosis of myelodysplasia was not clearly affirmed before acute myeloid leukemia (AML) state. ¤Only patient N34 experienced a prior history of solid tumor with prostatic neoplasm treated by hormonotherapy.

3058

haematologica | 2021; 106(12)

Frequent RUNX1 mutations in acute leukemia + pDC

Figure 1. Morphologies and immunophenotypes of populations of interest. (A to C) Representative morphologic aspects of peripheral blood smears from patient N1 (magnification 1,000X). (A) Blast cells are medium sized with a high nuclear cytoplasmic ratio, fine chromatin with proeminent nucleoli. Cytoplasm is basophilic with some rare azurophilic granulations. (B) Plasmacytoid dendritic cells (pDC) are smaller with more mature chromatin. The cytoplasm is less basophilic without granulation but sometimes pseudopodia and small vacuoles under the cytoplasmic membrane. (C) A blast cell (top), a pDC with pseudopodia (center) and a monocyte (bottom). (D to E) Representative immunophenotype after gating on FSC-A vs. FSC-H plus sides catter (SSC) vs. forward scatter (FSC) to select singlets, leucocytes and exclude debris (not shown). Lymphocytes in blue (CD45bright/SSCdim cells); immature blasts in black (CD34+ cells); pDC in pink (CD123bright); monocytes in green (CD123dim, CD33bright, CD64bright); cDC in orange (CD123+, CD33+ CD64dim). (D) Patient N8: acute myeloid leukemia with minimal differentiation (M0-AML) with a continuum of phenotypic acquisition of markers (arrows) from the immature blasts: downregulation of CD13 and CD33 to pDC or upregulation to monocytes and cDC; downregulation of CD117 and CD34 to pDC/monocytes/cDC. (E) Patient N12: M0-AML with Tdt+, HLA-DRbright, CD33-, cCD13+ immature blasts; CD7+ CD4+ CD56– pDC.

for normally distributed variables or the c2 test with Yates’ continuity correction for categorical variables. Correlations between quantitative variables were investigated by linear regression analyses. All statistical tests were two-sided, with a 5% alpha risk. Results are expressed as median (range). For further details see the Online Supplementary Appendix.

Results Patients Fifteen patients were included, mainly elderly men with a median age of 70 years (range, 52-87) and a sex ratio of haematologica | 2021; 106(12)

4:1 (Table 1). Cytological analysis of BM aspiration, associated with phenotypic data, found 11 cases of AML with minimal differentiation (M0-AML in French-AmericanBritish [FAB] classification) with a prior history of cytopenia in one of them; one AML without maturation (AML1); two acute myelomonocytic leukemia (M4-AML) with a prior history of CMML; and one M5-AML secondary to another myelodysplastic syndrome/myeloproliferative neoplasm (MDS/MPN) analyzed during progression under induction treatment. On BM aspiration, blast cells were observed in all cases, associated with pDC, described as smaller with a more mature chromatin, faint basophilic cytoplasm without granulation, sometimes small vacuoles 3059

L. Zalmai et al.

and short pseudopodia (Figure 1A to C). Lymph node involvement was reported in one patient, and skin lesions in four other patients (26.7%). Pathology analysis of skin lesions found myeloblast infiltration without pDC in patients N2 and N8, and pDC proliferation in patients N9 and N11. No clonal aberration was detected on karyotype for nine cases, while three others had trisomy 13 (N1, N9 and N11), two had chromosome 7 defects (N14, N16), and one had Y loss (N20). Two patients with normal karyotype were rearranged for KMT2A (N8, N35), and a deletion of EZH2 was detected by fluorescence in situ hybridization (FISH) in patient N16.

Immunophenotype identified different subpopulations in patients In contrast to BPDCN, cells of interest were heterogeneous, with a significant population of CD45low CD34+ immature myeloid blasts (44%; range, 10-80%) without markers of pDC commitment on the one hand, and an excess of CD4+ CD123+ HLA-DR+ cTCL1+ CD303+ pDC

(15%; range, 4-36%) on the other hand. Consequently, these cases more likely fit the description of pDC-AML (or MPDCP associated with AML according to the WHO classification)8 where pDC were in excess, greater than 4% (median 15%) in our cohort. Indeed, a median of 0.25% of total nucleated cells was detected for the pDC contingent in 15 normal BM aspiration and 0.24% in PB, with similar ranges of 0.02-0.95% and 0.17-0.53%, respectively. For 11 of 15 AML cases (73%), CD45low CD34+ immature blasts were more frequent than pDC, whereas pDC were preponderant in the remaining four cases (range, 10-26%). Monocytes (3%; range 1-10%) were also found in 14 cases, greater in acute myelomonocytic leukemia (M4AML) than in other cases. Interestingly, cDC were detected in patients N7 and N8 (4.8% and 19% of cells respectively) (Figure 2). The CD45low immature blasts were CD34+ (15 of 15, 100% of cases), CD117+ (11 of 15; 73%), TdT+ (five of nine; 56%), expressed myeloid markers such as CD13 (11 of 15, 73%) (Figure 1D and E) and/or CD33 (five of 15,

Figure 2. Immunophenotypic features of plasmacytoid dendritic cell-acute myeloid leukemia. Positive high on 100% of cells in red (>10x4), positive on 100% of cells in orange (10x2 to 10x4), partially positive in light yellow, negative (<20%) in green, not done in grey. CD15 and CD65 are both labeled by fluorescein isothiocyanate (FITC) in the same tube of our panel. ¤Percentage of cells corresponds to flow cytometry, quantification on the sample used for phenotyping, possibly diluted by peripheral blood. All cases exhibited more than 20% of blasts on bone marrow smears. $Analyses performed on sample obtained after induction of chemotherapy, 70 days after diagnosis. cCD3, CD3, CD19, cCD79a were negative for all cases and all fractions. pDC: plasmacytoid dendritic cells; cDC: conventional DC; Mono: monocytes; AML: acute myeloid leukemia; M0-AML: AML with minimal differentiation.

3060

haematologica | 2021; 106(12)

Frequent RUNX1 mutations in acute leukemia + pDC

33%); and less frequently CD15/65 (two cases) or myeloperoxidase (case case). Monocyte lineage markers were only expressed on blasts of M4/5-AML patients. T-cell markers were expressed in two cases (CD7 +/- CD5: patients N2 and N20) and the B-cell marker CD22 in two

other cases (N7 and N14) which was not sufficient to meet the definition of mixed-phenotype acute leukemia, as CD3/cCD3 or CD19/cCD79a were not expressed (Figure 2). CD123 was expressed in 12 cases of 15 (80%) at low level (MFIR: 15.58 [range, 2.5-63.1], MFI: 2,309 [range 465-

Figure 3. Expression of CD123 and cTCL1 on plasmacytoid dendritic cells from plasmacytoid dendritic cell-acute myeloid leukemia. (A) Comparison of mean fluorescence intensity ratio (MFIR) of CD123 between plasmacytoid dendritic cells (pDC) and immature CD34+ blasts in pDC-acute myeloid leukemia (pDC-AML). (B) Comparison of MFI of CD123 between pDC and immature CD34+ blasts in pDC-AML. (C) Comparison of MFIR of cTCL1 between pDC from pDC-AML, blastic pDC neoplasms (BPDCN) and non-neoplastic pDC from healthy donors. (D) Comparison of MFI of cTCL1 between pDC from pDC-AML, BPDCN and non-neoplastic pDC from healthy donors. P-values (unpaired Mann-Whitney test) are marked above.

haematologica | 2021; 106(12)

3061

L. Zalmai et al.

8,674]). In two cases, blasts expressed CD4, whereas CD56 was never expressed. The blastic population did not express markers strongly associated with the pDC (cTCL1, CD303 and CD304) or cDC (CD1c, CD11c) lineages. Associated pDC were constantly CD123+high, CD4+/+low and CD303+ (100%), with CD304 expression in most cases (13 of 15, 87%) and none of them expressed CD56. The immaturity marker CD34 was expressed in 33% of cases, whereas TdT in only one case out of nine tested.

The CD123 expression level was significantly higher on pDC than on the immature CD34+ blasts (MFIR: 172.9 [range, 12.5-482.6] vs. 15.58 [range, 2.5-63.1] on CD34+ blasts, P<0.0001; MFI: 20,836 [range, 10,755-45,746] vs. 2,309 [range, 465-8,674] on CD34+ blasts, P<0.0001) (Figure 3A and B). cTCL1 was expressed in pDC (nine of 12 cases, 75%) at a statistically lower level (MFIR: 5.3 [range, 0.6-22.6]; MFI: 2,473 [range, 487-20,684]) than in the 21 BPDCN cases (MFIR: 34 [range, 6.0-96.0], P<0.0001; MFI: 13,990 [range, 2,186-125,568], P=0.0006)

Figure 4. Mutation profile of plasmacytoid dendritic cell-acute myeloid leukemia. Mutations detected by next-generation sequencing with variant allele frequencies (VAF), or Sanger sequencing (especially for ASXL1). Abnormalities in plasmacytoid dendritic cells-acute myeloid leukemia (pDC-AML) and blastic pDC neoplasms (BPDCN) are depicted in: bright blue (monoallelic mutation); dark blue (biallelic mutation); white (absence of mutation); grey (not available). ¤Percentage of cells corresponds to flow cytometry, quantification on the sample used for phenotyping, possibly diluted by peripheral blood (all cases exhibited more than 20% of blasts on bone marrow smears). $Analyses performed on sample obtained after induction of chemotherapy, 70 days after diagnosis. £Fluorescence in situ hybridization (FISH) 7q36 on case N16: loss of EZH2 in 91 of 200 nuclei. ***VAF not available because ASXL1 c.1934dupG;p.Gly646TrpfsX12 was confirmed by Sanger sequencing. Mono: monocytes.

3062

haematologica | 2021; 106(12)

Frequent RUNX1 mutations in acute leukemia + pDC

and even than in normal pDC (MFIR: 16.9 [range, 3.545.5], P=0.0052; MFI: 8,106 [range, 3,389-3,1544], P=0.0192) (Figure 3C and D). Interestingly, four patients with M0-AML expressed myeloid, B-cell or T-cell markers on both CD34+ blasts and pDC, but with partial and lower expression on pDC than on CD34+ blasts: CD33 and CD22 for N7, CD13 for N1, CD33 for N36, CD5 and CD7 for N20 (Figure 2). A maturation continuum between immature blasts and pDC was observed in some cases (Figure 1D). cDC were identified by the expression of CD1c and CD11c without CD14 and CD64 expression in patients N7 and N8, and were sorted for molecular studies for patient N8 only.

Mutational profile The purity of each sorted fraction is depicted in Online Supplementary Table S3. The NGS panel was informative for all cases, with a number of detected mutations ranging from two to six. Fourteen genes were found to be mutated among the 70 explored (Figure 4), corresponding to transcription factors RUNX1 (11 of 15, 73%); epigenetic modifiers ASXL1 (five of 15, 33%), EZH2 (three of 15, 20%), TET2 (four of 15, 27%), DNMT3A (three of 15, 20%); genes involved in splicing SRSF2 (five of 15, 33%), SF3B1 (two of 15, 13%), U2AF1 (one of 15); RAS pathway CBL (three of 15, 19%), KRAS, PTPN11 (one of 15 each); cytokine signaling FLT3 (three of 15, 19%) as well as other genes, such as PFH6 and WT1 (one of 15 each) (Online Supplementary Table S4). Of note, only one case was mutated for NPM1 (patient N35) and none for CEBPA. In

BPDCN, 17 out of the 21 of the cohort were studied by NGS, with zero to five mutations per case on 17 genes: TET2 (nine of 17, 53%), ASXL1 (six of 17, 41%), ZRSR2 (four of 17, 24%), TP53 (three of 17, 18%), IKZF1, NRAS, SRSF2, IDH1 (two of 17 each, 12%), ZEB2, MET, ETV6, ATM, IKZF3, CXCR4, NOTCH2, KRAS, JAK2 (one of 17 each) (Figure 4). Mutations were systematically found in sorted CD34+ immature blasts, pDC, monocyte and cDC of the same sample, and were not detected in the T-cell fraction, thus confirmed to be a non-neoplastic subpopulation (Figure 4). Variant allele frequencies (VAF) were quite similar between cell fractions. However, VAF of the monocyte subpopulation were lower than in blasts and pDC in two cases (N13 and N36), which may indicate that this mutation is subclonal in monocytes (N13) or that there is a mixture of neoplastic and reactive non-neoplastic monocytes (same VAF difference in all detected mutations for N36). Some mutations were subclonal in both blast and monocyte fractions: KRAS for N19, FLT3 for N20 (Figure 4). Overall, the most frequently mutated genes were the transcription factor RUNX1, splicing genes (SRSF2, SF3B1, U2AF1) and epigenetic modifiers DNMT3A and TET2. Interestingly, RUNX1 mutations concerned all M0-AML, while none of the other cases were mutated (M4/5-AML and M1-AML). Consequently, despite the low number of cases, there was a significant association between the M0AML subtype and RUNX1 mutations (c2 with Yates' correction, X2=10.32, 1df, P=0.0013). The majority of RUNX1 mutations detected were frameshift (n=6) or stop gain (n=3), with a biallelic invalidation in patient N13. RUNX1-

Figure 5. Maturation model in plasmacytoid dendritic cell-acute myeloid leukemia. (A) Representative CD45/SSC dot plot of plasmacytoid dendritic cells-acute myeloid leukemia (pDC-AML), with four populations identified: immature CD34+ blasts in black, pDC in pink, monocytes in green and lymphocyte in blue, with morphologies of these populations depicted above. (B) The maturation model: immature blast cells are mainly proliferative without maturation, but at least part of them conserved MDP (macrophage-DC progenitor)-like potential of maturation leading to variable amounts of clonal pDC, monocytes and conventional DC (cDC).

haematologica | 2021; 106(12)

3063

L. Zalmai et al.

mutated cases involved mainly males (M/F=8/3), with a median age of 70 years (range, 55-87 years).

Discussion The World Health Organization 2017 classification clearly recognizes BPDCN as a form of pDC neoplasm, while MPDCP is still insufficiently described and probably underdiagnosed8. MPDCP mainly concerns CMML and more rarely MDS or AML. Our study focused on pDC-AML, using a phenotypic and mutation characterization on sorted population of 15 cases. Our gateway for the inclusion of cases is the pDC-like morphology of some blastic cells detected by cytologists in French hospitals. This point precludes us from determining the prevalence of pDC-AML among AML, due to a recruitment bias. We highlight in this study an excess of pDC in a context of AML in order to differentiate them from BPDCN and to precisely dtermine their molecular profile, not well-described so far. The clinical presentation of our cohort was close to that of BPDCN, with elderly patients and a clear male predominance, but skin lesions were less frequent (25% of cases vs. 90% in BPDCN), linked to either pDC (two cases) or myeloblast (two cases) infiltration. The excess of pDC in BM aspiration of pDC-AML was well over the pDC infiltrate in healthy donor BM and PB, considered to be below 1% of total nucleated cells24,26,27 and confirmed by our data. Moreover, pDC are frequently gathered on BM smears, compatible with aggregates and in the only case of BM biopsy, clustered nodules of pDC were present, as reported by pathologists with islands of pDC in MPDCP.13,21,24 In the two cases with prior history of CMML, pDC were not in high excess prior transformation on BM smears. However, in the absence of available sample, flow cytometry assays were not assessed at the CMML stage. Blasts exhibited an immature phenotype with expression of CD34, CD117 and TdT, which distinguishes these cases from BPDCN. They also frequently express myeloid markers, and the expression of CD123 is definitely weaker than pDC from the same samples or pDC blasts from BPDCN. Of note, strong pDC or cDC lineage markers (CD303, CD304, cTCL1, ILT7 and CD1c, CD141, CD11c) were never expressed on blasts. Focusing on the 11 M0AML cases, blasts are immature with markers of commitment to the myeloid lineage (CD13, CD33 without MPO) plus expression of lymphoid markers in some cases; this could be a sign of an original progenitor origin or oncogenic dysregulation in pDC-AML compared to other AML. Only four cases of pDC-AML showed higher infiltration of pDC than blasts: three at diagnosis (N2, N8 and N34) and the fourth at day 81 post-induction (N35). In these 15 cases, the phenotype of pDC is different from pDC in healthy donors and in BPDCN. CD56, is not expressed in our 15 cases, similarly to physiological mature pDC, only CD56+ when stimulated by Flt3L during maturation.28,29 Conversely CD56 is almost systematically expressed in BPDCN5,30,31 except for rare cases.7,32,33 Moreover, the pDC in pDC-AML always expressed CD303 (100%) while 20 to 30% of BPDCN are CD303- 6,31 and BPDCN express a lower intensity of CD303 than pDC of pDCAML (data not shown); cTCL-1 expression was substantially lower on pDC of pDC-AML compared to pDC of BPDCN and even pDC of healthy donors (Figure 3B). 3064

Finally, CD34 is sometimes expressed on pDC (33% of cases), whereas CD34 is almost never expressed on BPDCN blasts. This phenotype is closer to normal pDC in some respects (CD56–, CD303+, cTCL1+low). pDC lineage maturation is divided in three stages with progressive acquisition of CD303 and CD304 plus downregulation of CD34 and CD117 and gradual loss of CD13, CD33 or CD22 expression34. In our series, the expression of CD34 (six cases) and CD13, CD33 or CD22 (three cases) makes these pDC closer to the intermediate stage of pDC maturation (CD34+, CD303+, CD304+/-, CD123+). Moreover, aberrant markers of myeloid or lymphoid lineage (CD13, CD33, CD11b, CD22, CD7, CD5) on both blasts and pDC in the same patient support the idea of a common origin of these populations, with maturation of CD34+ blasts towards the pDC lineage by downregulation of CD34 and upregulation of CD123, CD303 and CD304. Reinforcing this hypothesis, we clearly show a maturation continuum between blasts and pDC or monocytes (Figure 1D) with CD34 downregulation and upregulation of CD45, pDC markers (CD123, CD304, CD303) on the pDC cells or CD14, CD64 on monocytes, as recently highlighted.14,15 Another study also described similar cases of leukemia associating immature CD34+ myeloid blasts and CD34+/- CD56– pDC, but considered these cases as an immature subgroup of BPDCN.35 Altogether, a variety of teams described similar cases under different denominations: MPDCP with myeloid neoplasm (MN), pDC proliferations associated with MN, AML/MN with PDC differentiation, AML/MDS-pDC, leukemia of ambiguous lineage, or immature group 1 of BPDCN.14–17,22,23,35,36 There is a need to refine this poorly defined MPDCP, where the “mature” denomination should be omitted because it only refers to the absence of the blastic morphology of BPDCN.8 Our data and others highlight that pDC are present at all stages of maturation in pDC-AML, even CD34+, because they keep their potential for maturation14,15 as already described in pDC-CMML.24 In addition, pDC-AML appear to be quite similar to pPDC-CMML, as clonal monocytes are also frequently detected in our cohort (14 of 15 AML). In this regard, all cases of MPDCP described so far fall within our definition of pDCAML/CMML/MDS. Interestingly, the expression of CD123 on both blasts and pDC opens up the potential to use new therapies targeting CD123, such as tagraxofusp, IMGN632 or CD123 CAR-T cells.37-40 Crucially, these pDC-AML may respond differently from other AML or BPDCN. Very few molecular data have been obtained to assert the clonal link between pDC and leukemic cells in CMML and AML.21-23 In our series, using sorted cell populations, we show that the same mutations are shared by blasts, pDC, monocytes and even cDC, thus confirming the neoplastic origin of pDC and their shared clonal origin with blasts, monocytes and cDC (for one tested case). Of note, no unique mutation was identified in pDC but not in the blast compartment, or vice versa, albeit with the limitation that only 70 genes were tested. Thus, no specific mutation arising during the lineage commitment was highlighted. Blasts seem to be broadly blocked in an immature undifferentiated stage in pDC-AML compared to pDC-CMML, and maintain their ability to mature towards pDC, monocytes or even cDC (Figure 5A and B). The immature CD34+ blasts could be a proliferation, at the very least, of haematologica | 2021; 106(12)

Frequent RUNX1 mutations in acute leukemia + pDC

a bipotent progenitor with DC (pDC and cDC) and with monocyte potential such as the MDP. The expression of CD22, CD2 and CD5 on pDC and blasts in some cases could also suggest that they derive from a granulocytemonocyte-lymphoid progenitor or from AXL+ SIGLEC6+ DC, recently identified.41 The mutation status obtained does not highlight specific genes for all cases. Many of them are frequently mutated in other myeloid malignancies, involved in epigenetics (ASXL1 and TET2), splicing (SRSF2, SF3B1, U2AF1) or the RAS pathway (CBL, KRAS, PTPN11). This mutation profile is only partially similar to BPDCN, because BPDCN can also be mutated for ETV6, TP53, ZEB2, MET, ATM, IKZF3, JAK, NOTCH2 and CXCR4, plus TET2 with a high frequency (Figure 4). Cases classified as M4/5-AML were always transformations of MDS/MPN and had a similar molecular profile including ASXL1, TET2, SRSF2, CBL or PTPN11 mutations, sometimes with additional mutations responsible for acute transformation (NPM1 mutation for N35). Remarkably, RUNX1 is the most frequently mutated gene in pDC-AML (73% of cases), as already described.17 Moreover, it only concerned M0-AML cases in our study and all cases of M0-AML exhibited this mutation (100%). This point is particularly puzzling considering that this prevalence is markedly different from the described epidemiology of 20-30% of RUNX1 mutations in M0AML42,43 and knowing that RUNX1 mutation has only been reported once in BPDCN.44 Although, this study only analyzed a small number of cases and the recruitment bias precludes us from determining the frequency of M0-AML and RUNX1 mutations in pDC-AML. Our 11 PDCP-M0AML cases were 71 years old on average, with a male/female sex ratio of 2.67, consistent with the RUNX1-mutated AML provisional entity.8,42 Unfortunately, given the advanced age of patients with palliative care and low number of cases, prognostic conclusions are impossible. RUNX1 mutations were not detected in the non-neoplastic T-cell fraction, demonstrating that these mutations are somatic. WGA prevented us from definitely obtaining copy number variations, but VAF were higher than 50% for six patients (N8, N9, N11, N12, N16, N19) and a seventh (N13) was double mutated for RUNX1, suggesting secondary alteration of RUNX1, as already described.45 Notably, inhibition of RUNX1 is considered to increase RUNX2 and RUNX3 protein levels, following a complementary compensation mechanism that maintains the entire RUNX family at a constant level.46 Then, RUNX1 invalidation could promote a RUNX2 switch and then a pDC commitment because RUNX2 plays a crucial role in pDC diffentiation.47,48 Further experiments are nevertheless required to confirm this hypothesis. To conclude, our study identifies a group of pDC-AML requiring a differential diagnosis with BPDCN. They are characterized by an immature myeloid population

References 1. Banchereau J, Briere F, Caux C, et al. Immunobiology of dendritic cells. Annu Rev Immunol. 2000;18767-811. 2. Jegalian AG, Facchetti F, Jaffe ES. Plasmacytoid dendritic cells: physiologic roles and pathologic states. Adv Anat Pathol. 2009;16(6):392-404.

haematologica | 2021; 106(12)

(CD34+, CD117+/- CD123+low without expression of pDC markers) associated with an excess of pDC (CD123+high, CD4+) that clearly differ from BPDCN neoplastic cells by no expression of CD56, the possible expression of CD34, higher expression of CD303 and lower expression of cTCL1. Molecular data show that these pDC are neoplastic and not reactive, and the mutational landscape of pDCAML appears distinct from BPDCN, notably with frequent RUNX1 mutations. Moreover, a continuous maturation pattern suggests that these pDC as well as monocytes could arise from the immature CD34+ blasts, potentially with MDP properties (Figure 5B). Finally, this study addresses (i) the frequency with which RUNX1-mutated AML are associated with an excess of pDC, (ii) the type of progenitors involved and (iii) its prognostic or therapeutic impact. These questions warrant investigation in an independent and larger cohort of AML. Disclosures No conflicts of interest to disclose. Contributions FGO designed the study; FGO and FR supervised the study; ED, ED, SB, ED, VB, MLG, DRW, OWB, VS, JF, SB, BD, CMR, PO, VD, MT, JR, MTR, MCJ, VR, ES and FGO procured patient specimens; SB performed cell sorting; TP performed anatomopathological analysis and MACR cytogenetics analysis; LZ, FR and LS performed the molecular experiments; PJV provided assistance in bioinformatics analysis; LZ and FR analyzed NGS data; FR performed statistical analysis; LZ, FR and FGO wrote the original manuscript; CP, CR, MC, FAD, SG, FJ, CF, PS and OA revised the manuscript and provided guidance and expertise. All authors provided input and approved the final version of the manuscript. Acknowledgments The authors would like to thank Hugues Faucheu for helping on the design panel, Véronique Yerly-Motta for helping on the NGS platform and Fiona Ecarnot for English proofreading. Funding This work was supported by Ligue régionale contre le Cancer (CCIRGE-BFC-2016), Fondation ARC (Aides Individuelles DOC20170505805) and Association Laurette Fugain (ALF 2018/08).

3. Naik SH, Sathe P, Park H-Y, et al. Development of plasmacytoid and conventional dendritic cell subtypes from single precursor cells derived in vitro and in vivo. Nat Immunol. 2007;8(11):1217-1226. 4. Onai N, Obata-Onai A, Schmid MA, Ohteki T, Jarrossay D, Manz MG. Identification of clonogenic common Flt3+M-CSFR+ plasmacytoid and conventional dendritic cell pro-

genitors in mouse bone marrow. Nat Immunol. 2007; 8(11):1207-1216. 5. Chaperot L, Bendriss N, Manches O, et al. Identification of a leukemic counterpart of the plasmacytoid dendritic cells. Blood. 2001;97(10):3210-3217. 6. Garnache-Ottou F, Feuillard J, Ferrand C, et al. Extended diagnostic criteria for plasmacytoid dendritic cell leukaemia. Br J

3065

L. Zalmai et al. Haematol. 2009;145(5):624-636. 7. Facchetti F, Cigognetti M, Fisogni S, Rossi G, Lonardi S, Vermi W. Neoplasms derived from plasmacytoid dendritic cells. Mod Pathol. 2016;29(2):98-111. 8. Swerdlow SH, Campo E, Harris NL, et al. WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. Revised 4th Edition. Lyon, France: International Agency for Research on Cancer; 2017. 9. Petrella T, Comeau MR, Maynadié M, et al. “Agranular CD4+ CD56+ hematodermic neoplasm” (blastic NK-cell lymphoma) originates from a population of CD56+ precursor cells related to plasmacytoid monocytes. Am J Surg Pathol. 2002;26(7):852-862. 10. Petrella T, Facchetti F. Tumoral aspects of plasmacytoid dendritic cells: what do we know in 2009? Autoimmunity. 2010; 43(3):210-214. 11. Vitte F, Fabiani B, Bénet C, et al. Specific skin lesions in chronic myelomonocytic leukemia: a spectrum of myelomonocytic and dendritic cell proliferations: a study of 42 cases. Am J Surg Pathol. 2012;36(9):13021316. 12. Dargent J-L, Delannoy A, Pieron P, Husson B, Debecker C, Petrella T. Cutaneous accumulation of plasmacytoid dendritic cells associated with acute myeloid leukemia: a rare condition distinct from blastic plasmacytoid dendritic cell neoplasm. J Cutan Pathol. 2011;38(11):893-898. 13. Orazi A, Chiu R, O’Malley DP, et al. Chronic myelomonocytic leukemia: the role of bone marrow biopsy immunohistology. Mod Pathol. 2006;19(12):1536-1545. 14. Hamadeh F, Awadallah A, Meyerson HJ, Beck RC. Flow cytometry identifies a spectrum of maturation in myeloid neoplasms having plasmacytoid dendritic cell differentiation. Cytometry B Clin Cytom. 2020; 98(1):43-51. 15. Huang Y, Wang Y, Chang Y, et al. Myeloid neoplasms with elevated plasmacytoid dendritic cell differentiation reflect the maturation process of dendritic cells. Cytom A. 2020;97(1):61-69. 16. Wang P, Feng Y, Deng X, et al. Tumor-forming plasmacytoid dendritic cells in acute myelocytic leukemia: a report of three cases and literature review. Int J Clin Exp Pathol. 2017;10(7):7285-7291. 17. Xiao W, Goldberg AD, Famulare C, et al. Acute myeloid leukemia with plasmacytoid dendritic cell differentiation: predominantly Secondary AML, enriched for RUNX1 mutations, frequent cross-lineage antigen expression and poor prognosis. Blood. 2018;132(Suppl 1):S2789. 18. Bodmer A, Menter T, Juskevicius D, et al. Sharing of a PTPN11 mutation by myelodysplastic bone marrow and a mature plasmacytoid dendritic cell proliferation provides evidence for their common myelomonocytic origin. Virchows Arch. 2017;470(4):469-473. 19. Horny HP, Feller AC, Horst HA, Lennert K. Immunocytology of plasmacytoid T cells: marker analysis indicates a unique phenotype of this enigmatic cell. Hum Pathol. 1987;18(1):28-32. 20. Angelot-Delettre F, Biichle S, Ferrand C, et al. Intracytoplasmic detection of TCL1--but not ILT7-by flow cytometry is useful for

3066

blastic plasmacytoid dendritic cell leukemia diagnosis. Cytom A. 2012;81(8):718-724. 21. Vermi W, Facchetti F, Rosati S, et al. Nodal and extranodal tumor-forming accumulation of plasmacytoid monocytes/interferon-producing cells associated with myeloid disorders. Am J Surg Pathol. 2004; 28(5):585-595. 22. Rickmann M, Krauter J, Stamer K, et al. Elevated frequencies of leukemic myeloid and plasmacytoid dendritic cells in acute myeloid leukemia with the FLT3 internal tandem duplication. Ann Hematol. 2011; 90(9):1047-1058. 23. Mohty M, Jarrossay D, Lafage-Pochitaloff M, 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. 24. Lucas N, Duchmann M, Rameau P, et al. Biology and prognostic impact of clonal plasmacytoid dendritic cells in chronic myelomonocytic leukemia. Leukemia. 2019;33(10):2466-2480. 25. Viailly P-J, Mareschal S, Bertrand P, et al. GenerateReports: an Ion Torrent plugin summarizing a whole NGS experiment for clinical interpretation. 2015. 26. Wang W, Khoury JD, Miranda RN, et al. Immunophenotypic characterization of reactive and neoplastic plasmacytoid dendritic cells permits establishment of a 10color flow cytometric panel for initial workup and residual disease evaluation of blastic plasmacytoid dendritic cell neoplasm. Haematologica. 2021 Apr 1;106(4):10471055. 27. Xiao W, Goldberg AD, Famulare CA, et al. Loss of plasmacytoid dendritic cell differentiation is highly predictive for post-induction measurable residual disease and inferior outcomes in acute myeloid leukemia. Haematologica. 2019;104(7):1378-1387. 28. Comeau MR, Van der Vuurst de Vries A-R, Maliszewski CR, Galibert L. CD123bright plasmacytoid predendritic cells: progenitors undergoing cell fate conversion? J Immunol. 1950 2002;169(1):75-83. 29. MacDonald KPA, Munster DJ, Clark GJ, Dzionek A, Schmitz J, Hart DNJ. Characterization of human blood dendritic cell subsets. Blood. 2002;100(13):4512-4520. 30. Feuillard J, Jacob M-C, Valensi F, et al. Clinical and biologic features of CD4(+)CD56(+) malignancies. Blood. 2002; 99(5):1556-1563. 31. Garnache-Ottou F, Vidal C, Biichlé S, et al. How should we diagnose and treat blastic plasmacytoid dendritic cell neoplasm patients? Blood Adv. 2019;3(24):4238-4251. 32. Julia F, Dalle S, Duru G, et al. Blastic plasmacytoid dendritic cell neoplasms: clinicoimmunohistochemical correlations in a series of 91 patients. Am J Surg Pathol. 2014;38(5):673-680. 33. Lamar E, Roggy A, Le Calvez G, et al. Blastic plasmacytoid dendritic cell neoplasm: report of a case with atypical cytology and immunophenotype. J Blood Disord. 2015;2(3):1033. 34. Martín-Martín L, Almeida J, HernándezCampo PM, Sánchez ML, Lécrevisse Q, Orfao A. Immunophenotypical, morphologic, and functional characterization of maturation-associated plasmacytoid dendritic cell subsets in normal adult human bone mar-

row. Transfusion. 2009; 49(8):1692-1708. 35. Martín-Martín L, López A, Vidriales B, et al. Classification and clinical behavior of blastic plasmacytoid dendritic cell neoplasms according to their maturation-associated immunophenotypic profile. Oncotarget. 2015;6(22):19204-19216. 36. Tsagarakis NJ, Kentrou NA, Papadimitriou KA, et al. Acute lymphoplasmacytoid dendritic cell (DC2) leukemia: results from the Hellenic Dendritic Cell Leukemia Study Group. Leuk Res. 2010;34(4):438-446. 37. Mani R, Goswami S, Gopalakrishnan B, et al. The interleukin-3 receptor CD123 targeted SL-401 mediates potent cytotoxic activity against CD34+CD123+ cells from acute myeloid leukemia/myelodysplastic syndrome patients and healthy donors. Haematologica. 2018;103(8):1288-1297. 38. Kovtun Y, Jones GE, Adams S, et al. A CD123-targeting antibody-drug conjugate, IMGN632, designed to eradicate AML while sparing normal bone marrow cells. Blood Adv. 2018;2(8):848-858. 39. Mardiros A, Dos Santos C, McDonald T, et al. T cells expressing CD123-specific chimeric antigen receptors exhibit specific cytolytic effector functions and antitumor effects against human acute myeloid leukemia. Blood. 2013;122(18):3138-3148. 40. Pemmaraju N, Lane AA, Sweet KL, et al. Tagraxofusp in blastic plasmacytoid dendritic-cell neoplasm. N Engl J Med. 2019; 380(17):1628-1637. 41. Villani A-C, Satija R, Reynolds G, et al. Single-cell RNA-seq reveals new types of human blood dendritic cells, monocytes, and progenitors. Science. 2017; 356(6335):eaah4573. 42. Gaidzik VI, Teleanu V, Papaemmanuil E, et al. RUNX1 mutations in acute myeloid leukemia are associated with distinct clinico-pathologic and genetic features. Leukemia. 2016;30(11):2282. 43. Roumier C, Eclache V, Imbert M, et al. M0 AML, clinical and biologic features of the disease, including AML1 gene mutations: a report of 59 cases by the Groupe Français d’Hématologie Cellulaire (GFHC) and the Groupe Français de Cytogénétique Hématologique (GFCH). Blood. 2003; 101(4):1277-1283. 44. Menezes J, Acquadro F, Wiseman M, et al. Exome sequencing reveals novel and recurrent mutations with clinical impact in blastic plasmacytoid dendritic cell neoplasm. Leukemia. 2014;28(4):823-829. 45. Antony-Debré I, Duployez N, Bucci M, et al. Somatic mutations associated with leukemic progression of familial platelet disorder with predisposition to acute myeloid leukemia. Leukemia. 2016; 30(4):999-1002. 46. Morita K, Suzuki K, Maeda S, et al. Genetic regulation of the RUNX transcription factor family has antitumor effects. J Clin Invest. 2017;127(7):2815-2828. 47. Kubota S, Tokunaga K, Umezu T, et al. Lineage-specific RUNX2 super-enhancer activates MYC and promotes the development of blastic plasmacytoid dendritic cell neoplasm. Nat Commun. 2019;10(1):1653. 48. Chopin M, Preston SP, Lun ATL, et al. RUNX2 mediates plasmacytoid dendritic cell egress from the bone marrow and controls viral immunity. Cell Rep. 2016; 15(4):866-878.

haematologica | 2021; 106(12)

ARTICLE

Acute Myeloid Leukemia

The IL1-IL1RAP axis plays an important role in the inflammatory leukemic niche that favors acute myeloid leukemia proliferation over normal hematopoiesis

Ferrata Storti Foundation

Bauke de Boer,1*° Sofia Sheveleva,1* Katja Apelt,1 Edo Vellenga,1 André B. Mulder,2 Gerwin Huls1 and Jan Jacob Schuringa1 1 Department of Experimental Hematology, Cancer Research Center Groningen, University Medical Center Groningen, University of Groningen, and 2Department of Laboratory Medicine, University Medical Center Groningen, University of Groningen, Groningen, the Netherlands

Haematologica 2021 Volume 106(12):3067-3078

°Current address: The Finsen Laboratory, Biotech Research and Innovation Center (BRIC), University of Copenhagen, Copenhagen, Denmark *BB and SS contributed equally as co-first authors.

ABSTRACT

pregulation of the plasma membrane receptor IL1RAP in acute myeloid leukemia (AML) has been reported but its role in the context of the leukemic bone marrow niche is unclear. Here, we studied the signaling events downstream of IL1RAP in relation to leukemogenesis and normal hematopoiesis. High IL1RAP expression was associated with a leukemic GMPlike state, and knockdown of IL1RAP in AML reduced colony-forming capacity. Stimulation with IL1b resulted in the induction of multiple chemokines and an inflammatory secretome via the p38 MAPK and NFkB signaling pathways in IL1RAP-expressing AML cells, but IL1b-induced signaling was dispensable for AML cell proliferation and NFkB-driven survival. IL1RAP was also expressed in stromal cells where IL1b induced expression of inflammatory chemokines and cytokines as well. Intriguingly, the IL1b-induced inflammatory secretome of IL1RAP-expressing AML cells grown on a stromal layer of mesenchymal stem cells affected normal hematopoiesis including hematopoietic stem/progenitor cells while AML cell proliferation was not affected. The addition of Anakinra, an Food and Drug Aministration-approved IL1 receptor antagonist, could reverse this effect. Therefore, blocking the IL1-IL1RAP signaling axis might be a good therapeutic approach to reduce inflammation in the bone marrow niche and thereby promote normal hematopoietic recovery over AML proliferation after chemotherapy.

Introduction The fate of both acute myeloid leukemia (AML) and normal hematopoietic stem cells (HSC) is critically dependent on interactions with the bone marrow (BM) niche.1-4 Recent data has also suggested that malignant cells can remodel the BM microenvironment into a leukemic BM niche favoring leukemogenesis over normal hematopoiesis.4-6 Plasma membrane (PM) proteins are first in line to respond to signals that arise from the BM niche. One of these PM proteins, interleukin-1 receptor accessory protein (IL1RAP), is specifically upregulated in stem/progenitor cells from chronic myeloid leukemia (CML) and AML patients but not on normal CD34+ hematopoietic stem/progenitor cells (HSPC).7-10 This has led to several studies that investigated the targetability of IL1RAP as treatment strategy of chronic myeloid leukemia (CML) and AML,9,11-13 but little is known regarding the cell-intrinsic role of IL1RAP in AML stem/progenitor cells. In addition, the canonical IL1RAP signaling axis in AML with respect to the leukemic BM niche has not been studied extensively. The IL1 family is part of the innate immunity that regulates local inflammatory responses, and its dysregulation may lead to autoinflammatory diseases often caused by excessive IL1b production.14 The IL1 family consists of seven ligands including IL1α, IL1b, IL18, IL33, IL36α, b and g that can bind different IL1 recep-

haematologica | 2021; 106(12)

Correspondence: JAN JACOB SCHURINGA j.j.schuringa@umcg.nl Received: April 9, 2020. Accepted: October 9, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.254987

3067

B. de Boer et al.

tors whereby the majority forms a dimer with the IL1RAP co-receptor.15 Activation of the IL1 receptor complex activates the MyD88/IRAK1/IRAK4/TRAF6/TAK1 signaling pathway, which in turn results in activation of NFkB and mitogen-activated protein kinases (MAPK), including p38.16 Many proteins in this pathway are often upregulated in myelodysplastic syndromes (MDS) and AML, which suggests an important role for this pathway in leukemogenesis.17-19 The IL1 signaling route can induce a variety of inflammatory cytokines and chemokines, which has been shown to be an important factor for development and maintenance of MDS.20 In AML, IL1 has been proposed to enhance proliferation and survival.21-23 Here, we studied the IL1–IL1RAP signaling axis in primary AML patients in the context of the BM niche and revealed that the IL1b-induced secretome impacts on leukemogenesis and most notably on normal hematopoiesis.

Methods Extensive details on the methods used can be found in the Online Supplementary Appendix.

Primary samples Neonatal cord blood (CB), mobilized peripheral blood stem cells (PBSC), normal bone marrow (NBM), mesenchymal stromal cells (MSC), MDS, and AML patient material was obtained as described in the Online Supplementary Methods. All healthy individuals and AML patients gave an informed consent in accordance with the Declaration of Helsinki at the University Medical Center Groningen (UMCG) and Martini Hospital Groningen, the Netherlands. All protocols were approved by the Medical Ethical Committee of the UMCG. Details of AML characteristics used in this study can be found in the Online Supplementary Table S1.

Cell (co-)cultures MSC co-cultures/triple-cultures with CB CD34+, PBSC CD34+ or AML CD34+ cells were performed in Gartner’s medium with the addition of 20 ng/mL granulocyte colony-stimulating factor (G-CSF), N-plate and IL3. Inhibition of the IL1-signaling pathway was established by the addition of 500 ng/mL Anakinra (Swedish Orphan Biovitrum BVBA). In case of triple co-cultures, CB CD34+ cells were transduced with pLKO eGFP to distinguish them from AML cells. Co-cultures were grown at 37˚C and 5% CO2 and demi-populated regularly, replacing 50-80% of the volume with fresh or conditioned Gartner’s medium. Suspension cells were used for further analysis.

Colony-forming cell assay The colony-forming capacity of CB CD34+ cells was evaluated in methylcellulose (1,600 mL) mixed with CM (900 mL) from MSC-AML co-cultures treated for 7 days with or without IL1b and Anakinra.

Lentiviral transfection For knockdown of IL1RAP, cells were transduced with a pLKO eGFP construct, containing short hairpins against IL1RAP sh1: 5’-TGGCCTTACTCTGATCTGGTATTGGACTA-3’, sh2: 5’-CGGGCATTAATTGATTTCCTACTATATTC-3’,10 or scrambled control 5’-TTCTCCGAACGTGTCACGTT-3’.

3068

Results IL1RAP is upregulated in acute myeloid leukemia patients and correlates with a leukemic granulocyte-monocyte progenitor cell signature IL1RAP expression was evaluated in NBM CD34+ cells (n=11) and blasts of AML, acute promyelocytic leukemia (APL) and MDS patients (CD34+ or SSClowCD45mid in case of NPM1 mutant AML with CD34 expression <1%). De novo AML patients (n=110), patients that developed AML from an MDS (s-AML) (n=27), APL patients (n=4), and MDS/MDS-excess blasts (MDS-EB) patients (n=13), all showed heterogeneous but on average significantly upregulated expression of IL1RAP (Online Supplementary Figure S1A and B). Immunofluorescent staining showed clear IL1RAP expression on the PM in AML cells representing different genetic subtypes with high IL1RAP expression (Figure 1A). For 42 AML patients, quantitative proteome data was generated previously7 and we evaluated cellular processes that were enriched in patients with either high or low IL1RAP expression. For 31 of these patients IL1RAP expression was also measured by flow cytometry independently and a significant correlation with our quantitative proteome data was observed (Online Supplementary Figure 1C). Gene ontology (GO) and gene set enrichment analysis (GSEA) was performed on a ranked list based on Pearson correlations of IL1RAP protein expression with the complete quantitative proteome and showed that high expression of IL1RAP was associated with the terms “mitochondrial translation elongation and termination”, “energy production via oxidative phosphorylation” and a “leukemic granulocyte-monocyte progenitor (L-GMP) signature”, whereas AML patients with low IL1RAP expression were associated with “regulation of RNA metabolic processes”, “gene expression” and a glycolysis-enriched HSC-like signature (Figure 1B to C). While IL1A and IL1B were expressed by AML cells at varying levels, no correlations were seen with IL1RAP expression, neither in our quantitative proteome data nor in various published transcriptome datasets (data not shown).24,25 Neither did we find positive correlations of IL1RAP high-expressing AML with RELA, encoding the NFkB transcription factor, MyD88, IRAK1, and TRAF6, all associated with the IL1-IL1RAP downstream pathway (data not shown).

IL1-induced IL1RAP signaling is associated with an inflammatory secretome Data generated in previous studies comparing gene expression profiles of primary AML and NBM CD34+ cells suggested that the IL1RAP pathway might be actively used in many AML patients since components of the IL1RAP-TAK1 signaling pathway were significantly upregulated in AML CD34+ cells while negative feedback proteins such as IL1R2, IL1RN and MARCH8 were significantly downregulated (Online Supplementary Figure S2A).24 In order to investigate the repertoire of targets downstream of the IL1-IL1RAP axis we performed genomewide transcriptome studies in CD34+ of AML#1 (for details of all AML used in this study, see the Online Supplementary Table S1), THP1, and K562 cells that were stimulated with IL1b for 1 hour (Online Supplementary Table S2). While the IL1RAP receptor can be activated by multiple different cytokines including IL1α, IL1b, IL33,

haematologica | 2021; 106(12)

IL1-IL1RAP axis in AML

Figure 1. High IL1RAP plasma membrane protein expression in acute myeloid leukemia is associated with a leukemic granulocyte-monocyte progenitor signature. (A) Immunofluorescent staining and expression measured by flow cytometry (red histogram) of interleukin-1 receptor accessory protein (IL1RAP) in multiple acute myeloid leukemia (AML) cell lines and one primary AML patient. The grey histogram indicates the unstained control. (B and C). Gene ontology analysis (C) and gene set enrichment analysis (GSEA) (D) on a ranked gene list based on label-free quantitative protein expression in 42 primary AML patient samples.7 Genes were ranked based on Pearson correlation with IL1RAP protein expression. Normalized enrichment score (NES) and false discovery rate (FDR) were used to determine significance. *P<0.05, ***P<0.001.

haematologica | 2021; 106(12)

3069

B. de Boer et al.

and IL36 depending on its co-receptor IL1R1, IL1RL1 or IL1RL2, and although, e.g., IL33 has also recently been shown to impact on HSC,26 we initially focused on IL1b in our studies. IL1b was chosen since the IL1R1 receptor appeared to be highly expressed in some AML subsets (data not shown) and due to its potential role in inflammation,27 which we wanted to investigate in more detail. AML#1 and THP1 cells both had high IL1RAP expression whereas K562 showed partial IL1RAP expression (Online Supplementary Figure S2B). We identified 299 genes that were >2-fold upregulated in at least two groups and 32 genes that were >2-fold upregulated in all three groups (Figure 2A). GO analysis on the combination of these genes (331) showed enrichment for genes associated with chemokine signaling, inflammation, response to IL1 and an anti-apoptotic signature (Figure 2B). In addition, GSEA of a ranked gene list of primary AML cells showed signif-

icant enrichment in IL1b-stimulated cells for processes associated with “inflammation”, “chemokine signaling”, “TNF signaling via p38”, “hypoxia” and “AML with a NPM1 mutation” (Figure 2C). We confirmed the upregulation of several of the identified genes in an independent set of IL1RAP+ primary AML samples (Figure 2D; Online Supplementary Figure S2C). We noted that K562 cells could be divided into a IL1RAP+ and IL1RAP– population. We sorted these populations to elucidate the pathways downstream of IL1RAP (Online Supplementary Figure S3A). Both of them showed similar growth kinetics, clonogenicity in CFC assays, and the IL1RAP expression remained stable over time (Online Supplementary Figure S3B to D). As expected, an upregulation of IL8 upon IL1b stimulation was only observed in IL1RAP+ cells (Figure 3A). The IL1b response was blocked with inhibitors against TAK1 and IKK, whereas inhibition

Figure 2. Legend on following page.

3070

haematologica | 2021; 106(12)

IL1-IL1RAP axis in AML

Figure 2. IL1-induced IL1RAP signaling is associated with an inflammatory secretome. (A) Transcriptome analysis of genes 2-fold upregulated in one primary acute myeloid leukemia (AML) patient (AML#1) and two AML cell lines upon stimulation with IL1b. (B) Gene ontology analysis on 331 genes that were 2-fold upregulated in at least two out of three groups (THP1, K562 and AML#1). (C) Gene set enrichment analysis analysis on a ranked gene list of AML#1. Genes were ranked from upregulated to downregulated upon stimulation with IL1b. Normalized enrichment score (NES) and false discovery rate (FDR) were used to determine significance. (D) Quantitative real-time polymerase chain reaction analysis of five primary AML patient samples ± IL1b stimulation. Bars indicate mean ± standard deviation of biological triplicates. Statistical analysis was performed by a one-tailed Student’s t-test. *P<0.05; ** P<0.01; ***P<0.001.

of JNK, p38 and MEK/ERK was less effective (Figure 3A). In accordance, K562 IL1RAP+ cells showed a strong increase in phosphorylation of p65 upon stimulation with IL1b, which could be partially reversed by inhibiting the NFkB pathway with an IKK inhibitor in a dose-dependent manner (Figure 3B to C). These data are in line with previous observations by Bosman et al. who studied the TAK1-NFkB axis.17 Next, we transduced K562 IL1RAP+ cells, K562 IL1RAP– cells (as negative controls), OCIAML3 cells, and THP1 cells with short hairpin RNA (shRNA) against IL1RAP and sorted transduced cells by green fluorescence protein (GFP) positivity (Figure 3D; Online Supplementary Figure S3E and G). Knockdown of IL1RAP did not result in impaired cell proliferation (Figure 3E; Online Supplemental Figure S3H). We observed reduced haematologica | 2021; 106(12)

colony-forming capacity in THP1 cells but not in OCIAML3 cells (Figure 3F). A trend towards reduced CFC capacity upon knockdown of IL1RAP was also observed in primary AML patients #8 and #9 as also observed previously (Figure 3G; Online Supplementary Figure S3I to J).28 Finally, we challenged K562, OCI-AML3 and THP1 cells by serum deprivation and stimulated them with IL1b to determine whether cell viability was controlled by the IL1-IL1RAP axis under stress conditions. Serum starvation-induced loss of viability, however, this was not rescued by addition of IL1b (Figure 3H).

Synergism of the IL1-IL1RAP signaling with other active signaling pathways in acute myeloid leukemia IL1RAP was recently described to be directly associated 3071

B. de Boer et al.

Figure 3. IL1-IL1RAP mediated activation of the NFkB signaling does not rescue proliferation under stress conditions, but IL1RAP knockdown results in reduced colony-forming capacity. (A) Quantitative real-time polymerase chain reaction (qRT-PCR) analysis in K562 IL1RAP+ and IL1RAP– cells treated with TAK1, NFkB, JNK, p38 and MEK inhibitors and subsequently stimulated with IL1b. Bars indicate mean ± standard deviation (SD) of technical triplicates. (B) Western blot of K562 IL1RAP+ and IL1RAP- treated with or without IL1b and/or IKK inhibitor (IKK inh). (C) Quantification of western blot in panel C, p-p65 was normalized to b-ACTIN. (D) Interleukin-1 receptor accessory protein (IL1RAP) mRNA levels measured by qPCR in K562 IL1RAP+ and IL1RAP- cells transduced with short hairpin (shRNA) including a non-targeting control (scr) and shRNA targeting IL1RAP (shI and shII). Bars indicate mean ± SD of technical triplicates. (E) Growth curves of K562 IL1RAP+ and IL1RAP- cells (n=3) ± knockdown of IL1RAP. (F) Colony-forming cell (CFC) output of OCI-AML3 and THP1 ± knockdown of IL1RAP. Bars indicate mean ± SD of technical duplicates. (G) CFC output of acute myeloid leukemia (AML) patient 8 (AML#8) and AML#9 ± knockdown of IL1RAP. Bars indicate mean ± SD of technical duplicates. (H) Growth curves after serum depletion of K562, OCI-AML3 and THP1 cells (n=3) ± IL1b. Statistical analysis in all panels was performed using a Student’s t-test. *P<0.05; **P<0.01; ***P<0.001.

with FLT3 (CD135) and c-kit (CD117) receptors.28 We analyzed co-expression of IL1RAP with signaling receptors including CD135, CD117 and IL3 receptor (CD123) in immature AML stem progenitor cells (CD34+ or SSClowCD117+ in case of CD34 expression <1%) by flow cytometry in a cohort of 124 primary AML patients of which four representative examples are shown (Figure 4A). Subsequently, patients were defined as single positive, double positive, or double negative when at least 50% of the cells resided within either one of the gates (Figure 4A and B). These analyses revealed that 21% of the patients were CD135+IL1RAP+, while 4% expressed IL1RAP without CD135 (Figure 4B); 29.8% and 28.1% of the patients were IL1RAP+CD117+ and IL1RAP+CD123+, respectively, and we did not identify patients that 3072

expressed IL1RAP without any detectable CD117 or CD123 (Figure 4B). Expression of IL1RAP, as quantified by flow cytometry (mean fluorescense intensity [MFI]), correlated significantly with CD123 expression, and to a lesser degree with CD135, but no significant correlations were found with CD117 (Figure 4C; Online Supplementary Figure 4A). These observations indicate that IL1RAP signaling can co-occur in cells that are also hardwired for FLT3 ligand (FLT3L), stem cell factor (SCF) and/or IL3-induced signal transduction, and in fact might influence those pathways as well. In addition, Muto et al. showed that MDS HSPC switch from canonical to non-canonical NFkB signaling in response to inflammatory signals like IL1b, which might also occur in AML cells.29 In order to investigate both hypotheses, we isolated CD34+ cells from AML haematologica | 2021; 106(12)

IL1-IL1RAP axis in AML

Figure 4. Synergism of the IL1-IL1RAP signaling with other active signaling pathways in acute myeloid leukemia. (A) Gating strategy of co-expressing Interleukin1 receptor accessory protein (IL1RAP), CD135, CD123 and CD117 measured by flow cytometry within four representative primary acute myeloid leukemia (AML) patients. (B) Pie-chart showing co-expression of IL1RAP with CD135, CD123 and CD117 in blasts of 124 primary acute myeloid leukemia (AML) patients. At least 50% of the total amount of cells in a specific group was used as a cutoff to include a patient in a certain group; otherwise, patients were called “unclassified”. (C) Pearson correlation of IL1RAP with CD135, CD123 and CD117 based on mean fluorescense intensity (MFI) (n=124). (D) Western blot of primary patient CD34+ blasts positive for IL1RAP, CD135, CD117 and CD123. Cells were stimulated with IL1b, FLT3L, stem cell factor (SCF), IL3 or a combination of these cytokines. (E) Quantification of western blot in panel D. p-p65 was normalized to total H3, pSTAT5 and p-p38 were normalized to b-ACTIN and pAKT was normalized to AKT. (F) Western blot of primary patient CD34+ blasts with low IL1RAP and CD135 expressing but positive for CD117 and CD123. Cells were stimulated with IL1b, FLT3L, SCF, IL3 or a combination of these cytokines. (G) Quantification of western blot in panel (F). p-p65 was normalized to total H3, pSTAT5 and p-p38 were normalized to b-ACTIN and pAKT was normalized to AKT. ***P<0.001.

haematologica | 2021; 106(12)

3073

B. de Boer et al.

Figure 5. Legend on following page.

3074

haematologica | 2021; 106(12)

IL1-IL1RAP axis in AML

Figure 5. The IL1-IL1RAP signaling pathway affects normal hematopoiesis but not acute myeloid leukemia cell growth in the context of human mesenchymal stromal cells. (A) Growth curve of acute myeloid leukemia (AML) patient 1 (AML#1) CD34+ cells in liquid culture (left) and on a stromal layer of mesenchymal stromal cells (MSC) (right) ±IL1b in different concentrations. The arrow indicates from whereon IL1b was added. (B) Growth curve of CB CD34+ cells in liquid culture (left) and on a stromal layer of MSC (right) ±IL1b in different concentrations. The arrow indicates from whereon IL1b was added. (C) Schematic overview of experimental setup of co-cultures and conditioned medium (CM) transferring. (D and E) Growth curve (left) and cumulative cell number on day 14 (right) of CB CD34+cells in liquid (D) and on a stromal layer of MSC (E) with the addition of CM from a MSC culture (CM MSC), AML#1 CD34+ liquid culture (CM AML), AML#1 CD34+ MSC co-culture (CM MSC + AML), and AML#1 CD34+ MSC co-culture with the addition of 10 ng/mL IL1b (CM MSC + AML + IL1b). CM was added at day 4 (arrow) and every following demi-population. (F) Growth curve (left) and cumulative cell number on day 21 (right) of CD34+ peripheral blood stem cells (PBSC) grown on MSC with the CM from AML#18 (co-)cultures. Treatment conditions were similar to the once described in the legend of Figure 5D and E, adding one condition including AML CD34+ co-culture with the addition of 10 ng/mL IL1b and 500 ng/mL Anakinra (CM MSC + AML + IL1b + Anakinra). CM was added at day 0 and at every following demi-population. (G) Colony-forming cell (CFC) assay of cord blood (CB) CD34+ treated with CM of IL1RAPhigh AML (AML#9 and AML#19) and IL1RAPlow AML (AML#20 and AML#21), which were cultured for 7 days on a MSC stromal-layer in the presence or absence of IL1b and Anakinra, before CM was harvested. Data of two biological duplicates are shown relative to the untreated condition. (H) Interleukin-1 receptor accessory protein (IL1RAP) expression on MSC measured by flow cytometry. (I) Quantitative real-time polymerase chain reaction of MSC stimulated with and without IL1b. Statistical analysis was performed using a Student’s t-test. (J) Cumulative cell number on day 21 of CD34+ PBSC grown on MSC (experimental setup identical to panel F) including conditions AML + IL1b and MSC + IL1b. Gartner’s and CM MSC + AML + IL1b (identical to panel F) has been added for direct comparison. (K) Growth curve (left) and cumulative cell number on day 11 (right) of CB CD34+ cells in triple co-culture with MSC and AML#16 CD34+ cells ±IL1b (L) Percentage of CB cells in triple co-culture with MSC and AML#22 ±IL1b at day 4. (M) Schematic model how AML cells might impact on normal hematopoiesis in the bone marrow niche, in part via the IL1-IL1RAP axis. Statistical analysis in all panels was performed using a Student’s t-test.* P<0.05; **P<0.01; ***P<0.001.

patients #10-13 and stimulated them with IL1b, FLT3L, SCF, IL3, or a combination thereof. Activation of signal transduction pathways was determined by western blotting for phosphorylated p65 (p-p65), phosphorylated p100 (p-p100), p52/p100, RelB, phosphorylated STAT5 (pSTAT5), phosphorylated p38 (p-p38), and phosphorylated AKT (pAKT), (Figure 4D to G; Online Supplementary Figure S4B to E). Figure 4D to E illustrates AML patient #10 that expressed high levels of IL1RAP, FLT3, CD117 and CD123 within the CD34+ blast compartment. Stimulation with IL1b resulted in downstream activation of the canonical NFkB-(p65) and p38 pathways, whereas we observed limited activation of non-canonical NFkB (p52/p100, RelB, and p-p100), IL3 activated the STAT5 signaling pathway, FLT3L and SCF both activated the PI3K-AKT signaling pathway. Co-stimulation of IL1b with SCF resulted in a slightly increased downstream activation of the PI3K-AKT signaling, although no additive effects were seen on pSTAT5 or p-p38 (Figure 4D to E). In AML#13, which expressed IL1RAP, CD123 and CD117 and low levels of CD135, we observed that IL1b led to activation of p-p38 and a moderate activation of only canonical NFkB, IL3 induced pSTAT5, but again the IL3 signaling was not further potentiated by co-stimulation with IL1b (Online Supplementary Figure 4B to C). The third example (AML#11) showed increased p-p65 levels upon IL1b stimulation whereas the non-canonical NFkB was active at baseline but not further enhanced upon IL1b stimulation as read out by p52/p100 levels (Figure 4F to G). AML#11 responded to FLT3L, SCF and IL3 stimulation, but co-stimulation with IL1b did not further enhance activation of any of these signaling pathways (Figure 4F and G). The fourth patient sample (AML#12) had limited IL1RAP expression and showed no activation of neither canonical nor non-canonical NFkB upon stimulation with IL1b, no additional activation of the p-p38 signal that was already highly activated at baseline, and no effects of co-stimulation with IL1b were seen on IL3-induced pSTAT5 (Online Supplemental Figure S4D to E). Further evaluation of the non-canonical NFκB pathway showed that some primary AML patients already have high baseline non-canonical NFκB activity compared to THP1 cells. We did not observe differences in baseline non-canonical NFkB activity in K562 IL1RAP+ and K562 IL1RAP- cells (Online Supplementary Figure S4F and G). Neither did we observe synergistic effects of IL1b with FLT3L or IL3 in THP1 cells on downstream phosphorylation of ERK (pERK), cJUN (phaematologica | 2021; 106(12)

cJUN) and AKT (pAKT) (Online Supplementary Figure S4H, I, K and L). Addition of the anti-IL1RAP monoclonal antibody (α-IL1RAP MAb), that could partly rescue IL1induced upregulation of IL8 and CXCL1, did not alter phosphorylation levels of AKT and cJUN upon stimulation with IL1b in combination with FLT3L or IL3 (Online Supplementary Figure S4J to L). In summary, both the canonical NFkB and p38 pathway can be activated by IL1b in primary AML patients whereby downstream canonical NFkB signaling might be correlated to IL1RAP expression levels. We did not observe synergism of IL1b with other signaling molecules including FLT3L, SCF and IL3 on downstream phosphorylation of p65, p38, STAT5, AKT, ERK and cJUN.

The IL1-IL1RAP signaling axis reduces proliferation of normal hematopoietic cells but does not affect acute myeloid leukemia cell growth We showed that IL1RAP is often upregulated in AML cells and that the receptor is functional, inducing an inflammatory gene expression signature. However, we found no evidence for a cell-intrinsic role for the IL1IL1RAP axis in controlling cell proliferation or survival under stress conditions. Therefore, we wondered whether the inflammatory secretome induced via the IL1-IL1RAP signaling route plays a role in inducing an inflammatory BM niche that favors AML cell proliferation over normal hematopoiesis. Therefore, IL1RAP-expressing AML CD34+ cells were grown in liquid culture conditions or on MSC, in the absence or presence of IL1b. The addition of IL1b had limited impact on the proliferative capacity of primary AML CD34+ cells, neither when grown in liquid culture conditions nor when grown in co-culture with MSC (Figure 5A). Similarly, IL1b did not affect the proliferation of CB CD34+ cells when grown in liquid culture conditions, however, a marked, dose-dependent, reduction in proliferation was noted when cultured on MSC (Figure 5B). From the MSC/AML cultures, the conditioned medium (CM) was harvested and transferred to CBderived or PB-derived CD34+ cell cultures to determine the effects on proliferation and differentiation (Figure 5C). CM was harvested every time cultures were demi-populated by taking a third of the volume and transferring it to CB- or PB-derived CD34+ cell cultures in a 1:1 ratio (v/v) (see the Online Supplementary Methods for more details) The addition of the CM of AML#1 grown in liquid culture conditions did not affect CB CD34+ growth, neither 3075

B. de Boer et al.

in liquid culture conditions nor when grown on MSC (Figure 5D-E). Possibly, this relates to the high levels of TGFb, BMP and angiopoietins that are known to be secreted by stromal cells. These factors negatively impact on the cell cycle and can preserve stemness, which would coincide with the observation that CD34+ cell populations were better maintained under stromal coculture conditions (Online Supplementary Figure 5C and D). CM harvested from MSC, only marginally reduced CB cell growth in liquid culture and not when grown on MSC (Figure 5D to E). Strikingly, the CM harvested from AML#1 grown on MSC negatively impacted on normal CB proliferation, both in liquid cultures and on MSC, which was even further aggravated when the AML cultures were treated with IL1b (Figure 5 D and E; Online Supplementary Figure S5A and B). The percentage of DAPI-positive CB cells was slightly increased upon treatment with CM from AML cells, compared to the addition of Gartner’s medium (Online Supplementary Figure 5B). Similar results on cell proliferation and survival were obtained using adult PB CD34+ cells that were cultured in presence of CM from a different AML (AML#18; Figure 5F; Online Supplementary Figure 5E). Importantly, the negative phenotype could be partially rescued by addition of Anakinra, an FDAapproved inhibitor of the IL1 receptor, and consequently, the IL1RAP pathway (Figure 5F and Online Supplementary Figure S5F to H). The CB CD34+ percentage, but not absolute CD34+ cell counts, were increased 3 days after addition of CM of AML co-cultures (Online Supplementary Figure S5C and D). We observed no clear differences in the CD34+CD38+/CD34+CD38– distribution after addition of CM to CB cells grown in liquid culture or on MSC, respectively (data not shown). For a more functional HSPC analysis, we performed a CFC assay with CB CD34+ cells supplemented with CM of AML-MSC co-cultures. Two high-expressing IL1RAP (AML#9 and #19) and two low-expressing IL1RAP (AML#20 and #21) AMLs were grown for 7 days on a stromal layer of MSCs with or without IL1b/Anakinra (Online Supplementary Figure S5I). IL8 was strongly upregulated by IL1b in the IL1RAP+ AML cells (AML#19), but not in the IL1RAP– AML cells (AML #20), and Anakinra abrogated this effect (Online Supplementary Figure S5F). At day 7, CM was harvested and added to the methylcellulose mixture in a 1:2 v/v ratio. Only the addition of CM from AMLs with a high IL1RAP expression resulted in a significant reduction of colony-forming potential of CB CD34+ cells, which was completely reversed by the addition of Anakinra (Figure 5G; Online Supplementary Figure 5J). Likely, the stromal cells play an important role in mediating the negative effects of IL1b on the proliferation of healthy CD34+ HSPCs. Surprisingly, MSCs expressed high levels of IL1RAP and were responsive to IL1b comparable to what was seen for IL1RAP+ AML cells (Figure 5H-I). Both the CM of AML cells as well as MSCs treated with IL1b affected cell proliferation of healthy HSPCs, which was further enhanced when CM from AML + MSC + IL1b was used (Figure 5J, Online Supplementary Figure 5H). This would argue that both the stromal cells as well as AML cells participate in generating an inflammatory environment. In addition, two triple co-cultures were performed with MSC, AML CD34+ cells of two IL1RAP+ AML patients (AML#16 and #22), and CB CD34+ cells (Figure 5J to K; Online Supplementary Figure S5K to L). We observed reduced CB proliferation in the presence of AML#16, 3076

which was further reduced by the addition of IL1b (Figure 5J). CB in co-culture with AML#22 without IL1b did not result in reduced CB proliferation, but this was the case when cultured in the presence of IL1b (Online Supplementary Figure S5K). We observed a slight increase of Annexin V+ cells at day 16, which might not account fully for the reduced normal hematopoiesis. These data propose a model in which interplay between AML cells and MSC results in an inflammatory secretome that impairs normal hematopoiesis (Figure 5K). This negative phenotype of normal hematopoiesis is partly IL1-induced and the addition of exogenous IL1b can aggravate this observed phenotype.

Discussion The fate of normal and leukemic stem cells critically depends on signals arising from the BM niche.30-32 Many of them initiate signal transduction in target cells by binding to PM receptors. The identification and functional analyses of such AML-specific PM proteins will help our understanding of leukemia initiation, progression and maintenance. Here, we studied IL1RAP, which was associated with a L-GMP like signature suggesting that cells from IL1RAP+ AML patients differ significantly in their origin, metabolic state, and cell cycle state, compared IL1RAP–/low patients.33,34 High IL1RAP expression in normal karyotype AML patients showed worsened overall survival suggesting that indeed these are different AML subtypes.35 Also at the subclonal level, IL1RAP expression is associated with different biology, as an IL1RAP+ NRASmutated subclone differed significantly from an IL1RAPWT1-mutated subclone, while both subclones contained similar founder mutations.7 The IL1RAP+ NRAS-mutated subclone was strongly enriched for L-GMP and inflammatory gene signatures. The connection between dysregulation of the Ras-pathway and IL1 signaling was previously investigated in non-small cell lung cancer in the context of GATA2 dependency.36 Combined, these data indicate that IL1RAP expression can be used to distinguish distinct biological characteristics of leukemic clones. A recent study indicated that IL1RAP can also co-dimerize with CD117 and CD135 receptors, which resulted in an amplification of survival and proliferation signaling in AML.28 In our dataset, FLT3 and CD123, but not CD117 expression, correlated well with IL1RAP expression. Stimulation of these receptors with FLT3L, IL3 and SCF showed heterogeneous downstream activation between different AML but no clear synergism was observed when FLT3, CD123 and CD117 were co-stimulated with IL1b. While the hypothesis that IL1RAP receptors can interact with other receptors and thereby affect signal transduction is certainly intriguing, further studies are required to resolve these issues. We showed that the IL1RAP pathway was functional in primary AML CD34+ cells and could be activated by IL1b. Various pathways were activated downstream, including canonical NFkB signaling. Muto et al. showed that MDS HSPC could switch to non-canonical NFkB signaling in response to inflammation, resulting in a competitive advantage over normal HSPC.29 In general, we observed rather low levels of proteins involved in active non-canonical NFkB signaling, which was not further increased upon stimulation with IL1b suggesting that this switch might be haematologica | 2021; 106(12)

IL1-IL1RAP axis in AML

less relevant for AML stem cells. Previously, studies showed reduced clonogenicity in AML cell lines as well as reduced engraftment capacity upon knockdown or inhibition of IL1RAP using antagonistic antibodies.28,35 We also found a reduction in CFC frequency upon knockdown of IL1RAP in THP1 cells or primary IL1RAP+ AML cells, but not in K562 or OCI-AML3 cells. It was also shown that IL1RAP-targeting antibodies resulted in reduced cell proliferation, however, we did not observe any difference in cell proliferation upon knockdown of IL1RAP.28,35 Possibly, a reduction in expression is dissimilar to an antibody-mediated block in signaling. Importantly, we did observe an IL1b–induced activation of canonical NFkB signaling in IL1RAP-expressing AML, but this did not result in NFkBdriven survival under stress conditions such as serum deprivation. Intriguingly, the IL1b-induced inflammatory secretome of AML cells grown on MSC affected normal hematopoietic proliferation and HSPC clonogenicity, while AML cells were much less affected. This observation was specific for IL1RAP-expressing AML cells that were cultured on stromal cells as CM of AML cells and MSC alone did not affect normal hematopoiesis. We observed only a mild increase in DAPI percentage or Annexin V positivity in CB CD34+ cells, suggesting that an increase in apoptosis does not fully explain the loss of cell growth. Recently, Waclawiczek et al. showed that transcriptionally remodeled MSC, due to the presence of AML cells, resulted in suppression of HSPC but did not affect their viability.37 Along the same lines, MirakiMoud et al. suggested that AML cells do not impair the survival of normal HSC but do inhibit their differentiation, from which HSC can recover once removed from the leukemic environment.38 Single-cell sequencing studies of the mouse BM provided a very detailed description of the cellular heterogeneity within the BM niche, which was remodeled upon stress or MLL-AF9 leukemia engraftment.39,40 This remodeling affected function and maturation of BM stromal cells resulting in the loss of signaling molecules known to be essential for normal hematopoiesis.39 Similarly, studies showed BM remodeling via exosome secretion, TGF-b, Notch and inflammatory signals.41,42 Together, these findings suggest that leukemic cells can impact on normal hematopoiesis in multiple ways. Likely, there are more proteins secreted by AML and/or MSC that also influence normal hematopoietic proliferation, as Anakinra could not fully rescue the negative phenotype. For example, Carter and colleagues showed that IL1b can result in a Cox2-dependent secretion of prostaglandin E2 (PGE2) by MSC, which ultimately resulted in b-catenin-mediated augmented chemotherapy resist-

References 1. Morrison SJ, Scadden DT. The bone marrow niche for haematopoietic stem cells. Nature. 2014;505(7483):327-334. 2. Baryawno N, Przybylski D, Kowalczyk MS, et al. A cellular taxonomy of the bone marrow stroma in homeostasis and leukemia. Cell. 2019;177(7):1915-1932.e16. 3. Rizo A, Vellenga E, G. dH, Schuringa JJ. Signaling pathways in self-renewing hematopoietic and leukemic stem cells: do all stem cells need a niche? Hum Mol Genet. 2006;15 Spec No 2:R210-R219. 4. Goulard M, Dosquet C, Bonnet D. Role of

haematologica | 2021; 106(12)

ance in AML cells.43 Besides inflammatory factors, GATA2 was also upregulated ~2.6-fold in AML cells upon IL1b stimulation. p38-dependent GATA2 activation has been associated with poor overall survival and increased transcriptional activation of IL1b and CXCL2.44,45 We hypothesize that this p38-dependent activation of GATA2 is part of a positive feedback loop in IL1RAP-expressing AML that results in an inflammatory niche. The formation of an inflammatory niche possibly plays a role in the early stages of leukemia development as well. It has been shown that hematopoietic clones harboring a pre-leukemia mutation in Tet methylcytosine dioxygenase 2 (TET2) can outgrow non-mutated clones after inflammatory stress, which in turn might be aggravated by the fact that TET2 knockout mice show increased levels of inflammatory proteins including IL1b, IL-6 and chemokines including Cxcl1-3 and Pf4.46,47 Upon aging, the BM niche changes and becomes more senescent and as a result, via a senescent-associated secretory phenotype (SASP), more inflammatory.48-50 Although the data are limited, it is enticing to consider that an inflammatory BM niche might accelerate clonal expansion and that the IL1IL1RAP signaling axis plays an important role already in early stages of leukemic initiation. Overall, our study contributes to the understanding of the role that plasma membrane receptors play in the leukemic BM niche. Such insights might aid further development of therapies aimed at specifically targeting factors that are essential for leukemogenesis. Inhibition of the IL1-IL1RAP signaling axis might be a good therapeutic approach to reduce inflammation in the BM niche and thereby promote normal hematopoietic recovery over AML proliferation after chemotherapy. Disclosures No conflicts of interest to disclose. Contributions BdB, SS and JJS conceived the study concept; BdB, SS, ABM and KA carried out experiments; GH and EV provided funding; BdB, SS and KA analyzed the data; BB, SS and JJS wrote the original draft; BdB, SS, GH, EV, ABM and JJS co-wrote, reviewed and edited the manuscript; JJS acquired funding and supervised the project. Funding This work was supported by a grant from the European Research Council (ERC-2011-StG 281474-huLSCtargeting) awarded to JJS. The UMCG/MPDI program is acknowledged for the scholarship awarded to SS.

the microenvironment in myeloid malignancies. Cell Mol Life Sci. 2018;75(8):13771391. 5. Duarte D, Hawkins ED, Akinduro O, et al. Inhibition of endosteal vascular niche remodeling rescues hematopoietic stem cell loss in AML. Cell Stem Cell. 2018;22(1):6477.e6. 6. Schepers K, Campbell TB, Passegue E. Normal and leukemic stem cell niches: insights and therapeutic opportunities. Cell Stem Cell. 2015;16(3):254-267. 7. de Boer B, Prick J, Pruis MG, et al. Prospective isolation and characterization of genetically and functionally distinct

AML subclones. Cancer Cell. 2018;34(4): 674-689.e8. 8. Bonardi F, Fusetti F, Deelen P, van GD, Vellenga E, Schuringa JJ. A Proteomics and transcriptomics approach to identify leukemic stem cell (LSC) markers. Mol Cell Proteomics. 2013;12(3):626-637. 9. Jaras M, Johnels P, Hansen N, et al. Isolation and killing of candidate chronic myeloid leukemia stem cells by antibody targeting of IL-1 receptor accessory protein. Proc Natl Acad Sci U S A. 2010;107(37):1628016285. 10. Barreyro L, Will B, Bartholdy B, et al. Overexpression of IL-1 receptor accessory

3077

B. de Boer et al. protein in stem and progenitor cells and outcome correlation in AML and MDS. Blood. 2012;120(6):1290-1298. 11. Askmyr M, Agerstam H, Hansen N, et al. Selective killing of candidate AML stem cells by antibody targeting of IL1RAP. Blood. 2013;121(18):3709-3713. 12. Agerstam H, Hansen N, von Palffy S, et al. IL1RAP antibodies block IL-1-induced expansion of candidate CML stem cells and mediate cell killing in xenograft models. Blood. 2016;128(23):2683-2693. 13. Agerstam H, Karlsson C, Hansen N, et al. Antibodies targeting human IL1RAP (IL1R3) show therapeutic effects in xenograft models of acute myeloid leukemia. Proc Natl Acad Sci U S A. 2015; 112(34):10786-10791. 14. Sims JE, Smith DE. The IL-1 family: regulators of immunity. Nat Rev Immunol. 2010; 10(2):89-102. 15. Garlanda C, Dinarello CA, Mantovani A. The interleukin-1 family: back to the future. Immunity. 2013;39(6):1003-1018. 16. Dinarello CA. Overview of the IL-1 family in innate inflammation and acquired immunity. Immunol Rev. 2018;281(1):8-27. 17. Bosman MC, Schepers H, Jaques J, et al. The TAK1-NF-kappaB axis as therapeutic target for AML. Blood. 2014;124(20):31303140. 18. Smith MA, Choudhary GS, Pellagatti A, et al. U2AF1 mutations induce oncogenic IRAK4 isoforms and activate innate immune pathways in myeloid malignancies. Nat Cell Biol. 2019;21(5):640-650. 19. Gasparini C, Celeghini C, Monasta L, Zauli G. NF-kappaB pathways in hematological malignancies. Cell Mol Life Sci. 2014; 71(11):2083-2102. 20. Li AJ, Calvi LM. The microenvironment in myelodysplastic syndromes: niche-mediated disease initiation and progression. Exp Hematol. 2017;55:3-18. 21. Hoang T, Haman A, Goncalves O, et al. Interleukin 1 enhances growth factordependent proliferation of the clonogenic cells in acute myeloblastic leukemia and of normal human primitive hemopoietic precursors. J Exp Med. 1988;168(2):463-474. 22. Delwel R, van Buitenen C, Salem M, et al. Interleukin-1 stimulates proliferation of acute myeloblastic leukemia cells by induction of granulocyte-macrophage colonystimulating factor release. Blood. 1989; 74(2):586-593. 23. Bradbury D, Bowen G, Kozlowski R, Reilly I, Russell N. Endogenous interleukin-1 can regulate the autonomous growth of the blast cells of acute myeloblastic leukemia by inducing autocrine secretion of GMCSF. Leukemia. 1990;4(1):44-47. 24. de Jonge HJ, Woolthuis CM, Vos AZ, et al. Gene expression profiling in the leukemic

3078

stem cell-enriched CD34(+) fraction identifies target genes that predict prognosis in normal karyotype AML. Leukemia. 2011;25(12):1825-1833. 25. Cancer Genome Atlas Research N, Ley TJ, Miller C, et al. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 2013; 368 (22):2059-2074. 26. Capitano ML, Griesenauer B, Guo B, Cooper S, Paczesny S, Broxmeyer HE. The IL-33 Receptor/ST2 acts as a positive regulator of functional mouse bone marrow hematopoietic stem and progenitor cells. Blood Cells Mol Dis. 2020;84:102435. 27. Dinarello CA. Overview of the IL-1 family in innate inflammation and acquired immunity. Immunol Rev. 2018;281(1):8-27. 28. Mitchell K, Barreyro L, Todorova TI, et al. IL1RAP potentiates multiple oncogenic signaling pathways in AML. J Exp Med. 2018;215(6):1709-1727. 29. Muto T, Walker CS, Choi K, et al. Adaptive response to inflammation contributes to sustained myelopoiesis and confers a competitive advantage in myelodysplastic syndrome HSCs. Nat Immunol. 2020; 21(5): 535-545. 30. Schepers K, Campbell TB, Passegue E. Normal and leukemic stem cell niches: insights and therapeutic opportunities. Cell Stem Cell. 2015;16(3):254-267. 31. Crane GM, Jeffery E, Morrison SJ. Adult haematopoietic stem cell niches. Nat Rev Immunol. 2017;17(9):573-590. 32. Mendez-Ferrer S, Bonnet D, Steensma DP, et al. Bone marrow niches in haematological malignancies. Nat Rev Cancer. 2020; 20(5):285-298. 33. Goardon N, Marchi E, Atzberger A, et al. Coexistence of LMPP-like and GMP-like leukemia stem cells in acute myeloid leukemia. Cancer Cell. 2011;19(1):138-152. 34. Ye M, Zhang H, Yang H, et al. Hematopoietic differentiation is required for initiation of acute myeloid leukemia. Cell Stem Cell. 2015;17(5):611-623. 35. Barreyro L, Will B, Bartholdy B, et al. Overexpression of IL-1 receptor accessory protein in stem and progenitor cells and outcome correlation in AML and MDS. Blood. 2012;120(6):1290-1298. 36. Kumar MS, Hancock DC, Molina-Arcas M, et al. The GATA2 transcriptional network is requisite for RAS oncogene-driven nonsmall cell lung cancer. Cell. 2012; 149(3):642-655. 37. Waclawiczek A, Hamilton A, RouaultPierre K, et al. Mesenchymal niche remodeling impairs hematopoiesis via stanniocalcin 1 in acute myeloid leukemia. J Clin Invest. 2020;130(6):3038-3050. 38. Miraki-Moud F, Anjos-Afonso F, Hodby KA, et al. Acute myeloid leukemia does not

deplete normal hematopoietic stem cells but induces cytopenias by impeding their differentiation. Proc Natl Acad Sci U S A. 2013;110(33):13576-13581. 39. Baryawno N, Przybylski D, Kowalczyk MS, et al. A cellular taxonomy of the bone marrow stroma in Hhmeostasis and leukemia. Cell. 2019;177(7):1915-1932.e16. 40. Tikhonova AN, Dolgalev I, Hu H, et al. The bone marrow microenvironment at singlecell resolution. Nature. 2019;569(7755):222228. 41. Kumar B, Garcia M, Weng L, et al. Acute myeloid leukemia transforms the bone marrow niche into a leukemia-permissive microenvironment through exosome secretion. Leukemia. 2018;32(3):575-587. 42. Schepers K, Pietras EM, Reynaud D, et al. Myeloproliferative neoplasia remodels the endosteal bone marrow niche into a selfreinforcing leukemic niche. Cell Stem Cell. 2013;13(3):285-299. 43. Carter BZ, Mak PY, Wang X, et al. An ARCregulated IL1beta/Cox-2/PGE2/betaCatenin/ARC circuit controls leukemiamicroenvironment interactions and confers drug resistance in AML. Cancer Res. 2019; 79(6):1165-1177. 44. Vicente C, Vazquez I, Conchillo A, et al. Overexpression of GATA2 predicts an adverse prognosis for patients with acute myeloid leukemia and it is associated with distinct molecular abnormalities. Leukemia. 2012;26(3):550-554. 45. Katsumura KR, Ong IM, DeVilbiss AW, Sanalkumar R, Bresnick EH. GATA factordependent positive-feedback circuit in acute myeloid leukemia cells. Cell Rep. 2016;16(9):2428-2441. 46. Cai Z, Kotzin JJ, Ramdas B, et al. Inhibition of inflammatory signaling in Tet2 mutant preleukemic cells mitigates stress-induced abnormalities and clonal hematopoiesis. Cell Stem Cell. 2018;23(6):833-849.e5. 47. Jaiswal S, Natarajan P, Silver AJ, et al. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017;377(2):111-121. 48. Gnani D, Crippa S, Della Volpe L, et al. An early-senescence state in aged mesenchymal stromal cells contributes to hematopoietic stem and progenitor cell clonogenic impairment through the activation of a proinflammatory program. Aging Cell. 2019; 18(3):e12933. 49. Farr JN, Xu M, Weivoda MM, et al. Targeting cellular senescence prevents agerelated bone loss in mice. Nat Med. 2017; 23(9):1072-1079. 50. Mendelson A, Frenette PS. Hematopoietic stem cell niche maintenance during homeostasis and regeneration. Nat Med. 2014; 20(8):833-846.

haematologica | 2021; 106(12)

ARTICLE

Acute Myeloid Leukemia

Networking for advanced molecular diagnosis in acute myeloid leukemia patients is possible: the PETHEMA NGS-AML project Claudia Sargas,1 Rosa Ayala,2 María Carmen Chillón,3 María J. Larráyoz,4 Estrella Carrillo-Cruz,5 Cristina Bilbao,6 Manuel Yébenes-Ramírez,7 Marta Llop,1 Inmaculada Rapado,2 Ramón García-Sanz,3 Iria Vázquez,4 Elena Soria,5 Yanira Florido-Ortega,6 Kamila Janusz,7 Carmen Botella,8 Josefina Serrano,7 David Martínez-Cuadrón,9,10 Juan Bergua,11 Mari Luz Amigo,12 Pilar Martínez-Sánchez,2 Mar Tormo,13 Teresa Bernal,14 Pilar Herrera-Puente,15 Raimundo García,16 Lorenzo Algarra,17 María J. Sayas,18 Lisette Costilla-Barriga,19 Esther Pérez-Santolalla,20 Inmaculada Marchante,21 Esperanza Lavilla-Rubira,22 Víctor Noriega,23 Juan M. Alonso-Domínguez,24 Miguel Á. Sanz,9,10 Joaquín Sánchez-Garcia,7 María T. Gómez-Casares,6 José A. Pérez-Simón,5 María J. Calasanz,4 Marcos GonzálezDíaz,3 Joaquín Martínez-López,2 Eva Barragán,1,10# and Pau Montesinos,9,10# on behalf of the PETHEMA group

Ferrata Storti Foundation

Haematologica 2021 Volume 106(12):3079-3089

Molecular Biology Unit, Hospital Universitari i Politècnic-IIS La Fe, Valencia; 2Hematology Department, Hospital Universitario 12 de Octubre, CNIO, Complutense University, Madrid; 3 Hospital Universitario de Salamanca (HUS/IBSAL), CIBERONC and Center for Cancer Research-IBMCC (USAL/CSIC), Salamanca; 4CIMA LAB Diagnostics-Universidad de Navarra, Pamplona; 5Hospital Universitario Virgen del Rocío, Instituto de Biomedicina (IBIS/CSIC/CIBERONC), Universidad de Sevilla, Sevilla; 6Hospital Universitario de Gran Canaria Dr. Negrín, Las Palmas de Gran Canaria; 7IMIBIC, Hematology, Hospital Universitario Reina Sofía, UCO, Córdoba; 8Hospital General Universitario de Alicante, Alicante; 9Hematology Department, Hospital Universitari i Politécnic-IIS La Fe, Valencia; 10 CIBERONC Instituto de Salud Carlos III, Madrid; 11Hospital Universitario San Pedro de Alcántara, Cáceres; 12Hospital Universitario Morales Messeguer, Murcia; 13Hematology Department, Hospital Clínico Universitario-INCLIVA, Valencia; 14Hospital Universitario Central de Asturias, Oviedo; 15Hospital Universitario Ramón y Cajal, Madrid; 16Hospital Universitari General de Castelló, Castellón; 17Hospital Universitario General de Albacete, Albacete; 18Hospital Univerisitario Dr. Peset, Valencia; 19Hospital Universitario Miguel Servet, Zaragoza; 20Hospital de Donosti, San Sebastián; 21Hospital Universitario Puerta del Mar, Cádiz; 22Complexo Hospitalario Lucus Augusti, Lugo; 23Complexo Hospitalario Universitario A Coruña, A Coruña and 24Hospital Universitario Fundación Jiménez Díaz, Madrid, Spain 1

EB and PM contributed equally as co-senior authors.

Correspondence: ABSTRACT

ext-generation sequencing (NGS) has recently been introduced to efficiently and simultaneously detect genetic variations in acute myeloid leukemia (AML). However, its implementation in the clinical routine raises new challenges focused on the diversity of assays and variant reporting criteria. In order to overcome this challenge, the PETHEMA group established a nationwide network of reference laboratories aimed to deliver molecular results in the clinics. We report the technical cross-validation results for NGS panel genes during the standardization process and the clinical validation in 823 samples of 751 patients with newly diagnosed or refractory/relapse AML. Two cross-validation rounds were performed in seven nationwide reference laboratories in order to reach a consensus regarding quality metrics criteria and variant reporting. In the pre-standardization cross-validation round, an overall concordance of 60.98% was obtained with a great variability in selected genes and conditions across laboratories. After consensus of relevant genes and optimization of quality parameters the overall concordance rose to 85.57% in the second cross-validation round. We show that a diagnostic network with harmonized NGS analysis and reporting in seven experienced laboratories is feasible in the context of a scientific group. This cooperative nationwide strategy provides advanced molecular diagnostic for AML patients of the PETHEMA group (clinicaltrials gov. Identifier: NCT03311815). haematologica | 2021; 106(12)

EVA BARRAGÁN GONZÁLEZ barragan_eva@gva.es Received: June 24, 2020. Accepted: November 3, 2020. Pre-published: November 12, 2020. https://doi.org/10.3324/haematol.2020.263806

3079

C. Sargas et al.

Introduction Acute myeloid leukemia (AML) is a heterogeneous disease characterized by a wide spectrum of molecular alterations that lead malignant transformation of normal hematopoietic cells.1 The relevance of chromosomal alterations and gene variants for diagnosis, risk stratification and choice of targeted therapies (i.e, FLT3 and IDH1/2 inhibitors) has remarkably increased the complexity of routine molecular diagnostic strategies.2–5 Next-generation sequencing (NGS) has been established as a new molecular diagnostic tool rapidly adopted by clinical laboratories, being able to simultaneously assess different genetic alterations such as rearrangements, single nucleotide variants (SNV), insertions-deletions (indels) and copy number variations (CNV) in a wide variety of genes.6 NGS gene panels have been preferentially adopted rather than whole genome or exome sequencing due to an easier results interpretation, lower cost and less time consumption, as well as higher read deep needed for low frequency variant detection. Compared to NGS, conventional single-gene approaches by polymerase chain reaction (PCR)7 are laborious, time-consuming and less efficient to detect minor clones, but they are still needed as rapid-screening tests for druggable variants.8 The new scenario for AML molecular diagnosis, requiring rapid screening by conventional PCR and comprehensive characterization by NGS, is a great challenge for molecular biology laboratories. For this purpose, the PETHEMA (Programa Español de Tratamientos en Hematología) group established a nationwide network involving seven central laboratories aimed to deliver molecular results to clinics for newly diagnosed and relapsed/refractory AML patients. The first step was to ensure appropriate logistic support, including geographical localization of highly skilled central laboratories strategically distributed according to population density and distance. The second step was to harmonize NGS and PCR techniques methodology and result reporting across the seven central laboratories, establishing consensus panel genes, quality metrics cutoffs and variant reporting criteria. In this work, we performed the first analysis of a NGSAML study (clinicaltrials gov. Identifier: NCT03311815), reporting the technical cross-validation results for NGS panel genes during the standardization process and the clinical validation in 823 samples of 751 patients with newly diagnosed or refractory/relapse AML.

relapsed/refractory AML (excluding acute promyelocytic leukemia) according to the World Health Organization criteria (2008), regardless of the treatment received, were eligible for the NGS-AML study. The Institutional Ethics Committee for Clinical Research of each institution approved this study. Written informed consent in accordance with the recommendations of the Declaration of Human Rights, the Conference of Helsinki, and institutional regulations were obtained from all patients.

Cross-validation The first cross-validation round was developed to evaluate the starting situation of reference laboratories (see the Online Supplementary Appendix). For this purpose, four samples harboring 24 variants were distributed from HULF (coordinator center) and each laboratory carried out NGS analysis according to their already implemented protocols. Reports were sent to the coordinator center to analyze the results. Taking into account the obtained results, the collaborative group established a set of relevant AML genes and minimum quality metrics criteria. Then, a second cross-validation round was designed to strengthen the established quality parameters, the consensus recommendations, and variant reporting for NGS analysis among the seven reference laboratories. Variant detection, variant allele frequency (VAF), dispersion among centers and variant reporting (clinically and non-clinically relevant variants) were assessed in six samples with 30 variants (five with a lower VAF than 5%). Reports were sent to the coordinator center to analyze the results.

Consensus genes establishment Thrirty genes were established as key genes for AML pathogenesis: ABL1, ASXL1, BRAF, CALR, CBL, CEBPA, CSF3R, DNMT3A, ETV6, EZH2, FLT3, GATA2, HRAS, IDH1, IDH2, JAK2, KIT, KRAS, MPL, NPM1, NRAS, PTPN11, RUNX1, SETBP1, SF3B1, SRSF2, TET2, TP53, U2AF1 and WT1. ASXL1, CEBPA, FLT3, IDH1, IDH2, NPM1, RUNX1, and TP53 were mandatory for their implication in clinical guidelines, targeted therapy and risk stratification. The remaining genes were recommended for NGS panels, according to laboratory features and sequencing panel options.

Sequencing platforms and panels The sequencing platform and panel were selected by each laboratory using the following criteria: i) to include all eight mandatory genes and, ii) to include the maximum number of other 22 relevant genes (sequencing platforms and panels data are shown in the Online Supplementary Appendix and Online Supplementary Table S1).

Clinical validation Methods Study design and reference laboratories This was a prospective, multi-center, non-interventional study, performed in seven Spanish PETHEMA central laboratories: Hospital Universitario La Fe (HULF, Valencia), Hospital Universitario de Salamanca (HUS, Salamanca), Hospital Universitario 12 de Octubre (H12O, Madrid), Hospital Universitario Virgen del Rocío (HUVR, Sevilla), Hospital Universitario Reina Sofía (HURS, Córdoba), Hospital Universitario de Gran Canaria Dr. Negrín (HUDN, Las Palmas de Gran Canaria) and CIMA LAB Diagnostics (UNAV, Pamplona) (see the Online Supplementary Appendix for further details).

Inclusion criteria All adult patients (≥18 years) with newly diagnosed or

3080

NGS was performed according to already implemented protocols and the consensus parameters established in both cross-validation rounds. Samples meeting the quality metrics criteria established in previous standardization rounds were considered in the clinical validation.

Statistical analyses All statistics were performed using SPSS version 22 (IBM, Armonk, NY, USA) and GraphPad Prism 4 (GraphPad, La Jolla, CA, USA) software programs. A P-value (P) <0.05 was considered statistically significant (see the Online Supplementary Appendix).

Results Cross-validation In the first cross-validation round (pre-standardization), haematologica | 2021; 106(12)

A centralized NGS diagnostic platform for AML

we observed a great diversity in the included genes in each NGS panel. Some AML relevant genes such as NPM1 and CEBPA were not studied, while other non-relevant genes for AML pathogenesis were included. The global error rate was 39.02% with a higher error rate in variants showing a VAF lower than 10% (77.04 ± 6.98% vs. 18.56 ± 29.24%, P<0.001) (Table 1). In the second round (post-standardization), the mean read depth was 4,032 (range, 1,463-4,532) with a median uniformity of 98.34%. The error rate for all variants was reduced to 14.43% (Table 2). In this round, the error rate was significantly higher in variants with a VAF lower than 5% (28.57 ± 14.28% vs. 12.27 ± 14.39%, P=0.031) (Figure 1). All centers provided a correct clinical classification of the detected variants. No differences in the error rate were observed between indels and SNV. Regarding the accuracy of VAF determination among the different centers, VAF standard deviation (SD) was higher in indel variants than SNV (5.44 ± 2.80 vs. 2.15 ± 2.03, P=0.001). After cross-validation results, uniformity (>85%) and mean read depth of 1,000X were established as quality control parameters for a valid NGS assay. Synonymous, intronic and polymorphic variants (minor allele frequency [MAF] ≥1% and/or included in the dbSNP database) were filtered out. VAF ≥5% was established as a cutoff value for variant reporting with the exception of pathogenic variants with strong clinical evidence which were reported with a VAF ≥1% (e.g, TP53 or FLT3). Variants accomplishing all these requirements were considered.

were obtained. No significant differences were observed when analyzing all the included variants or in any of the sub groups (Indel, SNV, variants with VAF ≤5% and variants with VAF >5%) (Online Supplementary Table S2).

Clinical validation From October 2017 to October 2019 a total of 823 samples from 751 AML patients were sent to the laboratory network. Disease status at sample collection was: newly diagnosis (DX) (n=639), refractoriness (RS) (n=82), and relapse (RP) (n=102). Patient characteristics are summarized in Table 3. NGS was performed according to already implemented protocols and the consensus parameters established in both cross-validation rounds.

Mutation distribution A total of 2,052 variants were reported in the 823 samples, with 90.81% of patients showing at least one mutated gene (Online Supplementary Figure S1A). The mean number of variants per sample was 2.49 (range, 0-8). Most patients had three variants (24.37%), followed by patients with two (21.04%) and one (20.77%) variants, respectively (Online Supplementary Figure S1B). A high frequency of variants in genes involved in signal transduction and epigenetic regulation was observed. FLT3 (24.06%: FLT3 internal tandem duplications [ITD] 16.52%, FLT3-point mutations [PM] 8.87%) was the most prevalent mutated gene followed by IDH (22.60%: IDH1 9.11%, IDH2 13.85%), DNMT3A (21.63%) and NPM1 (21.51%) (Online Supplementary Figure S2).

Platform performance The performance of the NGS platforms (Ion Torrent vs. Illumina) regarding the error rate and the VAF SD was assessed after the results of the standardization rounds

Co-mutations FLT3, NPM1 and DNMT3A were significantly co-mutated for all combinations (P<0.001). PTPN11 variants also

Table 1. First cross-validation round results.

Gene

Coding

Protein

Detected

Included

Error Rate

Mean VAF

NPM1 (NM_002520) IDH2 (NM_002168.3) DNMT3A (NM_022552) STAG2 (NM_001042749.2) RUNX1 (NM_001754.4) ASXL1 (NM_015338.5)

c.860_863dup c.419G>A c.2645G>A c.2124del c.736A>C c.1934dup

p.Trp288Cysfs*12 p.Arg140Gln p.Arg882His p.Leu708Phefs*9 p.Thr246Pro p.Gly646Trpfs*12

5 6 6 1 1 1

5 6 6 3 6 6

0.00% 0.00% 0.00% 66.67% 83.33% 83.33%

41.68% 44.73% 43.77% NA NA NA

18.90% 3.31% 1.92% NA NA NA

CEBPA (NM_004364.4) CEBPA (NM_004364.4) IDH2 (NM_002168.3) NRAS (NM_002524.4) EZH2 (NM_004456.4) EZH2 (NM_004456.4) DNMT3A (NM_022552) KMT2A (NM_001197104.1) GATA2 (NM_032638.4) ASXL1 (NM_015338.5)

c.68_78del c.895A>G c.419G>A c.37G>C c.952del c.1321G>A c.1961G>A c.3253G>A c.1084C>T c.1934dup

p.Pro23Glnfs*81 p.Ser299Gly p.Arg140Gln p.Gly13Arg p.Thr318Glnfs*3 p.Glu441Lys p.Gly654Asp p.Val1085Met p.Arg362* p.Gly646Trpfs*12

4 5 6 6 3 4 1 1 1 2

5 5 6 6 4 4 6 4 5 6

20.00% 0.00% 0.00% 0.00% 25.00% 0.00% 83.33% 75.00% 80.00% 66.67%

51.32% 45.32% 49.82% 46.32% 47.54% 50.13% NA NA NA 39.82%

7.17% 3.86% 5.22% 2.01% 2.16% 2.81% NA NA NA 4.41%

DNMT3A (NM_022552) TP53 (NM_000546.5) STAG2 (NM_001042749.2) CUX1 (NM_181552.4) ASXL1 (NM_015338.5)

c.2678G>C c.652_670del c.2858G>A c.1588A>C c.1934dup

p.Trp893Ser p.Val218fs p.Arg953Gln p.Lys530Gln p.Gly646Trpfs*12

6 5 1 1 1

6 6 3 3 6

0.00% 16.67% 66.67% 66.67% 83.33%

44.54% 67.58% NA NA NA

2.32% 19.42% NA NA NA

TP53 (NM_000546.5) EZH2 (NM_004456.4) ASXL1 (NM_015338.5)

c.392A>T c.553G>C c.1934dup

p.Asn131Ile p.Asp185His p.Gly646Trpfs*12

6 1 1

6 4 6

0.00% 75.00% 83.33%

47.33% NA NA

1.83% NA NA

Detected: number of centers which have detected the mutation; Included: number of centers which include each variant in its next-generation sequencing assay; Error Rate: number of centers which failed to detect the variant regarding the total of centers;VAF: variant allele frequency; SD: standard deviation of VAF establishment among centers; NA: not applicable; variants only were detected by one center.

haematologica | 2021; 106(12)

3081

C. Sargas et al.

Coding

Protein

Detected

Included

Error Rate

Mean VAF

CRV

NPM1 (NM_002520) FLT3 (NM_004119.2) FLT3 (NM_004119.2)

c.863_864insCCTG c.1801_1802ins30 c.2505T>A

p.Trp288Cysfs*12 p.Asp600_Leu601ins10 p.Asp835Glu

7 7 5

7 7 7

0.00% 0.00% 28.57%

34.88% 29.62% 2.46%

8.47% 9.32% 0.53%

NCRV

PHF6 (NM_032458.2) DNMT3A (NM_022552) NRAS (NM_002524.4)

c.548C>T c.2264T>C c.34G>A

p.Ser183Phe p.Phe755Ser p.Gly12Ser

4 7 4

6 7 7

33.33% 0.00% 42.86%

50.32% 42.10% 1.75%

2.23% 5.85% 0.30%

CRV

RUNX1 (NM_001754.4) IDH1 (NM_005896.3)

c.1306dupT c.394C>T

p.Ser436Phefs*164 p.Arg132Cys

6 7

7 7

14.29% 0.00%

43.24% 16.25%

4.33% 2.31%

NCRV

TET2 (NM_001127208.2) PHF6 (NM_032458.2) EZH2 (NM_004456.4) SRSF2 (NM_003016.4) JAK2 (NM_004972.3)

c.3866G>T c.346C>T c.2255G>C c.161C>T c.1849G>T

p.Cys1289Phe p.Arg116* p.*752Ser p.Ser54Phe p.Val617Phe

7 4 7 6 6

7 5 7 7 7

0.00% 20.00% 0.00% 14.29% 14.29%

43.02% 42.25% 10.39% 5.73% 2.73%

6.82% 1.90% 1.59% 0.66% 0.66%

CRV

FLT3 (NM_004119.2) FLT3 (NM_004119.2)

c.2028C>G c.2504A>C

p.Asn676Lys p.Asp835Ala

5 6

7 7

28.57% 14.29%

22.17% 5.40%

3.63% 0.62%

NCRV

SH2B3 (NM_005475.2)

c.557G>T

p.Ser186Ile

33.33%

56.40%

0.57%

CRV

PHF6 (NM_032458.2) EZH2 (NM_004456.4)

c.129_130insGG c.2212_2231del

p.Lys44Glyfs*38 p.Ala738Argfs*18

4 6

5 7

20.00% 14.29%

49.44% 29.49%

3.48% 6.96%

NCRV

NRAS (NM_002524.4) EZH2 (NM_004456.4)

c.35G>A c.796G>A

p.Gly12Asp p.Gly266Arg

7 4

7 7

0.00% 42.86%

17.08% 4.86%

2.33% 0.93%

CRV

ASXL1 (NM_015338.5) ASXL1 (NM_015338.5) TP53 (NM_000546.5)

c.1772dup c.1745_1758del c.916C>T

p.Tyr591* p.Pro582Argfs*32 p.Arg306*

7 6 6

7 7 7

0.00% 14.29% 14.29%

21.69% 13.96% 4.49%

2.32% 2.87% 0.28%

NCRV

SF3B1 (NM_012433.3)

c.1873C>T

p.Arg625Cys

14.29%

8.02%

1.06%

CRV

RUNX1 (NM_001754.4) ASXL1 (NM_015338.5)

c.593A>G c.2463_2478del

p.Asp198Gly p.Asp821Glufs*12

7 7

0.00% 0.00%

77.98% 40.47%

2.75% 8.29%

NCRV

Table 2. Second cross-validation round results.

Gene

ASXL1 (NM_015338.5) SF3B1 (NM_012433.3) CSF3R (NM_156039.3) CSF3R (NM_156039.3)

c.2537G>A c.2098A>G c.1853C>T c.2346dup

p.Ser846Asn p.Lys700Glu p.Thr618Ile p.Ser783Glnfs*6

7 6 6 3

7 7 7 7

0.00% 14.29% 14.29% 57.14%

54.70% 47.47% 47.06% 37.77%

6.82% 1.85% 1.50% 2.92%

Detected: number of centers which have detected the mutation; Included: number of centers which include each variant in its next-generation sequencing assay; Error Rate: number of centers which failed to detect the variant regarding the total of centers; VAF: variant allele frequency; SD: standard deviation of VAF establishment among centers; CRV: clinically-relevant variants; NCRV: non-clinically relevant variants.

associated with these genes (P<0.01 for NPM1 and P<0.05 for DNMT3A and FLT3). IDH variants associated with NPM1 (P=0.01 for IDH1 and P<0.01 for IDH2) as well as DNMT3A (P<0.01 for IDH1 and P<0.05 for IDH2). SRSF2 was strongly co-mutated with IDH2 and TET2 (P<0.01) (Figure 2). Exclusion analysis provided 17 MEGS which are defined as lists of exclusive mutated genes. NPM1, TP53, RUNX1 and KIT set was defined as the most significant mutually exclusive gen set (p-nominal: 2,25E-24). TP53 was present in all MEGS being the most exclusive gene. RUNX1 was highly exclusive with NPM1. CEBPA and NRAS were included in several MEGS of three blocks suggesting their exclusive nature (Online Supplementary Figure S3; Online Supplementary Table S3).

Variant allele frequency analysis Variants in genes involved in signaling pathways (NRAS, 3082

FLT3, KIT, KRAS, and PTPN11) showed lower median VAF, whereas genes involved in clonal hematopoiesis of indeterminate potential (CHIP) (ASXL1, DNMT3A and TET2) showed a VAF around 50%. TP53 variants showed the highest median VAF value (Figure 3A). According to the functional categories, epigenetically related genes and tumor suppressor genes were characterized by a high VAF while genes related to signaling pathways genes showed the lowest VAF (Figure 3B).

Mutational landscape according to disease stage NPM1 and PTPN11 variants were more frequent at diagnosis (NPM1 23.16% vs. 9.76%, P=0.004, and PTPN11 6.62% vs. 1.10%, P=0.048). RUNX1 and SF3B1 variants were more frequent in refractory as compared to diagnosis (RUNX1 30.49% vs. 16.43%, P=0.003 and SF3B1 9.76% vs. 4.23%, P=0.049). IDH1, IDH2 and WT1 variants were more frequent at relapse as compared to haematologica | 2021; 106(12)

A centralized NGS diagnostic platform for AML

Figure 1. Variant allele frequency of samples included in the second cross validation round. (A) Variant allele frequency (VAF) >5% variants and (B) VAF ≤5% variants. Black dots indicate VAF reported for each center. Red dots mean not detected variant. Mean VAF is represented by a horizontal bar and whiskers represent the standard deviation.

diagnosis (IDH1 14.71% vs. 8.14%, P=0.040; IDH2 21.57% vs. 12.52%, P=0.020, and WT1: 7.84% vs. 3.13%, P=0.043). KRAS and PTPN11 variants were more frequent in refractory as compared to relapse stage (KRAS 9.76% vs. 1.96%, P=0.025, and PTPN11 8.54% vs. 1.10%, P=0.028) (Figure 4A and B).

Age-related mutations At diagnosis, patients aged ≥65 years had more variants than younger (<65 years) patients (2.74 ± 0.81 vs. 2.18 ± 0.74 variants per patient, P<0.001). The following genes were more frequently mutated in patients aged ≥65 years vs. <65 years: ASXL1, EZH2, IDH2, JAK2, SF3B1, SRSF2, TET2, TP53, and U2AF1. FLT3-ITD and NPM1 mutations were more frequent in younger AML patients (Figure 5). In relapsed AML, ASXL1 (20.00% vs. 3.57%, P=0.011) and IDH variants (46.70% vs. 25%, P=0.035) were associated with patients aged ≥65 years (Online Supplementary Table S4).

Mutational stability in paired samples Paired samples at DX-RP (n=14) and DX-RS (n=20) were obtained to assess clonal evolution. The following were stable variants: NPM1 (100%, as no patients acquired or lost variants), TP53, IDH2 (one acquisition at RS and one at RP in each gene), DNMT3A (100% stable at RS and one acquisition at RP), and RUNX1 (stable at RP and one loss at RS). The following variants were unstable: activating signaling pathways genes such as FLT3, NRAS, KRAS, BRAF, KIT and PTPN11 (Online Supplementary Figure S4). Interestingly for targeted therapy, 26.47% of patients changed the mutational status of FLT3 at RS or RP (FLT3ITD: two gains and three losses; FLT3-PM: one gain and three losses). In all cases, the loss of function mutation of FLT3-PM was located on the Asp835 codon. haematologica | 2021; 106(12)

Clinically relevant mutations Overall, 72.30% of patients harbored at least one clinically relevant variant included in the AML clinical guidelines, clinical trials inclusion criteria or as a risk stratification biomarker (ASXL1, CEBPA, FLT3, IDH1/2, NPM1, RUNX1 and TP53) (Online Supplementary Figure S5). Moreover, druggable mutations were present in a significant proportion of patients (FLT3 in 21.14% and IDH1/2 in 22.60%).

NPM1 mutations NPM1 variants were found in 21.51% of samples being 78.53% type A (c.860_863dupTCTG), 6.21% type B (c.863_864insCATG), 6.21% type D (c.863_864insCCTG), and 9.04% had uncommon variants (Online Supplementary Figure S6; Online Supplementary Table S5).

FLT3 mutations FLT3 was the most frequently mutated gene (24.06%). FLT3 ITD was the most frequent FLT3 aberration (16.52%) followed by D835 and I836 variants (5.71%) and other variants (3.16%) (Online Supplementary Figure S7A). Other variants were mostly SNV (95.18%) located in the tyrosine kinase 1 domain (TKD1; 41.18%), juxtamembrane domain (JMD; 23.53%), tyrosine kinase 2 domain (TKD2; 20.59%), extracellular domain (ED; 11.76%) and kinase insert domain (KID 2.94%). No variants were detected in transmembrane (TMD) and C-terminal domains (CTD). (Online Supplementary Figure S7B). 84.75% of all FLT3 variants were targetable with FLT3 inhibitors and had a direct clinical impact in 21.14% of patients through targeted therapy or clinical trials.

CEBPA mutations CEBPA variants were found in 5.35% of samples, 3.52% were monoallelic variants and 1.82% were biallelic variants 3083

C. Sargas et al.

(both CEBPA alleles mutated). Two variants with similar VAF were reported as probably biallelic variants (Online Supplementary Table S6). Although CEBPA variants were distributed along the entire coding sequence, biallelic variants were frequently detected in BZIP (43.33%) and the N-terminal domain (30%). Monoallelic variants, were mostly detected in BZIP (44.83%) and almost equally distributed among TAD1 (13.79%), TAD2 (17.94%) and the N-terminal domain (6.90%).

IDH1 and IDH2 mutations IDH variants were detected in 22.60% of samples. In mutated IDH1 samples (9.11%), all variants were detected in the Arg132 codon. IDH2 (13.85%) was exclusively mutated in the Arg140 or Arg172 codons (84.20% and 15.79%, respectively). Three patients (1.61%) showed simultaneous variants in IDH1 and IDH2 (Online Supplementary Table S7). All IDH1 and IDH2 variants were targetable mutations by IDH inhibitors as no atypical variants were detected.

ASXL1, RUNX1 and TP53 mutations Overall, 39.00% of samples showed variants in at least one of them: 18.23% RUNX1, 14.70% TP53 and 12.39% ASXL1. In 26.12% of samples a variant detected in one of these genes was the only clinically relevant variant. Moreover, 28.19% of patients were classified to an unfavorable risk group according to ASXL1, RUNX1 and TP53 mutations, following European LeukemiaNet 2017 recommendations.

Discussion This study shows that a network platform involving many highly skilled laboratories can successfully deliver robust molecular data for AML patients. This strategy allows for testing NGS in the majority of newly diagnosed and relapsed/refractory AML patients involved in the PETHEMA studies, overcoming the current challenging needs for a high-standard diagnosis in cooperative groups. Our descriptive analysis performed in a large series of reallife patients depicts the complex molecular landscape of AML. In the last 5 years, NGS has irrupted as a potential routine tool for molecular diagnosis, allowing for precise and simultaneous detection of relevant variants in AML. However, this technique is still non-affordable for many institutions due to: i) remarkable cost as compared to conventional PCR tests, ii) batch of samples, ranging from eight to more than 30, to run the test, and consequently high time consumption, both making it difficult to report results in less than 710 days; and iii) the need of expensive machinery and highly-qualified teams for biostatistical and molecular biologists. Moreover, the majority of prognostic or druggable mutations can be rapidly and relatively easily detected by conventional PCR. In fact, from the mandatory NGS panel genes selected by the PETHEMA central laboratories (i.e, ASXL1, CEBPA, FLT3, IDH1, IDH2, NPM1, RUNX1, and TP53), a mutation screening by conventional PCR is still required for FLT3, IDH1, IDH2 and NPM1, as a positive result could allow for rapid implementation of targeted or risk-adapted therapeutic approaches.8 In addition, rapid PCR is also needed for core-binding factor (CBF), PMLRARA and BCR-ABL rearrangements. Under this scenario, 3084

Table 3. Demographic and baseline characteristics of the study population (n=751).

Characteristic

Mean

Age, years 62.5 <60 ≥60 Sex Male Female ECOG 0.9 0 1 2 3 4 Not available Type of AML De novo Secondary Not available WBC, ×109/L 31.2 ≤5 5-10 10-50 > 50 Not available BM blast cells, % 55 ≤ 30 30-70 > 70 Not available Creatinine, mg/dL 1.0 ≤ 1,2 > 1,2 Not available Cytogenetic risk Favorable Normal Intermediate Adverse Not available Therapeutic approach Intensive Non-intensive Clinical Trial Supportive care Not available

NGS population Median Range 65

8-93

0-4

8.4

0.3-305

2-100

0.87

0.23-3.78

n (%) 751 (100) 284 (38) 467 (62) 751 (100) 423 (56) 328 (44) 751 (100) 184 (25) 203 (27) 48 (6) 26 (3) 6 (1) 284 (38) 751 (100) 378 (50) 155 (21) 218 (29) 751 (100) 205 (27) 58 (8) 146 (19) 95 (13) 247 (33) 751 (100) 109 (15) 203 (27) 157 (21) 282 (38) 751 (100) 400 (53) 68 (9) 283 (38) 751 (100) 25 (3) 224 (30) 66 (9) 125 (17) 311 (41) 751 (100) 297 (40) 125 (17) 33 (4) 20 (3) 276 (37)

AML: acute myeloid leukemia, BM: bone marrow, WBC: white blood cell; ECOG: Eastern Cooperative Oncology Group.

our cooperative group designed a nationwide network involving seven central laboratories aimed to deliver homogeneous and comparable molecular results for newly diagnosed and relapsed/refractory AML patients. As far as we know, this is a different strategy as compared to other cooperative groups that usually rely on only one or two central laboratories for molecular diagnostics (e.g, British NCRI).9 Several reasons guided us to make this decision: i) the economic and work burden required to collect samples from the whole group, which covers a wide territory and population, was not affordable for a single laboratory; ii) a minimum referral population is required to permit an efficient diagnosis by studying the appropriate number of samples in every run; iii) the involvement many on-site teams in order haematologica | 2021; 106(12)

A centralized NGS diagnostic platform for AML

Figure 2. Circos diagram showing mutation concurrences in our cohort.

Figure 3. Variant allele frequency analysis. (A) Mean variant allele frequency (VAF) for individual genes and (B) mean VAF according functional categories. Each dot represents one variant, median VAF is represented by a horizontal bar and whiskers represent the interquartile range. ITD: internal tandem duplication; PM: point mutations.

haematologica | 2021; 106(12)

3085

C. Sargas et al. A

Figure 4. Distribution of gene alterations in acute myeloid leukemia samples. (A) Mutational landscape in the global cohort. Horizontal green bars: diagnosis, orange: refractory, and red: relapse.Vertical dark blue bars: positive (in global FLT3 row represents internal tandem duplications [ITD]), red: FLT3-D835/I836, green: other FLT3 point mutations (PM), orange: FLT3-ITD and D835/I836, light blue: FLT3-ITD and other FLT3-PM, light grey: negative, dark grey: not tested, yellow: biallelic variants in CEBPA. (B) Mutational prevalence according to disease stage. Diagnosis are represented as blue bars, refractory as red bars and relapse as green bars. *P<0.05, **P<0.01.

to create a true research network; and iv) the need to facilitate rapid delivery of samples while preserving closer and well established relationships between the sample referral institution and the assigned central laboratory. We can affirm that the PETHEMA model for centralized diagnosis has been successful collecting samples from 751 patients in roughly 1 year, enabling the use of this network in routine clinical practice and research. Our study demonstrates that harmonized and reliable NGS results can be achieved across several laboratories, even if they are using their own diagnostic platforms. As shown in the pre-standardization cross-validation round, an overall concordance of 60.98% was obtained with a great variability in selected genes and conditions across laboratories. After consensus of AML relevant genes and optimization of quality parameters (uniformity >85%; mean read depth of 1,000X) the overall concordance rose to 85.57% in the second cross-validation round. This was a remarkable achievement for all laboratories taking into account that low VAF (≤5%) variants were included in this second round. To the best of our knowledge, there are no similar studies reported in the literature for AML. Clinical validation of our AML cohort was consistent with previous reports. Roughly 91% of AML patients had at least one variant, and many harbored three, four and up to eight variants reflecting the heterogeneous AML mutational profile.10 FLT3, IDH1/2, DNMT3A and NPM1 were the most frequently mutated genes,11,12 and we showed that up to 73% of patients had variants with clinical implications for risk stratification or targeted therapy-based 3086

approaches (i.e, ASXL1, CEBPA, FLT3, IDH1/2, NPM1, RUNX1 and TP53).13 Moreover, ASXL1, RUNX1 and TP53 variants which are not easily analyzed with conventional molecular techniques,14–16 were the unique clinically relevant alteration detected in up to 28.19% of patients, highlighting that NGS-based mutational profiling seems crucial to categorize AML risk according to the European LeukemiaNet 2017 guidelines.3 As reported by other groups,17–20 elderly patients had a higher number of variants, which were enriched in spliceosome machinery, epigenetic regulators and in DNA repair (i.e, SRSF2, U2AF1, SF3B1, ASXL1, TET2, IDH2 and TP53). In line with previous studies, NPM1 variants were more frequent in younger AML patients, and we noticed a striking decrease of FLT3 variants in older patients.21 We can affirm that a lower number of older patients may benefit from tyrosine kinase inhibitors-based approaches,22 but more from novel IDH-inhibitors.23 In our experience, NGS has efficiently screened FLT3 gene variants, including less frequent variants, which could also be informative for therapeutic decisions.24,25 Furthermore, NGS is a promising tool to assess FLT3-ITD duplicated region, which could have prognostic impact regarding its location and extension.26 We also provide insights on clonal evolution and leukemogenesis: i) variants in signalling pathway genes (FLT3, KIT, RAS) had lower VAF, reflecting their role as late events,27; ii) genes related to CHIP showed higher VAF values,28 and iii) median VAF in AML patients with TP53 variants was above 50%, indicating the frequent loss of the wild-type TP53 allele. Recent studies suggest that a higher haematologica | 2021; 106(12)

A centralized NGS diagnostic platform for AML

Figure 5. Age-related mutational profile. Pvalues (P) for statistically significant results are shown. ITD: internal tandem duplication; PM: point mutations.

clonal size of TP53 variants, determined by VAF and chromosomal aberrations (del 17p), could discriminate patients with worse prognosis among TP53 mutated AML.29,30 Our study allowed to observe differences in the mutational profile of relapsed/refractory patients as compared to newly diagnosed/untreated subjects. Interestingly, RUNX1 variants were more frequent among refractory AML patients, which is consistent with previously reported poorer outcomes after intensive chemotherapy and its association with older age.31 The same was observed with SFRS2, depicting poorer outcomes when commutated with IDH232 which was highly associated in our cohort. NPM1 variants were significantly less frequent in refractory patients, as compared to newly diagnosed or relapsed, reflecting the known high complete remission rates in this setting.33,34 Interestingly, IDH variants were the most frequent alteration at relapse, suggesting that they are associated with higher relapse risk,35 but also have more possibilities to obtain an initial response with front-line therapies.18,36 We analyzed 35 patients with paired samples at diagnosis versus relapse/refractory setting, confirming some findings from scarce studies on clonal evolution: i) as a founder variant, NPM1 was very stable,37,38 ii) DNMT3A variants were very stable, probably due to its early acquisition and haematologica | 2021; 106(12)

preleukemic occurrence,39,40 as well as their persistence during remission and disease progression,41 and iii) activating signaling pathways genes were unstable (FLT3, NRAS, KRAS, BRAF, KIT and PTPN11). This is particularly relevant for the management of patients who acquire FLT3 mutations during relapse and refractoriness which may benefit from second generation inhibitors such as gilteritinib, available for the treatment of relapsed/refractory AML with FLT3 mutation.42,43 In line with the study by Kronke et al., we show that IDH2 variants were very stable, contrarily to IDH1, but numbers are low and should be cautiously interpreted.38 The main limitation of our study is that, apart from age and disease phase, baseline clinical characteristics, treatment patterns, and outcomes, were not available for a minor proportion of patients at the time of this interim analysis of the NGS-AML trial. This is why the results herein presented focused on the overall network building strategy, as well as in the cross-validation of samples and reporting harmonization, showing a mutational landscape of AML consistent with the current knowledge. We should mention that our diagnostic platform has many areas of improvement: i) time to NGS reporting, which is longer than conventional molecular analysis (approximately 2-3 3087

C. Sargas et al.

weeks), ii) need to promote larger participation of PETHEMA clinical sites, as a sizable proportion of patients are not yet benefiting from advanced laboratory centralization (especially in the relapse/refractory setting), and ii) budgetary vulnerability. In conclusion, the PETHEMA cooperative scientific group has adopted the reported nationwide strategy network with centralized NGS analyses. Sample and information exchange allowed us to unify analysis criteria and decrease reporting variability in order to offer reliable and consistent NGS results. This cooperative strategy has also been applied to rapid screening by conventional PCR and quantitive real-time PCR to measure residual disease, and is being expanded to other AML diagnostic areas (e.g., biobanking and multiparametric flow cytometry). Ongoing therapeutic guidelines (NCT01296178) and clinical trials (clinicaltrials gov. Identifier: NCT04230239, NCT04107727, NCT04112589, NCT04090736) by the PETHEMA group are benefiting from this diagnostic network. Disclosures No conflicts of interest to disclose. Contributions EB and PM conceived the study; CS, EB and PM analyzed, interpreted the data and wrote the paper; CS performed the statis-

References 1. Saultz J, Garzon R. Acute myeloid leukemia: a concise review. J Clin Med. 2016;5(3):33. 2. Arber DA, Orazi A, Hasserjian R, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood. 2016;127(20):2391-2405. 3. 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. 4. Hackl H, Astanina K, Wieser R. Molecular and genetic alterations associated with therapy resistance and relapse of acute myeloid leukemia. J Hematol Oncol. 2017;10(1):51. 5. Luppi M, Fabbiano F, Visani G, Martinelli G, Venditti A. Novel agents for acute myeloid leukemia. Cancers (Basel). 2018;10(11):429. 6. Riva L, Luzi L, Pelicci PG. Genomics of acute myeloid leukemia: the next generation. Front Oncol. 2012;2:40. 7. Kuo FC, Mar BG, Lindsley RC, Lindeman NI. The relative utilities of genome-wide, gene panel, and individual gene sequencing in clinical practice. Blood. 2017;130(4):433439. 8. Bacher U, Shumilov E, Flach J, et al. Challenges in the introduction of next-generation sequencing (NGS) for diagnostics of myeloid malignancies into clinical routine use. Blood Cancer J. 2018;8(11):113. 9. Macklin PS, Pillay N, Lee JL, et al. CM-Path Molecular Diagnostics Forum—consensus statement on the development and implementation of molecular diagnostic tests in the United Kingdom. Br J Cancer. 2019; 121(9):738-743. 10. Ley TJ, Miller C, Ding L, et al. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med.

3088

tical analyses; CS, RA, CC, MJL, EC, CB, MYR, MLL, IR, RGS, IVU, ES, YFO, KJ, CB, JS, DMC, JB, MA, PMS, MT, TB, PHP, RG, LA, MJS, LCB, EPS, IM, ELR, VN, JMA, MAS, JSG, MTGC, JAPS, MJC, MG, JML, EB and PM included data of patients treated in their institutions, reviewed the manuscript and contributed to the final draft. Acknowledgments The authors would like to thank María D. García, Carlos Pastorini, Rafael Vianney, and Mar Benlloch for data collection and management; and Data Science, Biostatistics and Bioinformatics Unit from IIS La Fe for its collaboration in statistical analysis. Funding This work was partially supported by a Celgene grant, the Subdirección General de Investigación Sanitaria (Instituto de Salud Carlos III, Spain) Spanish Ministry of Economy and Competitiveness: PI15/01706, PI16/00517 PI16/0665, PI16/01530, PI18/01340, PI18/01946 PI19/00730, PI19/01518, FI19/00059, Fundación Española de Hematología y Hemoterapia (FEHH) grant, CRIS against Cancer foundation 2018/001. CIBERONC-CB16/12/00233 and “Una manera de hacer Europa” (Innocampus; CEI-2010-1-0010), Instituto de Investigación Sanitaria La Fe (Contrato de Investigación postresidentes 2019-052-1)

2013;368(22):2059-2074. 11. Tyner JW, Tognon CE, Bottomly D, et al. Functional genomic landscape of acute myeloid leukaemia. Nature. 2018;562(7728):526-531. 12. 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. 13. Castelli G, Pelosi E, Testa U. Targeted therapies in the treatment of adult acute myeloid leukemias: current status and future perspectives. Int J Hematol Oncol. 2016;5(4):143-164. 14. Gelsi-Boyer V, Brecqueville M, Devillier R, Murati A, Mozziconacci MJ, Birnbaum D. Mutations in ASXL1 are associated with poor prognosis across the spectrum of malignant myeloid diseases. J Hematol Oncol. 2012;5:12. 15. Kohlmann A, Nadarajah N, Alpermann T, et al. Monitoring of residual disease by next-generation deep-sequencing of RUNX1 mutations can identify acute myeloid leukemia patients with resistant disease. Leukemia. 2014;28(1):129-137. 16. Barbosa K, Li S, Adams PD, Deshpande AJ. The role of TP53 in acute myeloid leukemia: challenges and opportunities. Genes Chromosomes Cancer. 2019;58(12): 875-888. 17. Dinardo CD, Ravandi F, Agresta S, et al. Characteristics, clinical outcome, and prognostic significance of IDH mutations in AML. Am J Hematol. 2015;90(8):732-736. 18. Medeiros BC, Fathi AT, DiNardo CD, Pollyea DA, Chan SM, Swords R. Isocitrate dehydrogenase mutations in myeloid malignancies. Leukemia. 2017;31(2):272281. 19. Prassek VV, Rothenberg-Thurley M, Sauerland MC, et al. Genetics of acute myeloid leukemia in the elderly: Mutation spectrum and clinical impact in intensively treated patients aged 75 years or older. Haematologica. 2018;103(11):1853-1861.

20. Metzeler KH, Herold T, RothenbergThurley M, et al. Spectrum and prognostic relevance of driver gene mutations in acute myeloid leukemia. Blood. 2016;128(5):686698. 21. Schneider F, Hoster E, Schneider S, et al. Age-dependent frequencies of NPM1 mutations and FLT3-ITD in patients with normal karyotype AML (NK-AML). Ann Hematol. 2012;91(1):9-18. 22. Daver N, Schlenk RF, Russell NH, Levis MJ. Targeting FLT3 mutations in AML: review of current knowledge and evidence. Leukemia. 2019;33(2):299-312. 23. Liu X, Gong Y. Isocitrate dehydrogenase inhibitors in acute myeloid leukemia. Biomark Res. 2019;7(1):22. 24. Williams AB, Nguyen B, Li L, et al. Mutations of FLT3/ITD confer resistance to multiple tyrosine kinase inhibitors. Leukemia. 2013;27(1):48-55. 25. Heidel F, Solem FK, Breitenbuecher F, et al. Clinical resistance to the kinase inhibitor PKC412 in acute myeloid leukemia by mutation of Asn-676 in the FLT3 tyrosine kinase domain. Blood. 2006;107(1):293300. 26. Liu SB, Dong HJ, Bao XB, et al. Impact of FLT3-ITD length on prognosis of acute myeloid leukemia. Haematologica. 2019; 104(1):e9-e12. 27. Hirsch P, Zhang Y, Tang R, et al. Genetic hierarchy and temporal variegation in the clonal history of acute myeloid leukaemia. Nat Commun. 2016;7(1):1-13. 28. Martignoles JA, Delhommeau F, Hirsch P. Genetic hierarchy of acute myeloid leukemia: from clonal hematopoiesis to molecular residual disease. Int J Mol Sci. 2018;19(12):3850. 29. Goel S, Hall J, Pradhan K, et al. High prevalence and allele burden-independent prognostic importance of p53 mutations in an inner-city MDS/AML cohort. Leukemia. 2016;30(8):1793-1795. 30. Rücker FG, Schlenk RF, Bullinger L, et al.

haematologica | 2021; 106(12)

A centralized NGS diagnostic platform for AML

TP53 alterations in acute myeloid leukemia with complex karyotype correlate with specific copy number alterations, monosomal karyotype, and dismal outcome. Blood. 2012;119(9):2114-2121. 31. Gaidzik VI, Bullinger L, Schlenk RF, et al. RUNX1 mutations in acute myeloid leukemia: results from a comprehensive genetic and clinical analysis from the AML study group. J Clin Oncol. 2011;29(10): 1364-1372. 32. Yoshimi A, Lin KT, Wiseman DH, et al. Coordinated alterations in RNA splicing and epigenetic regulation drive leukaemogenesis. Nature. 2019;574(7777):273-277. 33. Jain P, Kantarjian H, Patel K, et al. Mutated NPM1 in patients with acute myeloid leukemia in remission and relapse. Leuk Lymphoma. 2014;55(6):1337-1344. 34. Schneider F, Hoster E, Unterhalt M, et al. NPM1 but not FLT3-ITD mutations predict early blast cell clearance and CR rate in

haematologica | 2021; 106(12)

patients with normal karyotype AML (NKAML) or high-risk myelodysplastic syndrome (MDS). Blood. 2009;113(21):5250-5253. 35. Ok CY, Loghavi S, Sui D, et al. Persistent IDH1/2 mutations in remission can predict relapse in patients with acute myeloid leukemia. Haematologica. 2019;104(2):305311. 36. Megías-Vericat JE, Ballesta-López O, Barragán E, Montesinos P. IDH1-mutated relapsed or refractory AML: current challenges and future prospects. Blood Lymphat Cancer Targets Ther. 2019;9:19-32. 37. Cocciardi S, Dolnik A, Kapp-Schwoerer S, et al. Clonal evolution patterns in acute myeloid leukemia with NPM1 mutation. Nat Commun. 2019;10(1):2031. 38. Krönke J, Bullinger L, Teleanu V, et al. Clonal evolution in relapsed NPM1-mutated acute myeloid leukemia. Blood. 2013;122(1):100-108. 39. Bhatnagar B, Eisfeld AK, Nicolet D, et al.

Persistence of DNMT3A R882 mutations during remission does not adversely affect outcomes of patients with acute myeloid leukaemia. Br J Haematol. 2016;175(2):226236. 40. Hou HA, Kuo YY, Liu CY, et al. DNMT3A mutations in acute myeloid leukemia: stability during disease evolution and clinical implications. Blood. 2012;119(2):559-568. 41. Gaidzik VI, Weber D, Paschka P, et al. DNMT3A mutant transcript levels persist in remission and do not predict outcome in patients with acute myeloid leukemia. Leukemia. 2018;32(1):30-37. 42. Larrosa-Garcia M, Baer MR. FLT3 Inhibitors in acute myeloid leukemia: current status & future directions. Mol Cancer Ther. 2017;16(6):991-1001. 43. Short NJ, Kantarjian H, Ravandi F, Daver N. Emerging treatment paradigms with FLT3 inhibitors in acute myeloid leukemia. Ther Adv Hematol. 2019;10:2040620719827310.

3089

ARTICLE Ferrata Storti Foundation

Acute Myeloid Leukemia

A Pin1/PML/P53 axis activated by retinoic acid in NPM-1c acute myeloid leukemia Rita Hleihel,1,2* Hiba El Hajj,3* Hsin-Chieh Wu,4-6* Caroline Berthier,4-6 Hong-Hu Zhu,7 Radwan Massoud,1 Zaher Chakhachiro,8 Marwan El Sabban,2 Hugues de The4-6# and Ali Bazarbachi1,2# Department of Internal Medicine, American University of Beirut, Beirut, Lebanon; Department of Anatomy, Cell Biology and Physiological Sciences, American University of Beirut, Beirut, Lebanon; 3Department of Experimental Pathology, Immunology and Microbiology, Beirut, Lebanon; 4Université de Paris, INSERM UMR 944, CNRS UMR 7212, Equipe labellisée par la Ligue Nationale Contre le Cancer, IRSL, Hôpital St. Louis, Paris, France; 5Oncologie Moléculaire, Hôpital St. Louis, Paris, France; 6Collège de France, PSL University, CIRB, INSERM UMR 1050, CNRS UMR 7241, Paris, France; 7Department of Hematology, The First Affiliated Hospital, College of Medicine, Zhejiang University, Hangzhou, People’s Republic of China and 8Department of Pathology and Laboratory Medicine, American University of Beirut, Beirut, Lebanon 1 2

Haematologica 2021 Volume 106(12):3090-3099

*RH, HEH and HCW contributed equally as co-first authors. HDT and AB contributed equally as co-senior authors.

ABSTRACT

etinoic acid (RA) was proposed to increase survival of chemotherapy-treated patients with nucleophosmin-1 (NPM-1c)-mutated acute myeloid leukemia. We reported that, ex vivo, RA triggers NPM-1c degradation, P53 activation and growth arrest. PML organizes domains that control senescence or proteolysis. Here, we demonstrate that PML is required to initiate RA-driven NPM-1c degradation, P53 activation and cell death. Mechanistically, RA enhances PML basal expression through inhibition of activated Pin1, prior to NPM-1c degradation. Such PML induction drives P53 activation, favoring blast response to chemotherapy or arsenic in vivo. This RA/PML/P53 cascade could mechanistically explain RA-facilitated chemotherapy response in patients with NPM-1c mutated acute myeloid leukemia.

Correspondence: ALI BAZARBACHI bazarbac@aub.edu.lb Received: October 24, 2020. Accepted: May 3, 2021. Pre-published: May 27, 2021. https://doi.org/10.3324/haematol.2020.274878

3090

Introduction Nucleophosmin 1 (NPM-1) is a chaperone implicated in multiple processes, notably ribosomal biogenesis and growth control. NPM-1 alterations have been directly implicated in cancer development, through a variety of mechanisms, including chromosomal translocations and recurrent mutations.1 In acute myeloid leukemia (AML), the most prevalent one is a short nucleotide insertion that induces a frame shift in the C-terminus of the protein, yielding NPM-1c.2 NPM-1c has been demonstrated to have multiple properties, including inhibition of P53 and cytoplasmic sequestration of key regulatory proteins.3,4 AML with NPM-1 mutation is the most common subtype, accounting for more than one-third of AML. More than half of the patients ultimately relapse when treated with conventional chemotherapy. NPM-1associated cases of AML in relapsed or elderly patients unfit for chemotherapy represent a major unmet medical need. Retinoic acid (RA) is a hormone with multiple effects on development and tissue homeostasis. RA has a dual effect on stem cell fate and differentiation following modulation of transcription. High doses of RA have also been shown to inhibit Pin1, a protein-modifying enzyme involved in the activation of multiple growth suppressive pathways.5,6 RA demonstrated unambiguous clinical efficacy in a variety of conditions including neuroblastoma and acute promyelocytic leukemia (APL).7 In APL, RA directly targets the driving PML/RARA oncoprotein for degradation and produces complete remissions8 through activation of a PML/P53 senescence checkpoint.9,10 In other AML, in vitro studies suggested some efficacy of RA.11,12 However, contradictory results emerged from clinical trials investigating the efficacy of RA in non-APL AML.13 Whereas an Austrian-German study (AML HD98B) found that the addition of RA to intensive chemotherapy improved remission, event-free survival,

haematologica | 2021; 106(12)

RA/arsenic activate a PML/P53 axis in NPM-1c AML

and overall survival,13 a United Kingdom Medical Research Council trial failed to demonstrate any overall advantage of adding RA to chemotherapy.14,15 Intriguingly, the potential benefit of RA co-administration with chemotherapy seems restricted to patients bearing an NPM-1c mutation in the absence of an fms-like tyrosine kinase 3 (FLT3)-internal tandem duplication (ITD).16 Whether this efficacy reflects RA-induced AML differentiation, as observed in several non-APL AML primary patients’ cells or models,11,12 remains to be elucidated. Interestingly, NPM-1c is degraded upon administration of RA, suggesting that loss of NPM-1c expression may underlie, or at least contribute to, RA-driven differentiation and apoptosis ex vivo.17,18 Ex vivo, NPM-1c degradation was accelerated by coadministration of arsenic trioxide (ATO) which, like RA, may inhibit Pin1, a protein-modifying enzyme implicated in growth control through multiple mechanisms.5,6,19 The actual mechanism of RA-induced enhancement of the response to chemotherapy in NPM-1c- AML remains to be elucidated. PML (TRIM19) nucleates nuclear bodies, which are stress-responsive domains that have growth suppressive properties.20 In vivo, PML nuclear bodies are oxidative stress sensors that control P53 activation.21 PML plays a key role in the therapeutic response of APL.9,22-25 The expression of PML is altered in multiple tumor types, most often through PML protein loss upon activation of several degradation pathways, including Pin1.26-30 Interestingly, we previously reported impairment of PML nuclear body formation in NPM-1c-driven AML.17,18 Here, we unravel an unexpected role of PML in RA-initiated responses of NPM-1c AML cells. PML is required to initiate RA-driven NPM-1c degradation, P53 activation and cell death. Mechanistically, RA stabilizes PML through inhibition of overexpressed and activated Pin1, enforcing growth arrest. Such RA-induced activation of the PML/P53 signaling cascade enhances the activity of chemotherapy or arsenic in vivo. Our studies identify PML as an unsuspected actor downstream of RA in NPM-1c AML.

Methods Cell lines and treatments OCI-AML3 or OCI-AML2 AML cells (harboring the NPM-1c mutation without FLT3-ITD or wild type [wt] NPM-1 respectively) were grown in minimum essential medium-α (MEM-α) supplemented with 20% fetal bovine serum (FBS) and antibiotics. Cells were seeded at the density of 2x105/mL. RA (Sigma Aldrish) was used at a final concentration of 1 mM. The Pin1 inhibitor AG17724 (Sigma Aldrish) was used at 20 mM. Doxorubicin (Ebewe Pharma) and cytarabine (AraC) (Alexan, Ebewe Pharma) were each used at a final concentration of 1 mM. Cell growth was assessed using the trypan blue dye assay.

Patients’ cells Primary bone marrow blasts from AML patients were extracted following Ficoll separation and cultured in MEM-α supplemented with 20% FBS and antibiotics. Patients’ samples were collected following approval by the American University of Beirut Institutional Review Board and after patients had provided written informed consent in accordance with the Declaration of Helsinki. The patients’ characteristics are summarized in Online Supplementary Table S1. Two AML patients with NPM-1c mutations who were judged

haematologica | 2021; 106(12)

unfit for conventional chemotherapy received off-label compassionate RA (25 mg/m2 daily) and ATO (0.15 mg/kg daily) after informed consent.

CRISPR OCI-AML3 cell lines PML expression was abrogated by CRISPR-mediated excision. A guide RNA targeting PML (forward: 5’-GTCGGTGTACCGGCAGATTG; reverse: 5’-AATCTGCCGGTACACCGAC) was designed and cloned into pLAS5w.Ppuro-Cas9 plasmid for viral packaging. OCI-AML3 cells were infected with the corresponding viruses. Stable selection of knock-out cells was performed in the presence of 1 mg/mL puromycin, over a period of 2 weeks. Three OCI-AML3pml-/- clones were generated and tested in this study. Similarly, P53 extinction was performed using a guide RNA targeting P53 (forward: 5’-CCATTGTTCAATATCGTCCG; reverse: 5’CGGACGATATTGAACAATGG). Recombinant Cas9 protein was synthesized from IDT to form Alt-R CRISPR/Cas9 RNP. OCIAML3 cells were transiently transfected with Alt-R CRISPR/Cas9 RNP by using Nucleofector kit T (Amaxa) and applied program number X-01 in the nucleofactor device (Lonza). The stable CRISPR knock-out clones were cloned by serial dilution to generate a single-cell separation. DNA from individual clones was extracted and the region surrounding the Cas9 cutting site was amplified by polymerase chain reaction and verified by sequencing to ensure the deletion of the target genes. Two OCI-AML3P53-/clones were generated and tested in this study. Pin1 knock-down was perfomed using all-in-one plasmid pLAS5w.Ppuro-Cas9-gPin1 Pin1 (CCACCGTCACACAGTATTTAT). Lentiviruses were produced by transient transfection of HEK-293T cells. OCI-AML3 cells were infected by spinoculation for 90 min at 28,000 rpm at 32°C. Stable selection of knock-out cells was performed in the presence of 1 mg/mL puromycin. Single cells were sorted in 96-well plate to have stable clones. Knock-out stable clones were verified by immunofluorescence and western blot using a polyclonal anti-Pin1 antibody (Cell Signaling).

Statistical analysis Data are reported as means ± standard deviations. Statistical analysis was done using a Student t-test, P-values <0.05 are considered statistically significant.

Other methods Immunoblotting, RNA isolation and quantitative reverse transcriptase polymerase chain reaction, microarray analysis and gene set enrichment analysis, immunofluorescence and confocal microscopy, Pin1 activity assay, colony formation assay, xenograft animal studies, human CD45 staining and cell sorting are described in the Online Supplementary Methods.

Results PML-dependent NPM-1c degradation activates P53 We and others demonstrated that RA triggers NPM-1c degradation, P53 activation, apoptosis and induces PML nuclear body formation in NPM-1c-expressing AML cell lines.17,18 Since PML nuclear bodies may be implicated in therapy-induced catabolism of other oncoproteins,34 we were prompted to investigate any role of PML in RA-triggered NPM-1c degradation. We thus generated CRISPR PML OCI-AML3 cell lines (Figure 1A, Online Supplementary Figure S1A). In these OCI-AML3pml-/- cells, RA-mediated NPM-1c degradation was blocked (Figure 1A, Online Supplementary Figure S1A) and RA-induced cell death was abrogated (Figure 1B, Online Supplementary Figure S1B). 3091

R. Hleihel et al.

3092

Figure 1. PML-dependent NPM-1c degradation and P53 activation. (A) Western blot analysis of PML, P53 and NPM-1c was performed on extracts of OCIAML3, one clone of OCI-AML3pml-/- (OCI-AML3pml-/-#1) and one clone of OCI-AML3P53-/- (OCI-AML3P53-/-#1) after treatment with retinoic acid (RA) for 48 h. Densitometry histograms represent data from an average of five independent experiments. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test. (B) Cell growth (percent of control) was assessed using the trypan blue exclusion dye assay, in triplicate wells in OCI-AML3, one clone of OCIAML3pml-/- (OCI-AML3pml-/-#1) and one clone of OCIAML3P53-/- (OCI-AML3P53-/-#1) following treatment with RA for 48 h (n=3). (C) Western blot analysis of P53 and NPM-1c in OCI-AML3 and OCI-AML3pml-/- after treatment with RA for 2, 12 and 24 h as indicated. Densitometry histograms represent data from an average of three independent experiments. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test. (D) Western blot analysis of NPM-1c and P53 in primary blasts derived from five patients with NPM-1c acute myeloid leukemia (AML) and three NPM-1wt AML patients, after ex-vivo treatment with RA for 2 h. Densitometry histograms represent average P53 and NPM-1c expression levels in five NPM1c AML patients and three NPM-1wt AML patients. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test. (E) Western blot analysis of NPM-1c and P53 in primary blasts derived from five NPM-1c AML patients and three NPM-1wt AML patients, after exvivo treatment with RA for 48 h. Densitometry histograms represent average P53 and NPM-1c expression levels in five NPM-1c AML patients and three NPM-1wt AML patients. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test, *P<0.05; **P<0.01; ***P<0.001.

haematologica | 2021; 106(12)

RA/arsenic activate a PML/P53 axis in NPM-1c AML

Figure 2. Legend on following page.

haematologica | 2021; 106(12)

3093

R. Hleihel et al.

Figure 2. Retinoic acid targets PML/P53 through Pin1 inactivation. (A) Confocal microscopy of PML-nuclear bodies (NB) in primary blasts derived from one representative acute myeloid leukemia (AML) patient with wild-type (wt) NPM-1 and one representative AML patient with mutated NPM-1c, after ex-vivo treatment with retinoic acid (RA) or AF-17724 for 2 h as indicated. Histograms represente average number of PML NB per cell in two patients with NPM-1 wt and two patients with mutated NPM-1c. Statistical analysis was done using a Student t-test. (B) Western blot analysis of PML, P53 and total NPM-1 (NPM1-wt) in primary blasts derived from three patients with NPM-1c AML and three patients with NPM-1 wt AML, after ex-vivo treatment with 20 mM of AG17724 or 1 mM of RA for 2 h as indicated. (C) Western blot of PML and P53 in OCI-AML3 and OCI-AML2 cells following treatment with 20 mM of AG17724 for 2 h. Densitometry histograms represent an average of three independent experiments. (D) Colony-formation assays in methylcellulose of OCI-AML3, and OCI-AML3 Pin1-knock down (KD) cells, pre-treated with RA or AG17724 for 3 h (n=3). (E) Western blot analysis of PML, P53, P21, or Pin1 in OCI-AML3 and OCI-AML3 Pin1-KD after treatment with 1 mM of RA for 2 h. Densitometry histograms represent an average of three independent experiments. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test. (F) Western blot analysis of PML and P53 in OCI-AML3, OCI-AML3pml-/- and OCI-AML3P53-/- cells after treatment with 20 mM of AG17724 or 1 mM of RA for 2 h as indicated. Densitometry histograms represent an average of three independent experiments. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test, *P<0.05; **P<0.01; ***P<0.001.

When assessing the transcriptional effects of RA treatment in OCI-AML3 and control OCI-AML2 cells, a clear P53 signature was noted in OCI-AML3 cells (Online Supplementary Figure S1C-E), in line with RA-induced P53 stabilization.17,18 To assess any role of P53 in cell death upon RA exposure, we also generated CRISPR P53 OCI-AML3 (OCIAML3P53-/-) cell lines (Figure 1A, Online Supplementary Figure S1F). In this model, RA again failed to initiate cell death (Figure 1B, Online Supplementary Figure S1G), although it efficiently degraded NPM-1c (Figure 1A, Online Supplementary Figure S1F). Collectively, the RA-triggered, PML-facilitated NPM-1c degradation likely explains P53 activation and growth arrest of AML cells.

Retinoic acid targets P53 prior to NPM-1c loss Further investigating the kinetics of response to RA, we unexpectedly obtained evidence of rapid P53 stabilization prior to any significant NPM-1c loss, but with a requirement for PML expression (Figure 1C). Thus, NPM-1c loss is not the sole contributor to P53 activation. Remarkably, similar data were obtained upon ex vivo treatment of primary blasts derived from NPM-1c AML patients, in whom RA-triggered NPM-1c loss was only obtained after 48 h, while P53 stabilization was generally observed as soon as 2 h (Figure 1D, E). Critically, such RA-triggered P53 activation was solely observed in samples from patinets with NPM-1c AML (Figure 1D, E). Thus, delayed NPM-1c degradation is most unlikely to explain early P53 activation upon RA treatment.

Retinoic acid stabilizes PML through Pin1 inactivation RA also rapidly stabilized PML levels and induced PML nuclear body formation, solely in NPM-1c-positive patients' blasts or OCI-AML3 cells, with kinetics closely resembling those of P53 stabilization (Figure 2A, Online Supplementary Figure S2A-C). RA inconsistently enhanced PML gene expression, possibly through increased interferon production.35 Previous studies have shown that RA inhibits the Pin1 enzyme and the latter regulates PML stability.5,30,36 We thus compared the effects of RA and a Pin1 inhibitor (AG17724) on PML abundance, nuclear body formation and P53 activation. Strikingly, RA or AG17724 similarly stabilized PML or P53 levels and promoted nuclear body formation in NPM-1c AML patients’ blasts and OCI-AML3 cells (Figure 2A-C, Online Supplementary Figure S2A-C). In contrast, NPM-1-WT AML cells were unresponsive to RA and AG17724 (Figure 2A-C, Online Supplementary Figure S2A). Pin1 inhibition did not initiate NPM-1c degradation (Online Supplementary Figure S2D, E), implying that RAinduced PML stabilization is necessary, but not sufficient, to induce NPM-1c catabolism. Functionally, RA and AG17724 lead similarly to loss of clonogenic activity of OCI-AML3 cells in methyl-cellulose (Figure 2D). For a direct demonstration of the involvement of Pin1 in the response to RA, we 3094

generated an OCI-AML3 cell line with stable Pin1 downregulation by shRNA. Downregulation of Pin1 did not affect the viability of NPM-1c AML cells (Online Supplementary Figure S2F). Remarkably, PML and P53 or P21 activation by RA was abrogated following downregulation of Pin1 (Figure 2E), and RA-induced cell death was lost (Online Supplementary Figure 2F). Loss of clonogenic activity by RA or Pin1 inhibition was also abrogated (Figure 2D). Thus, RA-induced activation of P53 and resulting growth arrest are initiated by Pin1 inhibition.

PML is the primary target of retinoic acid and Pin1 inhibitors Pin1 directly controls both the stability of PML and P53 signaling.37 PML and P53 are highly cross-regulated: PML controls P53 activation, but P53 transcriptionally induces PML expression.38,39 To decipher the respective roles of PML and P53 in response to Pin1, we compared the responses to RA and AG17724 in OCI-AML3 and its pml-/- and P53-/derivatives. Both drugs upregulated PML levels in P53-/- cells, and no (or minimal) induction of P53 was observed in OCIAML3pml-/- cells (Figure 2F). These results establish that PML is the primary target of Pin1 inhibition in NPM-1c-expressing cells. These results support a model wherein RA inactivates Pin1, to stabilize PML, induce nuclear body formation, activate P53 and suppress growth. This does not exclude the possibility that P53 then constitutes a feed-forward amplification loop on PML expression.

NPM-1c-expressing cells exhibit high basal level and activity of Pin1 The implication of Pin1 inhibition in RA-mediated effects in NPM-1c-expressing cells prompted us to investigate Pin1 levels and activity. Strikingly, high Pin1 protein levels and activity were observed in NPM-1c-expressing cell lines and primary blasts (Figure 3A-C, Online Supplementary Figure S3A). Treatment with RA significantly reduced Pin1 activity (Figure 3C) without affecting Pin1 protein level (Online Supplementary Figure S3B). In line with this increased Pin1 protein level and activity in NPM-1c-expressing cells, both RA and AG17724 synergized with arsenic trioxide (ATO), and standard chemotherapy drugs used in AML, doxorubicin or cytarabine, to induce death of OCI-AML3 cells but not of OCI-AML2 cells (Figure 3D).

Retinoic acid and chemotherapy cooperate to clear NPM-1c-expressing cells in vivo The clinical benefit of co-administration of RA with chemotherapy seems restricted to AML bearing an NPM-1c mutation.16 We thus examined the possibility of in vivo cooperation between RA and standard-of-care AML therapy, anthracyclins and cytarabine, and examined any PML dependency. We used xenografts from OCI-AML3 or OCIhaematologica | 2021; 106(12)

RA/arsenic activate a PML/P53 axis in NPM-1c AML

AML3pml-/-, which were treated or not with RA for 1 week followed by either doxorubicin or cytarabine as single agents, for an additional 1 week. RA induced PML-dependent NPM-1c downregulation and human P53 stabilization in vivo (Figure 4A-D), while all three drugs ultimately induced human P53 phosphorylation (Figure 4A, B).

Importantly, RA synergized with doxorubicin or cytarabine to decrease abundance of human cells in treated mice, only in cells bearing intact PML (Figure 4E-H). These observations suggest that, at least in this model, RA cooperates with chemotherapy to decrease AML burden, resembling the clinical observations made in AML.

Figure 3. NPM-1c-expressing cells exhibit high Pin1 level and activity. (A) Western blot analysis of Pin1 in OCI-AML2, OCI-AML2-NPM-1wt, OCI-AML2-NPM-1c, OCIAML3, OCI-AML3pml-/- and OCI-AML3P53-/-. (B) Western blot analysis of NPM-1c and Pin1 in primary blasts derived from seven patients with acute myeloid leukemia (AML) with NPM-1c (p4, p2, p5, p8, p7, p1 and p6) and six AML patients with wild-type NPM-1 (NPM-1 wt) (p11, p13, p14, p10, p15 and p16). Densitometry histograms represent an average of Pin1 expression level in the seven tested NPM-1c AML patients and the six tested NPM-1 wt AML patients. Densitometry was performed using ImageJ software. Statistical analysis was done using a Student t-test. (C) Pin1 relative activity in OCI-AML2, OCI-AML3 and in primary blasts derived from three NPM-1c AML patients and two NPM-1 wt AML patients after treatment with 1 mM of RA, as indicated. Statistical analysis was done using a Student t-test. (D) Cell growth (percent of control) was assessed using the trypan blue exclusion dye assay in OCI-AML2 and OCI-AML3 cells following treatment with RA alone, AG-17724 alone or their combination with arsenic trioxide (ATO), cytarabine (Ara-C) and doxorubicin for 72 h as indicated (n=3). Statistical analysis was done using a Student t-test, *P<0.05; **P<0.01; ***P<0.001.

haematologica | 2021; 106(12)

3095

R. Hleihel et al.

3096

Figure 4. Retinoic acid and chemotherapy cooperate to clear NPM-1c-expressing cells in vivo. (A, B) Eight-week-old NSG mice were injected with 2x106 OCIAML3 cells intravenously. At day 21 after injection of the leukemic cells, retinoic acid (RA) was administered on a daily basis at the dose of 2.5 mg/g over a period of 1 week, followed by the administration of doxorubicin (2 mg/g) or cytarabine (AraC) (60 mg/g) twice a week over a period of 1 week. Mice were sacrificed, Bone marrow was harvested from femura and tibiae of xenografted mice and then stained with anti-human CD45 (hCD45) antibody. Western blot of NPM-1c, human P53, human P-P53 and PML in sorted hCD45-positive cells from BM harvested from untreated or treated NSG xenografted mice as indicated (2 mice per condition). (C, D) Eight-weekold NSG mice were intravenously injected with 2x106 OCI-AML3pml-/- cells. The same treatment regimen was followed as indicated above. Western blot of NPM-1c, human P53 and P-P53 in sorted hCD45-positive cells from bone marrow harvested from untreated or treated NSG xenografted mice as indicated. (E, F) Graphs showing the percentage of hCD45 cells in OCI-AML3 xenografted NSG mice treated as described above (7 mice in the untreated group and in the groups treated with doxorubicin alone or RA in combination with cytarabine [AraC], 9 mice in the group treated with RA alone or doxorubicin and RA, 8 mice in the group treated with cytarabine alone). (G, H) Graphs showing the percentage of hCD45 cells in OCI-AML3pml-/xenografted NSG animals treated as described above (3 mice in each condition).

haematologica | 2021; 106(12)

RA/arsenic activate a PML/P53 axis in NPM-1c AML

Figure 5. Retinoic acid and arsenic cooperate to clear NPM-1c-expressing cells. (A) Eight-week-old NSG mice were injected with 3x106 OCI-AML3 or OCI-AML3pml-/cells intravenously. From day 7 after injection of the leukemic cells, arsenic trioxide (ATO, 5 mg/g/day) and retinoic acid (RA, 2.5 mg/g) were administered intraperitoneally every other day, over a period of 4 weeks. Western blot of human P53 and NPM-1c in sorted hCD45 positive bone marrow (BM) cells from NSG mice xenografted with OCI-AML3 or OCI-AML3pml-/- cells, after in vivo treatment with ATO alone, RA alone or the combination of RA and ATO. (B) Eight-week-old NSG mice were injected with 3x106 OCI-AML3 or OCI-AML3pml-/- cells intravenously. From day 7 after injection of the leukemic cells, ATO and RA were administered every other day, over a period of 4 weeks intraperitoneally. At the end of treatment, bone marrow was harvested from femora and tibiae of xenografted mice and then stained with anti-hCD45 antibody. Graphs show the percentage of hCD45 cells in xenografted animals. (C) Treatment schedule in two patients with NPM-1c mutated acute myeloid leukemia (AML) treated with RA and ATO as indicated. Percentages of peripheral blood (PB) and bone marrow (BM) blasts are displayed. (D) Proposed model of the molecular mechanisms of the response of NPM-1c AML to RA. ITD: internal tandem duplication; NB: nuclear body.

The combination of retinoic acid and arsenic trioxide has clinical activity in mice and patients with acute myeloid leukemia RA and ATO trigger NPM-1c degradation, P53 activation and apoptosis in NPM-1c-expressing AML cell lines.17,18 RA upregulates PML, while ATO targets PML to enforce nuclear body formation and also inhibits Pin1.6 Thus, in principle, ATO could cooperate with RA through PML nuclear body targeting.19 To explore any in vivo relevance of these findings, we treated xenografts from OCI-AML3 or OCI-AML3pml-/- with RA and ATO for 1 week. RA/ATO induced PML-dependent NPM-1c degradation and human P53 stabilization in vivo (Figure 5A) and decreased abundance of human cells in treated mice, again solely in cells bearing intact PML (Figure 5B). The combination of RA/ATO is a very well-tolerated therapeutic association in APL.40 Two NPM-1c AML patients, unfit for conventional therapy, received this RA/ATO combination on an off-label compassionate basis. Blast clearance from peripheral blood and, to a lesser extent, from bone marrow was observed, although complete haematologica | 2021; 106(12)

remission was not achieved (Figure 5C). Longer follow-up after 2 months showed appearance of slowly growing AML cells. Thus, the RA/ATO combination may transiently target AML cells in some NPM-1c-mutated patients.

Discussion The basis for sensitivity of non-APL AML to RA was initially believed to be RA-induced differentiation.11,41 Here we report that PML constitutes an unsuspected actor downstream of RA and is required for its synergistic activity with other therapies in NPM-1c AML models. Previous ex vivo studies suggested that RA-driven NPM-1c degradation could be the molecular basis of the therapeutic activity of this drug through upregulation of ARF and resulting activation of P53. NPM-1c degradation should also correct multiple other phenotypes associated with NPM-1c, including sequestration of key regulators in the cytoplasm or transcriptional deregulation.4,42-45 Here, kinetic analysis of P53 activation upon RA treatment revealed that P53 3097

R. Hleihel et al.

upregulation actually precedes NPM-1c loss, which is suggestive of the existence of at least another pathway of P53 activation. Indeed we found that RA plays an essential role in growth arrest through Pin1 inhibition in both OCIAML3 and primary patients’ blasts of NPM-1c AML cells only. Downstream of Pin1 inhibition, we identified the PML and P53 growth suppressors as its essential downstream effectors.36,46 Analysis of pml-/- and P53-/- AML3 cells demonstrated that P53 is downstream of PML-triggered Pin1 or RA responses. Our discovery of the key role of PML downstream of RA-initiated Pin1 inhibition (Figure 5D) suggests that RA-initiated Pin1 inhibition would upregulate PML and promote PML nuclear body formation, ultimately driving P53/senescence. These results suggest that in NPM-1c-driven AML, but not in NPM1-WT AML, impairment of PML nuclear body formation is involved in leukemic transformation and that RA-mediated restoration of PML nuclear bodies contributes to the therapeutic effects. In NPM-1c-positive AML, elucidation of the respective contributions of PML nuclear bodies and NPM-1c degradation in the in vivo response, notably in combination with conventional chemotherapy, requires further investigations. Yet, the absence of an effect of RA on clonogenic activity of Pin1 downregulated AML cells favors an important role of Pin1 inhibition in biological response and not only early P53 activation (Figure 2D). This model has several feed-forward loops, all favoring anti-proliferative responses: RA-induced PML stabilization should facilitate PML-dependent NPM-1c degradation and P53 activation will enhance PML expression. Our results unravel a parallelism with the APL model: both involve oncoproteins that downregulate basal P53 signaling. In both, therapy response involves degradation of the driv-

References 1. Grisendi S, Mecucci C, Falini B, Pandolfi PP. Nucleophosmin and cancer. Nat Rev Cancer. 2006;6(7):493-505. 2. Falini B, Nicoletti I, Martelli MF, Mecucci C. Acute myeloid leukemia carrying cytoplasmic/mutated nucleophosmin (NPMc+ AML): biologic and clinical features. Blood. 2007;109(3):874-885. 3. Heath EM, Chan SM, Minden MD, Murphy T, Shlush LI, Schimmer AD. Biological and clinical consequences of NPM1 mutations in AML. Leukemia. 2017;31(4):798-807. 4. Kunchala P, Kuravi S, Jensen R, McGuirk J, Balusu R. When the good go bad: mutant NPM1 in acute myeloid leukemia. Blood Rev. 2018;32(3):167-183. 5. Wei S, Kozono S, Kats L, et al. Active Pin1 is a key target of all-trans retinoic acid in acute promyelocytic leukemia and breast cancer. Nat Med. 2015;21(5):457-466. 6. Kozono S, Lin YM, Seo HS, et al. Arsenic targets Pin1 and cooperates with retinoic acid to inhibit cancer-driving pathways and tumor-initiating cells. Nat Commun. 2018;9(1):3069. 7. de The H. Differentiation therapy revisited. Nat Rev Cancer. 2018;18(2):117-127. 8. de The H, Pandolfi PP, Chen Z. Acute promyelocytic leukemia: a paradigm for oncoprotein-targeted cure. Cancer Cell. 2017;32(5):552-560. 9. Ablain J, Rice K, Soilihi H, de Reynies A, Minucci S, de The H. Activation of a

3098

ing oncogene, PML nuclear body formation and P53 activation.8 The co-existence of other major oncogenes (epigenetic regulators, FLT3) in NPM-1c-positive AML most likely explains why RA/ATO is not curative on its own. Nevertheless, RA could favor the action of chemotherapy, possibly by reverting basal Pin1 activation and P53 inhibition. Our observations could explain the clinical benefit of the co-administration of RA with conventional chemotherapy in NPM-1c-positive AML.13,16 More broadly, the RA/ATO combination could target malignancies in which Pin1 and/or PML are deregulated.26,27 In this respect, clinical responses by solid tumors were observed in some RA/ATO-treated APL patients who had another synchronous malignancy,47,48 possibly reflecting activation of the RA/Pin1/PML/P53 axis elucidated by this study. Disclosures No conflicts of interest to disclose. Contributions RH, HCW, and CB performed experiments; ZHH treated patients; RM collected bone marrow samples; RH, HEH, HCW, CB, ZC, MES, HdT, and AB. analyzed results; RH, HEH, HCW, and CB created the figures; HEH, AB, and HdT designed the research and wrote the paper. Funding This work was supported by the American University of Beirut (AUB) and the Lebanese National Council for Scientific Research (CNRSL) (Group Research Proposal GRP AUB-CNRSL) (to HEH and AB); the Paris laboratory is supported by INSERM, CNRS, College de France, Université de Paris, Ligue Contre le Cancer, TRANSCAN, CAMELIA, The Sjoberg Foundation, and Foundation St. Joseph.

promyelocytic leukemia-tumor protein 53 axis underlies acute promyelocytic leukemia cure. Nat Med. 2014;20(2):167-174. 10. Lehmann-Che J, Bally C, Letouzé E, et al. Dual origin of relapses in retinoic-acid resistant acute promyelocytic leukemia. Nat Commun. 2018;9(1):2047. 11. Boutzen H, Saland E, Larrue C, et al. Isocitrate dehydrogenase 1 mutations prime the all-trans retinoic acid myeloid differentiation pathway in acute myeloid leukemia. J Exp Med. 2016;213(4):483-497. 12. Altucci L, Rossin A, Hirsch O, et al. Rexinoid-triggered differentiation and tumours selective apoptosis of AML by protein kinase-A-mediated de-subordination of RXR. Cancer Res. 2005;65(19):8754-8765. 13. Schlenk RF, Frohling S, Hartmann F, et al. Phase III study of all-trans retinoic acid in previously untreated patients 61 years or older with acute myeloid leukemia. Leukemia. 2004;18(11):1798-1803. 14. Burnett AK, Milligan D, Prentice AG, et al. A comparison of low-dose cytarabine and hydroxyurea with or without all-trans retinoic acid for acute myeloid leukemia and high-risk myelodysplastic syndrome in patients not considered fit for intensive treatment. Cancer. 2007;109(6):1114-1124. 15. Milligan DW, Wheatley K, Littlewood T, Craig JI, Burnett AK, NCRI Haematological Oncology Clinical Studies Group. Fludarabine and cytosine are less effective than standard ADE chemotherapy in highrisk acute myeloid leukemia, and addition

of G-CSF and ATRA are not beneficial: results of the MRC AML-HR randomized trial. Blood. 2006;107(12):4614-4622. 16. Schlenk RF, Dohner K, Kneba M, et al. Gene mutations and response to treatment with all-trans retinoic acid in elderly patients with acute myeloid leukemia. Results from the AMLSG Trial AML HD98B. Haematologica. 2009;94(1):54-60. 17. Martelli MP, Gionfriddo I, Mezzasoma F, et al. Arsenic trioxide and all-trans retinoic acid target NPM1 mutant oncoprotein levels and induce apoptosis in NPM1-mutated AML cells. Blood. 2015;125(22):3455-3465. 18. El Hajj H, Dassouki Z, Berthier C, et al. Retinoic acid and arsenic trioxide trigger degradation of mutated NPM1, resulting in apoptosis of AML cells. Blood. 2015;125 (22):3447-3454. 19. Mugoni V, Panella R, Cheloni G, et al. Vulnerabilities in mIDH2 AML confer sensitivity to APL-like targeted combination therapy. Cell Res. 2019;29(6):446-459. 20. Lallemand-Breitenbach V, de The H. PML nuclear bodies: from architecture to function. Curr Opin Cell Biol. 2018;52:154-161. 21. Niwa-Kawakita M, Ferhi O, Soilihi H, Le Bras M, Lallemand-Breitenbach V, de The H. PML is a ROS sensor activating P53 upon oxidative stress. J Exp Med. 2017;214(11): 3197-3206. 22. Zhu J, Koken MHM, Quignon F, et al. Arsenic-induced PML targeting onto nuclear bodies: implications for the treatment of acute promyelocytic leukemia. Proc Natl

haematologica | 2021; 106(12)

RA/arsenic activate a PML/P53 axis in NPM-1c AML

Acad Sci U S A. 1997;94(8):3978-3983. 23. Lehmann-Che J, Bally C, de The H. therapy resistance in APL. N Engl J Med. 2014;371(12):1171-1172. 24. Lehmann-Che J, Bally C, Letouze E, et al. Dual origin of relapses in retinoic-acid resistant acute promyelocytic leukemia. Nat Commun. 2018;9(1):2047. 25. Jeanne M, Lallemand-Breitenbach V, Ferhi O, et al. PML/RARA oxidation and arsenic binding initiate the antileukemia response of As2O3. Cancer Cell. 2010;18(1):88-98. 26. Koken MHM, Linares-Cruz G, Quignon F, et al. The PML growth-suppressor has an altered expression in human oncogenesis. Oncogene. 1995;10(7):1315-1324. 27. Gurrieri C, Capodieci P, Bernardi R, et al. Loss of the tumor suppressor PML in human cancers of multiple histologic origins. J Natl Cancer Inst. 2004;96(4):269-279. 28. Scaglioni PP, Yung TM, Cai LF, et al. A CK2dependent mechanism for degradation of the PML tumor suppressor. Cell. 2006;126 (2):269-283. 29. Wu HC, Lin YC, Liu CH, et al. USP11 regulates PML stability to control Notch-induced malignancy in brain tumours. Nat Commun. 2014;5:3214. 30. Yuan WC, Lee YR, Huang SF, et al. A Cullin3-KLHL20 ubiquitin ligase-dependent pathway targets PML to potentiate HIF-1 signaling and prostate cancer progression. Cancer Cell. 2011;20(2):214-228. 31. Fischer M. Census and evaluation of P53 target genes. Oncogene. 2017;36(28):39433956. 32. Lallemand-Breitenbach V, Jeanne M, Benhenda S, et al. Arsenic degrades PML or PML-RARalpha through a SUMO-triggered RNF4/ubiquitin-mediated pathway. Nat Cell Biol. 2008;10(5):547-555.

haematologica | 2021; 106(12)

33. Nabbouh AI, Hleihel RS, Saliba JL, et al. Imidazoquinoxaline derivative EAPB0503: a promising drug targeting mutant nucleophosmin 1 in acute myeloid leukemia. Cancer. 2017;123(9):1662-1673. 34. Dassouki Z, Sahin U, El Hajj H, et al. ATL response to arsenic/interferon therapy is triggered by SUMO/PML/RNF4-dependent Tax degradation. Blood. 2015;125(3):474482. 35. Stadler M, Chelbi-Alix MK, Koken MHM, et al. Transcriptional induction of the PML growth suppressor gene by interferons is mediated through an ISRE and a GAS element. Oncogene. 1995;11(12):2565-2573. 36. Reineke EL, Lam M, Liu Q, et al. Degradation of the tumor suppressor PML by Pin1 contributes to the cancer phenotype of breast cancer MDA-MB-231 cells. Mol Cell Biol. 2008;28(3):997-1006. 37. Mantovani F, Zannini A, Rustighi A, Del Sal G. Interaction of P53 with prolyl isomerases: healthy and unhealthy relationships. Biochim Biophys Acta. 2015;1850(10):20482060. 38. Pearson M, Carbone R, Sebastiani C, et al. PML regulates P53 acetylation and premature senescence induced by oncogenic Ras. Nature. 2000;406(6792):207-210. 39. de Stanchina E, Querido E, Narita M, et al. PML is a direct P53 target that modulates P53 effector functions. Mol Cell. 2004;13(4): 523-535. 40. Lo-Coco F, Di Donato L: GIMEMA, Schlenk RF: German-Austrian Acute Myeloid Leukemia Study Group and Study Alliance Leukemia. Targeted therapy alone for acute promyelocytic leukemia. N Engl J Med. 2016;374(12):1197-1198. 41. Altucci L, Rossin A, Hirsch O, et al. Rexinoid-triggered differentiation and

tumor-selective apoptosis of acute myeloid leukemia by protein kinase A-mediated desubordination of retinoid X receptor. Cancer Res. 2005;65(19):8754-8765. 42. Gu X, Ebrahem Q, Mahfouz RZ, et al. Leukemogenic nucleophosmin mutation disrupts the transcription factor hub that regulates granulomonocytic fates. J Clin Invest. 2018;128(10):4260-4279. 43. Haindl M, Harasim T, Eick D, Muller S. The nucleolar SUMO-specific protease SENP3 reverses SUMO modification of nucleophosmin and is required for rRNA processing. EMBO Rep. 2008;9(3):273-279. 44. Kuo ML, den Besten W, Thomas MC, Sherr CJ. Arf-induced turnover of the nucleolar nucleophosmin-associated SUMO-2/3 protease Senp3. Cell Cycle. 2008;7(21):33783387. 45. Yun C, Wang Y, Mukhopadhyay D, et al. Nucleolar protein B23/nucleophosmin regulates the vertebrate SUMO pathway through SENP3 and SENP5 proteases. J Cell Biol. 2008;183(4):589-595. 46. Lim JH, Liu Y, Reineke E, Kao HY. Mitogenactivated protein kinase extracellular signalregulated kinase 2 phosphorylates and promotes Pin1 protein-dependent promyelocytic leukemia protein turnover. J Biol Chem. 2011;286(52):44403-44411. 47. Jain P, Konoplev S, Benjamini O, Romagura J, Burger JA. Long-term control of refractory follicular lymphoma after treatment of secondary acute promyelocytic leukemia with arsenic trioxide (As2O3) and all-trans retinoic acid (ATRA). Blood Res. 2018;53(2):169-172. 48. Alsafadi S, Even C, Falet C, al. Retinoic acid receptor alpha amplifications and retinoic acid sensitivity in breast cancers. Clin Breast Cancer. 2013;13(5):401-408.

3099

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(12):3100-3106

Acute Promyelocytic Leukemia

Characteristics and outcome of patients with low-/intermediate-risk acute promyelocytic leukemia treated with arsenic trioxide: an international collaborative study Sabine Kayser,1,2 Richard F. Schlenk,2,3 Delphine Lebon,4 Martin Carre,5 Katharina S. Götze,6 Friedrich Stölzel,7 Ana Berceanu,8 Kerstin Schäfer-Eckart,9 Pierre Peterlin,10 Yosr Hicheri,11 Ramy Rahmé,12 Emmanuel Raffoux,12 Fatiha Chermat,12 Stefan W. Krause,13 Walter E. Aulitzky,14 Sophie Rigaudeau,15 Richard Noppeney,16 Celine Berthon,17 Martin Görner,18 Edgar Jost,19 Philippe Carassou,20 Ulrich Keller,21 Corentin Orvain,22,23,24 Thorsten Braun,25 Colombe Saillard,26 Ali Arar,27 Volker Kunzmann,28 Mathieu Wemeau,29 Maike de Wit,30 Dirk Niemann,31 Caroline Bonmati,32 Carsten Schwänen,33 Julie Abraham,34 Ahmad Aljijakli,35 Stéphanie Haiat,36 Alwin Krämer,37,38 Albrecht Reichle,39 Martina Gnadler,40 Christophe Willekens,41,42 Karsten Spiekermann,43 Wolfgang Hiddemann,43 Carsten Müller-Tidow,37 Christian Thiede,7 Christoph Röllig,7 Hubert Serve,44 Martin Bornhäuser,7 Claudia D. Baldus,45 Eva Lengfelder,46 Pierre Fenaux,12 Uwe Platzbecker1# and Lionel Adès12# Medical Clinic and Policlinic I, Hematology and Cellular Therapy, University Hospital Leipzig, Leipzig, Germany; 2NCT Trial Center, National Center of Tumor Diseases, German Cancer Research Center (DKFZ), Heidelberg, Germany; 3Department of Internal Medicine V, Heidelberg University Hospital, Heidelberg, Germany 4Service des Maladies du Sang GH Sud, Amiens, France; 5HCE Grenoble, Service d'Oncologie et Hématologie Pédiatrique, Grenoble, France; 6Department of Medicine III, Technical University of Munich, Munich, Germany; 7Department of Medicine I, University Hospital Carl-Gustav-Carus, Dresden, TU Dresden, Germany; 8Service d'Hématologie du Pr. Cahn Hopital Jean Minjoz, Besancon, France; 9Hospital Nord, Nurnberg, Germany; 10Service d'Hématologie, Hotel Dieu Nantes, France; 11Département d'Hématologie Clinique du Chu Saint Eloi Montpellier, Montpellier, France; 12Hôpital Saint Louis, Université Paris Diderot, Paris, France; 13Department of Internal Medicine 5 – Hematology/Oncology, University Hospital of Erlangen, Erlangen, Germany; 14Robert-Bosch-Hospital, Stuttgart, Germany; 15Service d'Hémato-Oncologie du Pr. Castaigne Hopital Andre Mignot le Chesnay, Versailles, France; 16University Hospital Essen, Essen, Germany; 17Service des Maladies du Sang Chru, Hopital Claude Huriez, Lille, France; 18Klinik für Hämatologie, Onkologie und Palliativmedizin, Klinikum Bielefeld Mitte, Bielefeld, Germany; 19University Hospital Aachen, Aachen, Germany; 20Department of Hematology, Centre Hospitalier Régional (CHR) Metz-Thionville, Metz, France; 21Department of Hematology, Oncology and Tumor Immunology, Charité-University Medical Center, Campus Benjamin Franklin, Berlin, Germany; 22Angers University Hospital, Maladies du Sang, Angers, France; 23Fédération Hospitalo-Universitaire Grand-Ouest Acute Leukemia, FHU-GOAL, France; 24Université d'Angers, INSERM, CRCINA, Angers, France; 25Hôpital Avicenne, Assistance Publique - Hôpitaux de Paris, Université Paris 13, Bobigny, France; 26 Hematology Department, Institut Paoli Calmettes, Marseille, France; 27Service d' Oncologie Médicale Hopital de la Source, Orleans, France; 28University Hospital of Würzburg, Würzburg, Germany; 29Service d'Hématologie, Hopital V. Provo, Roubaix, France; 30 Vivantes Klinikum Neukölln, Berlin, Germany; 31Gemeinschaftsklinikum Mittelrhein gGmbH, Koblenz, Germany; 32Division of Hematology, Hôpital de Brabois, Centre Hospitalier Universitaire de Nancy, Nancy, France; 33Klinikum Offenburg, Offenburg, Germany; 34Service d' Hématologie, Thérapie Cellulaire du Pr. Bordessoule, Hopital Universitaire Dupuytren, Limoges, France; 35Service d'Hématologie du Dr. Sutton Centre Hospitalier v. Dupouy, Argenteuil, France; 36Service d'Hematologie Clinique, CH Sud Francilien, Corbeil Essonnes, France; 37Department of Internal Medicine V, University Hospital of Heidelberg, Heidelberg, Germany; 38German Cancer Research Center (DKFZ) and Department of Internal Medicine V, University of Heidelberg, Heidelberg, Germany; 39 Department of Medicine III - Hematology and Internal Oncology, University Hospital Regensburg, Regensburg, Germany; 40St. Vincentius Kliniken, Abteilung für Hämatologie, Onkologie, Immunologie und Palliativmedizin, Karlsruhe, Germany; 41Département d'Hématologie, Gustave Roussy, Université Paris-Saclay, Villejuif, France; 42INSERM Unit 1287, Gustave Roussy Cancer Campus, Paris-Saclay University, Villejuif, France; 43 Department of Medicine III, University Hospital, Ludwig-Maximilians University (LMU) Munich, Munich, Germany; 44Department of Internal Medicine II, University Hospital of Frankfurt am Main, Germany; 45Department of Internal Medicine II, University Hospital of Kiel, Kiel, Germany and 46Department of Hematology and Oncology, University Hospital Mannheim, Mannheim, Germany. 1

Presented in part at the 21st Annual Meeting of the European Hematology Association in Frankfurt, Germany, June 10th, 2020.

Correspondence: SABINE KAYSER sabine.kayser@medizin.uni-leipzig.de Received: March 6, 2021. Accepted: April 26, 2021. Pre-published: May 27, 2021. https://doi.org/10.3324/haematol.2021.278722

UP and LA contributed equally as co-senior authors.

3100

haematologica | 2021; 106(12)

Outcome of APL patients after ATO-based treatment

ABSTRACT

he aim of this study was to characterize a large series of 154 patients with acute promyelocytic leukemia (median age, 53 years; range, 18-90 years) and evaluate real-life outcome after up-front treatment with arsenic trioxide and all-trans retinoic acid. All patients were included in the prospective NAPOLEON registry (NCT02192619) between 2013 and 2019. The acute promyelocytic leukemia was de novo in 91% (n=140) and therapy-related in 9% (n=14); 13% (n=20) of the patients were older than 70 years. At diagnosis bleeding/hemorrhage was present in 38% and thrombosis in 3%. Complete remission was achieved in 152 patients (99%), whereas two patients (1%) experienced induction death within 18 days after starting therapy. With a median follow-up of 1.99 years (95% confidence interval: 1.61-2.30 years) 1-year and 2-year overall survival rates were 97% (95% confidence interval: 94-100%) and 95% (95% confidence interval: 91-99%), respectively. Age above 70 years was associated with a significantly shorter overall survival (P<0.001) compared to that of younger patients. So far no relapses have been observed. Six patients (4%) died in complete remission at a median of 0.95 years after diagnosis (range, 0.18-2.38 years). Our data confirm the efficiency and durability of arsenic trioxide and all-trans retinoic acid therapy in the primary management of adults with low-/intermediate-risk acute promyelocytic leukemia in the real-life setting, irrespective of age.

Introduction Acute promyelocytic leukemia (APL), characterized by the balanced translocation t(15;17)(q22;q12) resulting in the fusion transcript PML-RARA, is a rare entity of acute myeloid leukemia, accounting for roughly 5-8% of cases of acute myeloid leukemia.1 With the introduction of all-trans retinoic acid (ATRA) and arsenic trioxide (ATO), the prognosis of APL has been transformed and this is now one of the most curable malignant diseases. Published data from a large multicenter phase III randomized trial (APL0406) on a direct comparison of ATO/ATRA versus ATRA in combination with idarubicin (AIDA) or mitoxantrone in adult patients with de novo, low-/intermediate-risk APL showed very promising results in favor of ATO/ATRA, with a 2-year event-free survival rate of 97% versus 86% (P=0.02).2 Within this trial, the early mortality rate as well as hematologic toxicities were significantly lower in patients treated with ATO/ATRA than in those treated with AIDA. In particular, the cumulative incidence of relapse after 50 months was only 1.9% after ATO/ATRA as compared to 13.9% after chemotherapy and ATRA.3 Moreover, none of the patients treated with ATO/ATRA developed a therapy-related myeloid neoplasm as compared to two patients in the chemotherapy/ATRA arm.3 Another Medical Research Council publication supported these results, with a 4-year event-free survival rate of 91% after ATO/ATRA as compared to 70% after chemotherapy and ATRA (P=0.002).4 The ATO/ATRA regimen was, however, associated with a higher frequency of grade 3 or 4 hepatic toxicity (44% vs. 3%; P<0.001). In all cases, toxic effects resolved with temporary discontinuation of ATO and/or ATRA3 and the chemotherapy-free regimen with ATO/ATRA has become standard first-line therapy in low/intermediate-risk de novo APL. However, apart from reports of the randomized Italian-German APL04062,3 and the AML17 trials,4 data on outcomes after up-front treatment with ATO/ATRA in the real-world setting are scarce. Previously published data from registries and populationbased studies indicated that outcomes of patients with APL were inferior to those reported in clinical trials, mainly due to higher early death rates.5-8 However, none of these studies included results after treatment with ATO/ATRA. Recently published results from the randomized Italianhaematologica | 2021; 106(12)

German APL0406 trial suggest, albeit within the context of highly selected patients and closely monitored therapy, that treatment with ATO/ATRA can reduce early deaths and improve long-term outcomes.2,3 Whether or not these results can be replicated in population-based settings is currently unclear. The objectives of our study were to characterize a large series of APL patients included in the prospective NAPOLEON registry and evaluate their outcome after upfront treatment with ATO/ATRA.

Methods Patients and treatment Data on 167 APL patients, reported within the NAPOLEON registry (ClinicalTrials.gov Identifier: NCT02192619) between 2013 and 2019 within two large European study groups (APL French group, n=89; Study Alliance Leukemia, n=78) were collected. This real-life registry included data on patients’ demographics, treatment, response, severe adverse events and follow-up. There were no criteria for exclusion of APL patients from this registry. However, for the current analysis the inclusion criterion was treatment with ATO/ATRA and exclusion criteria were age below 18 years (n=6), high-risk APL (white blood cell [WBC] count >10x109/L, at diagnosis, n=4) and lack of data on WBC count (n=3). The final study cohort consisted of 154 patients. Ideally, patients were included within the same (n=45) or following (n=32) day of APL diagnosis. Additionally 46 patients were included within the first 8 days after diagnosis (overall n=123; 80%), when the risk of early death is highest.8 Outcome was calculated using the date of initial diagnosis. None of the patients was treated within a clinical trial. The diagnosis of APL was based on genetic analysis as well as on French-American-British Cooperative Group criteria,9 and, after 2003, on revised International Working Group criteria.10 Chromosome banding analysis was performed using standard techniques, and karyotypes were described according to the International System for Human Cytogenetic Nomenclature.11 The diagnosis was confirmed by either reverse-transcriptase polymerase chain reaction (RT-PCR) or fluorescence in-situ hybridization (FISH) detection by standard methods. FLT3 mutation screening for internal tandem duplications (ITD) and point mutations within the tyrosine kinase domain was carried out as previously

3101

S. Kayser et al.

described.12,13 Data collection and analysis were approved by the local Institutional Review Boards.

Treatment All patients were treated with ATO/ATRA as induction therapy and with up to four consolidation cycles of ATO/ATRA.2 Eight patients additionally received 12 mg/m2 idarubicin (for up to 4 days) and two patients received 100 mg/m2 cytarabine for up to 2 days during induction with ATO/ATRA because of WBC counts rising above 10x109/L. One elderly patient (age, 72 years) was initially treated with decitabine (38 mg/day) for 5 days and was then switched to two cycles of ATO/ATRA after genetic analysis revealed t(15;17). Response was assessed according to Cheson et al.10

Statistical analyses Survival endpoints including overall survival and cumulative incidence of death in complete remission were defined according to the revised recommendations of the International Working Group.10 Induction death was defined as death occurring at any time during induction therapy before the achievement of complete remission. Comparisons of patients’ characteristics were performed with the Kruskal-Wallis rank sum test for continuous variables and Fisher exact test for categorical variables. The median follow-up time was computed using the reverse Kaplan-Meier estimate.14 The Kaplan-Meier method was used to estimate the distribution of overall survival.15 Confidence interval (CI) estimates for survival curves were based on the cumulative hazard function using the Greenwood formula for variance estimation. Log-rank tests were employed to compare survival curves between groups. The cumulative incidence of death in complete remission and the standard error were computed using the method described by Gray16 and included only patients attaining complete remission. All statistical analyses were performed with the statistical software environment R, version 3.3.1, using the R packages, and survival, version 2.39-5.17

Results In total, data on 154 low-/intermediate-risk APL patients, reported within the NAPOLEON registry between 2013 and 2019 within two large European study groups were included. The median age of these patients was 53 years (range, 18-90 years); 13% (n=20) were older than 70 years. APL was de novo in 91% (n=140) and therapy-related in 9% (n=14). Primary malignancies included solid cancers in 12 patients (breast cancer, n=5; female genital tract, n=2; prostate cancer, lung cancer, colon cancer, intraabdominal liposarcoma, and epidermoid cancer, n=1 each) and lymphoma in two patients. All patients received chemotherapy and/or radiation as treatment for the prior malignancy/neoplasm and were in remission at the time of APL diagnosis. The median latency period between diagnosis of the primary malignancy/neoplasm and the occurrence of therapyrelated APL was 2.80 years (range, 1.68-7.70 years). Information on extramedullary disease was available for 131 patients and was present in two (1.5%) patients (skin manifestation, hepatosplenomegaly; n=1 each). Comorbidities were present in 48 (58%) of 83 patients and included arterial hypertension (n=25), pulmonary disease (n=12; mild, n=5, moderate/severe, n=7), gastro-intestinal disease (n=9), renal disease (n=8; mild, n=4, moderate/severe, n=4), diabetes (n=8), psychiatric disturbance (n=8), cardiac/arrhythmias (n=6), rheumatologic dis3102

order (n=5), hepatic disease (n=3, mild, n=1, moderate/severe, n=2), cerebrovascular disease (n=3), and immunosuppression due to human immune-deficiency virus infection (n=2). In all patients APL was confirmed by cytogenetics and/or FISH/RT-PCR. Information on cytogenetics was available for 132 (86%) patients. In seven (5%) patients t(15;17) could only be detected by FISH and/or RT-PCR, while cytogenetics showed a normal karyotype. In 82 (66%) of the remaining 125 patients, the balanced t(15;17) translocation was the sole abnormality, whereas in 43 (34%) patients, the translocation was accompanied by additional cytogenetic abnormalities, most frequently t(15;17) within a complex karyotype, i.e., ≥2 cytogenetic abnormalities in addition to t(15;17) (n=16) or trisomy 8 (n=7). Regarding FLT3-ITD mutational status, information was available for 51 (33%) of the 154 patients. Of those, eight (16%) patients were FLT3-ITD-positive. Information on the PML-RARA transcript isoform (breakpoint cluster region, BCR) was available for 104 (67%) patients and 50 had the short isoform (BCR3). The patients’ baseline characteristics are shown in Table 1.

Response to induction therapy Overall, 152 patients (99%) achieved complete remission, whereas two patients (1%) experienced induction death within 18 days after starting therapy due to acute respiratory failure with bilateral pneumonia in one case and ischemic cardiomyopathy in the other. Both patients had de novo APL, Eastern Cooperative Oncology Group performance status ≤2 and no signs of hemorrhage or thrombosis at initial presentation. They were 64 and 75 years old. Cytogenetics in both cases revealed t(15;17) as the sole abnormality; no concurrent FLT3-ITD was present. None of the patients was refractory to treatment. In univariable analysis age above 70 years had no impact on response to induction therapy (P=0.24).

Survival analysis The median follow-up for survival was 2 years (95% CI: 1.61-2.30 years). Overall, the estimated 1-year and 2-year overall survival rates were 97% (95% CI: 94-100%) and 95% (95% CI: 91-99%), respectively (Figure 1). The overall survival rate was significantly lower in elderly patients (>70 years) than in younger patients (P<0.001) (Figure 2). None of the patients relapsed after ATO/ATRA treatment, thus overall survival and event-free survival were identical. Six (4%) patients died in remission at a median of 0.95 years after diagnosis (range, 0.18-2.38 years). Of these patients, three were below 70 years old. Causes of death were sepsis (n=2), bleeding/hemorrhage (n=1), pulmonary edema due to tachycardic atrial fibrillation (n=1), progress of an underlying intra-abdominal liposarcoma (n=1) as well as hepato-renal syndrome (n=1). The cumulative incidence of death in complete remission was significantly higher in older patients than in younger patients (P=0.001) (Figure 3). So far, none of the patients has developed a secondary neoplasm after treatment with ATO/ATRA.

Toxicity A total of 118 serious adverse events were reported in 60 patients, mainly infections (n=31), hepatobiliary events (n=21), neurotoxicity/neuropathy (n=9, including headache in 2 patients), and cardiac events (n=6, including QTc prolongation in 2 patients). Serious adverse events classified as haematologica | 2021; 106(12)

Outcome of APL patients after ATO-based treatment

related to differentiation syndrome were reported in seven patients. As prophylaxis for the differentiation syndrome prednisone at a dose of 0.5 mg/kg of body weight per day was administered from day 1 until the end of induction therapy.2 Toxicities are listed in Table 2. Due to the adverse event, ATRA was withdrawn in five patients, and restarted in four, requiring a dose reduction to 50% in one of the patients. ATO was withdrawn in two patients, and restarted in one; additionally, three patients had dose reductions. Most of the patients recovered from the serious adverse events (n=52), although eight patients died. Of the eight fatal serious adverse events, two occurred during induction within 18 days (acute respiratory failure with bilateral pneumonia, n=1; and ischemic cardiomyopathy, n=1) and six occurred in complete remission (infection/sepsis, n=2, bleeding/hemorrhage, n=1; pulmonary edema due to tachycardic atrial fibrillation n=1; progress of an underlying intraabdominal liposarcoma n=1, and hepato-renal syndrome n=1).

Discussion This is the largest real-life data analysis of first-line ATO/ATRA treatment in a prospective registry in low/intermediate-risk APL patients. It confirms a very high response rate as well as excellent overall survival with this new standard of care.2,3 Our data confirm that results can be replicated in a clinical registry, even within the context of non-selected patients and less closely monitored therapy than in the pivotal study. Indeed, on the basis of our registry data, the outcome of patients was identical to that reported after treatment with ATO/ATRA within the APL0406 trial.2,3 In contrast, the outcome of APL patients from previously published registries or population-based studies was inferior to that reported in clinical trials, mainly because of higher early death rates.5-8 However, these registries and population-based studies included results before treatment with ATO/ATRA had been approved by regulatory authorities.5-8 In our cohort the rate of induction death was extremely low, which might be in part attributable to improved supportive care and awareness of APL as a medical emergency,18 but also to early treatment with ATO/ATRA itself. Thus, effective strategies due to standardized guidelines along with consultative support and sharing of expertise seem to be well-taken and overcome induction mortality.19 Only two patients died during induction, in one case due to ischemic cardiomyopathy and in the other acute respiratory failure as a result of bilateral pneumonia. More impressively, none of the patients experienced induction death due to bleeding/hemorrhage or infections, which is in sharp contrast to previously published data on AIDA-based treatment.20,21 Nevertheless, we cannot rule out a potential selection bias since patients may also die before diagnosis. Highrisk APL (WBC count >10x109/L) at diagnosis is a known risk-factor for a bleeding diathesis.22 The European randomized intergroup study “APOLLO” is currently investigating idarubicin 12 mg/m2 on days 1 and 3 in addition to oral ATRA 45 mg/m2 twice daily on days 1-28 and ATO 0.15 mg/kg/day intravenously on days 5-28 followed by four cycles of ATO/ATRA consolidation therapy as compared to the standard chemotherapy/ATRA approach (ClinicalTrails.gov identifier: NCT02688140). The induction death rate in our analysis even compares haematologica | 2021; 106(12)

favorably with the recently published data from the Swedish Acute Leukemia Registry, which recorded an early death rate of 15% in newly diagnosed, low-/intermediaterisk APL (n=65) for the observation period 2009-2013.8 Thus, early treatment with ATO/ATRA seems to tackle the threat of induction death, again confirming recently pubTable 1. Characteristics of the 154 patients with acute promyelocytic leukemia at diagnosis. Gender: female Type of APL De novo Therapy-related ECOG performance status 0 1 2 3 4 Missing Bleeding/hemorrhage Overall Skin CNS bleeding Pulmonary Missing Thrombosis Missing BCR1/2 BCR3 Missing Cytogenetics Normal Sole abnormality Additional abnormalities - Complex* - Trisomy 8 Missing FLT3-ITD positive Missing FAB subtype M3 M3v Missing Median ITD allelic ratio Median age, years N. >70 years Median WBC, x 109/L Median platelets x 109/L Median hemoglobin, g/dL Missing Median BM blasts,** % Missing Median creatinine, mg/dL Missing

N=154

140 14

91 9

42 84 9 3 3 13

27 55 6 2 2 8

59 32 3 1 1 5 5 54 50 50

38 21 2 1 1 3 3 35 32 32

7 82 43 16/43 7/43 22 8/51 103

5 53 28 37 16 14 16 67

138 7 9

90 5 6

Value

Range

0.113 53 20 1.2 41 10 1 65 23 0.88 8

0.01-63 18-90 13% 0.2-10 2-210 3.5-14.8 0-95 0.49-8.14

APL; acute promyelocytic leukemia; ECOG: Eastern Cooperative Oncology Group; CNS: central nervous system; BCR, breakpoint cluster region; ITD: internal tandem duplication; FAB: French American British; WBC: white blood cell counts; BM: bone marrow. Percentages may not add to 100 because of rounding. *Two or more cytogenetic abnormalities in addition to t(15;17); **Blast cells included malignant promyelocytes. Percentages may not sum up to 100 due to rounding.

3103

S. Kayser et al.

lished results on ATO/ATRA from the randomized ItalianGerman APL0406 trial.2,3 However, part of the difference may be explained by the fact that the Swedish report contained data from a population-based registry, whereas for our data, selective reporting cannot be ruled out. The underlying patho-mechanism by which ATO in combination with ATRA exerts this effect remains elusive. The transcription factor PML-RARA behaves as an altered retinoic acid receptor with the ability to transmit oncogenic signals leading to accumulation of undifferentiated promyelocytes.23 ATRA induces disease remission in APL patients by triggering terminal differentiation of leukemic promyelocytes. In addition, ATO has been shown to contribute to degradation of the PML-RARA oncoprotein through binding the PML moiety.23,24 Regarding post-remission outcome, results of the randomized North American Leukemia Intergroup Study C9710 on 481 APL patients evaluating ATO in first-line therapy during consolidation demonstrated that ATO further reduced the risk of relapse and improved survival as compared to consolidation with daunorubicin/cytarabine.25

In addition, the randomized phase-III AML17 trial of the UK National Cancer Research Institute Acute Myeloid Leukaemia Working Group showed significantly better event-free survival and relapse-free survival after ATO/ATRA than after the AIDA-based regimen.4 In line with previously published results2,3 overall survival was excellent in our cohort. In addition, the death rate in remission was also very low at 4%, again arguing for the safety and efficacy of ATO/ATRA as consolidation treatment. In comparison, relatively high rates of deaths in remission, mainly due to infectious complications, were reported with AIDA treatment.20,21 Data on measurable residual disease was available for only a subgroup of patients before or after first consolidation (25%). Nevertheless, since no relapse was observed, our data as well as the currently up-dated APL recommendations of the European LeukemiaNet suggest that postconsolidation measurable residual disease monitoring can be avoided in this setting and performed only in high-risk patients (WBC count >10×109/L) in routine clinical practice.18

Figure 1. Kaplan Meier plot of overall survival. Green and red curves indicate upper and lower 95% confidence intervals, respectively.

Figure 2. Kaplan Meier plot of overall survival according to age. Red curve indicates age >70 years, black curve indicates age ≤70 years.

Table 2. Reported serious adverse events.

Serious adverse events Infection/sepsis Hepatobiliary Neurotoxicity/neuropathy Differentiation syndrome Cardiac, total QTc prolongation Pericarditis Hypertension Cardiac decompensation Ischemic cardiomyopathy Hematologic toxicity Hemorrhage/bleeding Thromboembolic complications Other Figure 3. Cumulative incidence of death according to age.

3104

N=118

31 21 9 7 6 2 1 1 1 1 5 4 4 31

26 18 8 6 5

4 3 3 26

Percentages may not sum up to 100 due to rounding.

haematologica | 2021; 106(12)

Outcome of APL patients after ATO-based treatment

Another positive finding is that none of the patients developed a secondary malignancy after ATO/ATRA, compared to 2% after chemotherapy,26 representing a major improvement in APL treatment, although we acknowledge that the median follow-up time of our cohort is relatively short. However, the reported latency period between the diagnosis of APL and the development of a secondary malignancy of 6.6 months (range, 3.8-7.6 months) as reported by Pagano et al. on behalf of the Gruppo Italiano Malattie Ematologiche dell'Adulto argues strongly for a true lower incidence after ATO/ATRA.27 The low relapse rate after ATO-based therapy in our cohort is in line with other reports.2,3,28 Confirming our previously published data,29 ATO/ATRA was also effective in elderly patients, although age above 70 years was associated with a shorter overall survival and higher cumulative incidence of death in complete remission in comparison with those outcomes in younger patients. Nonetheless, ATO/ATRA should not be withheld in older patients, particularly in light of its high efficacy and safety profile. The toxicity profiles were comparable to those previously published.2,3 The serious adverse events were mainly infections and elevation of liver enzymes. Hepatic toxicity has frequently been reported in studies of ATO, particularly in terms of increased liver enzymes.2,3,30 Although frequent, this complication is usually reversible and successfully managed with temporary discontinuation of ATO.31 Neurological toxicity, mainly consisting of peripheral neuropathy, has also been reported.2,3,30 This side effect is usually managed with temporary drug discontinuation.31 QTc prolongation is another common side effect of ATO. It can lead to torsade-de-pointes-type ventricular arrhythmia, which is potentially fatal.2 However, the reported rate of QTc prolongation was lower than previously published,2-4 as was the rate of differentiation syndrome,2-4 although the data might be biased due to incomplete reporting. In our series, most of the patients recovered from their serious adverse events, although the outcome was fatal in eight cases. An unresolved issue is whether or not central nervous system prophylaxis is needed in APL. Pharmacodynamic studies of plasma levels indicated that about one third of ATO crosses the blood–brain barrier, which suggests that a sufficient amount of the drug is available on site to prevent disease recurrence.32 In addition, no ATO accumulation or delayed toxicity was observed in patients followed up for over 10 years.34,35 However, none of our patients was reported to have central nervous system involvement at diagnosis.

References 1. Swerdlow SH, Campo E, Harris NL, et al. WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. Revised 4th edition. WHO Press, Geneva, Switzerland, 2017. 2. Lo-Coco F, Avvisati G, Vignetti M, et al. Retinoic acid and arsenic trioxide for acute promyelocytic leukemia. N Engl J Med. 2013;369(2):111-121. 3. 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 results of the

haematologica | 2021; 106(12)

Regarding biological characteristics, additional cytogenetic abnormalities were present in 28% of the patients, most frequently t(15;17) within a complex karyotype or trisomy 8, which is in line with published data.35,36 Although data were limited to a small number of patients, FLT3-ITD mutations were less frequent in our cohort than in published data (31%).37 To date, there are still conflicting data regarding the impact of additional chromosomal or genetic abnormalities on outcome in APL patients.38-46 In our large cohort, neither the presence of additional cytogenetic abnormalities nor FLT3-ITD had an impact on overall survival, suggesting that ATO/ATRA may abrogate the negative prognostic impact of FLT3ITD.46 The latter issue, however, should be interpreted with caution given the limited availability of data on FLT3 mutational status in our cohort. In conclusion, our real-life data showed excellent and sustained response rates after ATO/ATRA for first-line treatment of APL, confirming published results from the AP0406 trial.2,3 Compared to reported historical data, the induction death rate was extremely low despite a high bleeding/hemorrhage rate at diagnosis, probably attributable to improved supportive care and awareness of APL as a medical emergency, but also to early treatment with ATO/ATRA itself. Our results confirm the efficiency and sustainability of ATO/ATRA in the primary management of adults with low-/intermediate-risk APL in the real-life setting, irrespective of their age. Disclosures UP has received research support from TEVA. CT is chief executive officer and co-owner of AgenDix GmbH. All other authors declare that they have no conflicts of interest. Contributions SK and RFS were responsible for the concept of this paper, contributed to the literature search and data collection, analyzed and interpreted data, and wrote the manuscript. UP and LA designed the registry, contributed patients, interpreted data and critically revised the manuscript. DL, MC, KSG, FS, AB, KS-E, PP, YH, RR, ER, FC, SWK, WEA, SR, RN, CB, MG, EJ, PC, KK, CO, TB, CS, AA, VK, MW, MdW, DN, CB, CS, JA, AA, SH, AK, AR, MG, CW, KS, WH, CM-T, CR, HS, MB, CDB, EL, and PF contributed patients and critically revised the manuscript. CT performed research and critically revised the manuscript. All authors reviewed and approved the final manuscript. Acknowledgments We acknowledge publication support from Leipzig University.

randomized Italian-German APL0406 trial. J Clin Oncol. 2016;35(6):605-612. 4. 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. 5. Paulson K, Serebrin A, Lambert P, et al. Acute promyelocytic leukaemia is characterized by stable incidence and improved survival that is restricted to patients managed in leukaemia referral centres: a pan-Canadian epidemiological study. Br J Haematol. 2014;166(5):660-666. 6. 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. 7. 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):24222424. 8. 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. 9. Bennett JM, Catovsky D, Daniel MT, et al. Proposed revised criteria for the classification of acute myeloid leukemia. A report of the French-American-British Cooperative Group. Ann Intern Med. 1985;103(4):620-625.

3105

S. Kayser et al. 10. Cheson BD, Bennett JM, Kopecky KJ, et al. Revised recommendations of the International Working Group for Diagnosis, Standardization of Response Criteria, Treatment Outcomes, and Reporting Standards for Therapeutic Trials in Acute Myeloid Leukemia. J Clin Oncol. 2003;21 (24):4642-4649. 11. Mitelman F: ISCN: An International System for Human Cytogenetic Nomenclature. Basel, Switzerland: S. Karger; 1995. 12. Yokota S, Kiyoi H, Nakao M, et al. Internal tandem duplication of the FLT3 gene is preferentially seen in acute myeloid leukemia and myelodysplastic syndrome among various hematological malignancies. A study on a large series of patients and cell lines. Leukemia. 1997;11(10):1605-1609. 13. Thiede C, Steudel C, Mohr B, et al. Analysis of FLT3-activating mutations in 979 patients with acute myelogenous leukemia: association with FAB subtypes and identification of subgroups with poor prognosis. Blood. 2002;99(12):4326-4335. 14. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343-346. 15. Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481. 16. Gray RJ. A class of k-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat. 1988;16(3):1141-1154. 17. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria, 2014. 18. Sanz MA, Fenaux P, Tallman MS, et al. Management of acute promyelocytic leukemia: updated recommendations from an expert panel of the European LeukemiaNet. Blood. 2019;133(15):16301643. 19. Jillella AP, Kota VK. The global problem of early deaths in acute promyelocytic leukemia: a strategy to decrease induction mortality in the most curable leukemia. Blood Rev. 2018;32(2):89-95. 20. Lo-Coco F, Avvisati G, Vignetti M, et al. Front-line treatment of acute promyelocytic leukemia with AIDA induction followed by risk-adapted consolidation for adults younger than 61 years: results of the AIDA2000 trial of the GIMEMA group. Blood. 2010;116(17):3171-3179. 21. de la Serna J, Montesinos P, Vellenga E, et al. Causes and prognostic factors of remission induction failure in patients with acute promyelocytic leukemia treated with alltrans retinoic acid and idarubicin. Blood. 2008;11(7):3395-4302. 22. Testa U, Lo-Coco F. Prognostic factors in acute promyelocytic leukemia: strategies to define high-risk patients. Ann Hematol. 2016;95(5):673-680.

3106

23. Davison K, Mann KK, Miller WH Jr, Davison K. Arsenic trioxide: mechanisms of action. Semin Hematol. 2002;39(2 Suppl 1):3-7. 24. Lo-Coco F, Hasan SK. Understanding the molecular pathogenesis of acute promyelocytic leukemia. Best Pract Res Clin Haematol. 2014;27(1):3-9. 25. Powell BL, Moser B, Stock W, et al. Arsenic trioxide improves event-free and overall survival for adults with acute promyelocytic leukemia: North American Leukemia Intergroup Study C9710. Blood. 2010;116 (19):3751-3757. 26. Montesinos P, González JD, González J, et al. Therapy-related myeloid neoplasms in patients with acute promyelocytic leukemia treated with all-trans-retinoic acid and anthracycline-based chemotherapy. J Clin Oncol. 2010;28(24):3872-3879. 27. Pagano L, Pulsoni A. Second malignancy after treatment of acute promyelocytic leukemia: experience of GIMEMA trials. Blood. 2002;100(4):1514-1515. 28. Zhang Y, Zhang Z, Li J, et al. Long-term efficacy and safety of arsenic trioxide for firstline treatment of elderly patients with newly diagnosed acute promyelocytic leukemia. Cancer. 2013;119(1):115-125. 29. Kayser S, Rahmé R, Martínez-Cuadrón D, et al. Outcome of older (70 years) APL patients frontline treated with or without arsenic trioxide-an International Collaborative Study. Leukemia. 2020;34(9):2333-2341. 30. Ghavamzadeh A, Alimoghaddam K, Rostami S, et al. Phase II study of singleagent arsenic trioxide for the front-line therapy of acute promyelocytic leukemia. J Clin Oncol. 2011;29(20):2753-2757. 31. Kayser S, Schlenk RF, Platzbecker U. Management of patients with acute promyelocytic leukemia. Leukemia. 2018;32 (6):1277-1294. 32. Kiguchi T, Yoshino Y, Yuan B, et al. Speciation of arsenic trioxide penetrates into cerebrospinal fluid in patients with acute promyelocytic leukemia. Leuk Res. 2010;34 (3):403-405. 33. Hu J, Liu YF, Wu CF, et al. Long-term efficacy and safety of all-trans retinoic acid/arsenic trioxide-based therapy in newly diagnosed acute promyelocytic leukemia. Proc Natl Acad Sci U S A. 2009;106(9):3342-3347. 34. Zhu H, Hu J, Chen L, et al. The 12-year follow-up of survival, chronic adverse effects, and retention of arsenic in patients with acute promyelocytic leukemia. Blood. 2016;128(11):1525-1528. 35. Schlenk RF, Germing U, Hartmann F, et al. High-dose cytarabine and mitoxantrone in consolidation therapy for acute promyelocytic leukemia. Leukemia. 2005;19(6):978983. 36. Lou Y, Suo S, Tong H, et al. Characteristics and prognosis analysis of additional chromosome abnormalities in newly diagnosed

acute promyelocytic leukemia treated with arsenic trioxide as the front-line therapy. Leuk Res. 2013;37(11):1451-1456. 37. Noguera NI, Catalano G, Banella C, et al. Acute promyelocytic leukemia: update on the mechanisms of leukemogenesis, resistance and on innovative treatment strategies. Cancers. 2019;18;11(10):1591. 38. Kiyoi H, Naoe T, Yokota S, et al. Internal tandem duplication of FLT3 associated with leukocytosis in acute promyelocytic leukemia. Leukemia Study Group of the Ministry of Health and Welfare (Kohseisho). Leukemia. 1997;11(9):1447-1452. 39. Noguera NI, Breccia M, Divona M, et al. Alterations of the FLT3 gene in acute promyelocytic leukemia: association with diagnostic characteristics and analysis of clinical outcome in patients treated with the Italian AIDA protocol. Leukemia. 2002;16 (11):2185-2189. 40. Shih LY, Kuo MC, Liang DC, et al. Internal tandem duplication and Asp835 mutations of the FMS-like tyrosine kinase 3 (FLT3) gene in acute promyelocytic leukemia. Cancer. 2003;98(6):1206-1216. 41. Lucena-Araujo AR, Kim HT, Jacomo RH, et al. Internal tandem duplication of the FLT3 gene confers poor overall survival in patients with acute promyelocytic leukemia treated with all-trans retinoic acid and anthracycline-based chemotherapy: an International Consortium on Acute Promyelocytic Leukemia study. Ann Hematol. 2014;93(12): 2001-2010. 42. De Botton S, Chevret S, Sanz M, et al. Additional chromosomal abnormalities in patients with acute promyelocytic leukaemia (APL) do not confer poor prognosis: results of APL 93 trial. Br J Haematol. 2000;111(3):801-806. 43. Cervera J, Montesinos P, Hernández-Rivas JM, et al. Additional chromosome abnormalities in patients with acute promyelocytic leukemia treated with all-trans retinoic acid and chemotherapy. Haematologica. 2010;95(3):424-431. 44. Pantic M, Novak A, Marisavljevic D, et al. Additional chromosome aberrations in acute promyelocytic leukemia: characteristics and prognostic influence. Med Oncol. 2000;17(4):307-313. 45. Poiré X, Moser BK, Gallagher RE, et al. Arsenic trioxide in front-line therapy of acute promyelocytic leukemia (C9710): prognostic significance of FLT3 mutations and complex karyotype. Leuk Lymphoma. 2014;55(7):1523-1532. 46. Cicconi L, Divona M, Ciardi C, et al. PMLRARα kinetics and impact of FLT3-ITD mutations in newly diagnosed acute promyelocytic leukaemia treated with ATRA and ATO or ATRA and chemotherapy. Leukemia. 2016;30(10):1987-1992.

haematologica | 2021; 106(12)

ARTICLE

Bone Marrow Transplantation

Use of the HLA-B leader to optimize cord blood transplantation

Ferrata Storti Foundation

Effie W. Petersdorf,1,2 Ted Gooley,1 Fernanda Volt,3 Chantal Kenzey,3 Alejandro Madrigal,4 Caroline McKallor,1 Sergio Querol,5 Hanadi Rafii,3 Vanderson Rocha,3,6 Ryad Tamouza,3,7 Christian Chabannon,8,9 Annalisa Ruggeri3,9,10 and Eliane Gluckman3,11 1

Division of Clinical Research, Fred Hutchinson Cancer Research Center, Seattle, WA, USA; 2Department of Medicine, University of Washington, Seattle, WA, USA; 3Eurocord, Hôpital Saint Louis APHP, Institut de Recherche de Saint-Louis (IRSL) EA3518, Université de Paris, Paris, France; 4University College London Cancer Institute, Royal Free Campus, London, UK; 5Cell Therapy Services, Catalan Blood and Tissue Bank, Barcelona, Spain, 6Hospital das Clínicas and LIM31, Faculty of Medicine University of São Paulo, Brazil; 7INSERM U955, CHU Henri Mondor, Créteil, France, 8Institut PaoliCalmettes, INSERM CBT1409, Marseille, France; 9Cellular Therapy and Immunobiology Working Party of the European Society for Blood and Marrow Transplantation, Leiden, the Netherlands; 10Hematology and Bone Marrow Transplant Unit, IRCCS San Raffaele Scientific Institute, Milan, Italy and 11Monacord, International Observatory on Sickle Cell Disease, Centre Scientifique de Monaco, Monaco

Haematologica 2021 Volume 106(12):3107-3114

ABSTRACT

ord blood transplantation (CBT) can cure life-threatening blood disorders. The HLA-B leader affects the success of unrelated donor transplantation but its role in CBT is unknown. We tested the hypothesis that the HLA-B leader influences CBT outcomes in unrelated single-unit cord blood transplants performed by Eurocord/European Blood and Marrow Transplant (EBMT) centers between 1990 and 2018 with data reported to Eurocord. Among 4,822 transplants, 2,178 had one HLA-B mismatch of which 1,013 were HLAA and HLA-A and -DRB1 matched. The leader (methionine [M] or threonine [T]) was determined for each HLA-B allele in patients and units to define the genotype. Among single HLA-B-mismatched transplants, the patient/unit mismatched alleles were defined as leader-matched if they encoded the same leader, or leader-mismatched if they encoded different leaders; the leader encoded by the matched (shared) allele was determined. The risks of graft-versus-host disease, relapse, non-relapse mortality and overall mortality were estimated for various leader-defined groups using multi-variable regression models. Among the 1,013 HLA-A and -DRB1-matched transplants with one HLA-B mismatch, increasing numbers of cord blood unit M-leader alleles was associated with increased risk of relapse (hazard ratio [HR] for each increase in one Mleader allele 1.30, 95% Confidence Interval [CI]: 1.05-1.60, P=0.02). Furthermore, leader mismatching together with an M-leader of the shared HLA-B allele lowered non-relapse mortality (HR 0.44, 95% CI: 0.23-0.81; P=0.009) relative to leader matching and a shared T-leader allele. The HLA-B leader may inform relapse and non-relapse mortality risk after CBT. Future patients might benefit from the appropriate selection of units that consider the leader.

Introduction Cord blood transplantation (CBT) offers curative therapy for life-threatening blood disorders. Its distinct advantages are 2-fold. Similar to haploidentical donor transplantation, the less stringent level of histocompatibility required between the patient and the cord blood unit is an asset for patients of diverse ancestry who lack HLA-matched peripheral blood stem cell and marrow donors.1-11 Compared to other sources of allogeneic stem cells, a unique benefit of cord blood transplantation is the avoidance of risks to the donor. The lower risk of severe acute and

haematologica | 2021; 106(12)

Correspondence: EFFIE W. PETERSDORF epetersd@fredhutch.org Received: June 23, 2020. Accepted: October 15, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.264424

3107

E. W. Petersdorf et al.

chronic graft-versus-host disease (GvHD) in cord blood transplantation compared to other allograft sources is related to the limited numbers of mature donor T cells in naïve cord blood grafts.6 Despite lower rates of GvHD, disease recurrence after cord blood transplantation is not Table 1. Demographics of the study population. The effect of the HLAB leader in clinical outcome was assessed among the base population of 4,822 single cord blood unit transplants, the subset of 2,178 transplants with one HLA-B mismatch, and the subset of 1,013 HLA-A and -DRB1-matched transplants with one HLA-B mismatch.

Characteristics

Single cord blood unit transplants Base population One HLA-A and n = 4,822 HLA-B DRB1mismatch matched n = 2,178 with one HLA-B mismatch n = 1,013

Patient age, average yrs 17.1 (range) (0.1-70.7) Patient sex, n (%) Female 2,152 (45%) Male 2,649 (55%) Missing or N/A 21 (0%) Disease type, n (%) Malignant 3,577 (74%) Non-malignant 1,245 (26%) Disease status at transplantation *, n (%) low 1,189 (25%) intermediate 1,308 (27%) high 735 (15%) Missing or N/A 1,590 (33%) Patient CMV serology, n (%) negative 1,900 (39%) positive 2,599 (54%) Missing or N/A 323 (6%) HLA Matching, n (%) 6/6 934 (19%) 5/6 2,233 (46%) 4/6 1,536 (32%) Other 119 (2%) Year of transplantation, n (%) ≤2005 1269 (26%) >2005 3,553 (74%) Transplant type, n (%) MAC 3,558 (74%) RIC 1,088 (23%) Missing or N/A 176 (4%) Total body irradiation, n (%) no 2,988 (62%) yes 1,531 (32%) Missing or N/A 303 (6%) Use of ATG, n (%) no 1,454 (30%) yes 3,368 (70%) Total nucleated cells infused, 5.51 median (range) (0.20-94.9)

21.3 (0.1-70.7)

14.5 (0.1-69.0)

986 (45%) 461 (46%) 1,189 (55%) 551 (54%) 3 (0%) 1 (0%) 1,769 (81%) 766 (76%) 409 (19%) 247 (24%) 606 600 404 568

(28%) (28%) (19%) (26%)

258 280 165 310

(25%) (28%) (16%) (31%)

784 (36%) 381 (38%) 1,258 (58%) 567 (56%) 136 (6%) 65 (6%) 0 (0%) 0 (0%) 1,013 (47%) 1,013 (100%) 1,110 (51%) 0 (0%) 55 (3%) 0 (0%) 595 (27%) 302 (30%) 1,583 (73%) 711 (70%) 1,578 (72%) 760 (75%) 537 (25%) 222 (22%) 63 (3%) 31 (3%) 1,230 (56%) 599 (59%) 812 (37%) 378 (37%) 136 (6%) 36 (4%) 645 (30%) 288 (28%) 1,533 (70%) 725 (72%) 4.91 6.24 (0.37-81.4) (0.38-81.4)

*Disease status at transplantation: low risk (non-Hodgkin lymphoma [NHL] or Hodgkin disease [HD] in complete remission [CR] 1; acute leukemia in CR1; chronic myeloid leukemia [CML] in chronic phase; myelodysplastic syndrome [MDS]-refractory anemia [RA] with or without ring sideroblasts [RS]; refractory anemia with excess blasts [RAEB]1); intermediate risk (NHL or HD in ≥CR2; acute leukemia in ≥CR2; CML in accelerated phase; MDS-RAEB-2; RCMD; RCMD-RS; MDS associated with isolated del [5q], or unclassifiable MDS); high risk (NHL or HD untreated, refractory, partial remission or relapse; acute leukemia untreated, refractory or relapse; CML in blastic crisis; MDS transformed in acute leukemia; plasma cell disorders; chronic lymphocytic leukemia/lymphoma; secondary leukemia). N/A: not available.

3108

necessarily higher than that observed with other allograft sources.12 The precise immunological pathways that lead to these clinical outcomes remain to be elucidated; however, the innate immune system is an attractive mechanism because natural killer (NK) cells are the earliest population of cells to reconstitute after cord blood transplantation.13-16 Beneficial NK-associated responses have been demonstrated in transplantation and the extent to which they can be leveraged to lower posttransplant disease recurrence remains an important area of clinical investigation.8-10,17,18 A basic premise of transplantation is HLA compatibility between the graft and the patient. Recent evidence in unrelated donor transplantation implicates the HLA-B leader in graft-versus-host responses and demonstrates the importance of HLA that goes beyond the classic model based on matching under the framework of antigen presentation.19,20 As a peptide presented by classical class I molecules, the leader provides an antigenic stimulus to T cells. As a peptide that facilitates the cell-surface expression of HLA-E, it also influences NK allorecognition. The HLA-B leader is distinguished from the leaders of other classical class I genes due to a dimorphism that leads to peptides with methionine (M) or threonine (T) at the second position of the leader peptide.21 Leader-associated differential expression of HLA-E influences inhibitory NKG2A- and to a lesser extent activating NKG2C-mediated responses.22 In HIV-AIDs, the dimorphic HLA-B leader influences progression of the disease depending on the strength of HLA-E/NKG2 responses.23,24 Cord blood transplants are a unique population of pairs who may be mismatched concurrently at multiple HLA genetic loci and differ from unrelated donor transplants in whom the total number of HLA disparities must be limited to prevent life-threatening complications.19 The contribution of an HLA mismatch to clinical outcomes after cord blood and unrelated donor transplantation may depend on the specific gene(s) that is mismatched and on additive effects of multi-locus mismatching across the HLA region.19,25-27 Furthermore, the effect of the HLA-B leader on specific outcomes may depend on the mismatched HLA locus.19,20 The current study was designed to address an unmet need in cord blood transplantation for a better understanding of the clinical significance of the HLA barrier. We tested the hypothesis that the HLA-B leader informs outcome in a large, well-characterized cohort of single-unit cord blood transplants, beginning with an evaluation of all transplants and subsequently of transplants defined by matching at HLA-A, -B and -DRB1, the classic determinants of cord blood transplant outcomes. This hierarchy permitted the evaluation of the effect of the leader while minimizing contributions from other mismatched loci. The results may have important implications for future patients and advance understanding of the immunobiology of relapse.

Methods Study population and design The study population included 4,822 patients who received a single-unit unrelated cord blood transplant in an Eurocord/European Blood and Marrow Transplant (EBMT) center for the treatment of a blood disorder between 1990 and 2018 (Table 1; Online Supplementary Table S1). Data were collected by

haematologica | 2021; 106(12)

HLA-B leader and cord-blood transplantation

Figure 1. The clinical significance of the HLA-B leader was evaluated in cord blood transplantation according to HLA matching, leader genotype, and leader-matching. The study population consisted of 4,822 single-unit unrelated cord blood transplantations. The clinical significance of the HLA-B leader in outcome was analyzed in the entire population, among transplants with one HLA-B mismatch (regardless of HLA mismatching elsewhere), and in HLA-A and DRB1-matched transplants with one HLA-B mismatch. The impact of the HLA-B leader was analyzed according to the cord blood unit’s leader genotype (left pie), the leader match status (middle pie), and the leader match status together with the leader of the shared antigen (right pie).

Eurocord through the EBMT/Promise database following national and Joint Accreditation Committee International Society for Cellular Therapy - Europe & EBMT (JACIE) guidelines, and shared according to agreements between Eurocord and the Fred Hutchinson Cancer Research Center. We previously observed a strong effect of the HLA-B leader in single-locus HLA-B-mismatched unrelated transplants who were matched at HLA-A, -C, -DRB1 and -DQB1, an effect that neither depends on HLA-C KIR ligands nor on the level of typing for HLA-A, -C, -DRB1 or -DQB1 allele and antigen mismatches.19 Since cord blood units are selected on the basis of HLA-A, B, -DRB1 and most transplants in the current study lack highresolution typing including HLA-C, models did not adjust for these factors. We tested the hypothesis that the HLA-B leader dimorphism informs clinical outcome by examining three groups defined by their match status at HLA-A, -B and -DRB1: i) the entire study population of 4,822 patients (Online Supplementary Table S2); ii) a subset of 2,178 transplants with one allele or antigen HLA-B mismatch regardless of mismatching elsewhere to examine whether the leader provides information on well-tolerated HLA-B mismatches (Online Supplementary Table S3), and iii) a subset of 1,013 HLA-A and -DRB1-matched transplants with one allele or antigen HLA-B mismatch which permitted the potential effects of the leader to be examined without disparity from HLA-A and -DRB1 (Online Supplementary Table S4; Figure 1).

Study oversight Protocols were approved by the Institutional Review Boards of Eurocord and the National Institutes of Health Office for Human Research Protections. The funding agencies had no role

haematologica | 2021; 106(12)

in the study design, data collection and analysis, the decision to submit the manuscript for publication, or the preparation of the manuscript.

HLA typing Classical HLA alleles were genotyped in the study cohort according to standard criteria and individual transplant center guidelines.25-27 The rs1050458 single nucleotide polymorphism in exon one is defined for World Health Organization-recognized HLA-B alleles and antigens.28 The one-to-one relationship between the leader dimorphism and HLA-B allele and antigen and has been previously confirmed, permitting the assignment of methionine (M) (rs1050458T) or threonine (T) (rs1050458C) leader to each patient and cord blood unit based on typing performed at the time of transplantation (Figure 2).19 The leader genotype (TT, MT or MM) was determined for each patient and each cord blood unit. When the patient and cord blood unit are HLA-B-matched, they have, by definition, the same leader genotype. Transplants mismatched for one HLA-B allele were defined as leader-matched when the patient’s and unit’s mismatched alleles were both M or both T leader. Single HLA-B mismatches were defined as leadermismatched when the patient’s and unit’s mismatched alleles had different leaders (M leader in the patient’s mismatched allele with T leader in the unit’s mismatched allele, or vice versa). Among single HLA-B-mismatched transplants, the leader of the matched (shared) allele between the patient and unit was determined (M or T). For single HLA-B-mismatched transplants, leader match status was combined with the leader of the shared matched allele to define four mutually exclusive groups: leader-matched/share T; leader-matched/share M; leader-mismatched/share T, and leadermismatched/share M (Figure 2).

3109

E. W. Petersdorf et al.

Table 2 Leader match status and clinical Outcome in 1,013 single HLA-B-mismatched cord blood transplantations (matched at HLA-A and -DRB1). A total of 814 patients are evaluable for relapse and non-relapse mortality.

Clinical Endpoint Grades II-IV acute GvHD

Relapse

Mortality

Non-relapse mortality

Group*

Odds Ratio or Hazard Ratio

95% Confidence Interval

Leader-matched (220/604=36%) Leader mismatched (138/384=36%) Leader-matched (132/498=27%) Leader-mismatched (92/316=29%) Leader-matched (332/620=54%) Leader-mismatched (182/393=46%) Leader-matched (168/498=34%) Leader-mismatched (79/316=25%)

0.98

0.75-1.29

0.89

1.02

0.77-1.35

0.87

0.85

0.71-1.02

0.08

-a

0.72

0.55-0.94

0.02

*The numbers refer to the number of patients who developed the clinical endpoint out of the total number of evaluable patients (for whom data for the clinical endpoint was available). GvHD: graft-versus-host disease.

Statistical analysis We examined the association of the HLA-B leader genotype and leader matching with grades II-IV acute GvHD, relapse, death not preceded by relapse, disease-free survival and overall mortality. Logistic regression was used to assess associations with acute GvHD. For all other endpoints, Cox regression models were fit to compare the hazards or cause-specific hazards of failure, as appropriate, between HLA-B leader groups, and patients who did not fail by last contact or who failed from a competing risk were censored at last contact or at the time of the competing risk (as appropriate). Relapse was studied for patients with malignant disorders. Regression models were adjusted for patient age, year of transplantation, disease risk at transplantation (low, intermediate, high, non-malignant), total nucleated cell dose, use of anti-thymocyte globulin, intensity of the conditioning regimen (reduced-intensity versus ablative), use of total body irradiation, patient cytomegalovirus serologic status, and number of mismatches at HLA-A and -DRB1. Various interactions, as detailed in the Results, were examined by including appropriate factors in these regression models. Covariates with missing data were included in models by creating an additional category to reflect the missing value of the appropriate covariate. If outcome data were missing for a particular patient, such a patient was excluded from the appropriate regression analysis. Two-sided P-values from Cox regression models were obtained from the Wald test. Several comparisons were made, all focused on refinements of the concept of leader matching versus leader mismatching. No adjustments were made to the P-values associated with the fitted regression models, although the impact of multiple comparisons is reduced because the examined endpoints are correlated with one another. For individual models with more than two categories, global tests of significance were also conducted.

population, 81% were leader-matched and 19% were leader-mismatched. The distribution among transplants with one HLA-B mismatch was 1,027 (47%) leader-matched/share T, 311 (14%) leader-matched/share M, 640 (29%) leader-mismatched/share T, and 200 (9%) leader-mismatched/share M.

Leader genotype and clinical outcome We tested the hypothesis that the HLA-B leader genotype of the patient and/or the unit may influence the clinical outcome by examining the impact of the patient’s leader genotype separately from the cord blood unit’s leader genotype among the entire population of 4,822 transplants. Although the leader genotype of the patient did not exhibit obvious correlation with clinical outcome (Online Supplementary Tables S2, S3 and S4), the cord blood unit’s leader genotype was associated with the risk of relapse among the 1,013 HLA-A and -DRB1-matched transplants with one HLA-B allele mismatch (Online Supplementary Table S4). In this subgroup, transplantation from MT cord blood units had an estimated hazard ratio (HR) of 1.29 (95% Confidence Interval [CI]: 0.96-1.73) relative to TT cord blood units and MM had an estimated HR of 1.69 (95% CI: 1.03-2.75) for relapse relative to TT. The probability of observing these or more extreme differences under the null hypothesis that the risks of relapse are identical across the three leader genotype groups is 0.06. In addition, if one models the number of M leaders as a continuous linear variable ranging from 0 to 2 (addressing the possible existence of an increase in risk with increasing number of M leader alleles), the probability of a difference in relapse risk that is as extreme or more extreme than what was observed is 0.02 (HR 1.30 for each increase of one M leader allele, 95% CI: 1.05-1.60). Interestingly, transplantation of MT and MM units did not show a clear impact on GvHD risk but MM units may be associated with a lower risk of non-relapse mortality (HR 0.46, 95% CI: 0.23-0.95; Online Supplementary Table S4), although the global test of significance of the unit leader yielded P=0.11.

Results Leader matching and clinical outcome HLA-B leader genotype and leader matching The frequency of leader genotypes among patients (293 [6%] MM; ,1724 [36%] MT; 2,805 [58%] TT) were similar to the frequencies among the cord blood units (295 [6%] MM; 1828 [38%] MT; 2699 [56%] TT) and HLA-matched and HLA-mismatched unrelated donor transplants (Figure 1).19,20 Among the entire study

3110

The dimorphism of the HLA-B leader defines two lineages of HLA-B alleles.19 In HLA-B-mismatched transplantation, the patient’s and cord blood unit’s mismatched alleles may be from the same (leader-matched) or different (leader-mismatched) lineage. The impact of leader matching was examined in the 1,013 HLA-A and -DRB1-matched transplants with a single HLA-B mismatch.

haematologica | 2021; 106(12)

HLA-B leader and cord-blood transplantation

Table 3. Impact of leader match status and leader of the shared allotype on clinical outcome in 1,013 single HLA-B-mismatched cord blood transplantations (matched at HLA-A and -DRB1). A total of 814 patients are evaluable for relapse and non-relapse mortality.

Clinical Endpoint Grades II-IV acute GvHD

Relapse

Mortality

Non-relapse mortality

Leader Match Status/ Leader of Shared Antigen*

Odds Ratio or Hazard Ratio

95% Confidence Interval

P (Global P)

Leader-matched/T (169/464=36%) Leader-matched/M (51/140=36%) Leader-mismatched/T (100/299=33%) Leader-mismatched/M (38/85=45%) Leader-matched/T (98/386=25%) Leader-matched/M (34/112=30%) Leader-mismatched/T (70/248=28%) Leader-mismatched/M (22/68=32%) Leader-matched/T (248/479=52%) Leader-matched/M (84/141=60%) Leader-mismatched/T (146/306=48%) Leader-mismatched/M (36/87=41%) Leader-matched /T (127/386=33%) Leader-matched/M (41/112=37%) Leader-mismatched/T (68/248=27%) Leader-mismatched/M (11/68=16%)

(global P 0.37)

1.05

0.70-1.57

0.83

0.90

0.66-1.23

0.49

1.42

0.87-2.29

0.16

(global P 0.94)

1.14

0.76-1.74

0.54

1.03

0.75-1.43

0.84

1.08

0.66-1.77

0.77

(global P 0.10)

1.15

0.90-1.48

0.26

0.94

0.76-1.15

0.52

0.71

0.50-1.01

0.06

(global P 0.02)

1.18

0.82-1.69

0.37

0.84

0.62-1.13

0.25

0.44

0.23-0.81

0.009

Relative to leader-matched transplants, leader-mismatched transplants had lower non-relapse mortality (HR 0.72, 95% CI: 0.550.94; P=0.02) (Table 2); overall mortality was lower but not statistically significantly so. When the leader of the shared HLA-B allele is considered alongside the leader match status, there was a statistically significant difference in the risk of non-relapse mortality across the resultant four groups (Table 3, global P=0.02). This result was largely driven by transplantation from leader-mismatched units that share an M leader allele with the patient compared to transplantation from leader-matched units that share a T leader allele (Table 3; HR 0.44, 95% CI: 0.23-0.81; P=0.009). Not surprisingly, the comparison of overall mortality for these same groups showed a difference in the same direction as that for non-relapse mortality; however, the magnitude of the difference was smaller and was not statistically significant) (Table 3). Other pairwise comparisons of non-relapse mortality that may be of particular interest include leader-mismatched units that share a T leader allele with the patient compared to transplantation from leader-mismatched units that share an M leader allele (HR 1.92, 95% CI: 1.01-3.65; P=0.05) and leadermatched units that share an M leader allele with the patient compared to transplantation from leader-mismatched units that share an M leader allele (HR 2.70, 95% CI: 1.38-5.29; P=0.004). In summary, the HLA-B leader genotype of the cord blood unit may inform outcome after transplantation from HLA-B-mismatched units. Among patients transplanted from HLA-A and

haematologica | 2021; 106(12)

DRB1-matched units with one HLA-B mismatch, leader mismatching and presence of a matched M-leader allele may be associated with lower non-relapse mortality. Units with MT and MM leader genotypes may be associated with a higher risk of relapse; however, the higher relapse risk conferred by the genotype together with lower non-relapse mortality associated with leader mismatching appeared to have no demonstrable impact on overall survival.

Discussion An unmet need of CBT is a better understanding of the factors that influence outcome. The current study was designed to test a series of novel hypotheses regarding the significance of the dimorphic HLA-B leader in CBT. Interest in the HLA-B leader is founded on its pivotal role in both adaptive and innate immune pathways. A role for the HLA-B leader in HLA-B-mismatched unrelated donor transplantation in GvHD provides an approach for understanding the immunogenicity of M and T leader alleles in a way that may bridge T- with NK-cell biology. The effects of HLA mismatching are well-tolerated in CBT, permitting a higher overall degree of disparity between the patient and cord blood unit. The majority of CBT are mismatched at HLA-B. With over 7,000 known 3111

E. W. Petersdorf et al.

Figure 2. The dimorphic HLA-B leader is assigned to patients and cord blood units based on its one-to-one association with the HLA-B allele. (A) Each HLA-B allele is defined by its leader peptide. (B) Among patients and cord blood units with one HLA-B mismatch, the leader of the patient’s and unit’s mismatched alleles, and the leader of the matched shared allele is determined. Four groups are defined by the match status of the leader (leader-matched; leader-mismatched) and by the leader of the shared allele (methionine [M] or threonine [T]).

3112

haematologica | 2021; 106(12)

HLA-B leader and cord-blood transplantation

HLA-B alleles,28 the sheer number of polymorphisms presents a challenge in understanding the immunogenetic basis of transplant outcome. We surmised that if HLA-B alleles are described by two lineages, one family that encodes M-leaders and one that encodes T leaders, that this dichotomy may aid the identification of HLA-B mismatches that influence GvHD and relapse. In CBT, HLAC and high-resolution matching influence clinical outcome.25-27 A limitation of the current analysis is the lack of HLA-C and high-resolution typing and although the basic principle that the leader informs cord blood transplant outcomes can be readily tested, the results of the current study require validation in an independent cohort. We identified two potential associations between the HLA-B leader and cord blood outcomes among HLA-A, -DRB1-matched transplants with one HLA-B mismatch. First, as the number of M-leader(s) in the cord blood unit increased, the risk of relapse increased. Secondly, nonrelapse mortality was lower when the mismatched alleles were from different lineages and when the matched allele encoded an M leader than when the alleles were from the same lineage and the matched allele contributed a T leader even after adjustment for various factors including the use of ATG. Interestingly, HLA-B leader genotype and matching did not definitively correlate with GvHD risk, and in this way, the impact of the leader on outcome differs between cord blood and unrelated donor transplantation.19,20 The associations of the HLA-B leader to outcome were observed among HLA-A and -DRB1-matched transplants with one HLA-B mismatch, and future validation of these findings in large independent populations is warranted. The current study was not designed to identify the putative mechanisms of leader-associated effects on relapse and mortality and there many be several potential reasons for the different associations observed in CBT and unrelated donor transplantation. Unrelated transplantation and CBT procedures differ with respect to GvHD prophylaxis regimens as well as graft characteristics. GvHD has been reported to be lower in children who have received a cord blood or HLA-identical sibling bone marrow transplant.6 Similarly, the incidence of relapse and grades III-IV acute GvHD were lower in cord blood compared to matched or mismatched unrelated donor transplantation for leukemia.12 Compared to unrelated grafts, immature cord blood cells might be less likely activated during the alloreaction following transplant while NK-cell activation appears early after cord blood transplantation mitigating the risk of relapse.29 Possible mechanisms for higher relapse with lower non-relapse mortality among M-leader transplants could be related to the role of M-leader associated HLA ligands in both T-cell and NK-cell alloreactivity. M leaders influence inhibitory NK alloresponses.30,31 The exon 1-encoded leader is physicially linked to exons 2 and 3 that define the peptide-binding region of HLA-B molecules, and the peptide-binding region of M leader linked allotypes have striking differences when compared to T leader linked allotypes. Non-relapse mortality might include infectious causes of death for which both NK and T cells may play critical roles. NK surveillance in de novo and reactivation of viral infections after cord blood transplantation may shape relapse through the enhancement of NK maturation.32-36 We hypothesize that inhibitory NK responses involved in M leader associated relapse might occur concurrently with T-cell recognition of viral or bacterial peptide/HLA-B complexes. These mechanisms could haematologica | 2021; 106(12)

manifest as higher relapse with lower non-relapse mortality. Together with previous studies that have implicated NK KIR and HLA ligand interactions in relapse,8,18 the observations from the current study provide additional evidence for a potential role of the innate immune system in CBT outcomes. Future studies in double CBT may provide new information on key maternal-fetal interactions underpinning immune reactivity as a potential mechanism for the lower GvHD and relapse rates observed in CBT compared to adult hematopoietic stem cells. Although recent data suggest that more comprehensive HLA matching between the patient and the cord blood unit is associated with improved outcomes,25,26,37 a major advantage of cord blood transplantation is the less stringent requirement for matching than is needed in unrelated donor transplantation. In this way, both cord blood transplantation and haploidentical transplantation significantly advance the availability of these curative modalities. The inherently low GvHD rates coupled with the ability to lower relapse through the judicious selection of cord blood units based on the HLA-B leader, enhances the curative potential of CBT and provides rationale to explore the leader in haploidentical transplantation. The general clinical applicability of our findings to enhance CBT is facilitated by the one-to-one correlation of the leader to each serologically-defined HLA-B determinant, a feature that significantly simplifies the selection of mismatched units when matched units are not available, as the assignment of the leader requires only serologic-level definition of HLA-B. The observations from the current analysis suggest that when an MM or MT patient has multiple HLA-A, -DRB1-matched cord blood units, selection of leader-mismatched units with a shared M leader allele may help to lower non-relapse mortality. The HLA-B leader might also have implications for NK immunotherapy.38-40 The frequencies of TT, MT and MM in cord blood units and patients in the current study, together with those observed in unrelated donor transplants and over 2 million US volunteer donors, suggest that patients will have choices. Disclosures EG reports grant from the Centre Scientifique de Monaco. EWP, TG, CM report grants from the National Institutes of Health. TG reports Kiadis Pharma and Regimmune. No other disclosures were reported. Contributions EWP and EG designed the study; EG, FV, CK, AM, CM, SQ, VR, RT, CC, HR and AR assembled the data; TG performed statistical analysis; EWP drafted the manuscript. All authors critically reviewed, edited the manuscript and approved the final version. Acknowledgments We are indebted to the clinical transplant teams who performed the transplants described in the study (Online Supplementary Table S1). Funding Supported by a grant to Eurocord (to EG) from the Centre Scientifique de Monaco; grants from the National Institutes of Health, USA (AI069197 to EWP, TG, CMK; CA100019 to EWP, TG, CMK; CA18029 to EWP and TG; CA72978 to EWP; CA015704 to TG). 3113

E. W. Petersdorf et al.

References 1. Gluckman E, Broxmeyer HA, Auerbach AD, et al. Hematopoietic reconstitution in a patient with Fancon’s anemia by means of umbilical-cord blood from an HLA-identical sibling. N Engl J Med. 1989; 321(17):1174-1178. 2. Broxmeyer HE, Douglas GW, Hangoc G, et al. Human umbilical cord blood as a potential source of transplantable hematopoietic stem/progenitor cells. Proc Natl Acad Sci U S A. 1989;86(10):3828-3832. 3. Wagner JE, Broxmeyer HE, Byrd RL, et al. Transplantation of umbilical cord blood after myeloablative therapy: analysis of engraftment. Blood. 1992;79(7):1874-1881. 4. Gluckman E, Rocha V, Boyer-Chammard A, et al. Outcome of cord-blood transplantation from related and unrelated donors. Eurocord Transplant Group and the European Blood and Marrow Transplantation Group. N Engl J Med. 1997;337(6):373-381. 5. Rubinstein P, Carrier C, Scaradavou A, et al. Outcomes among 562 recipients of placental-blood transplants from unrelated donors. N Engl J Med. 1998;339(22):15651577. 6. Rocha V, Wagner JE, Sobocinski KA, et al. Graft-versus-host disease in children who have received a cord-blood or bone marrow transplant from an HLA-identical sibling. Eurocord and International Bone Marrow Transplant Registry Working Committee on Alternative Donor and Stem Cell Sources. N Engl J Med. 2000; 342(25):1846-1854. 7. Eapen M, Rubinstein P, Zhang MJ, et al. Outcomes of transplantation of unrelated donor umbilical cord blood and bone marrow in children with acute leukemia: a comparison study. Lancet. 2007; 369(9577):1947-1954. 8. Willemze R, Rodrigues CA, Labopin M, et al. KIR-ligand incompatibility in the graftversus-host direction improves outcomes after umbilical cord blood transplantation for acute leukemia. Leukemia. 2009; 23(3):492-500. 9. Brunstein CG, Wagner JE, Weisdorf DK, et al. Negative effect of KIR alloreactivity in recipients of umbilical cord blood transplant depends on transplantation conditioning intensity. Blood. 2009;113(22):56285634. 10. Tanaka J, Morishima Y, Takahashi Y, et al. Effects of KIR ligand incompatibility on clinical outcomes of umbilical cord blood transplantation without ATG for acute leukemia in complete remission. Blood Cancer J. 2013;3(11):e164. 11. Barker JN, Boughan K, Dahi PB, et al. Racial disparities in access to HLA-matched unrelated donor transplants: a prospective 1312patient analysis. Blood Adv. 2019;3(7):939944. 12. Milano F, Gooley T, Wood B, et al. Cordblood transplantation in patients with minimal residual disease. N Engl J Med. 2016;375(10):944-953. 13. Small TN, Papadopoulos EB, Boulad F, et al. Comparison of immune reconstitution

3114

after unrelated and related T-cell-depleted bone marrow transplantation: effect of patient age and donor leukocyte infusions. Blood. 1999;93(2):467-480. 14. Thomson BG, Robertson KA, Gowan D, et al. Analysis of engraftment graft-versushost disease, and immune recovery following unrelated donor cord blood transplantation. Blood. 2000;96(8):2703-2711. 15. Ruggeri A, Peffault de Latour R, Carmagnat M, et al. Outcomes, infections, and immune reconstitution after double cord blood transplantation in patients with highrisk hematological diseases. Transpl Infect Dis. 2011;13(5):456-465. 16. Jacobson CA, Turki AT, McDonough SM, et al. Immune reconstitution after double cord blood stem cell transplantation comparison with unrelated peripheral blood stem cell transplantation. Biol Blood Marrow Transplant. 2012;18(4):565-574. 17. Garfall A, Kim HT, Sun L, et al. KIR ligand incompatibility is not associated with relapse reduction after double umbilical cord blood transplantation. Bone Marrow Transplant. 2013;48(7):1000-1002. 19. Sekine T, Marin D, Cao K, et al. Specific combinations of donor and recipient KIRHLA genotypes predict for large differences in outcome after cord blood transplantation. Blood. 2016;128(2):297-312. 19. Petersdorf E, Carrington M, O’hUigin C, et al. Role of HLA-B exon 1 in graft-versushost disease after unrelated haemopoietic cell transplantation: a retrospective cohort study. Lancet Haematol. 2020;7(1):e50-e60. 20. Petersdorf EW, Stevenson P, Bengtsson M, et al. HLA-B leader and survivorship after HLA-mismatched unrelated donor transplantation. Blood. 2020;136(3):362-369. 21. Horowitz A, Djaoud Z, Nemat-Gorgani N, et al. Class I HLA haplotypes form two schools that educate NK cells in different ways. Sci Immunol. 2016;1(3):eaag1672. 22. Braud V, Jones EY, McMichael A. The human major histocompatibility complex class Ib molecule HLA-E binds signal sequence-derived peptides with primary anchor residues at positions 2 and 9. Eur J Immunol. 1997;27(5):1164-1169. 23. Merino AM, Song W, He D, et al. HLA-B signal peptide polymorphism influences the rate of HIV acquisition but not viral load. J Infect Dis. 2012;205(12):1797-1805. 24. Ramsuran V, Naranbhai V, Horowitz A, et al. Elevated HLA-A expression impairs HIV control through inhibition of NKG2Aexpressing cells. Science. 2018; 359(6371):86-90. Erratum in: Science. August 2, 2019. 25. Eapen M, Klein JP, Sanz GF, et al. Effect of donor-recipient HLA matching at HLA A, B, C, and DRB1 on outcomes after umbilical-cord blood transplantation for leukaemia and myelodysplastic syndrome: a retrospective analysis. Lancet Oncol. 2011;12(13):1214-1221. 26. Eapen M, Wang T, Veys PA, et al. Allelelevel HLA matching for umbilical cord blood transplantation for non-malignant diseases in children: a retrospective analysis. Lancet Haematol. 2017;4(7):e325-e333. 27. Rocha V, Gluckman E. Improving outcomes of cord blood transplantation: HLA match-

ing, cell dose and other graft- and transplantation –related factors. Br J Haematol. 2009;147(2):262-274. 28. HLA Informatics Group, The Anthony Nolan Research Institute. HLA nomenclature. http://hla.alleles.org/wmda/ index. html. Accessed March 6, 2020. 29. Ando T, Tachibana T, Tanaka M, et al. Impact of graft sources on immune reconstitution and survival outcomes following allogeneic stem cell transplantation. Blood Adv. 2020;4(2):408-419. 30. Luevano M, Daryouzeh M, Alnabhan R, Querol S, Khakoo S, Madrigal A, Saudemont A. The unique profile of cord blood natural killer cells balances incomplete maturation and effective killing function upon activation. Hum Immunol. 2012; 73(3):248-257. 31. Bjorkstrom NK, Riese P, Heuts F, et al. Expression patterns of NKG2A, KIR, and CD57 define a process of CD56dim NKcell differentiation uncoupled from NK-cell education. Blood. 2010;116(19):3853-3864. 32. Foley B, Cooley S, Verneris MR, et al. Cytomegalovirus reactivation after allogeneic transplantation promotes a lasting increase in educated NKG2C+ natural killer cells with potent function. Blood. 2012;119(11):2665-2674. 33. Della Chiesa M, Falco M, Bertaina A, et al. Human cytomegalovirus infection promotes rapid maturation of NK cells expressing activating killer Ig-like receptor in patients transplanted with NKG2C-/umbilical cord blood. J Immunol. 2014;192(4):1471-1479. 34. Guma M, Angulo A, Vilches C, GomezLozano N, Malats N, Lopez-Botet M. Imprint of human cytomegalovirus infection on the NK cell receptor repertoire. Blood. 2004;104(12):3664-3671. 35. Elmaagacli AH, Steckel NK, Koldehoff M, et al. Early human cytomegalovirus replication after transplantation is associated with a decreased relapse risk: evidence for a putative virus-versus-leukemia effect in acute myeloid leukemia patients. Blood. 2011;118(5):1402-1412. 36. Green ML, Leisenring WM, Xie H, et al. CMV reactivation after allogeneic HCT and relapse risk: evidence for early protection in acute myeloid leukemia. Blood. 2013; 122(7):1316-1324. 37. Yokoyama H, Morishima Y, Fuji S, et al. Impact of HLA allele mismatch at HLA-A, B, -C, and -DRB1 in single cord blood transplantation. Biol Blood Marrow Transplant. 2020;26(3):519-528. 38. Knorr DA, Bachanova V, Verneris MR, Miller JS. Clinical utility of natural killer cells in cancer therapy and transplantation. Semin Immunol. 2014;26(2):161-172. 39. Dahlberg CI, Sarhan D, Chrobok M, Duru AD, Alici E. Natural killer cell-based therapies targeting cancer: possible strategies to gain and sustain anti-tumor activity. Front Immunol. 2015;6:605. 40. Liu E, Marin D, Banerjee P, et al. Use of CAR-transduced natural killer cells in CD19-positive lymphoid tumors. N Engl J Med. 2020;382(6):545-553.

haematologica | 2021; 106(12)

ARTICLE

Cell Therapy & Immunotherapy

PVRIG is a novel natural killer cell immune checkpoint receptor in acute myeloid leukemia

Ferrata Storti Foundation

Jessica Li,1,2 Sarah Whelan,3 Maya F. Kotturi,3 Deborah Meyran,1,2,4 Criselle D’Souza,1,2 Kyle Hansen,3 Spencer Liang,3 John Hunter,3 Joseph A. Trapani1,2# and Paul J. Neeson1,2# 1 Cancer Immunology Program, Peter MacCallum Cancer Center Melbourne, Victoria, Australia; 12Sir Peter MacCallum Department of Oncology, The University of Melbourne, Melbourne, Victoria, Australia; 3Compugen, USA, Inc., South San Francisco, CA, USA and 4 Université de Paris, INSERM, U976 HIPI Unit, Institut de Recherche Saint-Louis, Paris, France #

Haematologica 2021 Volume 106(12):3115-3124

JAT and PJN contributed equally as co-senior authors.

ABSTRACT

his study explored the novel immune checkpoint poliovirus receptor-related immunoglobulin domain-containing (PVRIG) in acute myeloid leukemia (AML). We showed that AML patient blasts consistently expressed the PVRIG ligand (poliovirus receptor-related 2, PVRL2). Furthermore, PVRIG blockade significantly enhanced naural killer (NK)-cell killing of PVRL2+, poliovirus receptor (PVR)lo AML cell lines, and significantly increased NK-cell activation and degranulation in the context of patient primary AML blasts. However, in AML patient bone marrow, NK-cell PVRIG expression levels were not increased. In order to understand how PVRIG blockade might potentially be exploited therapeutically, we investigated the biology of PVRIG and revealed that NK-cell activation resulted in reduced PVRIG expression on the cell surface. This occurred whether NK cells were activated by tumor cell recognition, cytokines (interleukin 2 [IL-2] and IL-12) or activating receptor stimulation (CD16 and NKp46). PVRIG was present at higher levels in the cytoplasm than on the cell surface, particularly on CD56bright NK cells, which further increased cytoplasmic PVRIG levels following IL-2 and IL-12 activation. PVRIG was continually transported to the cell surface via the endoplasmic reticulum and Golgi in both unstimulated and activated NK cells. Taken together, our findings suggest that anti-PVRIG blocking antibody functions by binding to surface-bound PVRIG, which undergoes rapid turnover in both unstimulated and activated NK cells. We conclude that the PVRIG-PVRL2 immune checkpoint axis can feasibly be targeted with PVRIG blocking antibody for NK-mediated immunotherapy of PVRL2+ AML.

Correspondence: PAUL NEESON paul.neeson@petermac.org Received: May 19, 2020. Accepted: October 12, 2020. Pre-published: November 5, 2020.

Introduction Poliovirus receptor-related immunoglobulin domain-containing (PVRIG) has recently been identified as an immune checkpoint molecule with potential for therapeutic development.1 In humans, PVRIG is expressed on T cells (predominantly CD8+ T cells) and natural killer (NK) cells, but not on B cells, monocytes or neutrophils.1 PVRIG binds to a single ligand, poliovirus receptor-related 2 (PVRL2, also known as CD112 or Nectin-2), and exerts an inhibitory effect on cytotoxic lymphocyte activity, likely via an ITIM-like motif in its intracellular domain.1-3 PVRL2 is an adhesion molecule involved in the formation of cell-cell junctions, and is overexpressed in various cancers.4-8 PVRL2 is also a ligand of the co-activating receptor DNAX accessory molecule 1 (DNAM-1)9,10 and weakly binds another inhibitory receptor, T- cell immunoreceptor with Ig and ITIM domains (TIGIT).11-13 Recently, Whelan et al. demonstrated the inhibitory effect of PVRL2 was predominantly mediated by PVRIG and not TIGIT.3 DNAM-1 and TIGIT (but not PVRIG) also bind to a closely related molecule, poliovirus receptor (PVR, also known as CD155 or Necl-5).9,11,12 Competition studies have determined that PVR has higher affinity for TIGIT than DNAM-1, and PVRL2 has a higher affinity for PVRIG than DNAM-

haematologica | 2021; 106(12)

https://doi.org/10.3324/haematol.2020.258574

3115

J. Li et al.

1, suggesting that the inhibitory signal is dominant.1,11 PVRIG inhibitory function was shown using antiPVRIG blocking antibodies. Xu et al. demonstrated that PVRIG blocking antibodies significantly increased NK-cell cytotoxicity against breast cancer cell lines in vitro, an effect that was further enhanced when used in combination with TIGIT blocking antibodies.2 An independent group using a different anti-PVRIG antibody similarly showed that PVRIG blockade enhanced T-cell cytotoxicity against melanoma and pancreatic cancer cell lines, which was also augmented by combination with TIGIT blockade.3 Notably, Whelan et al. demonstrated that T cells isolated from patient tumors and activated via CD3 increased interferon g production in response to combination PVRIG/TIGIT blockade.3 PVRIG blockade also reduced tumour burden in a mouse model when combined with anti-PDL1.14 On the basis of these data, a human IgG4 anti-PVRIG blocking antibody is currently undergoing phase I clinical trials in patients with advanced solid tumors.15 As PVRIG is present on both T cells and NK cells, blocking PVRIG provides the opportunity to augment both major cytotoxic effector cell types. Although many studies focused on the capacity for immune checkpoint blockade to enhance T-cell responses, the contribution of NK cells should not be overlooked. For instance, tumors often downregulate human leukocyte antigen (HLA) class I to evade CD8+ T-cell recognition.16 However, this simultaneously removes the ligand for killer cell immunoglobulinlike receptors (KIRs) on NK cells, rendering tumors more sensitive to NK cell-mediated killing.17 Reducing the inhibitory signal from KIR has also been shown to be effective in controlling acute myeloid leukemia (AML). AML is an aggressive disease in which myeloid progenitor cells proliferate uncontrollably, and which is frequently treated with allogeneic hematopoietic stem cell transplant (alloHSCT) when patients relapse after front-line chemotherapy. In a seminal study of allo-HSCT patients, mismatches between KIR on donor NK cells and recipient HLA was a key predictor of survival.18,19 Recipients lacking HLA ligands for one or more of the KIR expressed by the donor experienced graft-versus-host NK alloreactivity, which was significantly associated with a lower relapse rate.19,20 Given the pivotal role of NK cells in AML, strategies to enhance NK-cell activity could provide significant benefit for patients with AML, who have a 5-year survival rate of less than 30% with current treatments.21 This study aimed to determine whether PVRIG blockade could be used to enhance NK-cell responses against AML. Using healthy donor and AML patient blood or bone marrow samples, we evaluated the expression of PVRIG and PVRL2 on NK cells and AML blasts respectively. We also investigated whether PVRIG blockade could enhance NK-cell-mediated killing of AML blasts, and the kinetics of PVRIG surface expression to reveal when the target is expressed following AML target cell activation of NK cells.

Recombinant human IL-2, IL-12, IL-15 and IL-18 were purchased from Peprotech. Monensin (GolgiStop, BD Biosciences) and brefeldin A (eBioscience) were both used at 1:1,000. Antibodies used for flow cytometry staining are listed in the Online Supplementary Table S1. SKBR3, KG1a, K562, ML-2, THP-1 and Kasumi-1 cell lines were maintained in RPMI 1640 (Gibco) supplemented with Glutamax, penicillin, streptomycin and 10% (or 20% for Kasumi-1) fetal calf serum (FCS). AML-193 cell line was maintained in Iscove's Modified Dulbecco's Media supplemented with 5% FCS, 5 mg/mL transferrin, 5 mg/mL insulin and 2 ng/mL granulocyte-macrophage colony-stimulating factor. All cell lines tested negative for mycoplasma.

Acute myeloid leukemia patient and healthy donor bone marrow samples All patient and healthy donor samples were obtained under ethics approval from the Peter MacCallum Cancer Center human ethics committee (HREC approval numbers 01/14 and 10-61). Cryopreserved AML patient diagnostic bone marrow samples were obtained from the Cancer Collaborative Biobank (Metro South Health, Queensland, Australia). Patient clinical characteristics are summarised in the Online Supplementary Table S2. Healthy donor bone marrow samples were obtained from Royal Melbourne Hospital (Melbourne, Australia) or purchased from AllCells (Alameda, California). All bone marrow samples were used for flow cytometry staining immediately after thawing.

Healthy donor peripheral blood mononuclear cells and natural killer cells Peripheral blood mononuclear cells (PBMC) were isolated from healthy donor buffy coats (Australian Red Cross Blood Service) by density gradient (Ficoll-Paque, GE Healthcare Life Science) and cryopreserved. One day prior to experiments, PBMC were thawed and treated with DNase I (Merck) for 15 minutes at 37ºC. Where required, NK cells were isolated by negative selection using a human NK Cell Isolation Kit (Miltenyi Biotec) according to the manufacturer’s instructions (except antibodies and beads were used at half the recommended concentration). The purity of NK cells as determined by flow cytometry was > 95%. Bulk PBMC or isolated NK cells were incubated in media containing 25 U/mL IL-2 overnight at 37ºC before use in assays.

Natural killer cell stimulation Isolated NK cells were incubated at 37ºC alone, with the specified combination of cytokines, with target cells at a 1:1 ratio, or in wells precoated (overnight 4ºC) with agonistic antibodies against CD16, NKp46, 2B4 or NKG2D. After 24 hr, cells were washed and stained with LIVE/DEAD Fixable Yellow (ThermoFisher) followed by antibodies against CD56, CD16, CD69, PVRIG, TIGIT and DNAM-1. For analysis of short term kinetics, cells were incubated at 37ºC with the indicated stimuli, and at the specified time points were transferred to 4ºC. Cells for the 0 time point were kept at 4ºC. Upon completion of all time points, cells were stained with LIVE/DEAD Fixable Yellow followed by antibodies against CD56, CD16, CD69 and PVRIG.

Other methods Methods

Details of experimental procedures (flow cytometry, Chromium release assay, degranulation assay) are provided in the Online Supplementary Methods.

Reagents and cell lines Anti-PVRIG and anti-TIGIT blocking antibodies were provided by Compugen, USA, Inc. Anti-DNAM-1 (11A8), anti-CD16 (3G8), anti-NKp46 (9E2), anti-2B4 (eBioC1.7) and anti-NKG2D (1D11) purified antibodies were purchased from Biolegend.

3116

Data analysis Flow cytometry data were analysed using FlowJo software (BD), and statistical analysis was performed in Prism (GraphPad). Figure 4G was created with BioRender.com.

haematologica | 2021; 106(12)

PVRIG blockade in acute myeloid leukemia

Results PVRIG blockade enhances natural killer cell killing of PVRL2hiPVRlo acute myeloid leukemia cells In order to assess whether PVRIG blockade could enhance NK-cell responses against AML, we utilized the AML cell line KG1a. When co-cultured with healthy donor PBMC and PVRIG blocking antibody, a significant

increase in KG1a cell death was observed compared to the untreated control (Figure 1A). In contrast, TIGIT blocking antibody did not significantly enhance KG1a target cell lysis, and KG1a lysis in the presence of combined antiPVRIG and anti-TIGIT was comparable to PVRIG blockade alone (Figure 1A). In order to compare across donors with variable baseline killing, we calculated the NK:target ratio required for 10% KG1a lysis for each donor. PVRIG

J H

Figure 1. Blockade of PVRIG enhanced natural killer cell killing of tumor cell lines. (A) Percentage lysis of KG1a cells after 4-hour co-culture with peripheral blood mononuclear cells (PBMC) in the presence of anti-PVRIG, anti-TIGIT, anti-PVRIG + anti-TIGIT, or isotype antibodies, measured by 51Cr release assay. Representative data (mean ± standard deviation [SD] of triplicates) of four experiments shown. (B) Natural killer (NK):target ratio required to achieve 10% lysis, determined by nonlinear regression of curves plotted as in (A). Each symbol represents an individual donor, n=4. NK-cell expression of (C) CD69 and (D) CD107a after 4-hour co-culture of PBMC with KG1a (8:1 E:T ratio) in the presence of the indicated blocking antibodies or isotype control antibody. Representative data (mean ± SD of triplicates) of two experiments is shown. (E) Percentage lysis of KG1a cells after 4-hour co-culture with PBMC in the presence of anti-PVRIG or isotype antibodies, with or without 4 mM EGTA. Representative data (mean ± SD of triplicates) of two experiments is shown. (F) Percentage lysis of SKBR3 cells after 4-hour co-culture with PBMC in the presence of anti-PVRIG, anti-TIGIT, anti-PVRIG + anti-TIGIT, or isotype antibodies, measured by 51Cr release assay. Representative data (mean ± SD of triplicates) of three experiments is shown. (G) NK:target ratio required to achieve 10% lysis, determined by non-linear regression of curves plotted as in (F). Each symbol represents an individual donor (n=3). NK-cell expression of (H) CD69 and (I) CD107a after 4-hour co-culture of PBMC with SKBR3 (2:1 E:T ratio) in the presence of the indicated blocking antibodies or isotype control antibody. Representative data (mean ± SD of triplicates) of two experiments is shown. (J) Expression of PVRL2 and PVR (red histograms) on SKBR3 and KG1a cells compared with isotype control stain (grey histograms). NK:target ratios in (A, E and F) were calculated using % of NK cells in PBMC determined by flow cytometry. Significance was determined by repeated measures one-way ANOVA (A, B, F and G) or one-way ANOVA (C, D, H and I) with Holm-Sidak’s multiple comparisons test, not significant (ns) P>0.05, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. PVRIG: poliovirus receptor-related immunoglobulin domain-containing; TIGIT: T- cell immunoreceptor with Ig and ITIM domains; MFI: mean fluorescence intensity.

haematologica | 2021; 106(12)

3117

J. Li et al.

blockade significantly decreased the NK:target ratio required to reach 10% KG1a lysis, while TIGIT blockade had only a minor effect (Figure 1B). PVRIG blockade also significantly increased NK-cell activation and degranulation, as measured by CD69 and CD107a staining respectively (Figure 1C and D). TIGIT blockade had minimal effect on both NK-cell activation and degranulation and combined PVRIG and TIGIT blockade showed no benefit over PVRIG blockade alone (Figure 1C and D). The activating receptor DNAM-1 was important for recognition of targets, as blocking DNAM-1 significantly inhibited NKcell activation and degranulation (Figure 1C ,D, H and I). KG1a target cell death was perforin-dependent, as it was completely blocked when free calcium was complexed with EGTA (Figure 1E). PVRIG blockade was clearly more effective than TIGIT blockade for enhancing NK-cell responses against KG1a but did not enhance lysis of the breast cancer cell line

SKBR3. Rather, significantly more target cell death was observed with TIGIT blockade, or combined PVRIG and TIGIT blockade (Figure 1F). Pooled data from three donors suggested TIGIT blockade, but not PVRIG blockade, decreased the NK:target ratio required for 10% lysis, but the difference did not reach statistical significance (Figure 1G). TIGIT blockade significantly enhanced NK-cell activation and degranulation, whereas PVRIG blockade had minimal effect on activation and a much smaller effect on degranulation (Figure 1H and I). Combined PVRIG and TIGIT blockade enhanced NK-cell activation and degranulation cytotoxicity even further, suggesting that PVRIG blockade can have an additive effect to TIGIT blockade (Figure 1H and I). NK cells from all healthy donors tested expressed both PVRIG and TIGIT (Online Supplementary Figure S1). Although the levels of expression (particularly of TIGIT) varied, all donors were consistently more responsive to

Figure 2. PVRIG and its ligand PVRL2 are expressed in acute myeloid leukemia patient bone marrow. Expression of (A) poliovirus receptor-related 2 (PVRL2) (B) poliovirus receptor (PVR) or (C) poliovirus receptor-related immunoglobulin domain-containing (PVRIG) on blasts or immune cell types in the bone marrow of acute myeloid leukemia (AML) patients (n=19-20) or healthy donors (n=13). Open triangles mark the patient shown in (D to F). Representative histograms of (D) PVRL2 and (E) PVR on AML blasts, or (F) PVRIG on natural killer (NK) cells in the bone marrow of an AML patient. Overlay histograms of test (red) and isotype control stains (grey) are shown. NK-cell expression of (G, I and K) CD69 and (H, J and L) CD107a after 4-hour co-culture of healthy donor peripheral blood mononuclear cells (PBMC) with AML patient bone marrow (8:1 E:T ratio) in the presence of the indicated blocking antibodies or isotype control antibody (mean ± standard deviation of triplicates, n= 3 patients). Significance was determined by one-way ANOVA with Holm-Sidak’s multiple comparisons test, not significant (ns) P>0.05, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; MFI: mean fluorescence intensity.

3118

haematologica | 2021; 106(12)

PVRIG blockade in acute myeloid leukemia

PVRIG blockade against KG1a targets, and more responsive to TIGIT blockade against SKBR3 targets. Because TIGIT binds preferentially to PVR, while PVRIG binds exclusively to PVRL2, we explored whether differential expression of the ligands on KG1a and SKBR3 could explain the different NK-cell responses to PVRIG or TIGIT blockade. Indeed, we observed that while both SKBR3 and KG1a cells have high expression of PVRL2, KG1a expressed far less PVR than SKBR3 (Figure 1J). This suggests that tumors expressing high levels of PVRL2 but low levels of PVR are more likely to inhibit immune cells via PVRIG, whereas when both ligands are present, inhibition via TIGIT appears to predominate. This trend was also observed with other AML cell lines (Online Supplementary Figure S2). Both AML-193 and Kasumi-1 cells were PVRL2+PVRlo (Online Supplementary Figure S2A) and were killed at significantly higher levels in the context of antiPVRIG with PBMC from healthy donors (Online Supplementary Figure S2B and C). By contrast, ML-2 and THP-1 AML cells were PVRL2+PVR+ (Online Supplementary Figure S2A) and were significantly increased killed in the presence of anti-TIGIT rather than anti-PVRIG (Online Supplementary Figure S2D and E).

Acute myeloid leukemia patient bone marrow contains PVRL2hiPVRlo blasts and PVRIG+ natural killer cells We next examined the expression of PVRIG, PVR and PVRL2 in AML patient bone marrow. Using multicolour flow cytometry, we distinguished various lymphoid (CD3–CD56+ NK cells, CD3+CD56+ NKT cells, CD3+CD8+ T cells, CD3+CD8– T cells), myeloid (SSChiCD14+CD11b+ monocytes) and blast (CD45loSSCint) populations collected at diagnosis (Online Supplementary Figure S3). Primary AML blasts were PVRL2+PVRlo, and PVR was expressed at higher levels on CD14+CD11b+ monocytes (Figure 2A, B, D and E). In healthy donor bone marrow samples, the CD45loSSCint immature myeloid population (gated as per Online Supplementary Figure S3) were also PVRL2+PVRlo, suggesting that high PVRL2 expression is a feature of normal myeloblasts (Figure 2A and B). Nonetheless, the PVRL2hiPVRlo phenotype of the AML blasts suggested they would be a good target for PVRIG blockade, provided that patient effector cells expressed PVRIG. T cells and NK cells expressed PVRIG in all AML patients tested, with higher expression in NK, NKT and CD8+ T cells and lower expression in CD8- T cells (Figure 2C and F). There was no statistical difference in PVRIG expression levels on bone

Figure 3. PVRIG expression on natural killer cells is decreased upon activation. (A) Poliovirus receptor-related immunoglobulin domain-containing (PVRIG) expression on isolated natural killer (NK) cells after 24-hour cultured alone, or co-cultured with K562 or KG1a cells (1:1 ratio). Shown is percentage change in PVRIG mean fluorescence intensity (MFI) relative to NK alone, each point represents an individual donor (n=3), bars represent mean ± standard deviation (SD). (B) PVRIG expression on isolated NK cells after 24-hour incubation alone or in the presence of 100 U/mL interleukin 2 (IL-2) and 10 ng/mL IL-12. Shown is percentage change in PVRIG MFI relative to NK alone, each point represents an individual donor (n=3), bars represent mean ± SD. (C) Expression of PVRIG and CD69 on isolated NK cells after 24-hour incubation with 25 U/mL IL-2, 100 U/mL IL-2, or combinations of IL-2 (100 U/mL), IL-12 (10 ng/mL), IL-15 (50 ng/mL) and IL-18 (50 ng/mL), as indicated. Representative data of two independent experiments is shown. Expression of (D) CD69 or (E) PVRIG on isolated NK cells after 24-hour incubation with the indicated plate-bound antibodies. Shown is percentage change in PVRIG MFI relative to isotype. Each donor is represented by a distinct symbol (n=3), bars represent mean ± SD. (F) Expression of PVRIG vs. CD69 on isolated NK cells after 24-hour incubation with indicated stimuli, as in (C) and (E). Significance determined by one-way ANOVA with Holm-Sidak’s multiple comparisons test (A, Dand E) or Student’s t-test (B), not significant (ns) P>0.05, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

haematologica | 2021; 106(12)

3119

J. Li et al.

marrow immune subsets of AML patients compared with healthy donors (Figure 2C). PVRIG and PVRL2 expression levels varied considerably amongst patients, but this did not correlate with AML subtype or the percentage of bone marrow blasts (Online Supplementary Figure S4). In order to assess whether PVRIG blockade could enhance NK-cell killing of AML patient blasts, we co-cultured healthy donor PBMC with bone marrow from an AML patient with a high percentage (>90%) of PVRL2hiPVRlo AML blasts. In this context, NK cells showed significantly increased CD69 expression and degranulation with PVRIG or TIGIT blockade. In addition, combined PVRIG and TIGIT blockade was associated

with significantly higher NK-cell activation and degranulation (Figure 2G and H). Similar results were obtained for a further two AML patients tested (Figure 2I to L). This indicates that PVRIG blockade or combination PVRIG/TIGIT blockade could enhance NK-cell cytotoxicity against PVRL2+ tumor targets in AML patients.

PVRIG expression on natural killer cells is modulated by activation In order to understand why PVRIG was not upregulated on AML patient NK cells, we explored the mechanisms regulating NK cell PVRIG expression. In order to do this, we activated healthy donor NK cells for 24 hours via co-culture

Figure 4. Modulation of DNAM-1 and TIGIT expression on natural killer cells upon activation. (A to C) T-cell immunoreceptor with Ig and ITIM domains (TIGIT) or (D to F) DNAX accessory molecule 1 (DNAM-1) expression on isolated natural killer cells after 24-hour culture with (A and D) K562 or KG1a cells (1:1 ratio); (Band E) 100 U/mL interleukin 2 (IL-2) and 10 ng/mL IL-12; or (C and F) with the indicated plate-bound antibodies. Percentage change in mean fluorescence intensity (MFI) relative to natural killer (NK) alone is shown, each point represents an individual donor (n=3), bars represent mean ± standard deviation (SD). Significance determined by one-way ANOVA with Holm-Sidak’s multiple comparisons test (A, C, D and F) or Student’s t-test (B and E), not significant (ns) P>0.05, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. (G) Model of PVRIG-TIGIT-DNAM-1 modulation upon NK-cell activation. Left: upon recognition of tumor cells expressing poliovirus receptor (PVR) and poliovirus receptor-related 2 (PVRL2), NK cells increase TIGIT but lose expression of both poliovirus receptor-related immunoglobulin domain-containing (PVRIG) and DNAM-1. This loss of DNAM-1 may result from a tumor-intrinsic mechanism of immune escape, whereby tumor cells expressing DNAM-1 ligands induce loss of DNAM-1 expression on immune cells. Right: upon activation via cytokines such as IL-2 and IL-12, or by stimulation of activating receptors such as CD16, NK cells decrease expression of PVRIG while increasing expression of TIGIT and DNAM-1. The increased expression of DNAM-1 relative to the decreased expression of PVRIG may serve to push the balance within the cell towards more activating signaling. In this way, PVRIG may act to constitutively dampen NK responsiveness in the steady state and is lost upon activation to lower the activation threshold of the NK cell.

3120

haematologica | 2021; 106(12)

PVRIG blockade in acute myeloid leukemia

with tumor targets, NK-cell activating cytokines or via agonistic antibodies to NK-cell activating receptors. NK cells consistently decreased PVRIG expression after interaction with target cells, although to a much greater degree with K562 than KG1a cells (Figure 3A). This may be because K562 lacks human leukocyte antigen (HLA) class I, resulting in greater activation of the NK cells. We next investigated whether activation of NK cells via cytokines would also cause loss of PVRIG. Indeed, NK cells stimulated with IL-2 and IL-12 had significantly decreased PVRIG expression (Figure 3B). Using IL-2 or combinations of IL-2, IL-12, IL-15 and IL-18, NK cells were increasingly activated, as meas-

ured by CD69 levels. Interestingly, NK-cell surface PVRIG and CD69 levels were inversely correlated in NK cells undergoing cytokine-mediated activation (Figure 3C). Stimulation of NK cells with plate-bound agonistic antibodies against the activating receptors CD16, NKp46, 2B4 and NKG2D also resulted in differing levels of activation (Figure 3D) and a concomitant decrease in NK-cell PVRIG levels (Figure 3E). When PVRIG and CD69 expression was examined in individual NK cells, there was a trend for more activated NK cells (higher CD69) to express proportionally less PVRIG after anti-CD16 or anti-NKp46 stimulation (Figure 3F). In contrast, following cytokine stimulation NK cells

Figure 5. Differential regulation of surface and intracellular PVRIG by CD56dim and CD56bright natural killer cells. (A) Poliovirus receptor-related immunoglobulin domain-containing (PVRIG) expression on isolated natural killer (NK) cells (gated on CD56dim or CD56bright subsets) after 24-hour stimulation with 100 U/mL IL-2 and 10 ng/mL IL-12, measured by surface or total (intracellular + surface) staining. Representative of two to three experiments. (B and C) Expression of surface PVRIG on isolated NK cells (gated on CD56dim or CD56bright) after 24-hour culture with (B) 100 U/mL interleukin 2 (IL-2 )and 10 ng/mL IL-12; or (C) with K562 or KG1a cells (1:1 ratio). Shown is percentage change in mean fluorescence intensity (MFI) relative to NK alone, bars represent mean ± standard deviation (SD) of three experiments. PVRIG expression on NK cells gated on (D) CD56dim or (E) CD56bright subsets, measured by surface or total (intracellular + surface) staining after 24hour stimulation with 100 U/mL IL-2 and 10 ng/mL IL-12. Shown is the percentage change in PVRIG MFI relative to NK alone, bars represent mean ± SD of two to three experiments. Significance determined by Student’s t-test, not significant (ns) P>0.05, *P<0.05, **P<0.01.

haematologica | 2021; 106(12)

3121

J. Li et al.

increased CD69 and decreased PVRIG expression more uniformly (Figure 3F).

TIGIT and DNAM-1 expression on natural killer cells is modulated by activation TIGIT and DNAM-1 are NK-cell receptors within the same receptor-ligand axis as PVRIG.9-12 We investigated whether a similar modulation of TIGIT and DNAM-1 occurred following NK-cell activation. In contrast to PVRIG, TIGIT expression was increased after stimulation of NK cells with target cells, cytokines or agonistic antibodies (Figure 4A to C), whereas changes in DNAM-1 levels were dependent on the stimulus. DNAM-1 was reduced by interaction with target cells (Figure 4D), but increased

after activation with IL-2 and IL-12 or anti-CD16 (Figure 4E to F). Overall, our results indicated that the expression levels for different NK receptors are regulated differently, depending on the stimulus. PVRIG was consistently decreased and TIGIT increased upon NK-cell activation, regardless of the stimulus, and the magnitude of change was correlated with the level of activation. On the other hand, DNAM-1 expression decreased upon target recognition, but was increased by activation via cytokines or agonistic antibodies to activating receptors (Figure 4G). The loss of DNAM-1 may result from a form of immune evasion, which has previously been described to occur on contact with PVR+ tumor cells.22 DNAM-1 increase in response to stimulation via cytokines or activation receptors could

Figure 6. PVRIG is constitutively trafficked to the naural killer cell surface. (A) Poliovirus receptor-related immunoglobulin domain-containing (PVRIG) and (B) CD69 expression on isolated natural killer (NK) cells incubated alone, with K562 cells (1:1 ratio), or with plate-bound anti-CD16 antibody at 37ºC for the indicated time points. Representative data (mean ± standard deviation [SD] of duplicates) of two experiments is shown. (C and D) PVRIG expression on isolated NK cells incubated either (C) alone or (D) with plate-bound anti-CD16 antibody at 37ºC for the indicated time points, in the presence or absence of monensin (mon) or brefeldin A (BFA). Representative data (mean ± SD of duplicates) of two experiments is shown. Significance determined by multiple t-tests with Holm-Sidak’s correction, *P< 0.05 compared with NK alone (A and B) or untreated (C and D).

3122

haematologica | 2021; 106(12)

PVRIG blockade in acute myeloid leukemia

be a means to increase the net activation signal, which, in conjunction with decreased PVRIG expression, could serve to lower the activation threshold of the NK cell.

Intracellular PVRIG does not decrease upon activation We next investigated whether PVRIG levels are regulated differently in the two principal peripheral blood NK-cell subsets, CD56dimCD16+ and CD56brightCD16–. After IL-2 and IL-12 stimulation, the CD56dim subset showed decreased PVRIG surface expression, as was seen previously for unfractionated NK cells (Figure 5A and B). This was unsurprising given that CD56dim comprise 90-95% of all circulating NK cells. By contrast, CD56bright NK cells showed no loss of PVRIG (Figure 5A and B). This was not due to failed activation, as both subsets upregulated CD69 equally (data not shown). Interestingly, this distinct pattern of regulation of PVRIG occurred only with cytokine stimulation (Figure 5B), but not following interaction with target cells, which caused both subsets to downregulate PVRIG (Figure 5C). We next determined whether an intracellular pool of PVRIG exists, or if it is solely expressed on the cell surface. Using permeabilized cells, we detected total (surface plus intracellular) PVRIG levels. For both CD56dim and CD56bright NK-cell subsets, total PVRIG staining was far greater than surface staining, indicating that a pool of PVRIG is present in the cytosol (Figure 5A). Interestingly, while CD56dim NK cells lost their surface PVRIG upon activation with IL-2 and IL-12, the total amount of PVRIG was unchanged compared with untreated NK cells (Figure 5D). This suggested that PVRIG lost from the cell surface upon activation was internalized. Alternatively, the total PVRIG level could be maintained despite decreased surface expression by synthesis of new molecules, which then egress to the cell surface. The latter appeared to be the case for CD56bright NK cells, which maintained PVRIG surface expression and increased total PVRIG levels upon activation by IL-2 and IL-12 (Figure 5E).

Natural killer cell surface PVRIG levels are maintained by continuous trafficking to the cell surface In order to further explore the mechanisms by which NK cells regulate surface PVRIG levels, we first assessed the kinetics of PVRIG loss from the cell surface under different stimulation conditions using a short term time course. NK cells co-cultured with K562 cells showed loss of PVRIG expression within 1-2 hours, at which time CD69 began to be upregulated (Figure 6A-B). Stimulation of NK cells with anti-CD16 caused a similar level of PVRIG loss and activation, although K562 appeared to be the stronger stimulus at early time points (Figure 6A and B). Within the 4 hours assessed, IL-2 and IL-12 stimulation had no appreciable effect on PVRIG expression, and minimal effect on activation (Online Supplementary Figure S5), indicating that cytokine stimulation influences PVRIG levels more slowly. Next, we used monensin and brefeldin A to determine whether intracellular trafficking via the endoplasmic reticulum (ER) and Golgi network was important for maintaining NK-cell surface PVRIG. Both monensin and brefeldin A inhibit the trafficking of molecules to the cell surface, by disrupting the Golgi apparatus, trans-Golgi network or endosomal network.23-26 Untreated NK cells maintained constant PVRIG surface levels over 4 hours. However, the addition of either monensin or brefeldin A resulted in significant loss of surface PVRIG levels (Figure 6C). Interestingly, the presence of monensin haematologica | 2021; 106(12)

or brefeldin A also caused greater loss of PVRIG in NK cells undergoing activation via anti-CD16 stimulation (Figure 6D). These results suggest there is trafficking of PVRIG molecules to the cell surface, via the ER and Golgi, in both untreated and activated NK cells (Figure 6E). In summary, PVRIG is downregulated on the NK-cell surface following activation by tumor targets, anti-CD16 or cytokines. Furthermore, a pool of PVRIG is present in the NK-cell cytoplasm, and cell surface PVRIG is maintained by trafficking to the surface. Taken together, our findings suggest that anti-PVRIG blocking antibodies enhanced NK-cell killing of AML target cells by blocking PVRIG present on the NK-cell surface. This resulted in decreased PVRL2-PVRIG mediated inhibition, and a decreased threshold for NK-cell activation and increased AML blast killing.

Discussion Enhancing the activity of NK cells following HSCT may be beneficial for AML patients. NK cells are the first lymphoid cells to be reconstituted after HSCT, reaching normal levels within 1 month after transplant, much earlier than T cells.27,28 However, their capacity to kill residual leukemic blasts can be limited by the interaction of NKinhibitory receptors with ligands in the tumor microenvironment.27,29,30 Thus, blocking inhibitory receptors such as PVRIG could potentially be useful after HSCT to enhance NK-cell activity to delay or prevent relapse. In this study, we showed PVRIG blockade enhanced human NK-cell activity against PVRL2hiPVRlo AML target cells. AML blasts in patient bone marrow were PVRL2hiPVRlo, suggesting PVRIG blockade may increase NK-mediated killing of AML blasts. The AML blast PVRL2hiPVRlo phenotype is consistent with previous studies in AML patients.31 Our study is the first to report NK cell PVRIG expression in AML patient bone marrow. NK-cell PVRIG expression was not upregulated in AML patients. Our subsequent analysis suggested PVRIG upregulation is not required for PVRIG blockade to be effective. Even though interaction with AML cells caused loss of PVRIG from the NK-cell surface, PVRIG molecules trafficked via the ER-Golgi network and were then expressed on the cell surface. This suggests that, over time, a far greater amount of PVRIG is available on the cell surface to be blocked by anti-PVRIG antibodies than is detected at a single time point. In contrast to PVRIG, other NK-cell immune checkpoint receptors, such as TIGIT, are upregulated with activation.2,11 Despite sharing the same ligand as TIGIT, the modulation of PVRIG does not follow this model, suggesting it could have a distinct biological function. It is possible that PVRIG acts as a regulator to keep NK cells in check in the steady state. The downregulation of PVRIG in response to cytokines or target recognition would then allow greater activation of NK cells in response to inflammatory stimuli. Our data showed NK cell PVRIG was present at higher levels in the cytoplasm than on the cell surface; this intracellular PVRIG was not decreased by activation. This cytoplasmic pool of PVRIG could represent either newly synthesized or recycled protein. A recent study by Whelan et al.3 examined PVRIG expression on isolated human T cells, and observed a similar trend for loss of PVRIG expression immediately after acti3123

J. Li et al.

vation. However, sustained activation of T cells with antigen, IL-2 and IL-7 resulted in increased PVRIG expression by day 11. This study reveals unique aspects of PVRIG biology that should be considered when determining potential indications for its therapeutic use. Our results suggest that PVRIG blockade may still have a therapeutic effect, provided the tumor cells express the PVRL2hiPVRlo phenotype. Furthermore, when AML blasts express both PVRL2 and PVR, combination PVRIG and TIGIT blockade may also induce an effective NK-cell mediated anti-tumor response. These findings also have broader implications for the study of other checkpoint receptors. As more novel receptors are identified as potential targets, they should not be assumed to have the same biology as previously established immune checkpoints, and their potential effi-

References 1. Zhu Y, Paniccia A, Schulick AC, et al. Identification of CD112R as a novel checkpoint for human T cells. J Exp Med. 2016;213(2):167-176. 2. Xu F, Sunderland A, Zhou Y, Schulick RD, Edil BH, Zhu Y. Blockade of CD112R and TIGIT signaling sensitizes human natural killer cell functions. Cancer Immunol Immunother. 2017;66(10):1367-1375. 3. Whelan S, Ophir E, Kotturi MF, et al. PVRIG and PVRL2 Are induced in cancer and Inhibit CD8(+) T-cell function. Cancer Immunol Res. 2019;7(2):257-268. 4. Li M, Qiao D, Pu J, Wang W, Zhu W, Liu H. Elevated Nectin-2 expression is involved in esophageal squamous cell carcinoma by promoting cell migration and invasion. Oncol Lett. 2018;15(4):4731-4736. 5. Oshima T, Sato S, Kato J, et al. Nectin-2 is a potential target for antibody therapy of breast and ovarian cancers. Mol Cancer. 2013;12:60. 6. Miao X, Yang ZL, Xiong L, et al. Nectin-2 and DDX3 are biomarkers for metastasis and poor prognosis of squamous cell/adenosquamous carcinomas and adenocarcinoma of gallbladder. Int J Clin Exp Pathol. 2013;6(2):179-190. 7. Liang S, Yang Z, Li D, et al. The clinical and pathological significance of Nectin-2 and DDX3 expression in pancreatic ductal adenocarcinomas. Dis Markers. 2015; 2015:379568. 8. Bekes I, Lob S, Holzheu I, et al. Nectin-2 in ovarian cancer: how is it expressed and what might be its functional role? Cancer Sci. 2019;110(6):1872-1882. 9. Bottino C, Castriconi R, Pende D, et al. Identification of PVR (CD155) and Nectin-2 (CD112) as cell surface ligands for the human DNAM-1 (CD226) activating molecule. J Exp Med. 2003;198(4):557-567. 10. Tahara-Hanaoka S, Shibuya K, Onoda Y, et al. Functional characterization of DNAM-1 (CD226) interaction with its ligands PVR (CD155) and nectin-2 (PRR-2/CD112). Int Immunol. 2004;16(4):533-538. 11. Yu X, Harden K, Gonzalez LC, et al. The surface protein TIGIT suppresses T cell activa-

3124

cacy should not necessarily be measured by the same parameters. Disclosures No conflicts of interest to disclose.

Contributions JL, DM and CD performed research, analyzed data and wrote the manuscript; SW, MK, KH, DM, KH, SL and JH provided academic input, shared key reagents, and wrote the manuscript; JAT and PJN supervised the project and wrote the manuscript. Funding Research funding for this project was provided by Compugen Inc. In the prior 36 months, PN also received research funding from BMS, Roche Genentech and Allergan.

tion by promoting the generation of mature immunoregulatory dendritic cells. Nat Immunol. 2009;10(1):48-57. 12. Stanietsky N, Simic H, Arapovic J, et al. The interaction of TIGIT with PVR and PVRL2 inhibits human NK cell cytotoxicity. Proc Natl Acad Sci U S A. 2009; 106(42):1785817863. 13. Deuss FA, Gully BS, Rossjohn J, Berry R. Recognition of nectin-2 by the natural killer cell receptor T cell immunoglobulin and ITIM domain (TIGIT). J Biol Chem. 2017;292(27):11413-11422. 14. Murter B, Pan X, Ophir E, et al. Mouse PVRIG Has CD8(+) T cell-specific coinhibitory functions and dampens antitumor immunity. Cancer Immunol Res. 2019;7(2): 244-256. 15. NCT03667716: COM701 in subjects with advanced solid tumors. [cited 22 Jan 2020]; Available from: https:// ClinicalTrials.gov/ show/NCT03667716 16. Aptsiauri N, Ruiz-Cabello F, Garrido F. The transition from HLA-I positive to HLA-I negative primary tumors: the road to escape from T-cell responses. Curr Opin Immunol. 2018;51:123-132. 17. Guillerey C, Smyth MJ. Cancer Immunosurveillance by natural killer cells and other innate lymphoid cells. In: Zitvogel L, Kroemer G, eds. Oncoimmunology: A Practical Guide for Cancer Immunotherapy. Cham: Springer International Publishing, 2018:163-180. 18. Ruggeri L, Capanni M, Casucci M, et al. Role of natural killer cell alloreactivity in HLAmismatched hematopoietic stem cell transplantation. Blood. 1999;94(1):333-339. 19. Ruggeri L, Capanni M, Urbani E, et al. Effectiveness of donor natural killer cell alloreactivity in mismatched hematopoietic transplants. Science. 2002;295(5562):20972100. 20. Ruggeri L, Mancusi A, Capanni M, et al. Donor natural killer cell allorecognition of missing self in haploidentical hematopoietic transplantation for acute myeloid leukemia: challenging its predictive value. Blood. 2007;110(1):433-440. 21. Howlader N, Noone A, Krapcho M, et al. SEER Cancer Statistics Review, 1975-2016, National Cancer Institute. Bethesda, MD.

Available from https://seer.cancer.gov/ csr/1975_2016/, based on November 2018 SEER data submission, posted to the SEER web site, April 2019. 22. Carlsten M, Norell H, Bryceson YT, et al. Primary human tumor cells expressing CD155 impair tumor targeting by downregulating DNAM-1 on NK cells. J Immunol. 2009;183(8):4921-4930. 23. Mollenhauer HH, Morre DJ, Rowe LD. Alteration of intracellular traffic by monensin; mechanism, specificity and relationship to toxicity. Biochim Biophys Acta. 1990;1031(2):225-246. 24. Tartakoff AM. Perturbation of vesicular traffic with the carboxylic ionophore monensin. Cell. 1983;32(4):1026-1028. 25. Fujiwara T, Oda K, Yokota S, Takatsuki A, Ikehara Y. Brefeldin A causes disassembly of the Golgi complex and accumulation of secretory proteins in the endoplasmic reticulum. J Biol Chem. 1988;263(34):1854518552. 26. Klausner RD, Donaldson JG, LippincottSchwartz J. Brefeldin A: insights into the control of membrane traffic and organelle structure. J Cell Biol. 1992;116(5):1071-1080. 27. Nguyen S, Dhedin N, Vernant JP, et al. NKcell reconstitution after haploidentical hematopoietic stem-cell transplantations: immaturity of NK cells and inhibitory effect of NKG2A override GvL effect. Blood. 2005;105(10):4135-4142. 28. Mancusi A, Ruggeri L, Velardi A. Haploidentical hematopoietic transplantation for the cure of leukemia: from its biology to clinical translation. Blood. 2016;128 (23):2616-2623. 29. Muntasell A, Ochoa MC, Cordeiro L, et al. Targeting NK-cell checkpoints for cancer immunotherapy. Curr Opin Immunol. 2017;45:73-81. 30. Carlsten M, Jaras M. Natural killer cells in myeloid malignancies: immune surveillance, NK cell dysfunction, and pharmacological opportunities to bolster the endogenous NK cells. Front Immunol. 2019;10:2357. 31. Sanchez-Correa B, Gayoso I, Bergua JM, et al. Decreased expression of DNAM-1 on NK cells from acute myeloid leukemia patients. Immunol Cell Biol. 2012; 90(1):109-115.

haematologica | 2021; 106(12)

ARTICLE

Chronic Lymphocytic Leukemia

SF3B1-mutated chronic lymphocytic leukemia shows evidence of NOTCH1 pathway activation including CD20 downregulation Federico Pozzo,1 Tamara Bittolo,1 Erika Tissino,1 Filippo Vit,1,2 Elena Vendramini,1 Luca Laurenti,3 Giovanni D’Arena,4 Jacopo Olivieri,5 Gabriele Pozzato,6 Francesco Zaja,6 Annalisa Chiarenza,7 Francesco Di Raimondo,7 Antonella Zucchetto,1 Riccardo Bomben,1 Francesca Maria Rossi,1 Giovanni Del Poeta,8 Michele Dal Bo1 and Valter Gattei1 1

Clinical and Experimental Onco-Hematology Unit, Centro di Riferimento Oncologico di Aviano (CRO) IRCCS, Aviano; 2Department of Life Science, University of Trieste, Trieste; 3 Hematology Institute, Catholic University of the Sacred Heart, Fondazione Policlinico A. Gemelli, Rome; 4Hematology Service, S. Luca Hospital, Vallo Della Lucania; 5Clinica Ematologica, Centro Trapianti e Terapie Cellulari "Carlo Melzi", Azienda Sanitaria Universitaria Integrata di Udine, Udine; 6Department of Internal Medicine and Hematology, Maggiore General Hospital, University of Trieste, Trieste; 7Division of Hematology, Ferrarotto Hospital, Catania and 8Division of Hematology, S. Eugenio Hospital and University of Tor Vergata, Rome, Italy

Ferrata Storti Foundation

Haematologica 2021 Volume 106(12):3125-3135

ABSTRACT

hronic lymphocytic leukemia (CLL) is characterized by low CD20 expression, in part explained by an epigenetic-driven downregulation triggered by mutations of the NOTCH1 gene. In the present study, by taking advantage of a wide and well-characterized CLL cohort (n=537), we demonstrate that CD20 expression is downregulated in SF3B1-mutated CLL to an extent similar to NOTCH1-mutated CLL. In fact, SF3B1-mutated CLL cells show common features with NOTCH1mutated CLL cells, including a gene expression profile enriched in NOTCH1-related gene sets and elevated expression of the active intracytoplasmic NOTCH1. Activation of the NOTCH1 signaling and downregulation of surface CD20 in SF3B1-mutated CLL cells correlate with overexpression of an alternatively spliced form of DVL2, a component of the Wnt pathway and negative regulator of the NOTCH1 pathway. These findings were confirmed by separately analyzing the CD20dim and CD20bright cell fractions from SF3B1-mutated cases as well as by DVL2 knockout experiments in CLL-like cell models. Together, the clinical and biological features that characterize NOTCH1-mutated CLL may also be recapitulated in SF3B1-mutated CLL, contributing to explain the poor prognosis of this CLL subset and providing the rationale for expanding therapies based on novel agents to SF3B1-mutated CLL.

Correspondence: FEDERICO POZZO federico.pozzo@cro.it VALTER GATTEI vgattei@cro.it Received: June 11, 2020. Accepted: October 6, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.261891

Introduction Chronic lymphocytic leukemia (CLL) is characterized by pronounced clinical and biological heterogeneity.1 Therapy regimens that include monoclonal antibodies against CD20 are widely used, both as single agents and in combination2,3 although, in specific CLL subsets, the efficacy of anti-CD20 therapy may be reduced by the peculiar dimmer expression of CD20 in CLL4-6 compared to other lymphoproliferative disorders.7 A number of factors have been variously associated with CD20 regulation, e.g., NF-kB signaling, the CXCR4 pathway, B-cell receptor (BCR) signaling, histone deacetylases and activity of multiple transcription factors (such as IRF4, NF-kB, PU.1, OCT1/2);8 in addition, CD20 expression in CLL can be also affected by mutations of the NOTCH1 gene whose presence, detected in up to 25% of cases,9 has been associated with clinical resistance to anti-CD20 immunotherapy both in clinical trials and real-world scenarios.10-12 We previously demonstrated that the

haematologica | 2021; 106(12)

3125

F. Pozzo et al.

reduction of CD20 expression in NOTCH1-mutated CLL cells could be due to a NOTCH1 mutation-driven epigenetic dysregulation involving histone deacetylases.5,13 Although genetic mutations represent the main contributor to the aberrant activation of the NOTCH1 pathway in CLL,14 a mutation-independent activation was also reported for a significant fraction of NOTCH1-unmutated cases.15 In that setting, however, nothing is known regarding the mechanism(s) behind the activation of NOTCH1 and the concomitant modulation of CD20 expression. Another recurrently mutated gene in CLL is the RNA splicing factor 3b subunit 1 (SF3B1), found mutated in about 10% CLL cases.16-19 SF3B1 is a key component of the splicing machinery, responsible of recognizing the branch-point sequences in proximity of the 3′ splice site (acceptor site) allowing intron removal from precursor messenger RNA. Mutations are predicted to alter the protein’s tertiary structure, hampering the correct high-affinity recognition of the branch-point sequences and resulting in the selection of alternative 3ʹ splice sites. This leads to aberrantly spliced transcripts, gain/change/loss-of-function variants, novel stop codons or downregulation of gene expression through nonsense-mediated decay.20-23 In CLL, mutations of SF3B1 have been shown to induce transcriptome-wide alterations of splicing patterns, resulting in an increased frequency of alternative 3′ splice site selection, with functional consequences on several pathways such as DNA damage, telomere maintenance and, possibly, NOTCH1 signaling,18,19,23,24 although the actual impact of SF3B1 mutations in the pathobiology of CLL remains to be fully elucidated. In the present study, by taking advantage of a large cohort of primary CLL cases, we demonstrated that SF3B1-mutated CLL have features of NOTCH1 pathway activation and NOTCH1-dependent CD20 downregulation.

mutated for NOTCH1 and SF3B1 if the variant allele frequency (VAF) exceeded 1%9 or 5%, respectively; in the presence of concurrent mutations, cases were assigned to the SF3B1-mutated category.

Methods

Results

Primary chronic lymphocytic leukemia cells

SF3B1 and NOTCH1 mutational status in chronic lymphocytic leukemia

This study is part of a comprehensive CLL characterization approved by the Internal Review Board of the Centro di Riferimento Oncologico di Aviano (approval n. IRB-05-2010 and IRB-05-2015) upon informed consent in accordance with the Declaration of Helsinki, and included peripheral blood samples from 537 CLL patients (Online Supplementary Table S1).1,25 Primary CLL cells were separated by Ficoll-Hypaque (GE Healthcare, Uppsala, Sweden) density gradient centrifugation and either used directly or cryopreserved until use. All studies were performed on highly purified cells (>95% pure). CLL cases were characterized for IGHV mutational status, the main cytogenetic abnormalities, CD49d expression and mutational status of TP53 as described elsewhere.26,27 In 382/537 cases, time-to-first-treatment (TTFT) data were available (median TTFT 30 months, 95% confidence interval: 27-34 months).

Next-generation sequencing Mutational status of NOTCH1 and SF3B1 was assessed by next-generation sequencing. An amplicon-based strategy, with at least 2000X coverage, was used for NOTCH1, covering the whole of exon 34 and part of the 3′ untranslated region.9 SF3B1 mutational status was assessed using the Nextera XT DNA Library Prep Kit (Illumina, San Diego, CA, USA) covering exons 10 to 17. For the purpose of the present study, CLL samples were considered

3126

Evaluation of CD20 expression CD20 expression was investigated in the whole cohort by flow cytometry, using an anti-CD20 PE-Cy7 antibody (clone L27, BD Biosciences, Milan, Italy) in the neoplastic CD19+/CD5+ and in the normal CD19+/CD5- compartments. Samples were acquired on a FACSCanto II cell analyzer calibrated with CS&T calibration beads (BD Biosciences) and processed with FACSDiva (BD Biosciences) or FlowJo software (FlowJo LLC, Ashland, OR, USA). A complementary analysis of CD20 distribution was implemented as follows: (i) a linear gate (P1) spanning from the peak value of the histogram and comprising the whole CD20bright population was defined; and (ii) the percentage of the P1-gated population was subtracted from the residual (100-%P1) population.28

Gene expression profiling For gene expression profiling (GEP) experiments, total RNA was labeled, hybridized on oligonucleotide microarray slides (SurePrint G3 Human GE v2 8x60K) and analyzed as previously described.29 Microarray data are available at the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE137024.

Statistical analyses Statistical analyses were performed using Medcalc (Medcalc Software Ltd., Ostend, Belgium) and R (www.r-project.org). Data were compared using a two-sided Mann-Whitney rank-test and are presented with Tukey box-and-whisker plots. For correlation analyses, the Spearman rho correlation coefficient was employed. A P-value smaller than 0.05 was considered statistically significant and is represented with asterisks: *P≤0.05, **P≤0.01, ***P≤0.001. Further details are provided in the Online Supplementary Methods.

In a cohort of 537 unselected CLL cases, we investigated the mutational status of SF3B1 and NOTCH1 by nextgeneration sequencing. SF3B1 mutations were present in 48/537 cases (8.9%), with a VAF ranging from 5.0 to 53.0% (mean 32.0%) (Online Supplementary Table S1) and all clustered in the previously reported canonical hotspots.16,19,23 The residues surrounding Lys700 were the most frequently mutated (23/48, 48%), followed by Gly742 (13/48, 27%) (Figure 1A). Multiple mutations were rare, with one case having two concomitant mutations and one case having three of them (cases #3647 and #3301, respectively) (Online Supplementary Table S1). SF3B1 mutations were associated with TP53 mutations (c2 P=0.0313) and with unmutated IGHV status (P=0.0006), but not with specific IGHV alleles. At difference from previous reports,30,31 SF3B1 mutations did not show significant associations with classical chromosomal abnormalities (Online Supplementary Table S2). NOTCH1 mutations were present in 89/537 cases (16.6%) and concomitant mutations were more common, with 19 cases (19/89, 21.0%) bearing more than one lesion, increasing the overall mutation frequency to 21.6% (116 mutations in 537 cases). The VAF ranged from 1.6 to haematologica | 2021; 106(12)

SF3B1 mutations and NOTCH1 activation in CLL

84.0% (mean 33.6%) (Online Supplementary Table S1) with 65 canonical c.7541-7542delCT mutations, 35 other truncating mutations (nonsense or frameshift), 12 mutations in the 3′-untranslated region and four missense mutations (Figure 1B). Co-occurrence of NOTCH1 and SF3B1 mutations was infrequent, being detected in only six cases

(1.1%) (Online Supplementary Table S1), suggesting two independent events (c2 P=0.4268) (Online Supplementary Table S2). The remaining 406 cases were unmutated in both genes (wild-type, WT). TTFT data were available for 382/537 cases: the TTFT intervals for SF3B1-mutated and NOTCH1-mutated cases

Figure 1. Distributions of mutations in the SF3B1 and NOTCH1 genes and impact on time to first treatment. (A) Distribution of mutations of the SF3B1 gene. The mutated amino acids are indicated and grouped by hotspot. Symbols indicate the different type and number of mutations. The pie chart represents the mutation frequency of each hotspot. (B) Distribution of mutations of the NOTCH1 gene. The mutated amino acids are indicated. Symbols indicate the different type and number of mutations. The pie chart represents the frequency of each mutation type. delCT: c.7541-7542delCT; Ns/Fs: nonsense or frameshift; 3′UTR: 3′ untranslated region; Ms: missense (C) Kaplan-Meier survival analysis for time to first treatment in 382 cases of chronic lymphocytic leukemia with mutations of SF3B1 (n=35), NOTCH1 (n=53) or neither (WT, n=294; P<0.0001 for both comparisons, log-rank test).

haematologica | 2021; 106(12)

3127

F. Pozzo et al.

were similar and significantly shorter than those of cases lacking all these mutations (P≤0.0001 in all pairwise comparisons with WT) (Figure 1C), in keeping with previously reported studies,11,16,19,32 irrespective of IGHV status (Online Supplementary Figure S1A) or concomitant SF3B1/NOTCH1 mutations (Online Supplementary Figure S1B).

SF3B1 and NOTCH1 showed a trend for a progressive expansion of the CD20dim fraction (median 19.3; P=0.0466 vs. WT) (Online Supplementary Figure S2F). Although excluded upfront from the analyses, the lower expression of CD20 was also confirmed in the trisomy 12 subset in NOTCH1-mutated CLL5,13 and with a trend in the context of the few SF3B1-mutated cases (Online Supplementary Figure S2G).

SF3B1-mutated chronic lymphocytic leukemia displays lower expression of CD20 than wild-type cases and similar to NOTCH1-mutated cases

SF3B1-mutated cases show evidence of activation of the NOTCH1 pathway

Low CD20 expression represents a hallmark of CLL compared to other lymphoproliferative disorders or normal B cells (Online Supplementary Figure 2A).7 We previously showed that a reduction of CD20 could be due to NOTCH1 mutation-driven epigenetic dysregulation involving histone deacetylases.5,13 To expand this observation, here we investigated whether low CD20 expression was also associated with SF3B1 mutations in CLL, possibly via a non-mutational activation of the NOTCH1 pathway.15 In keeping with previous observations,5,33 in the whole cohort of 537 cases, CLL bearing trisomy 12 (121/537 cases, 22.5%) had significantly higher expression of CD20 (P<0.0001) (Online Supplementary Figure S2A) and a strong association with NOTCH1 mutations (36/121, 29.8%, c2 P<0.0001); conversely, SF3B1 mutations were infrequent (6/121, 5.0%, c2 P=0.0815) (Online Supplementary Figure S2B, Online Supplementary Table S2). Therefore, to avoid confounding effects, for the main subsequent analyses we considered only cases without evidence of trisomy 12 (416/537 cases), irrespectively of other concomitant chromosomal abnormalities. CD20 expression, determined by mean fluorescence intensity, was lower in SF3B1-mutated cases than in WT cases (SF3B1-mutated, 42 cases, median 7396; WT, 327 cases, median 9538; P=0.0024) (Figure 2A), without differences from NOTCH1-mutated cases 47 cases, median 8438; P=0.3748) (Figure 2A), whose CD20 expression was lower than that of WT cases (P=0.0248), as reported previously.5,13 CD20 expression is known to be heterogeneous and dispersed in CLL, often spanning a 2- to 3-log range of fluorescence intensity within the CD19+/CD5+ pathological population.34,35 We, therefore, implemented a complementary strategy of analysis, evaluating the percentage of the “CD20dim fraction”, defined as the differential population fraction between the two sides of the fluorescence distribution with respect to the mode (Online Supplementary Figure S2C).28 As a statistically robust measure of skewness, this method allowed us to discriminate cases with homogeneous or heterogeneous CD20 expression (Online Supplementary Figure S2D), with an inverse correlation with mean fluorescence intensity (rho= -0.417, P<0.001) (Figure 2B) and was consistent over time in sequential blood samples of treatment-naïve CLL cases (Online Supplementary Figure S2E). As shown in Figure 2C, the magnitude of the expansion of the CD20dim fraction was clearly increased in CLL cases with mutations of SF3B1 (median 15.5) or NOTCH1 (median 13.4) compared to WT cases (median 8.2; P<0.001 for both comparisons) while there were no evident differences between the SF3B1 mutational hotspots (Figure 2C, right panel). Again, no difference was found between NOTCH1-mutated and SF3B1-mutated cases (P=0.6365). Of note, when analyzed separately, the six cases with concomitant mutations of

To evaluate whether activation of the NOTCH1 pathway also occurs in primary SF3B1-mutated CLL, we performed GEP on 28 cases, 13 WT versus nine SF3B1-mutated (VAF: range 21-48%, mean 37%) or six NOTCH1-mutated (VAF: range 17-51%, mean 30%) cases, all with unmutated IGHV and devoid of trisomy 12. In the SF3B1-mutated category, 585 array probes (502 upregulated and 83 downregulated) corresponding to 443 known genes (402 upregulated and 41 downregulated) were differentially expressed (Online Supplementary Table S3) compared to WT samples. On the other hand, when comparing the NOTCH1-mutated and the WT categories, we identified 2,097 differentially expressed array probes (1,147 upregulated and 950 downregulated) (Online Supplementary Table S4), corresponding to 1,274 known genes (840 upregulated and 434 downregulated). When this NOTCH1 gene signature was applied to all samples, SF3B1-mutated cases clustered with the NOTCH1-mutated cases (Online Supplementary Figure S3A), suggesting the presence of a common underlying signature. In fact, by merging the 443 SF3B1-mutated differentially expressed genes with the 1,274 NOTCH1-mutated differentially expressed genes, 419/443 (94.6%) were shared and concordantly regulated in the two signatures (390 upregulated and 29 downregulated) (Figure 3A, B). Consistently, when the SF3B1 signature was applied to all samples, NOTCH1-mutated cases clustered together with SF3B1-mutated cases (Online Supplementary Figure S3B). GEP data were externally validated by quantitative reverse transcriptase polymerase reaction (RT-qPCR), selecting four genes related to the NOTCH1 pathway (CD300A, IL1R2, HEY1, HES4) (Online Supplementary Figure S3C). Although in these signatures the differential expression of the single probe dedicated to MS4A1 (the gene encoding CD20) in our GEP platform (see Online Supplementary Methods) was not statistically significant upon correction for the false discovery rate (Online Supplementary Tables S3 and S4), a re-evaluation of MS4A1 transcript expression by RT-qPCR confirmed its significantly lower levels in NOTCH1-mutated versus WT cases (P=0.001) and in SF3B1-mutated versus WT cases (P<0.001) (Online Supplementary Figure S3D). Gene set enrichment analysis of NOTCH1-mutated versus WT cases revealed a significant enrichment in 11 out of 17 NOTCH1-related datasets (Online Supplementary Table S5) linked to NOTCH1 transcriptional activity and overexpression (Figure 3C, Online Supplementary Figure S4A). Interestingly, one of the most significantly enriched datasets was a recently published CLL-specific NOTCH1 signature.15 The gene set enrichment analysis, using the same 17 NOTCH1-related datasets, when repeated on SF3B1-mutated versus WT cases, identified ten datasets as significantly enriched (Figure 3D, Online Supplementary Figure S4B, Online Supplementary Table S5), including the

3128

haematologica | 2021; 106(12)

SF3B1 mutations and NOTCH1 activation in CLL

CLL-specific NOTCH1 signature,15 again suggesting the presence of an active NOTCH1 pathway; it is noteworthy that all ten datasets were included in the 11 datasets enriched in the NOTCH1-mutated cases. Finally, using western blotting, we evaluated the expression of the NOTCH1 intracytoplasmic domain (NICD), the cleaved active form of NOTCH1 in whole cell lysates of 34 CLL cases, of which 13 SF3B1-mutated among those with the highest mutational burden and seven NOTCH1-mutated cases. We detected a distinct NICD band in all CLL cases with mutations of NOTCH1, while the other samples showed a varying degree of NICD staining, in agreement with previous observations of NOTCH1 pathway activation independent of NOTCH1 mutations.15 To determine a cutoff for NICD positivity, we defined the threshold as the lowest densitometric intensity among NOTCH1-mutated samples (Figure 3E, dashed line). In agreement with GEP data, positive NICD staining could be detected in the majority (8/13) of SF3B1-mutated cases but in only 2/15 WT cases (Online Supplementary Figure S5A), in keeping with observations of a mutation-independent activation of the NOTCH1 pathway in CLL.15 The NICD intensity correlated positively with the mutational burden for both NOTCH1 and SF3B1 mutations further suggesting, for these events, a modulating role on NOTCH1 signaling (Figure 3E) generally more elevated in NICD-positive samples, as determined by RT-

qPCR of the NOTCH1 target genes DTX1 and CD300A (Online Supplementary Figure S5B).

SF3B1 mutations consistently induce alternative splicing of DVL2 In CLL, mutations of SF3B1 have been shown to induce transcriptome-wide splicing alterations in several genes including DVL2, a key component of the Wnt pathway, reported to act as a negative regulator of NOTCH1.36-40 Using next-generation sequencing in 73 primary CLL cases, we first evaluated SF3B1-induced splicing alterations of DVL2 (Figure 4A) and, as a control, two other highly differentially spliced genes, namely GCC2 and MAP3K7 (Online Supplementary Figure S6A).23 For all three genes, the exact reported splicing defects were highly enriched in all SF3B1-mutated cases (P<0.0001) (Online Supplementary Figure S6B) confirming that these alterations are highly consistent both within and between cohorts. The expression of alternatively spliced DVL2 (altDVL2 hereafter) was then further investigated by RT-qPCR in a wider cohort of 222/537 CLL cases with available RNA, 35 of which with a SF3B1 mutation, 32 with a NOTCH1 mutation and 155 WT (Figure 4B). RT-qPCR results were highly concordant with next-generation sequencing data (rho=0.827, P<0.0001) (Online Supplementary Figure S6C). Expression of altDVL2 was significantly increased in the presence of SF3B1 mutations (P<0.001) (Figure 4B) and

Figure 2. CD20 expression is reduced in chronic lymphocytic leukemia cases with SF3B1 mutations. (A) Expression of CD20 determined by flow cytometry in SF3B1-mutated (SF3B1, n=42), NOTCH1-mutated (NOTCH1, n=47) or unmutated (WT, n=327) cases of chronic lymphocytic leukemia (CLL) without trisomy 12. MFI: mean fluorescence intensity. (B) Correlation between CD20 MFI and the CD20dim fraction in the whole cohort of non-trisomy 12 cases (n=416). The Spearman rank correlation coefficient (rho) and P-value are reported. (C) Percentage of the CD20dim fraction in SF3B1-mutated (SF3B1, n=42), NOTCH1-mutated (NOTCH1, n=47) or unmutated (WT, n=327) CLL cases. Right panel: percentage of the CD20dim fraction in SF3B1-mutated cases with respect to the mutated hotspot. Data are shown by Tukey box and whisker plots. Outliers indicate data outside the 1.5 interquartile range. *P≤0.05, **P≤0.01, ***P≤0.001, n.s. not significant, as determined by a two-sided Mann-Whitney test.

haematologica | 2021; 106(12)

3129

F. Pozzo et al.

highly correlated with the mutational burden (rho=0.694, P<0.0001) (Figure 4C). This analysis also evidenced that the increase in altDVL2 was independent of the type of SF3B1 mutation, in terms of both mutational hotspot (Figure 4D) or amino-acid residue affected (Online Supplementary Figure S6D). Conversely, there was no significant difference between WT cases and cases bearing SF3B1 or NOTCH1 mutations in the expression of total DVL2 (Online Supplementary Figure S6E). Of note, we detected a low frequency of alternative splicing for DVL2 and the other genes also in WT cases (Online Supplementary Figure S6B, right inset), suggesting that the

3130

altDVL2 form may be an underrepresented native splicing variation, particularly favored by a splicing machinery containing the mutated form of SF3B1.

Alternative splicing of DVL2 correlates with NOTCH1 signaling and CD20 expression We then investigated whether altDVL2 could influence NOTCH1 signaling differently from wild-type DVL2. To do so, we overexpressed an exogenous DVL2, either wildtype or alternate, into a HEK293 NOTCH1-dependent luciferase reporter cell line; in this context, the transfection of wild-type DVL2 but not of altDVL2 was able to

Figure 3. The NOTCH1 pathway is activated in SF3B1-mutated cases of chronic lymphocytic leukemia. (A) Venn diagram of the 1,274 differentially expressed known genes in NOTCH1-mutated (n=6) versus wild-type (WT) (n=13) cases and the 443 differentially expressed known genes in SF3B1mutated (n=9) versus WT (n=13) cases. (B) Correlation of the fold change of 419 probes shared between the NOTCH1-mutated and the SF3B1-mutated signatures. The Spearman rank correlation coefficient (rho) and P-value are reported. (C, D) Gene set enrichment analysis enrichment plots tracking the HALLMARK_NOTCH_SIGNALING gene set, significantly enriched in NOTCH1mutated (C) and SF3B1-mutated (D) chronic lymphocytic leukemia. NES: normalized enrichment score. (E) Correlation between NOTCH1 intracytoplasmic domain (NICD) expression, in arbitrary units, and SF3B1 or NOTCH1 variant allele frequency; the solid line represents the correlation within the SF3B1mutated cases with the Spearman rank correlation coefficient (rho) and Pvalue reported; the dashed line represents the limit for NICD positivity, defined as the lowest NICD intensity within the NOTCH1-mutated cases (see also Online Supplementary Figure S5A). a.u.; arbitrary units; VAF: variant allele frequency.

haematologica | 2021; 106(12)

SF3B1 mutations and NOTCH1 activation in CLL

repress the promoter activity (P=0.0244 and P=0.96, respectively) (Online Supplementary Figure S7A), in agreement with previous observations23 and consistent with a defective inhibitory role of altDVL2 against the NOTCH1dependent transcriptional machinery. Similarly, after depletion of wild-type DVL2 by siRNA transfection in the SF3B1-wild-type CLL-like MEC1 cell line, NOTCH1 signaling was found to be increased, as demonstrated by HES1 upregulation (P=0.0176) at 24 h; in addition, CD20 expression was reduced at both protein (P=0.0062) and transcript (P=0.0129) levels (Figure 5A). Taken together, these data corroborate the inhibitory role of DVL2 on the NOTCH1 pathway, counteracted by the presence of its alternate splicing isoform. In agreement with this observation, SF3B1-mutated CLL cases also showed high expression of the NOTCH1 target gene DTX1 (median SF3B1-mutated 0.0029, median WT 0.0013, P=0.0089; rho=0.369, P=0.0348), similar to NOTCH1-mutated cases (median NOTCH1-mutated 0.0033, median WT 0.0013, P=0.002) (Figure 5B). Notably,

in SF3B1-mutated cases, the levels of DTX1 correlated directly with the levels of altDVL2, in keeping with the hypothesis of elevated NOTCH1 signaling in the presence of a less functioning alternatively spliced DVL2 (Figure 5C). In turn, the expression of DTX1 correlated with reduced CD20 expression; as summarized in Figure 5D and Online Supplementary Figure S7B. SF3B1-mutated and NOTCH1-mutated CLL cases showed a progressively enriched double-positive CD20dim/DTX1 population compared to WT cases (WT vs. SF3B1-mutated, P=0.024; WT vs. NOTCH1-mutated, P<0.001; SF3B1-mutated vs. NOTCH1-mutated, P=0.362; c2 test for proportions) (Online Supplementary Figure S7B). Accordingly, by performing cell sorting on nine SF3B1-mutated CLL cases, all with clonal SF3B1 mutational burden (VAF 10-50%), we isolated two populations characterized by either high or low CD20 protein and transcript expression. Evaluation of altDVL2 by next-generation sequencing in the two subfractions revealed an increased frequency of the alterna-

Figure 4. SF3B1 mutations induce alternative splicing of DVL2. (A) Representation of the DVL2 splicing events determine from next-generation sequencing analysis; arches indicate wild-type (blue) and altered (red) splicing, black bars indicate coverage; the position of primers and the probe for quantitative reverse transcriptase polymerase chain reaction analysis is also reported. (B) Expression of DVL2 alternative splicing (altDVL2) in unmutated (WT, n=155), SF3B1-mutated (SF3B1, n=35), or NOTCH1-mutated (NOTCH1, n=32) cases of chronic lymphocytic leukemia. Data are shown by Tukey box and whisker plots. Outliers indicate data outside the 1.5 interquartile range. ***P≤0.001, n.s. not significant, as determined by a two-sided Mann-Whitney rank-test. (C) Correlation of altDVL2 expression with SF3B1 mutational burden. VAF: variant allele frequency. The Spearman rank correlation coefficient (rho) and P-value are reported. (D) altDVL2 expression in relation to SF3B1 mutations according to the mutational hotspot and burden (n=35). WT cases (n=155) are reported for comparison. Diamonds indicate SF3B1-mutated cases with a VAF <10%. Horizontal marks represent the median.

haematologica | 2021; 106(12)

3131

F. Pozzo et al.

tively spliced isoform in the CD20low fraction compared to the CD20high fraction in most cases (P=0.0078), paralleled by a higher expression of DTX1 (P=0.05). Finally, when challenged in a rituximab-driven complement-dependent cytotoxicity assay, CLL cells from either SF3B1-mutated or NOTCH1-mutated samples displayed a lower relative lysis compared to cells from WT samples (Online Supplementary Figure S7D), in agreement with the dose-dependent response of the assay from CD20 expression levels (rho=0.745, P=0.0085).

Discussion The CD20 molecule is one of the preferred therapeutic targets for CLL as testified by the widespread use of humanized anti-CD20 antibodies in different therapeutic regimens,2,41 although a dim expression of CD20 is among the peculiar features of CLL and part of the so-called Royal Marsden score for the differential diagnosis of CLL from other B-cell lymphoproliferative disorders.25,42 This notwithstanding, very little is still known about the biological instances that define CD20 expression and determine the stark difference between CLL cells and normal B cells, despite a number of mechanisms having been proposed. Among the many, a few stand out including BCR/CXCR4 signaling,43 NF-kB signaling and epigenetic/transcription factors such as histone deacetylases.8,44 Another of these factors is the presence of mutations of the NOTCH1 gene,5 an emerging novel predictive factor for response to anti-CD20 immunotherapy in patients treated with immuno-chemotherapeutic combinations.11,12 In the present study, by taking advantage of a large cohort of CLL cases, we show that mutations of the SF3B1 gene represent another factor associated with a reduction of CD20 expression through a mutation-independent activation of the NOTCH1 pathway. In our cohort, the mutational profile of SF3B1 was consistent with published data, with every mutation falling inside the canonical hotpots.16-19 SF3B1 mutations are consistently associated with an increased rate of aberrant splicing events throughout the transcriptome with varying intensity.23 By analyzing the relative abundance of different splicing variants, we invariably detected a strong increase of aberrantly spliced transcripts on multiple genes in SF3B1-mutated cases, independently of the location of the mutated residue and with a high correlation with the mutational burden. One of these genes was DVL2, whose alternatively spliced form, altDVL2, consisted in an inframe loss of 24 amino acids within exon 11. DVL2 is a key mediator of the Wnt pathway and has demonstrated the ability to act as a negative regulator of the NOTCH1 pathway (Online Supplementary Figure S6A) by binding the NOTCH1-related transcription factor RBPJ.39 and/or the cleaved active form of NOTCH1 (NICD) itself.38 Through co-transfection experiments of exogenous DVL2 in a NOTCH1/RBPJ-dependent luciferase reporter system, we could confirm that the in vitro induction of DVL2 is capable of silencing NOTCH1 signaling, whereas the induction of altDVL2 did not result in proficient silencing, as previously suggested.23 Here, we moved further to primary CLL cells and, using GEP, showed that SF3B1-mutated cases share a gene signature with NOTCH1-mutated cases which drove an unsupervised co3132

clustering of the two categories, with respect to NOTCH1/SF3B1-unmutated cases. In addition, several NOTCH1-specific gene sets were enriched in SF3B1-mutated CLL, further suggesting an active commitment of the NOTCH1 transcriptional machinery in this CLL subset; importantly, one of these gene sets was a custom gene set derived from a CLL-specific NOTCH1 gene signature.15 Consistently, the NICD was clearly detectable in protein lysates of SF3B1-mutated cases, with expression levels often comparable to those observed in NOTCH1-mutated cases. The increased NOTCH1 signaling positively correlated with the expression of altDVL2 and with an elevated CD20dim fraction in SF3B1-mutated CLL cases. These data were corroborated by cell sorting experiments of cell fractions with different CD20 expression levels in the context of SF3B1-mutated CLL cells, and by functional experiments of DVL2 knock-out in a SF3B1-wild-type CLL cell model. While the NOTCH1 pathway has for long been recognized as active in CLL,14 only few mechanisms are known to explain its activation state. NOTCH1 mutations represent the main but not the sole contributor, as there is evidence of active NOTCH1 signaling occurring in peripheral blood cells of a fraction of NOTCH1-unmutated CLL cases,15 although without a molecular explanation. Our present study may help to sort out at least a fraction of these cases, suggesting that the constitutively activated NOTCH1 pathway can indeed occur in the context of SF3B1-mutated CLL, possibly as the result of diminished repression of the dynamic association between NOTCH1 and RBPJ45 due to higher levels of the inefficient spliced form of DVL2.15,38,39 Wang et al.23 speculated that the numerous changes induced by SF3B1 mutations in the CLL transcriptome may allow neoplastic cells to diversify their evolutionary capacity, most likely through subtle alterations of many gene transcripts rather than through a single fatal lesion. This was hinted by the fact that SF3B1 mutations, usually a later event in CLL evolution,46 may not induce per se neoplastic transformation in B cells but rather drive a more aggressive and adaptive phenotype. This line of reasoning is in keeping with the hypothesis that the activation of the NOTCH1 pathway in CLL may reflect deregulated expression of a physiological signal, required for B lymphocyte maturation and differentiation.15,47 The exploitation of the NOTCH1 pathway by SF3B1mutated CLL cells may well reflect such a strategy by which the proliferative advantages that characterize NOTCH1-mutated CLL cells can be incorporated in the complex transcriptomic reshaping occurring in SF3B1-mutated CLL, thus contributing to explain the poor prognosis of this CLL subset.11,16,19,32 Also in keeping with the data presented here, it is likely that the documented activation of NOTCH1 in SF3B1-mutated CLL was primarily mediated by the direct involvement of the NICD through the canonical pathway.45 However, the hypothesis of a NICD-independent activation cannot be excluded firsthand, as Notch-independent activity of the transcription factor RBPJ has been documented.48 From a clinical standpoint, the biological similarity of NOTCH1- and SF3B1-mutated CLL could contribute to explain the poor response of these CLL subsets to chemoimmunotherapy.49,50 As documented by the CLL8 trial for NOTCH1-mutated cases, the clinical behavior of SF3B1-mutated CLL was worse that that of WT CLL, haematologica | 2021; 106(12)

SF3B1 mutations and NOTCH1 activation in CLL

Figure 5. NOTCH1 signaling and DVL2 silencing correlate with reduced CD20 expression. (A) Expression of DVL2, HES1, MS4A1 determined by quantitative reverse transcriptase polymerase chain reaction (RT-qPCR) analysis and surface CD20 determined by flow cytometry in MEC1 cells, transfected with negative control (Neg ctrl, n=5) or siRNA for DVL2 (n=5). Data are shown by Tukey box and whisker plots. MFI: mean fluorescence intensity. (B) DTX1 expression in wild-type (WT; n=155), SF3B1-mutated (n=35) or NOTCH1-mutated (n=32) cases of chronic lymphocytic leukemia (CLL). (C) Correlation between altDVL2 expression and DTX1 expression in SF3B1-mutated (n=35) cases, as determined by RT-qPCR analysis. The Spearman rank correlation coefficient (rho) and P-value are reported for SF3B1-mutated cases. WT cases (n=115) are reported for comparison. (D) Correlation between DTX1 expression determined by RT-qPCR and the CD20dim fraction. Density plots (log10 density) show clustering of SF3B1-mutated and NOTCH1-mutated cases as double positive DTX1high/%CD20dim (see also Online Supplementary Figure S7A). (E) Left panel: representative dot plot showing the gating strategy for CD20 expression in the CD19+/CD5+ CLL population; right panel: evaluation of altDVL2 by nextgeneration sequencing or DTX1 expression by RT-qPCR in the CD20high and CD20low subpopulations of sorted samples (n=9). Data are shown as dot-and-line diagrams of the fold increase over the CD20high fraction. *P≤0.05, **P≤0.01, ***P≤0.001, n.s. not significant, as determined by a two-sided Mann-Whitney rank-test or paired Wilcoxon signed-rank test.

haematologica | 2021; 106(12)

3133

F. Pozzo et al.

although observed in the context of both FCR (fludarabine, cyclophosphamide and rituximab) and FC (fludarabine and cyclophosphamide) regimens,11 in keeping with the presence of additional mechanism(s) of immunechemo-resistance occurring in SF3B1-mutated CLL cells.23,24 Collectively, our data may provide the biological basis to promote, for SF3B1-mutated CLL patients, the use of novel biological agents, whose activity does not need the addition of anti-CD20 drugs51,52 or of latest-generation anti-CD20 molecules53 which operate independently of specific genetic lesions.54 Disclosures No conflicts of interests to disclose. Contributions FP contributed to design the study, performed the research, analyzed and interpreted the data and wrote the manuscript; TB performed the research; ET, FV, EV, AZ, RB and FMR con-

References 1. Hallek M. Chronic lymphocytic leukemia: 2017 update on diagnosis, risk stratification, and treatment. Am J Hematol. 2017;92(9):946-965. 2. Falchi L, Ferrajoli A, Jacobs I, Nava-Parada P. An evidence-based review of anti-CD20 antibody-containing regimens for the treatment of patients with relapsed or refractory chronic lymphocytic leukemia, diffuse large B-cell lymphoma, or follicular ymphoma. Clin Lymphoma Myeloma Leuk. 2018;18(8):508-518. 3. Brown JR, Cymbalista F, Sharman J, Jacobs I, Nava-Parada P, Mato A. The role of rituximab in chronic lymphocytic leukemia treatment and the potential utility of biosimilars. Oncologist. 2018;23(3):288296. 4. Golay J, Lazzari M, Facchinetti V, et al. CD20 levels determine the in vitro susceptibility to rituximab and complement of Bcell chronic lymphocytic leukemia: further regulation by CD55 and CD59. Blood. 2001;98(12):3383-3389. 5. Pozzo F, Bittolo T, Arruga F, et al. NOTCH1 mutations associate with low CD20 level in chronic lymphocytic leukemia: evidence for a NOTCH1 mutation-driven epigenetic dysregulation. Leukemia. 2016;30(1):182189. 6. van Meerten T, Hagenbeek A. CD20-targeted therapy: a breakthrough in the treatment of non-Hodgkin's lymphoma. Neth J Med. 2009;67(7):251-259. 7. Prevodnik VK, Lavrencak J, Horvat M, Novakovic BJ. The predictive significance of CD20 expression in B-cell lymphomas. Diagn Pathol. 2011;6:33-636. 8. Pavlasova G, Mraz M. The regulation and function of CD20: an "enigma" of B-cell biology and targeted therapy. Haematologica. 2020;105(6):1494-1506. 9. D'Agaro T, Bittolo T, Bravin V, et al. NOTCH1 mutational status in chronic lymphocytic leukaemia: clinical relevance of subclonal mutations and mutation types. Br J Haematol. 2018;182(4):597-602. 10. Dal Bo M, Del Principe MI, Pozzo F, et al. NOTCH1 mutations identify a chronic lymphocytic leukemia patient subset with

3134

tributed to perform the research; LL, GDA, JO, FDR, AC, FZ, GP and GDP provided well characterized biological samples and contributed to writing the manuscript; MDB contributed to design the study; VG designed the study, interpreted the data, wrote and edited the manuscript. Funding MDB received funds via the Progetto Giovani Ricercatori, GR2011-02351370 and Progetto Ricerca Finalizzata PE-201602362756, Ministero della Salute, Rome, Italy. AZ received funds via the Progetto Ricerca Finalizzata RF-2018-12365790, Ministero della Salute, Rome, Italy. VG received funds via the Associazione Italiana Ricerca Cancro (AIRC), Investigator Grant IG-2018 (21687); Linfo-check - Bando ricerca - contributo art. 15, comma 2, lett b) LR 17/2014; Associazione Italiana contro le Leucemie, Linfomi e Mielomi (AIL), Venezia Section, Pramaggiore Group, Italy; and “5x1000 Intramural Program”, Centro di Riferimento Oncologico, Aviano, Italy. EV was supported by the Fondazione Umberto Veronesi, Post-doctoral Fellowships-year 2019

worse prognosis in the setting of a rituximab-based induction and consolidation treatment. Ann Hematol. 2014;93(10):17651774. 11. 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. 12. Tausch E, Beck P, Schlenk RF, et al. Prognostic and predictive role of gene mutations in chronic lymphocytic leukemia: results from the pivotal phase III study COMPLEMENT1. Haematologica. 2020;105(10):2440-2447. 13. Bittolo T, Pozzo F, Bomben R, et al. Mutations in the 3' untranslated region of NOTCH1 are associated with low CD20 expression levels chronic lymphocytic leukemia. Haematologica. 2017; 102(8): e305-e309. 14. Rosati E, Sabatini R, Rampino G, et al. Constitutively activated Notch signaling is involved in survival and apoptosis resistance of B-CLL cells. Blood. 2009; 113(4):856-865. 15. Fabbri G, Holmes AB, Viganotti M, et al. Common nonmutational NOTCH1 activation in chronic lymphocytic leukemia. Proc Natl Acad Sci U S A. 2017;114(14):E2911E2919. 16. Rossi D, Bruscaggin A, Spina V, et al. Mutations of the SF3B1 splicing factor in chronic lymphocytic leukemia: association with progression and fludarabine-refractoriness. Blood. 2011;118(26):6904-6908. 17. Puente XS, Pinyol M, Quesada V, et al. Whole-genome sequencing identifies recurrent mutations in chronic lymphocytic leukaemia 50. Nature. 2011;475(7354):101105. 18. Nadeu F, Clot G, Delgado J, et al. Clinical impact of the subclonal architecture and mutational complexity in chronic lymphocytic leukemia. Leukemia. 2017;32(3):645653. 19. Quesada V, Conde L, Villamor N, et al. Exome sequencing identifies recurrent mutations of the splicing factor SF3B1 gene in chronic lymphocytic leukemia. Nat Genet. 2011;44(1):47-52. 20. Kesarwani AK, Ramirez O, Gupta AK, et al. Cancer-associated SF3B1 mutants recog-

nize otherwise inaccessible cryptic 3' splice sites within RNA secondary structures. Oncogene. 2017;36(8):1123-1133. 21. Alsafadi S, Houy A, Battistella A, Popova T, Wassef M, Henry E, et al. Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage. Nat Commun. 2016;7:10615. 22. Cretu C, Schmitzova J, Ponce-Salvatierra A, et al. Molecular architecture of SF3b and structural consequences of its cancer-related mutations. Mol Cell. 2016;64(2):307319. 23. Wang L, Brooks AN, Fan J, et al. Transcriptomic characterization of SF3B1 mutation reveals its pleiotropic effects in chronic lymphocytic leukemia. Cancer Cell. 2016;30(5):750-763. 24. Te Raa GD, Derks IA, Navrkalova V, et al. The impact of SF3B1 mutations in CLL on the DNA-damage response. Leukemia. 2015;29(5):1133-1142. 25. Matutes E, Owusu-Ankomah K, Morilla R, et al. The immunological profile of B-cell disorders and proposal of a scoring system for the diagnosis of CLL. Leukemia. 1994;8(10):1640-1645. 26. Dal Bo M, Bulian P, Bomben R, et al. CD49d prevails over the novel recurrent mutations as independent prognosticator of overall survival in chronic lymphocytic leukemia. Leukemia. 2016;30(10):20112018. 27. Bulian P, Shanafelt TD, Fegan C, CD49d is the strongest flow cytometry-based predictor of overall survival in chronic lymphocytic leukemia. J Clin Oncol. 2014; 32(9):897-904. 28. Bickel DR. Robust estimators of the mode and skewness of continuous data. Comput Stat Data Anal. 2002;39(2):153-163. 29. Pozzo F, Bittolo T, Vendramini E, et al. NOTCH1-mutated chronic lymphocytic leukemia cells are characterized by a MYCrelated overexpression of nucleophosmin 1 and ribosome-associated components. Leukemia. 2017;31(11):2407-2415. 30. Wang L, Lawrence MS, Wan Y, et al. SF3B1 and other novel cancer genes in chronic lymphocytic leukemia. N Engl J Med. 2011;365(26):2497-2506. 31. Hernandez JA, Hernandez-Sanchez M, Rodriguez-Vicente AE, et al. A low fre-

haematologica | 2021; 106(12)

SF3B1 mutations and NOTCH1 activation in CLL

quency of losses in 11q chromosome is associated with better outcome and lower rate of genomic mutations in patients with chronic lymphocytic leukemia. PLoS One. 2015;10(11):e0143073. 32. Oscier DG, Rose-Zerilli MJ, Winkelmann N, et al. The clinical significance of NOTCH1 and SF3B1 mutations in the UK LRF CLL4 trial. Blood. 2013;121(3):468-475. 33. Tam CS, Otero-Palacios J, Abruzzo LV, et al. Chronic lymphocytic leukaemia CD20 expression is dependent on the genetic subtype: a study of quantitative flow cytometry and fluorescent in-situ hybridization in 510 patients. Br J Haematol. 2008; 141(1):36-40. 34. Almasri NM, Duque RE, Iturraspe J, Everett E, Braylan RC. Reduced expression of CD20 antigen as a characteristic marker for chronic lymphocytic leukemia. Am J Hematol. 1992;40(4):259-263. 35. Marti GE, Faguet G, Bertin P, et al. CD20 and CD5 expression in B-chronic lymphocytic leukemia. Ann N Y Acad Sci. 1992; 651:480-483. 36. Lu D, Zhao Y, Tawatao R, et al. Activation of the Wnt signaling pathway in chronic lymphocytic leukemia. Proc Natl Acad Sci U S A. 2004;101(9):3118-3123. 37. Axelrod JD, Matsuno K, ArtavanisTsakonas S, Perrimon N. Interaction between Wingless and Notch signaling pathways mediated by dishevelled. Science. 1996;271(5257):1826-1832. 38. Munoz-Descalzo S, Sanders PG, Montagne C, Johnson RI, Balayo T, Arias AM. Wingless modulates the ligand independent traffic of Notch through Dishevelled. Fly

haematologica | 2021; 106(12)

(Austin ). 2010;4(3):182-193. 39. Collu GM, Hidalgo-Sastre A, Acar A, et al. Dishevelled limits Notch signalling through inhibition of CSL. Development. 2012; 139(23):4405-4415. 40. Hayward P, Kalmar T, Arias AM. Wnt/Notch signalling and information processing during development. Development. 2008;135(3):411-424. 41. Hallek M. On the architecture of translational research designed to control chronic lymphocytic leukemia. Hematology Am Soc Hematol Educ Program. 2018;(1):1-8. 42. Delgado J, Matutes E, Morilla AM, et al. Diagnostic significance of CD20 and FMC7 expression in B-cell disorders. Am J Clin Pathol. 2003;120(5):754-759. 43. Pavlasova G, Borsky M, Seda V, et al. Ibrutinib inhibits CD20 upregulation on CLL B cells mediated by the CXCR4/SDF-1 axis. Blood. 2016;128(12):1609-1613. 44. Tomita A. Genetic and epigenetic modulation of CD20 expression in B-cell malignancies: molecular mechanisms and significance to rituximab resistance. J Clin Exp Hematop. 2016;56(2):89-99. 45. Castel D, Mourikis P, Bartels SJ, Brinkman AB, Tajbakhsh S, Stunnenberg HG. Dynamic binding of RBPJ is determined by Notch signaling status. Genes Dev. 2013;27(9):1059-1071. 46. Landau DA, Carter SL, Stojanov P, et al. Evolution and impact of subclonal mutations in chronic lymphocytic leukemia. Cell. 2013;152(4):714-726. 47. Arruga F, Vaisitti T, Deaglio S. The NOTCH pathway and its mutations in mature B cell malignancies. Front Oncol. 2018;8:550.

48. Wang H, Zou J, Zhao B, et al. Genomewide analysis reveals conserved and divergent features of Notch1/RBPJ binding in human and murine T-lymphoblastic leukemia cells. Proc Natl Acad Sci U S A. 2011;108(36):14908-14913. 49. Thompson PA, Tam CS, O'Brien SM, et al. Fludarabine, cyclophosphamide, and rituximab treatment achieves long-term diseasefree survival in IGHV-mutated chronic lymphocytic leukemia. Blood. 2016;127(3):303309. 50. Rossi D, Terzi-di-Bergamo L, De Paoli L, et al. Molecular prediction of durable remission after first-line fludarabine-cyclophosphamide-rituximab in chronic lymphocytic leukemia. Blood. 2015;126(16):1921-1924. 51. Burger JA, Sivina M, Jain N, et al. Randomized trial of ibrutinib vs ibrutinib plus rituximab in patients with chronic lymphocytic leukemia. Blood. 2019; 133(10):1011-1019. 52. Ten Hacken E, Valentin R, Regis FFD, et al. Splicing modulation sensitizes chronic lymphocytic leukemia cells to venetoclax by remodeling mitochondrial apoptotic dependencies. JCI Insight. 2018;3(19): e121438. 53. Goede V, Fischer K, Busch R, et al. Obinutuzumab plus chlorambucil in patients with CLL and coexisting conditions. N Engl J Med. 2014;370(12):11011110. 54. Tausch E, Schneider C, Robrecht S, et al. Prognostic and predictive impact of genetic markers in patients with CLL treated with obinutuzumab and venetoclax. Blood. 2020;135(26):2402-2412.

3135

ARTICLE Ferrata Storti Foundation

Hematopoiesis

Inhibition of the anti-apoptotic protein MCL-1 severely suppresses human hematopoiesis Sheila Bohler,1,2 Sehar Afreen,1 Juncal Fernandez-Orth,1 Eva-Maria Demmerath,1 Christian Molnar,1,2,3 Ying Wu,1,2 Julia Miriam Weiss,1 Venugopal Rao Mittapalli,1 Lukas Konstantinidis,4 Hagen Schmal,4 Mirjam Kunze5 and Miriam Erlacher1,6,7 Department of Pediatrics and Adolescent Medicine, Division of Pediatric Hematology and Oncology, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg; 2Faculty of Biology, University of Freiburg, Freiburg; 3Spemann Graduate School of Biology and Medicine (SGBM), University of Freiburg, Freiburg; 4 Department of Orthopedics and Trauma Surgery, Medical Center, Faculty of Medicine, Albert-Ludwigs-University of Freiburg, Freiburg; 5Department of Obstetrics and Gynecology, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg; 6German Cancer Consortium (DKTK), Freiburg, and 7German Cancer Research Center (DKFZ), Heidelberg, Germany 1

Haematologica 2021 Volume 106(12):3136-3148

ABSTRACT

Correspondence: MIRIAM ERLACHER miriam.erlacher@uniklinik-freiburg.de Received: March 17, 2020. Accepted: November 6, 2020. Pre-published: November 26, 2020. https://doi.org/10.3324/haematol.2020.252130

3136

H3-mimetics inhibiting anti-apoptotic BCL-2 proteins represent a novel and promising class of antitumor drugs. While the BCL-2 inhibitor venetoclax is already approved by the Food and Drug Administration, BCL-XL and MCL-1 inhibitors are currently in early clinical trials. To predict side effects of therapeutic MCL-1 inhibition on the human hematopoietic system, we used RNA interference and the small molecule inhibitor S63845 on cord blood-derived CD34+ cells. Both approaches resulted in almost complete depletion of human hematopoietic stem and progenitor cells. As a consequence, maturation into the different hematopoietic lineages was severely restricted and CD34+ cells expressing MCL-1 shRNA showed a very limited engraftment potential upon xenotransplantation. In contrast, mature blood cells survived normally in the absence of MCL-1. Combined inhibition of MCL-1 and BCL-XL resulted in synergistic effects with relevant loss of colony-forming hematopoietic stem and progenitor cells already at inhibitor concentrations of 0.1 mM each, indicating “synthetic lethality” of the two BH3mimetics in the hematopoietic system.

Introduction BH3-mimetics represent a novel and very promising group of anticancer drugs, with venetoclax being the first compound approved by the Food and Drug Administration (FDA).1 They act by directly inhibiting anti-apoptotic BCL-2 proteins that prevent the intrinsic apoptosis pathway and thereby ensure survival of every human cell. BCL-2 and its homologs BCL-XL, MCL-1, BFL1/A1 and BCL-W bind to and inhibit BAX and BAK, two downstream pro-apoptotic effector BCL-2 proteins which, upon activation, lead to permeabilization of the outer mitochondrial membrane.2 As a consequence, cytochrome c is released into the cytosol, a process regarded as a “point of no return” for the initiation of apoptosis. In the cytosol, cytochrome c together with APAF1 and procaspase 9 molecules form a large complex termed an apoptosome, in which caspase 9 is activated. Caspase 9 then activates the effector caspases 3, 6 and 7 which eventually degrade vital cellular structures and execute cell death. This whole process is regulated by upstream pro-apoptotic proteins that also belong to the large BCL-2 family but share only the BH3 (BCL2 homology 3) domain with the other members of the anti-apoptotic BCL-2 family. BH3-only proteins are upregulated or activated upon given stress signals and then bind and inhibit the anti-apoptotic BCL-2 proteins. As a consequence, BAX and BAK are released, leading to apoptosis.3 As their name indicates, BH3-mimetics imitate the mode of action of the BH3-only proteins.4 While BH3-only proteins are tightly regulated and only activated upon lethal stress

haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

signals, BH3-mimetics are able to bypass this mode of activation. Similar to BH3-only proteins, every BH3-mimetic available so far has specific binding affinities to one or more anti-apoptotic BCL-2 proteins (Online Supplementary Figure S1A). Navitoclax (ABT-263)5 and its intravenously used precursor drug, ABT-737,6 bind to BCL-2, BCL-XL and BCL-W. The drug showed good efficacy against non-small lung carcinoma and hematologic malignancies.7,8 However, its side effects on the hematopoietic system precluded its full clinical exploration and FDA approval. This indicated that a combined inhibition of more than one pro-survival BCL-2 protein might impede survival of healthy body cells. Later, a BCL-2-specific inhibitor called venetoclax (ABT-199) found its way into clinical trials.9 Thanks to the much less severe side effects, it was approved by the FDA in 2016 as a second-line treatment for chronic lymphocytic leukemia (CLL) with 17p deletion, and in 2019 for the treatment of all adult CLL and small lymphocytic lymphoma patients.1 For acute myeloid leukemia (AML), venetoclax was FDA-approved only in combination with hypomethylating agents.1 Unfortunately, as for other cytotoxic drugs, tumor cell resistance poses a major problem to the efficacy of venetoclax. Primary resistance is present when tumor cells require anti-apoptotic BCL-2 proteins other than BCL-2 for survival. Naturally, only lymphocytes10 and melanocytes11 are dependent on BCL-2 expression, as shown in BCL-2 knockout mice. This might explain why venetoclax is most effective in mature lymphoma while most other tumors show primary resistance. Such primary resistance to venetoclax can also be caused by overexpression of pro-survival proteins other than BCL-2, such as BCL-XL and/or MCL-1.12,13 As shown for CLL, these BCL2 homologs can be induced by signals from the tumor microenvironment.14 Secondary resistance, in contrast, is acquired by tumor cells to escape previously effective BCL-2 inhibition. Several mechanisms, such as BCL-2 mutations which strongly lower venetoclax affinity,15 have been implicated in the development of secondary venetoclax resistance.16 Alternatively, BCL-XL and MCL-1 overexpression was noted in relapsed CLL patients who had been previously treated with venetoclax.17,18 Therefore, the development and administration of MCL-1/BCL-XL inhibitors are much needed to overcome primary and secondary venetoclax resistance. MCL-1 inhibitors, in particular, are eagerly awaited by oncologists since this protein plays an essential role in many tumor types (e.g., AML, multiple myeloma, non-small cell lung carcinoma).19 MCL-1 was first identified during the differentiation of monocytes to macrophages in ML-1, a human myeloid leukemia cell line.20 Three isoforms of the gene have been reported; the most abundant anti-apoptotic MCL-1 long (MCL-1L)20 and two shorter pro-apoptotic isoforms (MCL1 short, MCL-1 extra short).21,22 In addition, a truncated isoform was shown to localize at the mitochondrial matrix where it facilitates mitochondrial fusion and ATP synthesis.23 Genetic Mcl-1 deletion in mice revealed its essential role in many tissues, both during embryogenesis and in adult mice. Specifically, constitutive MCL-1 deficiency resulted in peri-implantation embryonic lethality,24 while targeted deletion in the fetal hematopoietic system resulted in loss of stem cells.25 When one Mcl-1 allele was deleted in adult mice, hematopoietic stem and progenitor cells (HSPC) were depleted, leading to the death of the haematologica | 2021; 106(12)

animals within 2-3 weeks.26 With regard to human HSPC, there is only indirect evidence for the essential role of MCL-1, given by the BH3 profiling method: Mitochondria isolated from human CD34+ cells were highly sensitive to NOXA BH3 peptides, which typically correlated with MCL-1 dependency.27 Recently, a potent and specific MCL-1 inhibitor, S63845, was developed. This compound can efficiently kill a variety of tumor cell types such as multiple myeloma, lymphomas, leukemias and primary AML cells as well as to some extent solid cancers.28 Treatment of mice with S63845 resulted in only a few side effects in vivo,28 which was rather unexpected considering the many roles of MCL-1 during development and for tissue homeostasis. Here, we extended these studies to human cells and focused on the hematopoietic system. Understanding hematotoxicity of novel anticancer drugs is crucial since suppression of hematopoiesis accounts for most treatment-related morbidity and mortality. By using two different shRNA sequences and the MCL-1 inhibitor S63845, we consistently found that MCL-1 expression is essential for the survival of human stem and progenitor cells, especially during early stages of differentiation. In contrast, mature blood cells are less sensitive to MCL-1 inhibition. Of note, combined inhibition of MCL-1 and BCL-XL was synergistic and already low concentrations of both drugs resulted in profound stem and progenitor cell depletion.

Methods Lentiviruses A pLeGO-hU6 lentiviral vector with huU6 promoter and green fluorescent protein (GFP) expression was used to generate shRNA expressing lentiviruses (Online Supplementary Table S1),29 CD34+ cells were transduced with the lentivirus (2x MOI 10, 24 h each) and knockdown efficiencies were determined 24 h later.

Isolation and culture of human CD34+ cells Umbilical cord blood and bone marrow were obtained immediately after birth or from patients (age: 44-90 years) undergoing orthopedic surgery, respectively. Informed consent was obtained and the ethics committee approved the study. CD34+ cells were isolated (by magnetic activated cell sorting) from mononuclear cells to a purity >90%. Cells were used either immediately or stored in liquid nitrogen (CS10 freezing medium, Sigma) for later use. Cells were cultured in serum-free StemPro-34 medium supplemented with embryonic stem cell fetal bovine serum (ESFBS), penicillin/streptomycin (P/S; Invitrogen), stem cell factor (SCF), FMS-like tyrosine kinase 3 ligand (FLT3L) (200 ng/mL each), thrombopoietin (TPO; 100 ng/mL) and interleukin 3 (IL-3; 20 ng/mL; Immunotools/Peprotech). Where indicated, the BH3mimetics S63845 (SynMedChem), A-1155463 and ABT-199 (Sellekchem) were added.

Apoptosis assay CD34+ cells were subjected to cytokine deprivation or treated with etoposide (VP16), tunicamycin, taxol, thapsigargin and brefeldin A (BFA). After 0, 24 and 48 h, cells were stained with annexin V (Biolegend) and 7-aminoactinomycin D (7-AAD; Sigma-Aldrich) to detect apoptosis. The percentage of specific apoptosis was calculated as: 100 x (% living cells under control condition - % living cells under treatment)/ % living cells under control condition. Control condition was culture with ES-FBS and cytokines.

3137

S. Bohler et al.

Colony-forming assays and differentiating culture One thousand CD34+ cells were seeded in MethoCult SFH4436 medium. After 10-11 days, colony types (identified by light microscopy) and total cell counts were determined. The percentages of HSPC, erythroid and myeloid cells were determined by flow cytometry (Online Supplementary Table S2).

conditions and transferred onto polyvinylidene difluoride membranes. The antibodies used were: MCL-1 (D2W9E) rabbit monoclonal antibody, BCL-XL (54H6), BCL-2 (D17C4), BFL1/A1 (D1A1C), α/b-tubulin or b-actin (13E5) (all rabbit, Cell Signaling). A peroxidase-coupled goat anti-rabbit IgG secondary antibody was used (sc-2004, Santa Cruz).

Proliferating culture

Statistics

CD34+ cells were cultured for 5-11 days in StemPro-34 medium supplemented with 10% ES-FBS, SCF, FLT3L, TPO and IL-3.. The medium was refreshed every 3 days. Cells were analyzed for GFP and immature populations (Online Supplementary Table S2).

Statistical analyses were performed using the unpaired MannWhitney test in GraphPad Prism 7 software. P values less than 0.05 were considered statistically significant. Synergy of BCL-XL and MCL-1 inhibitors was calculated using the Bliss synergy score and the program SynergyFinder (https://synergyfinder.fimm.fi).32

Quantitative reverse transcription polymerase chain reaction

Results

RNA was isolated by a Quick RNA Micro Prep kit (Zymo Research) and reversely transcribed to cDNA (QuantitecReverse transcription kit, QIAgen). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed by using BIO-RAD (CFX96 Touch) a RT-PCR detection system and SYBR Green master mix (Thermofisher). Expression of the gene of interest was normalized to the expression of either 18S or 36B4.

Reverse transcriptase multiplex ligation dependent probe amplification RNA samples from CD34+ cells cultured under different conditions were obtained by a Quick RNA Micro Prep kit (Zymo Research). Reverse transcriptase multiplex ligation dependent probe amplification (RT-MLPA) was performed according to the manufacturer’s instructions (MRC Holland, R011-C1). The resultant amplicons were separated by capillary electrophoresis (ABI-3130xl Genetic Analyzer) and Sequence Pilot (JSI Medical Systems) was used for the analysis. The sum of all peaks was taken as 100%, and the values of the single peaks were normalized accordingly.

Xenotransplantation All experiments were performed after approval from the local ethics committee and in compliance with German law. Rag2−/−gc−/− mice were kept under specific pathogen-free conditions. Newborn mice were sub-lethally irradiated with 2.5 Gy.30 After 6 h, the progeny of 1x105 transduced or untransduced human CD34+ cells were injected intrahepatically. Mice were sacrificed for analysis after 8 weeks.

Flow cytometry Single cell suspensions obtained from colony-forming assays or hematopoietic organs from mice were surface stained with monoclonal antibodies: CD34 PE-Cy7(581), CD38 APC (HIT2), CD10 PE/APC (HI10a), CD45RA PerCP-Cy5.5 (HI100), CD90 APC-Cy7 (5E10), CD117 PE-Cy7 (104D2), CD71 APC (CY1G4), CD33 PE (WM53), CD14 APC (M5E2), CD115 BV421 (4D21E4), CD15 PeCy5 (W6D3), CD66b PerCP-Cy5.5 (G10F5), CD19 PE-Cy7 (HIB19), IgM APC-Cy7 (MHM-88), CD45 Biotin (HI30), CD45 PE-Cy7/V500 (30-F11) (Biolegend), and CD235a BV421 (HIR2) (BD Biosciences). Streptavidin PerCP-Cy5.5/V450 (Biolegend) was used as a secondary antibody. BD LSRFortessa and FlowJo were used for flow cytometry and analyses, respectively. The gating strategy was published earlier.31

Western blot Purified proteins were size fractioned by 12% sodium dodecylsulfate polyacrylamide gel electrophoresis under reducing

3138

MCL-1 knockdown sensitizes human CD34+ cells selectively to endoplasmic reticulum stress Putative shRNA sequences specific for MCL-1 were tested in transfected HEK293T cells. For shRNA delivery into human CD34+ cells, a lentiviral system that allowed stable expression of the shRNA and GFP was used. Two different shRNA sequences, both binding to exon 3 of the human MCL-1 gene, were identified to reduce MCL1 mRNA expression to 25-55% in HEK293T and by 4155% in human cord blood-derived CD34+ HSPC (Figure 1A). Knockdown of the Luciferase (Luci) gene was used as a negative control. Efficient MCL-1 knockdown was confirmed on a protein level in HEK293T and CD34+ cells and showed, at least in CD34+ cells, no relevant differences between the two shRNA sequences used (Figure 1B, C). Transduction efficiency was similar for the different viruses (Online Supplementary Figure S1B) but CD34+ cells transduced with shRNA specific for MCL-1 showed increased apoptosis rates 24 h after transduction (Figure 1D). The surviving cells were cultured and treated for 24 h and 48 h with different cytotoxic drugs including the DNA damaging agent etoposide, the mitotic spindle inhibitor taxol and different compounds inducing endoplasmic reticulum (ER) stress (i.e., tunicamycin, thapsigargin and brefeldin A) (Figure 1E, F). Interestingly, MCL1 inhibition selectively increased the sensitivity of CD34+ cells to ER stress (Figure 1E, F). While transduced GFP+ cells expressing MCL-1 shRNA were killed in a dosedependent manner (Online Supplementary Figure S2A), GFP+ cells expressing Luci shRNA or non-transduced GFP- cells did not undergo increased cell death upon ER stress (Online Supplementary Figure S2A, B). These findings indicate that MCL-1 inhibition increased cellular sensitivity in a stress-dependent and cell-intrinsic manner. Since knockdown of MCL-1 was not complete in CD34+ cells, we hypothesized that ER stress itself reduced MCL-1 levels, thereby leading to critical depletion of this protein once combined with the gene knockdown. Indeed, we found significant and dose-dependent downregulation of MCL-1 mRNA when CD34+ cells were treated with tunicamycin (Figure 1G). At the same time, tunicamycin led to downregulation of BCL-XL mRNA and upregulation of the BH3-only proteins PUMA and BMF (Online Supplementary Figure S2C). Upregulation of CHOP and PERK mRNA confirmed the presence of ER stress (Online Supplementary Figure S2D). In summary, ER stress shifts the BCL-2 equilibrium towards apoptosis, which is initiated once levels of haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

Figure 1. MCL-1 inhibition selectively sensitizes human CD34+ cells to endoplasmic reticulum stress. (A) HEK293T cells were transfected with plasmids expressing shRNA specific for Luciferase (shLuci) or human MCL-1 (shM#3 or shM#4). MCL-1 mRNA expression was determined in sorted GFP+ cells and normalized to the 36B4 reference gene. Bars represent mean ± standard error of mean (SEM); n=5 from five independent experiments. Human cord blood-derived CD34+ cells were transduced with the corresponding lentiviruses. GFP+ cells were sorted 24 h after transduction, and knockdown efficiency of MCL-1 was determined by quantitative reverse transcription polymerase chain reaction (qRT-PCR). mRNA expression was normalized to 18S expression. Bars represent mean ± SEM; n=2-4 from four independent experiments. Mann-Whitney test: *P<0.05. (B, C) MCL-1 protein levels were determined in HEK293T (B) and CD34+ (C) GFP+ cells. (D) Measurements of apoptosis in CD34+ cells 24 h after lentiviral transduction revealed that 14-19% of cells undergo apoptosis early after MCL-1 depletion. Bars represent mean ± SEM; n=4 from four independent experiments. (E, F) Transduced CD34+ cells (transduction efficiency 45-65%) were cultured either under optimal conditions (cytokines and 10% serum) or under conditions of stress; in the presence of serum but deprived of cytokines, etoposide (0.5 mg/mL), taxol (0.125 mg/mL), tunicamycin I and II (0.5 and 1 mg/mL, respectively), thapsigargin (3 mM) or brefeldin A (BFA; 0.5 mg/mL). Apoptosis in GFP+ cells was determined by flow cytometry using annexin V and 7-AAD staining 24 h (E) and 48 h (F) later. Bars represent mean ± SEM; n=3-8 from eight independent experiments. Mann-Whitney test: *P<0.05, **P<0.01, ***P<0.001. (G) RNA was isolated from CD34+ cells treated with increasing concentrations of tunicamycin and used for qRT-PCR. MCL-1 mRNA expression was normalized to 18S expression. Bars represent mean ± SEM; n=4 from four independent experiments. Mann-Whitney test: *P<0.05.

haematologica | 2021; 106(12)

3139

S. Bohler et al.

either MCL-1 (shown here) or BCL-XL (documented earlier)31 are further reduced by RNA interference.

Human hematopoietic stem and progenitor cells show impaired colony formation and differentiation upon MCL-1 knockdown To test the effect of MCL-1 knockdown on colony formation and differentiation of human CD34+ cells, we cultured 1,000 transduced and sorted GFP+ cells for 11 days in MethoCult medium containing the cytokines SCF, IL3, IL-6, erythropoietin (EPO), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colonystimulating factor (GM-CSF), insulin and transferrin. The

number of colonies arising was significantly lower when MCL-1 expression was inhibited (Figure 2A). All colony types were affected, indicating that all multipotent and lineage-committed progenitor cell types were lost to a similar degree (Figure 2B). In addition, fewer cells could be harvested from plates (Figure 2C). Flow cytometry revealed that all cell types were reduced in number when MCL-1 was depleted (Online Supplementary Figure S3A). Since we noticed a relevant toxicity of the sorting procedure on lentivirally-transduced cells, we repeated the experiment using unsorted cells. With this approach, we could directly compare transduced GFP+ with untransduced GFP- cells. Colony numbers and types were similar

Figure 2. MCL-1 is essential for all hematopoietic progenitor cells. (A-C) Lentivirally transduced human CD34+ cells were sorted for GFP expression. GFP+CD34+ cells were seeded in MethoCult medium (1,000 cells each). (D-G) Alternatively, unsorted cells were plated. (A, D) After 11 days of culture, colonies were counted by light microscopy. (B, E) Based on morphological findings, the following colony types were identified by light microscopy: GEMM: granulocytic-erythroid-megakaryocyticmonocytic, GM: granulocytic–monocytic, G: granulocytic, E: erythroid, M: monocytic. (C, F) Cells were dissolved from the semisolid medium and counted by a hemocytometer. (G) The different hematopoietic cell types were determined by flow cytometry. The percentages of GFP+ cells are shown within each of the following cell populations: HSC; hematopoietic stem cells (CD34+CD38-CD45RA-CD90+), MPP: multipotent progenitors (CD34+38–CD45RA-CD90-), GM: granulocytic–monocytic progenitors (CD34+CD33+CD115+), CFU-G: colony forming unit-granulocytes (CD34-CD33+CD15+CD115-), M: monocytes (CD34-CD33+CD14+CD115-), immE: immature erythrocytes (CD71hiCD235a-), matE: mature erythrocytes (CD71+CD235a+). Bars represent mean ± standard error of mean, n=6 (A-C) from six independent experiments and n=4 (D-G) from four independent experiments. Mann-Whitney test: *P<0.05, **P<0.01.

3140

haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

in all groups (Figure 2E, F) and cell numbers were not consistently reduced when MCL-1 was downregulated (Figure 2G). However, while transduction rates were comparable (Online Supplementary Figure S3B), MCL-1depleted GFP+ cells were selectively lost during the 11 days of MethoCult culture, indicating their selective disadvantage (Figure 2H, Online Supplementary Figure S3B). Loss of MCL-1 affected all analyzed immature and mature cell types in a similar manner, confirming that all progenitor cell types were dependent on MCL-1 expression (Figure 2H, Online Supplementary Figure S3C). To analyze whether MCL-1-depleted cells were lost immediately or progressively over time, we cultured them for only 5 days in MethoCult medium. At this early time point, only mild loss of GFP+ cells was observed, independently of the lentivirus used (Figure 3A, left). This indicates that progenitor cells became more susceptible to MCL-1 inhibition once they progressed in their differentiation process. Accordingly, immature CD34+ cells and specifically hematopoietic stem cells and multipotent progenitors were enriched in the first days of culture (Figure 3B, Online Supplementary Figure S3D, E). To inhibit differentiation but foster proliferation, CD34+ cells were cultured in the presence of the cytokines SCF (200 ng/mL), FLT3L (200 ng/mL), TPO (100 ng/mL) and IL-3 (20 ng/mL). CD34 and GFP expression was measured after 5 and 11 days. Under this condition, the percentage of CD34+ cells remained high (Figure 3B, right) and GFP+ cells were not depleted in a relevant manner (Figure 3A, right). To exclude that expression of shRNA and, consequently, MCL-1 knockdown was different in the two culture conditions, we measured MCL-1 mRNA after 5 days of culture. While we observed stable MCL-1 knockdown by shRNA #3, the shRNA #4 showed less consistent results

with re-expression of MCL-1 mRNA (Online Supplementary Figure S3F). However, no difference in knockdown efficiency was observed between the two culture conditions indicating that the dependence of HSPC on MCL-1 did indeed change during proliferation and differentiation, respectively. To understand why MCL-1 dependence was so different under the two culture conditions, we determined the composition of all BCL-2 proteins by RT-MLPA. As controls, we used freshly isolated CD34+ cells. Interestingly, both MCL-1 and BCL-XL were expressed at higher levels in cells stimulated to differentiate for 4 days (Figure 3C). Among the pro-apoptotic BCL-2 proteins, PUMA was highly upregulated under both culture conditions (i.e., differentiation and proliferation conditions) while BIM, BID and BAK1 were selectively upregulated under differentiation conditions (Figure 3D, E). Thus, it is possible that differentiation is associated with stronger pro-apoptotic signals that need to be counteracted by higher MCL-1 and BCLXL levels. Based on its binding affinities, it is conceivable that MCL-1 expression is required to counteract BIMmediated activation of BAK1.

MCL-1 inhibition severely restricts hematopoietic stem and progenitor cell engraftment in xenografted mice In order to determine the effects of MCL-1 inhibition on the engraftment potential of human CD34+ HSPC, untransduced and transduced cells were intrahepatically transplanted into sublethally irradiated Rag2-/-gc-/- mice. To reduce cell stress prior to transplantation, we waived the sorting procedure and transplanted GFP+ cells together with GFP- cells. After 8 weeks, xenografted mice were sacrificed and human engraftment was analyzed. Lentivirally transduced cells had a reduced potential to engraft but

Figure 3. MCL-1 expression is more important for differentiating than for proliferating CD34+ cells. (A) Sorted CD34+GFP+ cells were subjected to either differentiating or proliferating culture conditions. To induce differentiation, cells were cultured in semisolid MethoCult plates for 5 days (3,000 cells seeded per plate) or 11 days (1,000 cells seeded per plate). To foster proliferation, cells were cultured in stem cell medium containing 10% serum and stem cell factor, FLT3L (200 ng/mL each), thrombopoietin (100 ng/mL) and interleukin-3 (20 ng/mL). After 5 and 10 days of culture, the percentages of GFP+ cells were determined by flow cytometry. (B) The fraction of CD34+ cells was determined within GFP+ cells at each time point. (C-E) Untransduced CD34+ cells were subjected to the two different culture conditions and harvested at the indicated time points. mRNA was used for reverse transcriptase-multiplex ligation dependent probe amplification designed to determine levels of apoptosis genes. Results are shown for anti-apoptotic BCL-2 proteins (C), BH3-only proteins (D) and pro-apoptotic effector proteins (E). Freshly isolated CD34+ cells were used as controls. Bars represent mean ± standard error of mean, n=4 from four independent experiments. Mann-Whitney test: *P<0.05, **P<0.01.

haematologica | 2021; 106(12)

3141

S. Bohler et al.

there were no differences between the viruses used (Figure 4A). As known from this model system, most cells differentiated into CD19+ B cells while fewer CD33+ myeloid cells and almost no CD3+ T cells arose (Figure 4B and data not shown). Cells expressing the control shRNA (Luci) engrafted and contributed to all lineages (Figure 4C, D). In contrast, cells expressing shRNA specific for MCL-1 showed only very poor engraftment (Figure 4C). In line with the important role of MCL-1 for survival of immature progenitors with multipotent potential, all cell types found in the xenografted mice were equally affected (Figure 4D).

Stem and progenitor cells from neonates and adults are equally sensitive to MCL-1 or BCL-XL inhibition

MCL-1 inhibition limits the survival of immature but not mature hematopoietic cells To determine the effects of MCL-1 inhibition on more mature types of hematologic cells, we used the specific MCL-1 inhibitor S63845. First, we treated freshly isolated immature CD34+ and mature CD34- cells with increasing doses of the inhibitor. While CD34+ cells were moderately sensitive when compared to cancer cell lines (e.g., IC in most multiple myeloma cell lines <0.1 mM),28 no apoptosis was induced after 24 h and 48 h in mature CD34- blood cells even when very high doses of inhibitor were used (Figure 5A, B). In a second approach, we let untreated CD34+ cells differentiate for 11 days in MethoCult medium. Cells were 50

then isolated and put into stem cell medium containing 10% ES-FBS and cytokines (SCF, TPO, FLT3L, IL-3). Different concentrations of the MCL-1 inhibitor were added, and cell numbers were determined after 24 h. Again, differentiated CD34- cells were much less sensitive than immature CD34+ cells (Figure 5C, D). As a consequence, only a mild and non-significant reduction in cell numbers was noted (Figure 5C) and both myeloid and erythroid cells were depleted only to a minor and non-significant degree (Figure 5D).

Our experiments indicate a strong dependence of human HSPC on MCL-1 expression, which is not unexpected considering the high relevance of MCL-1 for survival of murine HSPC.25,26 However, other authors described an overall good tolerability of human HSPC to MCL-1 inhibitors.33-35 One reason for this discrepancy could be that we used CD34+ cells derived from cord blood while CD34+ cells derived from the bone marrow of aged persons were used in other studies.33 We therefore compared these two cell types with regards to protein levels of MCL-1 and other anti-apoptotic BCL-2 family members. Bone marrow CD34+ cells were obtained from patients with orthopedic problems (age range: 44 to 90 years). While MCL-1 and

Figure 4. Human CD34+ cells lacking MCL-1 show poor engraftment in xenografted mice. (A-D) Lentivirally transduced or untransduced human hematopoietic stem and progenitor cells (HSPC) were transplanted intrahepatically into newborn Rag2−/−gc−/− mice after sub-lethal irradiation. Mice were sacrificed 8 weeks after transplantation and bone marrow (BM) and spleen populations were analyzed. By using antibodies specific for human or murine CD45, percentage human engraftment was determined (A). The various human hematopoietic populations were determined within the huCD45+ cells using flow cytometry (B). GFP expression was determined in huCD45+ cells (C) and in each of the subpopulations (D). Bars represent mean ± standard error of mean, n =9-13 from eight independent experiments. Mann–Whitney test was performed; *P<0.05, **P<0.01.

3142

haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

BCL-XL levels were identical in both cell types and A1 was not expressed, BCL-2 levels were higher in adult HSPC than in HSPC from neonates (Figure 6A). Next, we performed comparative functional studies using the MCL-1 inhibitor S63845, the BCL-XL inhibitor A-1155463 and the BCL-2 inhibitor ABT-199. By treating bone marrow immature CD34+ and differentiated CD34cells with the MCL-1 inhibitor for 24 h and 48 h, we obtained similar results as those with cord blood cells (compare Figure 6B with Figure 5A, B). Chemical inhibition of MCL-1 also confirmed our RNA interference experiments: when we added the inhibitor S63845 to cord blood CD34+ cells cultured in MethoCult medium, colony formation was impeded in a dose-dependent manner (compare Figure 6C with Figure 2). The numbers of immature cell types, including hematopoietic stem cells, multipotent progenitors and mixed lymphoid progenitors, were significantly reduced (Figure 6C, right panel). Similarly, all emerging erythroid and myeloid cells (Figure 6C, right panel) were depleted in a dose-dependent manner. Importantly, there was no difference in MCL-1 inhibitor sensitivity in CD34+ cells derived from cord blood (Figure 6C) or bone marrow (Figure 6D). We have shown earlier that BCL-XL, too, is important for keeping human cord blood CD34+ cells alive,31 a finding that was unexpected considering its dispensable role for mouse HSPC.36-38 We now extended our studies to bone marrow-derived cells and showed that CD34+ cells were

Synthetic lethality of MCL-1 and BCL-XL inhibitors in human hematopoietic stem and progenitor cells We noted a striking functional homology of MCL-1 and BCL-XL for survival of human HSPC and concluded that combined inhibition of both anti-apoptotic proteins could result in complete depletion of colony-forming stem and progenitor cells. To test this, we treated cord blood-derived CD34+ cells with increasing doses of the

more sensitive than CD34- cells to the BCL-XL inhibitor A1155463 (Figure 7A). We compared colony-forming potential CD34+ cells derived from cord blood or bone marrow and observed no difference between the two cell types: BCL-XL inhibition resulted in significant reduction of colony formation (Figure 7B) and cell numbers (Figure 7C), independently of the source of the human HSPC. Notably, effects on colony formation were less pronounced than those caused by MCL-1 inhibition. As documented earlier, immature CD34+ cells (Figure 7D) as well as mature erythroid cells (Figure 7E) were more severely affected than myeloid cells (Figure 7F). Finally, the BCL-2 inhibitor ABT-199 did not negatively affect survival or colony formation of cord blood- or bone marrow-derived CD34+ cells, even when used at the high concentration of 1 mM (Online Supplementary Figure S4). This finding is consistent with the relatively mild myelosuppressive effects of venetoclax observed in clinical trials.39

Figure 5. Survival of mature hematopoietic cells is independent of MCL-1. (A-B) Freshly isolated cord blood was subjected to density gradient centrifugation and mononuclear cells were divided into CD34+ and CD34- cells using magnetic activated cell sorting technology. Both cell fractions were treated with the indicated concentrations of the MCL-1 inhibitor S63845. After 24 h (A) and 48 h (B), percentages of living cells were determined by flow cytometry using annexin V/7-AAD. Bars represent mean ± standard error mean (SEM); n=3-6 from six independent experiments. (C-D) CD34+ cells were differentiated in MethoCult culture. After 11 days, differentiated cells were isolated and treated with S63845 for 24 h. Total cell numbers (C) were determined and erythroid and myeloid cell populations (D) were analyzed by flow cytometry (n=5 from 5 independent experiments). Bars represent mean ± SEM. Mann-Whitney test: *P<0.05, **P<0.01, ***P<0.001.

haematologica | 2021; 106(12)

3143

S. Bohler et al.

Figure 6. MCL-1 is essential for survival of both cord blood- and bone marrow-derived human CD34+ cells. (A) Protein levels of the anti-apoptotic BCL-2 family members MCL-1, BCL-XL, BCL-2 and BFL1/A1 were determined in CD34+ cells isolated from both cord blood and bone marrow. b-actin served as a loading control. (B) Freshly isolated bone marrow was subjected to density gradient centrifugation and mononuclear cells were divided into CD34+ and CD34- cells using magnetic activated cell sorting technology. Both cell fractions were treated with the indicated concentrations of the MCL-1 inhibitor S63845. After 24 h (upper panel) and 48 h (lower panel), percentages of living cells were determined by flow cytometry using annexin V/7-AAD. Bars represent mean ± standard error of mean (SEM); n=6-7 from seven independent experiments. (C-D) Human CD34+ cells isolated from either cord blood (C) or bone marrow (D) were differentiated in MethoCult medium containing 0.1 or 1 mM S63845. As controls, untreated and dimethylsulfoxide (DMSO)-treated cells were used (n=7 from seven independent experiments). After 11 days, total colony numbers (left) and total cell numbers (middle) were determined using light microscopy and hemocytometry, respectively. Different immature and differentiated cell types were determined by flow cytometry, and their absolute cell numbers were calculated (right). The following cell types were studied; HSC: hematopoietic stem cells (CD34+CD38-CD45RA-CD90+), MPP: multipotent progenitors (CD34+38–CD45RA-CD90-), MLP: mixed lymphoid progenitors (CD34+CD38CD45RA+CD10+) GM: granulocytic–monocytic progenitors (CD34+CD33+CD115+), M: monocytes (CD34-CD33+CD14+CD115-), immE: immature erythrocytes (CD71hiCD235a-), matE: mature erythrocytes (CD71+CD235a+). Bars represent mean ± SEM. Mann-Whitney test: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

3144

haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

MCL-1 inhibitor S63845, together with increasing doses of the BCL-XL inhibitor A-1155463. Apoptosis induction was determined 24 h later (Figure 8A). Using SynergyFinder (https://synergyfinder.fimm.fi) a doseresponse matrix was calculated (Figure 8B). The resulting Bliss score of 21.26 indicates strong synergy between the two inhibitors. Synthetic lethality was confirmed in colony-forming assays, both with cord blood- and bone marrow-derived CD34+ cells. Already at concentrations of 0.1 mM each, the drug combination resulted in a substantial loss of colony-forming cells (compare Figure 8C, D with Figures 6B, C and 7). To determine the number of immature cells with selfrenewal potential that survived BCL-XL and/or MCL-1 inhibition, we used 10,000 cells isolated from primary colonies for serial colony-forming assays. Interestingly, only BCL-XL inhibition in the first plating resulted in depletion of progenitor cells able to form colonies in the

second plating. However, there was a synergistic effect when this BCL-XL inhibition was combined with MCL-1 inhibition (Online Supplementary Figure S5).

Discussion Because of the narrow spectrum of cancer entities susceptible to venetoclax, specific MCL-1 and BCL-XL inhibitors are eagerly awaited by oncologists. However, observations made in genetically modified mice indicate that inhibition of MCL-1 or BCL-XL could have more severe side effects than BCL-2 inhibition. Mice deficient for either MCL-1 or BCL-XL have severe developmental phenotypes while BCL-2-deficient mice lack lymphocytes and melanocytes but are otherwise normal.10,40,41 Hematopoietic toxicity of anticancer drugs is responsible for most therapy-related morbidity and mortality and a

Figure 7. BCL-XL inhibition impedes colony formation and survival of erythroid cells. (A) Freshly isolated bone marrow was subjected to density gradient centrifugation and mononuclear cells were divided into CD34+ and CD34- cells using magnetic activated cell sorting technology. Both cell fractions were treated with the indicated concentrations of the BCL-XL inhibitor A-1155463. After 24 h, percentages of living cells were determined by flow cytometry using annexin V/7-AAD. Bars represent mean ± standard error of mean (SEM); n=3-4 from four independent experiments. (B-F) Human CD34+ cells, either derived from cord blood or bone marrow, were differentiated in MethoCult culture containing 0.5 or 1.5 mM A-1155463. As controls, untreated and dimethylsulfoxide (DMSO)-treated cells were used (n=6-7 from seven independent experiments). After 11 days, total colony numbers (B) and total cell numbers (C) were determined using light microscopy and hemocytometry, respectively. Different cell types were determined by flow cytometry, and their absolute cell numbers were calculated (D-F). The following cell types were studied; immature CD34+ cells, monocytes (CD34-CD33+CD14+CD115-) and mature erythrocytes (CD71+CD235a+).

haematologica | 2021; 106(12)

3145

S. Bohler et al.

common reason for treatment delays or dose reductions. We, therefore, consider it important to generate preclinical data to evaluate the hematotoxicity profile of such novel drugs. Recently, we described the detrimental effects of BCL-XL inhibition on human HSPC and erythroid progenitors.31 Here, we dissected the effects of MCL-1 inhibition on different immature and mature types of hematologic cells. By using an RNA interference approach and the specific small molecule inhibitor S63845, we consistently found that MCL-1 expression is crucial for multipotent stem and progenitor cells, as well as for myeloid progenitors, while erythroid progenitors are less susceptible to MCL-1 inhibition. During later stages of blood cell differentiation, MCL-1 becomes dispensable for cell survival. Interestingly, we noted re-expression of MCL-1 mRNA after some days of culture, when shRNA #4 was used, while the shRNA #3 resulted in stable knockdown. Nevertheless, the resulting phenotypes were strikingly similar indicating that the loss of stem and multipotent progenitors occurs early after MCL-1 depletion and cannot be compensated by later MCL-1 re-expression. Looking more closely at the stem and progenitor cell compartment, we noted that cells that specifically enter

the differentiation process are highly dependent on MCL1 expression, while proliferating CD34+ cells remain fairly resistant. What is the reason for this difference? It is possible that the differentiation process is associated with increased stress levels reflected by accumulation of activated BH3-only proteins. Alternatively, it is possible that the cytokines TPO, FLT3L, SCF and IL3, which induce cell proliferation and are used for CD34+ cell culture, not only induce proliferation but also confer CD34+ cells with survival signals, thereby rendering them independent of MCL-1 expression. Indeed, we have shown earlier that these cytokines induce BCL-XL mRNA upregulation and at the same time repress expression of the pro-apoptotic BCL-2 proteins BIM and BMF.30 Also in this study, BIM mRNA levels were lower and BCL-XL mRNA levels higher when cells were cultured in the presence of TPO, FLT3L, SCF and IL-3. Delbridge et al. recently showed that MCL-1 expression in murine HSPC is critically required to counteract PUMA-induced apoptosis. While mice lacking only one Mcl-1 allele in the hematopoietic system rapidly succumbed to bone marrow failure, additional deletion of both Puma alleles rescued all animals.26 However, our in vitro studies showed strong upregulation of PUMA mRNA in human CD34+ cells irrespective of the

Figure 8. Synergistic action of MCL-1 and BCL-XL inhibitors on human hematopoietic stem and progenitor cells. (A-B) Cord blood CD34+ cells were treated with increasing doses of the inhibitors S63845 and A-1155463 to determine synergism between the drugs. (A) Apoptosis was measured 24 h after treatment using annexin V/7-AAD. The percent specific apoptosis was calculated. (B) A dose-response matrix was determined using the web application SynergyFinder (n=4 from four independent experiments). (C-D) Human CD34+ cells, derived from either cord blood or bone marrow, were differentiated in MethoCult culture (1,000 cells seeded per plate) in the presence of a combination of the MCL-1 inhibitor S63845 and the BCL-XL inhibitor A-1155463 (0.1 mM or 1 mM of each). After 10 days, total colony numbers (C) and total cell numbers (D) were determined by light microscopy and hemocytometry, respectively. Bars represent mean ± standard error of mean; n=6-7 from six independent experiments. Mann-Whitney test: *P<0.05, **P<0.01 ***P<0.001, ****P<0.0001.

3146

haematologica | 2021; 106(12)

MCL-1 in human hematopoiesis

culture conditions and without being reflected by their MCL-1 addiction. This indicates that although there is a conserved role of MCL-1 for survival of murine and human HSPC, its function in human cells might not be restricted to inhibition of PUMA. We wondered whether also the most immature stem cells, those able to self-renew, depend on MCL-1 expression. While it is one limitation of in vitro model systems that dormant stem cells cannot be analyzed reliably, we were able to show by serial colony-forming assays that cells able to self-renew were dependent on BCL-XL rather than on MCL-1 expression. This is in contrast to findings published by Campbell et al., who attributed the loss of immature stem and progenitor cells upon MCL-1 inhibition to their reduced propensity to self-renew.42 While human cord blood-derived HSPC were severely affected by MCL-1 inhibition in our hands, other authors claimed good tolerability of S63845 in non-malignant hematopoietic cells.33-35 We hypothesized that this discrepancy was caused by cell-intrinsic differences between cord blood HSPC used in our study and bone marrow HSPC derived from aged persons used in other studies.33 We therefore compared CD34+ cells isolated from cord blood and bone marrow with regards to protein expression of anti-apoptotic BCL-2 proteins and their susceptibility to different BH3-mimetics. Adult HSPC had higher levels of BCL-2. Nevertheless, BCL-2 inhibition had no negative effects on colony-forming potential of aged HSPC. MCL-1 and BCL-XL levels were similar between the two types of HSPC while BFL1/A1 was not expressed. Importantly, inhibition of either MCL-1 or BCL-XL significantly impeded colony formation of both neonatal and adult HSPC and no cell type-specific difference could be noted. What do our results imply? First, the resistance of mature blood cells to S63845 suggests that immediate hematologic side effects might be mild in patients treated with MCL-1 inhibitors. The depletion of immature progenitor cells can, however, be associated with a relevant risk of severe cytopenias, although these might not occur immediately. To avoid excessive or even permanent bone marrow damage, repeated bone marrow analyses might be useful to detect hypocellularity as early as possible. Second, MCL-1 and BCL-XL inhibitors have synergistic rather than additive effects on human CD34+ HSPC. Synthetic lethality of two BH3-mimetics was shown for multiple tumors: The combination of BCL-2 and MCL-1 inhibitors showed synergistic effects in a vast variety of malignancies, including mantle cell lymphoma, T-cell prolymphocytic leukemia, multiple myeloma, high-risk Bcell acute lymphoblastic leukemia, AML and melanoma. BCL-XL and MCL-1 inhibitors were successfully com-

References 1. Venclexta FDA Approval History. https://www.drugs.com/history/venclexta.html. 2. Kollek M, Muller A, Egle A, Erlacher M. Bcl2 proteins in development, health, and disease of the hematopoietic system. FEBS J. 2016;283(15):2779-2810. 3. Labi V, Erlacher M, Kiessling S, Villunger A. BH3-only proteins in cell death initiation, malignant disease and anticancer therapy. Cell Death Differ. 2006;13(8):1325-1338.

haematologica | 2021; 106(12)

bined in multiple myeloma, melanoma, prostate cancer and multiple pediatric tumors.35,43-50 These observations make the combination of different BH3-mimetics very attractive. However, our results point strongly towards synthetic lethality also in healthy tissues. Specifically, combined BCL-XL and MCL-1 inhibition might be detrimental to healthy hematopoietic tissue. This fact should be kept in mind when new clinical trials are designed. Most data available so far, including our own, were acquired either in vitro or in artificial mouse models.33-35,51 Clinical data gathered from already initiated phase I trials (e.g., NCT03218683, NCT02675452, NCT03465540, NCT02979366) will provide better insight into both the anticancer efficacy of MCL-1 inhibitors and their frequent side effects.52 Importantly, some trials using small molecule MCL-1 inhibitors were placed on hold by the FDA in september 2019 because of cardiac toxicity.53 No data on hematologic toxicities have been published yet. Based on the very promising preclinical data28,34,54 one could speculate that the overall benefit-to-risk profile of MCL-1 inhibitors will be favorable, especially for tumors otherwise refractory to chemotherapy. In case of severe irreversible hematopoietic damage created by MCL-1 inhibitors, these compounds could still be used within high-dose chemotherapy regimens given prior to autologous or allogeneic hematopoietic stem cell transplantation. Disclosures No conflicts of interest to disclose. Contributions SB, SA, JFO, EMD, CM, YW, JMW and VRM designed and performed experiments; SB, SA, JFO, EMD, YW and ME analyzed and interpreted data; SB, SA and ME wrote the manuscript; LK, HS and MK provided human samples. Acknowledgments We thank the members of the DFG-FOR2036 consortium for insightful discussions, Nora Kaltenbach and Caroline Ambs for excellent technical assistance, the animal care-takers of the Center for Experimental Models and Transgenic Services (CEMT) for animal care and the Lighthouse Fluorescence Technologies Core Facility, Freiburg for cell sorting and maintenance of flow cytometers. We are grateful to Heike Pahl for providing us lentiviral constructs and Albert Gründer for support during initial cloning. Funding This work was supported by grants from the German Academic Exchange Service (DAAD) (funding program 57048249 to SA) and the German Research Foundation (DFGFOR2036, ER599/3-1 and ER599/3-2 to ME).

4. 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. 5. Park CM, Bruncko M, Adickes J, et al. Discovery of an orally bioavailable small molecule inhibitor of prosurvival B-cell lymphoma 2 proteins. J Med Chem. 2008;51(21):6902-6915. 6. Oltersdorf T, Elmore SW, Shoemaker AR, et al. An inhibitor of Bcl-2 family proteins induces regression of solid tumours. Nature. 2005;435(7042):677-681.

7. Roberts AW, Seymour JF, Brown JR, et al. Substantial susceptibility of chronic lymphocytic leukemia to BCL2 inhibition: results of a phase I study of navitoclax in patients with relapsed or refractory disease. J Clin Oncol. 2012;30(5):488-496. 8. Wilson WH, O'Connor OA, Czuczman MS, et al. Navitoclax, a targeted high-affinity inhibitor of BCL-2, in lymphoid malignancies: a phase 1 dose-escalation study of safety, pharmacokinetics, pharmacodynamics, and antitumour activity. Lancet Oncol. 2010;11(12):1149-1159.

3147

S. Bohler et al. 9. Souers AJ, Leverson JD, Boghaert ER, et al. ABT-199, a potent and selective BCL-2 inhibitor, achieves antitumor activity while sparing platelets. Nat Med. 2013;19(2):202208. 10. Veis DJ, Sorenson CM, Shutter JR, Korsmeyer SJ. Bcl-2-deficient mice demonstrate fulminant lymphoid apoptosis, polycystic kidneys, and hypopigmented hair. Cell. 1993;75(2):229-240. 11. Yamamura K, Kamada S, Ito S, Nakagawa K, Ichihashi M, Tsujimoto Y. Accelerated disappearance of melanocytes in bcl-2-deficient mice. Cancer Res. 1996;56(15):3546-3550. 12. Punnoose EA, Leverson JD, Peale F, et al. Expression profile of BCL-2, BCL-XL, and MCL-1 predicts pharmacological response to the BCL-2 selective antagonist venetoclax in multiple myeloma models. Mol Cancer Ther. 2016;15(5):1132-1144. 13. Wuilleme-Toumi S, Robillard N, Gomez P, et al. Mcl-1 is overexpressed in multiple myeloma and associated with relapse and shorter survival. Leukemia. 2005;19(7):12481252. 14. Jayappa KD, Portell CA, Gordon VL, et al. Microenvironmental agonists generate de novo phenotypic resistance to combined ibrutinib plus venetoclax in CLL and MCL. Blood Adv. 2017;1(14):933-946. 15. Birkinshaw RW, Gong JN, Luo CS, et al. Structures of BCL-2 in complex with venetoclax reveal the molecular basis of resistance mutations. Nat Commun. 2019;10(1):2385. 16. Blombery P. Mechanisms of intrinsic and acquired resistance to venetoclax in B-cell lymphoproliferative disease. Leuk Lymphoma. 2020;61(2):257-262. 17. 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. 18. Herling CD, Abedpour N, Weiss J, et al. Clonal dynamics towards the development of venetoclax resistance in chronic lymphocytic leukemia. Nat Commun. 2018;9(1): 727. 19. Beroukhim R, Mermel CH, Porter D, et al. The landscape of somatic copy-number alteration across human cancers. Nature. 2010;463(7283):899-905. 20. Kozopas KM, Yang T, Buchan HL, Zhou P, Craig RW. MCL1, a gene expressed in programmed myeloid cell differentiation, has sequence similarity to BCL2. Proc Natl Acad Sci U S A. 1993;90(8):3516-3520. 21. Bae J, Leo CP, Hsu SY, Hsueh AJ. MCL-1S, a splicing variant of the antiapoptotic BCL-2 family member MCL-1, encodes a proapoptotic protein possessing only the BH3 domain. J Biol Chem. 2000;275(33):2525525261. 22. Kim JH, Sim SH, Ha HJ, Ko JJ, Lee K, Bae J. MCL-1ES, a novel variant of MCL-1, associates with MCL-1L and induces mitochondrial cell death. FEBS Lett. 2009;583(17):27582764. 23. Perciavalle RM, Stewart DP, Koss B, et al. Anti-apoptotic MCL-1 localizes to the mitochondrial matrix and couples mitochondrial fusion to respiration. Nat Cell Biol.

3148

2012;14(6):575-583. 24. Rinkenberger JL, Horning S, Klocke B, Roth K, Korsmeyer SJ. Mcl-1 deficiency results in peri-implantation embryonic lethality. Genes Dev. 2000;14(1):23-27. 25. Opferman JT, Iwasaki H, Ong CC, et al. Obligate role of anti-apoptotic MCL-1 in the survival of hematopoietic stem cells. Science. 2005;307(5712):1101-1104. 26. Delbridge AR, Opferman JT, Grabow S, Strasser A. Antagonism between MCL-1 and PUMA governs stem/progenitor cell survival during hematopoietic recovery from stress. Blood. 2015;125(21):3273-3280. 27. Vo TT, Ryan J, Carrasco R, et al. Relative mitochondrial priming of myeloblasts and normal HSCs determines chemotherapeutic success in AML. Cell. 2012;151(2):344-355. 28. Kotschy A, Szlavik Z, Murray J, et al. The MCL1 inhibitor S63845 is tolerable and effective in diverse cancer models. Nature. 2016;538(7626):477-482. 29. Roelz R, Pilz IH, Mutschler M, Pahl HL. Of mice and men: human RNA polymerase III promoter U6 is more efficient than its murine homologue for shRNA expression from a lentiviral vector in both human and murine progenitor cells. Exp Hematol. 2010;38(9):792-797. 30. Labi V, Bertele D, Woess C, et al. Haematopoietic stem cell survival and transplantation efficacy is limited by the BH3only proteins Bim and Bmf. EMBO Mol Med. 2013;5(1):122-136. 31. Afreen S, Bohler S, Muller A, et al. BCL-XL expression is essential for human erythropoiesis and engraftment of hematopoietic stem cells. Cell Death Dis. 2020;11(1):8. 32. Ianevski A, He L, Aittokallio T, Tang J. SynergyFinder: a web application for analyzing drug combination dose-response matrix data. Bioinformatics. 2020;36 (8):2645. 33. Moujalled DM, Pomilio G, Ghiurau C, et al. Combining BH3-mimetics to target both BCL-2 and MCL1 has potent activity in preclinical models of acute myeloid leukemia. Leukemia. 2019;33(4):905-917. 34. Fiskus W, Cai T, DiNardo CD, et al. Superior efficacy of cotreatment with BET protein inhibitor and BCL2 or MCL1 inhibitor against AML blast progenitor cells. Blood Cancer J. 2019;9(2):4. 35. Ramsey HE, Fischer MA, Lee T, et al. A novel MCL1 inhibitor combined with venetoclax rescues venetoclax-resistant acute myelogenous leukemia. Cancer Discov. 2018;8(12):1566-1581. 36. Motoyama N, Wang F, Roth KA, et al. Massive cell death of immature hematopoietic cells and neurons in Bcl-x-deficient mice. Science. 1995;267(5203):1506-1510. 37. Motoyama N, Kimura T, Takahashi T, Watanabe T, Nakano T. bcl-x prevents apoptotic cell death of both primitive and definitive erythrocytes at the end of maturation. J Exp Med. 1999;189(11):1691-1698. 38. Delbridge AR, Aubrey BJ, Hyland C, et al. The BH3-only proteins BIM and PUMA are not critical for the reticulocyte apoptosis caused by loss of the pro-survival protein BCL-XL. Cell Death Dis. 2017;8(7):e2914. 39. 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. 40. Nakayama K, Nakayama K, Negishi I, Kuida K, Sawa H, Loh DY. Targeted disruption of Bcl-2 alpha beta in mice: occurrence of gray hair, polycystic kidney disease, and lymphocytopenia. Proc Natl Acad Sci U S A. 1994;91(9):3700-3704. 41. Matsuzaki Y, Nakayama K, Nakayama K, et al. Role of bcl-2 in the development of lymphoid cells from the hematopoietic stem cell. Blood. 1997;89(3):853-862. 42. Campbell CJ, Lee JB, Levadoux-Martin M, et al. The human stem cell hierarchy is defined by a functional dependence on Mcl-1 for self-renewal capacity. Blood. 2010;116(9): 1433-1442. 43. Prukova D, Andera L, Nahacka Z, et al. Cotargeting of BCL2 with venetoclax and MCL1 with S63845 is synthetically lethal in vivo in relapsed mantle cell lymphoma. Clin Cancer Res. 2019;25(14):4455-4465. 44. Smith VM, Lomas O, Constantine D, et al. Dual dependence on BCL2 and MCL1 in Tcell prolymphocytic leukemia. Blood Adv. 2020;4(3):525-529. 45. Mukherjee N, Amato CM, Skees J, et al. Simultaneously inhibiting BCL2 and MCL1 is a therapeutic option for patients with advanced melanoma. Cancers (Basel). 2020;12(8):2182. 46. Abdul Rahman SF, Muniandy K, Soo YK, et al. Co-inhibition of BCL-XL and MCL-1 with selective BCL-2 family inhibitors enhances cytotoxicity of cervical cancer cell lines. Biochem Biophys Rep. 2020;22: 100756. 47. Slomp A, Moesbergen LM, Gong JN, et al. Multiple myeloma with 1q21 amplification is highly sensitive to MCL-1 targeting. Blood Adv. 2019;3(24):4202-4214. 48. Algarin EM, Diaz-Tejedor A, Mogollon P, et al. Preclinical evaluation of the simultaneous inhibition of MCL-1 and BCL-2 with the combination of S63845 and venetoclax in multiple myeloma. Haematologica. 2020;105(3):e116-e120. 49. Lee EF, Harris TJ, Tran S, et al. BCL-XL and MCL-1 are the key BCL-2 family proteins in melanoma cell survival. Cell Death Dis. 2019;10(5):342. 50. Arai S, Jonas O, Whitman MA, Corey E, Balk SP, Chen S. Tyrosine kinase inhibitors Increase MCL1 degradation and in combination with BCLXL/BCL2 inhibitors drive prostate cancer apoptosis. Clin Cancer Res. 2018;24(21):5458-5470. 51. Brennan MS, Chang C, Tai L, et al. Humanized Mcl-1 mice enable accurate preclinical evaluation of MCL-1 inhibitors destined for clinical use. Blood. 2018;132(15): 1573-1583. 52. Hird AW, Tron AE. Recent advances in the development of Mcl-1 inhibitors for cancer therapy. Pharmacol Ther. 2019;198:59-67. 53. www.ashclinicalnews.org; last accessed Nov 1, 2019. 54. Tron AE, Belmonte MA, Adam A, et al. Discovery of Mcl-1-specific inhibitor AZD5991 and preclinical activity in multiple myeloma and acute myeloid leukemia. Nat Commun. 2018;9(1):5341.

haematologica | 2021; 106(12)

ARTICLE

Iron Metabolism & its Disorders

Cell-specific expression of Hfe determines the outcome of Salmonella enterica serovar Typhimurium infection in mice

Ferrata Storti Foundation

Manfred Nairz,1* Christoph Metzendorf,2,3* Maja Vujic-Spasic,2,3,4+ Anna-Maria Mitterstiller,1+ Andrea Schroll,1 David Haschka,1 Alexander Hoffmann,1,5 Laura von Raffay,1 Richard Sparla,2,3 Christian W. Huck,6 Heribert Talasz,7 Patrizia L. Moser,8 Martina U. Muckenthaler2,3# and Günter Weiss1,5# 1 Department of Internal Medicine II, Infectious Diseases, Immunology, Rheumatology, Pneumology, Medical University of Innsbruck, Innsbruck, Austria; 2Department of Pediatric Hematology, Oncology and Immunology, University of Heidelberg, INF 350, Heidelberg, Germany; 3Molecular Medicine Partnership Unit, Heidelberg, Germany; 4 Institute of Comparative Molecular Endocrinology, Ulm University, Ulm, Germany; 5 Christian Doppler Laboratory for Iron Metabolism and Anemia Research, Medical University of Innsbruck, Innsbruck, Austria; 6Institute for Analytical Chemistry and Radiochemistry, University of Innsbruck, Innsbruck, Austria; 7Biocenter, Division of Clinical Biochemistry, Medical University of Innsbruck, Innsbruck, Austria and 8Institute of Pathology, INNPATH, Innsbruck, Austria

Haematologica 2021 Volume 106(12):3149-3161

*MN and CM contributed equally as co-first authors +MV-S and A-MM contributed equally as co-second authors # MM and GW contributed equally as co-senior authors

ABSTRACT

utations in HFE cause hereditary hemochromatosis type I hallmarked by increased iron absorption, iron accumulation in hepatocytes and iron deficiency in myeloid cells. HFE encodes an MHC-I like molecule, but its function in immune responses to infection remains incompletely understood. Here, we investigated putative roles of Hfe in myeloid cells and hepatocytes, separately, upon infection with Salmonella Typhimurium, an intracellular bacterium with iron-dependent virulence. We found that constitutive and macrophage-specific deletion of Hfe protected infected mice. The propagation of Salmonella in macrophages was reduced due to limited intramacrophage iron availability for bacterial growth and increased expression of the anti-microbial enzyme nitric oxide synthase-2. By contrast, mice with hepatocyte-specific deletion of Hfe succumbed earlier to Salmonella infection because of unrestricted extracellular bacterial replication associated with high iron availability in the serum and impaired expression of essential host defense molecules such as interleukin-6, interferon-g and nitric oxide synthase-2. Wild-type mice subjected to dietary iron overload phenocopied hepatocyte-specific Hfe deficiency suggesting that increased iron availability in the serum is deleterious in Salmonella infection and underlies impaired host immune responses. Moreover, the macrophage-specific effect is dominant over hepatocytespecific Hfe-depletion, as Hfe knockout mice have increased survival despite the higher parenchymal iron load associated with systemic loss of Hfe. We conclude that cell-specific expression of Hfe in hepatocytes and macrophages differentially affects the course of infections with specific pathogens by determining bacterial iron access and the efficacy of antimicrobial immune effector pathways. This may explain the high frequency and evolutionary conservation of human HFE mutations.

Introduction Most patients with hereditary hemochromatosis (HH) show homozygous C282Y missense mutations in the gene HFE.1,2 They are hallmarked by parenchymal iron deposition particularly in hepatocytes, cardiomyocytes and pancreatic

haematologica | 2021; 106(12)

Correspondence: MARTINA U. MUCKENTHALER martina.muckenthaler@med.uni-heidelberg.de GÜNTER WEISS guenter.weiss@i-med.ac.at Received: December 3, 2019. Accepted: September 30, 2020. Pre-published: October 13, 2020. https://doi.org/10.3324/haematol.2019.241745

3149

M. Nairz et al.

acinar cells, leading to organ damage. Conversely, monocytes and macrophages are iron-deficient in type I HH.3-5 An allelic frequency of approximately 5-10% renders the HFE C282Y missense mutation the most common genetic defect in individuals of Northwestern European ancestry. It has been hypothesized that the mutation may protect from iron-deficiency and/or infections; thus conferring an evolutionary advantage to healthy heterozygous carriers.6,7 Several mechanisms by which the HFE protein controls systemic iron balance have been proposed: early studies have shown that HFE, in association with b2-microglobulin, directly interacts with transferrin receptor-1 (TFR1) on the cell surface8 and lowers its affinity for transferrin-bound iron (TBI). Once the iron regulatory hormone hepcidin had been discovered, it became apparent that the HFE C282Y mutation causes systemic hepcidin deficiency and its consequences.9 The HFE mutation disrupts iron-inducible BMP/SMAD (for bone morphogenetic protein/suppressor of mothers against decapentaplegic) signaling and prevents appropriate hepcidin transcription.1,2 The relative lack of hepcidin then causes unrestricted dietary iron absorption by the duodenum and increased iron export from iron-recycling macrophages due to the stabilization of the iron exporter ferroportin (FPN)-1.10,11 As a result, iron accumulates in parenchymal cells where it causes tissue damage by toxic radical formation.12 HFE is an MHC-I like protein, but so far it has remained unclear whether it plays a role in immune function and/or host-pathogen interaction. HFE-deficient monocytes and macrophages are iron poor.3,4,13 Possible explanations include reduced TBI uptake, increased iron export via FPN1 as a consequence of decreased hepcidin levels or increased synthesis of the siderophore-iron binding peptide lipocalin (LCN)-2.3,5,13 For almost all bacteria, iron is essential as it stimulates growth and thus impacts on the course and outcome of many infectious diseases.14 However, iron requirements, iron uptake strategies and proliferation kinetics may greatly vary between bacterial species, possibly explaining species-specific effects on infection outcomes.15 In mice, constitutive Hfe deficiency partially protects from S. enterica Typhimurium (S. Tm.) infection.13 By contrast, Hfe deficient mice are more susceptible to Mycobacterium avium infection.16 Furthermore, human monocyte-derived macrophages from patients with HH limit iron availability for intracellular Mycobacterium tuberculosis, resulting in an improved control of infection.17 On the other hand, individuals with HH type I are highly susceptible to infection with Yersinia species, whose virulence is iron-dependent, as documented by case reports of human subjects and by mouse models.18,19 These diverse outcomes are counterintuitive given that all three pathogens, Salmonella, Mycobacterium and Yersinia, share a predominately intracellular lifestyle pointing to the importance of cell- and tissue-specific iron distribution for susceptibility to these infections.20 Because Hfe exerts contrasting effects in different infectious diseases, we asked whether Hfe plays a cell typespecific role during infection and whether this is linked to alterations of tissue iron distribution or associated with iron-independent effects of Hfe. We herein demonstrate that macrophage-specific deletion of Hfe (LysMCre+ Hfefl/fl) recapitulates the antibacterial phenotype of constitutive Hfe-/- mice in response to S. Tm. infection. By contrast, 3150

exclusive deletion of Hfe in hepatocytes (AlfpCre+ Hfefl/fl) is associated with an adverse outcome of S. Tm. infection. These contrasting cell type-specific effects of Hfe-deficiency correlate with bacterial iron availability and anti-microbial effector immune functions. Our findings support the idea that Hfe controls iron concentrations in the microenvironment thus differentially affecting immune effector mechanisms and bacterial growth in intra- and extracellular compartments.

Methods Salmonella infection in vivo All infection experiments were performed according to the guidelines of the Medical University of Innsbruck and the Austrian Ministry for Science and Education based on the Austrian Animal Testing Act of 1988 (approvals BMWFW-66.011/0074-C/GT/2007, 66.011/0154II/3b/2010 and 66.011/0031-WF/V/3b/2015). Male mice were used at 12-16 weeks of age and infected by intraperitoneal (i.p.) injection with 500 colony forming units (CFU) of S. Tm. diluted in 200 mL of phosphate buffered saline (PBS). Unless otherwise specified, S. Tm. Wild-type (WT), strain ATCC 14028s was used for the experiments. Where appropriate, mice were fed an iron adequate control diet (C1000 from Altromin containing 180 mg per g) or an ironenriched diet (C1038 from Altromin supplemented with 25 mg carbonyl iron per g). After 3 weeks, mice were infected by i.p. injection with 500 CFU of S. Tm. diluted in 200 mL of PBS as detailed in the Online Supplementary Methods.

In vitro experiments The isolation of bone marrow-derived macrophages (BMDM) was performed as detailed in the Online Supplementary Methods.

RNA extraction and quantitative real-time polymerase chain reaction Preparation of total RNA, reverse transcription and quantification of mRNA expression by quantitative Taqman real-time polymerase chain reaction (qRT-PCR) was performed as described.21 Results were first normalized using the housekeeping gene Hprt and then divided by the means of the control group (WT Hfe+/+ or Cre– mice as appropriate) to obtain expression data that is relative to the respective control group. Sequences of primers and probes are listed in the Online Supplementary Methods.

Measurement of iron and protein concentrations Measurement of tissue iron concentrations has been described in detail.22 The serum iron concentration was quantified using the QuantiChrom Iron Assay Kit (BioAssay Systems). Intracellular iron concentrations were determined in adherent bone marrow macrophages by atomic absorption spectrometry as described.23 The quantification of protein levels in sera and tissues is detailed in the Online Supplementary Methods.

Statistical analysis Statistical analysis was carried out using a GraphPad Prism statistical package and Microsoft Excel. We determined significance by unpaired two-tailed Student’s t-test or Mann-Whitney test to assess data, where only two haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

groups existed. For the comparison of organ bacterial loads and mRNA expression, data were log-transformed prior to Student’s t-test. ANOVA with Bonferroni correction was used when more than two groups existed. Survival was compared by log-rank test. Generally, P-values less than 0.05 were considered significant.

Results Hepatocyte-specific Hfe deletion stimulates extracellular growth of Salmonella Typhimurium We previously reported that mice lacking Hfe in all cell types (Hfe-/- mice) were partially protected from S. Tm. Infection.13,20 Consistently, we could recapitulate this finding in a different strain of Hfe-/- mice in which exons 3-524 rather than exons 2-313 of Hfe were deleted. We found that also these Hfe-/- mice survived significantly longer (Figure 1A) and carried reduced numbers of bacteria in spleen, liver and serum in response to S. Tm infection when compared to Hfe+/+ littermates (Figure 1B to D). In order to delineate in which cell type the absence of Hfe confers protection from infection, we next analyzed mice with selective Hfe-deficiency in hepatocytes (referred to as AlfpCre+ Hfefl/fl). Previous analyses of the AlfpCre+ Hfefl/fl line showed an iron phenotype comparable to Hfe-/- mice,25 with elevated iron levels in serum and liver and iron deficiency in the spleen. AlfpCre+ Hfefl/fl and control mice (AlfpCre- Hfefl/fl) were infected with S. Tm. and survival time was monitored for 14 days (336 hours). Unexpectedly and in contrast to the previous findings in Hfe-/- mice, we observed significantly shortened survival in the AlfpCre+ Hfe-/- mice (Figure 1E). Bacterial burden in spleen and liver was not substantially altered, when compared to control mice (Figure 1F and G). By contrast, the number of bacteria circulating in the serum was significantly higher in AlfpCre+ Hfefl/fl mice (Figure 1H). This finding suggested that Hfe-deficiency in hepatocytes does not confer protection against S. Tm. infection related death but even aggravates the infection phenotype.

Macrophage-specific Hfe-deletion phenocopies the protective effect of constitutive Hfe deletion in mice infected with Salmonella Typhimurium We next tested the response to S. Tm. infection in mice lacking Hfe in myeloid cells (referred to as LysMCre+ Hfefl/fl)25 in comparison to control mice (LysMCre– Hfefl/fl). Interestingly, macrophage-specific Hfe depletion fully recapitulated the protective effect observed in Hfe-/- mice, including prolonged survival (Figure 1I) and reduced bacterial load in spleen, liver and serum (Figure 1J to L). Importantly, the alleles required for tissue-specific recombination to generate the cell type-specific Hfe-depletion models, LysMCre (macrophage-specific Cre-recombinase expression) and AlfpCre (hepatocyte-specific Cre-recombinase expression) alone had no effect on survival and bacterial burden in the spleen, liver and serum (Online Supplementary Figure S1A to D and 26), excluding non-specific effects of the Cre-recombinases. We conclude that the lack of Hfe in myeloid cells is sufficient to protect mice from S. Tm. infection related consequences. This finding demonstrates an important extra-hepatic function of Hfe in vivo. haematologica | 2021; 106(12)

Salmonella-infection of LysMCre+ Hfefl/fl mice causes iron depletion in macrophages In order to understand the mechanism underlying divergent disease outcomes of S. Tm. infection in AlfpCre+ Hfefl/fl and LysMCre+ Hfefl/fl mice, we analyzed iron-related parameters. Iron localization was detected in tissue sections of Salmonella-infected mice by Prussian blue staining and tissue iron levels were quantified by colorimetric measurement. S. Tm.-infected Hfe-/- (Figure 2A and B) and LysMCre+ Hfefl/fl mice (Figure 2G and H) showed reduced iron levels in the spleen consistent with the protective phenotype observed in these mouse strains. This was not apparent in infected AlfpCre+ Hfefl/fl mice (Figure 2D and E). By contrast, infected Hfe-/- (Online Supplementary Figure S2A and B) and AlfpCre+ Hfefl/fl mice (Online Supplementary Figure S2C and D) showed hepatocellular iron accumulation, while liver iron levels were normal in infected LysMCre+ Hfefl/fl mice (Online Supplementary Figure S2E and F). Importantly, the reduction of splenic iron levels in infected Hfe-/- and LysMCre+ Hfefl/fl mice correlated with diminished intracellular iron levels in bone marrow macrophages (Figure 2C and I), while AlfpCre+ Hfefl/fl bone marrow macrophages had a normal iron content (Figure 2F). This finding suggests that upon Salmonella infection, macrophages lacking Hfe show reduced iron levels.

High serum iron in AlfpCre+ Hfefl/fl mice allows for increased proliferation of Salmonella Hfe-/- and AlfpCre+ Hfefl/fl mice infected with S. Tm. WT for 72 hours showed elevated serum iron levels compared to Hfe+/+ or AlfpCre– Hfefl/fl mice, respectively (Figure 3A and B). In contrast, serum iron levels in infected LysMCre+ Hfefl/fl mice were comparable to infected LysMCre– Hfefl/fl mice (Figure 3C). Notably, in the setting of Salmonella infection, serum levels of hepcidin-1 were not different between the mouse strains (Figure 3D to F). Moreover, Salmonella-infected Hfe-/- mice presented with increased serum concentrations of the siderophore-capturing peptide Lcn2 (Figure 3G) while hepatocyte-specific (Figure 3H) or macrophage-specific (Figure 3I) Hfe deletion did not affect serum Lcn2 levels. Thus, the presence of Hfe in hepatocytes is necessary and sufficient to limit serum iron levels both in steady state25 and in response to S. Tm. infection. Moreover, hepcidin-1 induction in response to Salmonella infection is appropriate in mice lacking Hfe. In contrast, the enhanced production of Lcn2 is only observed in the complete absence of Hfe13 suggesting that different cell-types mediate iron- and immune-regulatory effects of Hfe.

Salmonella iron acquisition pathways differently affect extracellular proliferation S. Tm. is a bacterial pathogen with dual lifestyle. First, S. Tm. is able to persist and replicate extracellularly, e.g., on contaminated food and surfaces, in the gut lumen and in the serum. Early after the invasion of a murine host, S. Tm. preferentially infects macrophages to propagate intracellularly. We therefore investigated how serum iron availability in Hfe-/-, AlfpCre+ Hfefl/fl and LysMCre+ Hfefl/fl mice may affect bacterial proliferation. We spiked RPMI medium with 10% of serum from uninfected mice of all three strains and inoculated spiked samples with bacteria. We used S. Tm. WT and isogenic mutants lacking either single or all three major bacterial iron uptake systems (enterobactin, feo and sitABCD).27 In addition, we includ3151

M. Nairz et al.

ed serum-spiked RPMI treated with 100 mM desferasirox (DFX) to deplete the medium of chelatable iron. Alternatively, we added 100 mM FeSO4 to saturate any iron-binding factors (e.g. transferrin, lactoferrin and Lcn2). S. Tm. WT growth was only inhibited by serum from Hfe-/- mice, possibly due to the presence of high Lcn2.13 In serum from AlfpCre+ Hfefl/fl mice, bacterial growth was strongly enhanced, while it was not affected in serum from LysMCre+ Hfefl/fl (Figure 4A). In addition, growth of S. Tm. WT was strongly restricted by the presence of the iron chelator DFX and enhanced by the addition of FeSO4, independent of the Hfe status of the mice the sera were derived from (Figure 4A). In liquid cultures of iron uptake mutant S. Tm. strains, growth was most pronouncedly

inhibited in the case of the triple mutant (entC, feo and sitABCD deletion). Importantly, we saw that the iron-rich serum of AlfpCre+ Hfefl/fl mice facilitated extracellular growth of S. Tm., an effect that was reduced by the lack of all three iron uptake systems (Figure 4B) or abolished by iron chelation (Figure 4A). Notably, the addition of recombinant murine Lcn2 reduced the growth of S. Tm. in a dose-dependent fashion, even though it did not abolish the differences between AlfpCre– Hfefl/fl and AlfpCre+ Hfefl/fl mice (Figure 4C). This suggests that in the presence of high Lcn2 concentrations, iron uptake pathways of Salmonella not targeted by Lcn2, such as the feo and sitABCD systems, are able to compensate. Apparently, feo and sitABCD can maintain a sufficient supply of iron for

Figure 1. Cell-type specific effect of Hfe deletion on the course of systemic Salmonella infection. Survival (A, E and I) and bacterial load in spleen (B, F and J), liver (C, G and K) and serum (D, H and L) of Hfe-/- (A-D) mice and mice lacking Hfe in hepatocytes (AlfpCre+ Hfefl/fl in E-H) or macrophages (LysMCre+ Hfefl/fl in I-L), respectively, compared to matched controls. Mice were infected with 500 colony forming units (CFU) of S. enterica serovar Typhimurim by intraperitoneal injection and monitored for 14 days (336 hours ). Data represent two independent experiments. Statistics: survival data between control and mutant mice were compared using the Log-rank (Mantel-Cox) Test. n=18 for Hfe+/+, n=16 for Hfe-/-, n=13 for AlfpCre- Hfefl/fl, n=9 for AlfpCre+ Hfefl/fl, n=16 for LysMCre- Hfefl/fl, n=15 for LysMCre+ Hfefl/fl. Log CFU data of tissue bacterial load of randomly selected mice euthanized after 72 hours of Salmonella infection were compared using student t-test. CFU data of serum bacterial load were compared by Mann-Whitney test. n=12 for Hfe+/+, n=12 for Hfe-/-, n=20 for AlfpCre- Hfefl/fl, n=14 for AlfpCre+ Hfefl/fl, n=10 for LysMCre- Hfefl/fl, n=10 for LysMCre+ Hfefl/fl.

3152

haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

bacteria when enterobactin incorporation is blocked by Lcn2. Thus, the proliferation advantage of S. Tm. in extracellular compartments of AlfpCre+ Hfefl/fl mice is a specific effect of increased iron availability even though it is not linked to a specific bacterial iron uptake pathway.

Hfe does not affect the phagocytic activity of macrophages Lower bacterial numbers in Hfe-/- and LysMCre+ Hfefl/fl macrophages could theoretically be explained by altered phagocytosis. Therefore, we next compared the phagocytic capacity of bone marrow-derived macrophages isolated from WT and Hfe-/- mice. However, differences were

not detected, suggesting that Hfe in macrophages does not affect the phagocytic capacity of macrophages (Online Supplementary Figure S3).

Cell type-specific Hfe deletions differentially affect iron homeostasis and anti-microbial immune gene and protein expression in spleen and liver So far, our results indicate that lack of Hfe in macrophages is sufficient to suppress intracellular growth of S. Tm., while hepatocyte-specific Hfe depletion supports iron-dependent extracellular growth of Salmonella. However, the finding that Hfe-/- and AlfpCre+ Hfefl/fl mice show comparable serum iron levels while resulting in

Figure 2. Reduced iron content in spleen and bone marrow macrophages in the absence of Hfe. Spleen sections of Hfe-/- mice (A), AlfpCre+ Hfefl/fl mice (D) and LysMCre+ Hfefl/fl mice (G) infected for 72 hours with Salmonella were stained by Prussian blue to analyze iron distribution. Scale bars: 200 mM. Iron content in infected spleen (B, E and H) and bone marrow macrophages (C, F and I) was measured and normalized for protein content. Data were compared by Mann-Whitney test. n=12 for Hfe+/+, n=12 for Hfe-/-, n=13-20 for AlfpCre– Hfefl/fl, n=11-14 for AlfpCre+ Hfefl/fl, n=10 for LysMCre– Hfefl/fl, n=10 for LysMCre+ Hfefl/fl.

haematologica | 2021; 106(12)

3153

M. Nairz et al.

contrasting infection outcomes suggested dominant effects of Hfe in myeloid cells. In order to identify the responsible mechanisms, we monitored gene response patterns of iron and immune genes in spleens and livers of S. Tm.-infected mice. As expected, mRNA expression of ferritin heavy chain (H-Ft) was significantly decreased in the spleen (hallmarked by iron deficiency) and increased in the liver (hallmarked by iron overload) of infected Hfe-/- mice. However, H-Ft remained unchanged in the other infected Hfe-models (Tables 1 and 2). Likewise, in livers of Hfe-/- and AlfpCre+ Hfefl/fl mice, we found significantly reduced expression of Tfr1 (2x and 1.3x in Hfe-/- and AlfpCre+ Hfefl/fl, respectively), consistent with hepatic iron overload (Table 2). We next studied the expression of central immune genes involved in the control of infection with intramacrophage

bacteria. Importantly, mRNA expression of Il-6 and Ifn-g was decreased in the spleen in AlfpCre+ Hfefl/fl, but neither in Hfe-/- nor LysMCre+ Hfefl/fl mice (Table 1). Il-10 was decreased in livers of Hfe-/- mice and Tnf was increased in livers of LysMCre+ Hfefl/fl mice (Table 2). By contrast, Nos2 (for nitric oxide synthase-2, AKA inducible Nos) expression was increased in mice lacking Hfe either globally or in macrophages, specifically, and decreased in AlfpCre+ Hfefl/fl mice (Table 1). Therefore, reduced macrophage iron levels selectively promote the expression of splenic Nos2, whereas high serum iron has a broader inhibitory effect on antimicrobial host responses in the spleen. Importantly, the protein levels of iron and immune genes mirrored the mRNA expression levels in both spleen (Figure 5) and liver (Online Supplementary Figure S4). For

Table 1. Gene expression in spleen of S. Tm. injected mice, 72 h post infection. Gene expression was normalized to expression of Hprt and is relative to the respective control group (means +/- SD).

Gene Iron genes Dmt1 Fpn1 Hamp1 H-Ft Tfr1 Hmox1 Immune genes Lcn2 Nos2 phox-p47 TNF IL-6 IL-10 IFN-g

WT (n = 10)

Hfe-KO (n = 12)

AlfpCre– (n = 20)

AlfpCre+ (n = 14)

LysMCre– (n = 15)

LysMCre+ (n = 15)

1 +/- 1.06 1 +/- 0.38 1 +/- 1.11 1 +/- 0.22 1 +/- 0.43 1 +/- 0.68

2.85 +/- 3.1 1.2 +/- 0.88 1.4 +/- 1.58 0.41 +/- 0.2 1.23 +/- 1.41 0.77 +/- 0.61

ns ns ns **** ns ns

1 +/- 0.42 1 +/- 0.26 1 +/- 1.02 1 +/- 0.72 1 +/- 0.57 1 +/- 0.67

0.98 +/- 0.5 1.28 +/- 0.44 1.12 +/- 1.15 0.63 +/- 0.32 1.55 +/- 1.05 0.9 +/- 0.37

ns * ns ns ns ns

1 +/- 0.55 1 +/- 0.46 1 +/- 0.95 1 +/- 0.68 1 +/- 0.84 1 +/- 0.65

0.73 +/- 0.33 0.87 +/- 0.49 1.14 +/- 0.58 0.9 +/- 0.62 1.16 +/- 0.87 0.79 +/- 0.53

ns ns ns ns ns ns

1 +/- 1.69 1 +/- 1.06 1 +/- 0.96 1 +/- 0.66 1 +/- 0.73 1 +/- 0.81 1 +/- 0.71

1.1 +/- 0.89 4.23 +/- 2.09 0.88 +/- 0.49 1.29 +/- 2.29 0.74 +/- 1.04 2.35 +/- 2.93 2.74 +/- 3.18

ns *** ns ns ns ns ns

1 +/- 1.72 1 +/- 0.81 1 +/- 0.55 1 +/- 0.62 1 +/- 0.31 1 +/- 1.12 1 +/- 0.28

0.61 +/- 0.61 0.37 +/- 0.23 0.61 +/- 0.25 0.82 +/- 0.71 0.6 +/- 0.36 0.9 +/- 0.75 0.45 +/- 0.29

ns ** * ns *** ns ****

1 +/- 0.72 1 +/- 0.83 1 +/- 0.31 1 +/- 0.5 1 +/- 0.66 1 +/- 0.79 1 +/- 0.72

1.43 +/- 0.85 1.87 +/- 1.2 1.14 +/- 0.33 1.02 +/- 0.38 1.02 +/- 0.5 1.2 +/- 0.63 1.17 +/- 0.66

ns * ns ns ns ns ns

Statistics: unpaired, two-sided student t-test; *P<0.05; **P<0.01; ***P<0.005; ****P<0.0001; WT: wild-type; ns: not significant.

Table 2. Gene expression in livers of S. Tm. injected mice, 72 h post infection. Gene expression was normalized to expression of Hprt and is relative to the respective control group (means +/- SD).

Gene Iron genes Dmt1 Fpn1 Hamp1 H-Ft Tfr1 Hmox1 Immune genes Lcn2 Nos2 phox-p47 TNF IL-6 IL-10 IFN g

WT (n = 10)

Hfe-KO (n = 12)

AlfpCre– (n = 20)

AlfpCre+ (n = 14)

LysMCre– (n = 15)

LysMCre+ (n = 15)

1 +/- 0.24 1 +/- 0.21 1 +/- 0.59 1 +/- 0.24 1 +/- 0.23 1 +/- 0.29

1.06 +/- 0.64 1.08 +/- 0.47 1.51 +/- 1.05 1.59 +/- 0.35 0.49 +/- 0.19 0.88 +/- 0.44

ns ns ns *** **** ns

1 +/- 0.42 1 +/- 0.2 1 +/- 0.46 1 +/- 0.26 1 +/- 0.26 1 +/- 0.37

0.98 +/- 0.68 1.61 +/- 0.63 0.91 +/- 0.46 1 +/- 0.33 0.75 +/- 0.36 1.32 +/- 0.79

ns *** ns ns * ns

1 +/- 0.57 1 +/- 0.38 1 +/- 0.45 1 +/- 0.37 1 +/- 0.49 1 +/- 0.49

1.08 +/- 0.68 1.08 +/- 0.34 1.01 +/- 0.36 0.83 +/- 0.37 0.77 +/- 0.46 0.99 +/- 0.46

ns ns ns ns ns ns

1 +/- 0.9 1 +/- 1.34 1 +/- 0.56 1 +/- 1.02 1 +/- 0.17 1 +/- 0.15 1 +/- 0.24

11.8 +/- 6.3 1.38 +/- 2.4 1.19 +/- 0.79 1.33 +/- 1.5 0.94 +/- 0.27 0.83 +/- 0.17 1.12 +/- 0.25

**** ns ns ns ns * ns

1 +/- 1.96 1 +/- 0.86 1 +/- 0.55 1 +/- 0.72 1 +/- 0.56 1 +/- 1.05 1 +/- 0.47

2.36 +/- 3.81 0.96 +/- 1.04 0.99 +/- 0.47 1.14 +/- 1.23 0.99 +/- 0.86 1.47 +/- 1.36 0.98 +/- 0.69

ns ns ns ns ns ns ns

1 +/- 2.32 1 +/- 1.06 1 +/- 0.66 1 +/- 0.51 1 +/- 1.1 1 +/- 0.85 1 +/- 0.49

1.15 +/- 1.7 1.17 +/- 1.25 1.05 +/- 0.62 1.65 +/- 0.93 1.11 +/- 1.02 0.76 +/- 0.75 0.97 +/- 0.57

ns ns ns * ns ns ns

Statistics: unpaired, two-sided student t-test; *P<0.05; **P< 0.01; ***P< 0.005; ****P<0.0001; WT: wild-type; ns: not significant.

3154

haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

Hepatocyte-specific deletion of Hfe impairs cytokine formation

instance, H-Ft protein expression was lower in the spleen of Hfe-/- mice compared to Hfe+/+ littermates (Figure 5B). Furthermore, splenic Nos2 protein levels were higher in Hfe-/- and LysMCre+ Hfefl/fl mice and lower in AlfpCre+ Hfefl/fl mice as compared to their respective counterparts expressing Hfe (Figure 5F). In addition, H-Ft (Online Supplementary Figure S4B) was higher and Tfr1 protein (Online Supplementary Figure S4C) was lower in both Hfe-/- mice and AlfpCre+ Hfefl/fl mice in comparison to the corresponding controls.

haematologica | 2021; 106(12)

Finally, we aimed at better understanding why AlfpCre+ Hfefl/fl mice succumb to early death. Based on the result of organ-specific immune gene analyses (Tables 1 and 2), we assessed the levels of these circulating mediators of immunity as well as markers of liver synthesis and damage: C3 is a complement factor produced by hepatocytes and essential for the activation of the membrane attack complex which then destroys bacterial cell walls.

Figure 3. Serum iron and hepcidin-1 levels are differentially affected by Hfe. Serum iron (A to C), hepcidin-1 (D and F) and Lcn2 (G to I) concentrations of the mice infected for 72 hours were compared by Mann-Whitney test. n=9-12 for Hfe+/+, n=9-12 for Hfe-/-, 20 for AlfpCre– Hfefl/fl, n=11-13 for AlfpCre+ Hfefl/fl, n=10-15 for LysMCre– Hfefl/fl, n=10-15 for LysMCre+ Hfefl/fl.

3155

M. Nairz et al.

However, C3 levels were not affected by high serum or parenchymal iron (Figure 6A). IL-6, the key cytokine for the initiation of the acute-phase response, was reduced in the serum (Figure 6B) but not in the liver (Online Supplementary Figure 4E) of AlfpCre+ Hfefl/fl mice (Figure 6B). This and the unaltered production of hepcidin-1 (Figure 3E and Table 2) suggest that the acute-phase response is intact in AlfpCre+ Hfefl/fl mice. Moreover, glutamate-pyruvate transaminase (GPT) was reduced in AlfpCre+ Hfefl/fl mice, ruling out increased iron-induced hepatic injury (Online Supplementary Figure S4A). Rather, a specific defect in the pro-inflammatory cytokine output was associated

with the poor outcome of Salmonella-infected AlfpCre+ Hfefl/fl mice. Specifically, serum IFN-g concentrations (Figure 6C) were significantly lower in AlfpCre+ Hfefl/fl mice and may have contributed to the insufficient induction of cellular effector mechanisms such as Nos2 in the spleen (Table 1).

Increased serum iron due to dietary iron overload enhances extracellular Salmonella growth and reduces IFN-g levels In order to further investigate whether the reduced cytokine production in AlfpCre+ Hfefl/fl mice and may have

A Figure 4. Bacterial proliferation is affected by Hfe. RPMI was spiked with 10% sera of naïve mice of the indicated genotypes. Spiked RPMI was inoculated with wild-type (WT) S. enterica Typhimurium (S. Tm.) and its isogenic derivatives mutated in one of three or all three iron uptake systems (entC, sitABCD, feo). Where applicable, deferasirox (DFX), ferrous sulfate (FeSO4) and recombinant murine Lcn2 (rmuLcn2) was added. Liquid cultures were assessed for extracellular bacterial proliferation (G to I) using the optical density at 600 nm (OD600). **P<0.01, ***P<0.001 for the comparison between mouse genotypes, #P<0.05, ##P<0.01 and ###P<0.001 for the comparison to solvent (Ctrl) or the S. Tm. WT strain as applicable. n=4-6 independent experiments.

3156

haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

Figure 5. Figure continued on following page.

haematologica | 2021; 106(12)

3157

M. Nairz et al.

Figure 5. Cell-type specific effects of Hfe on protein expression in the spleen. Spleen homogenates were prepared to quantify the expression of iron and immune relevant proteins by enzyme-linked immunosorbent assay. Protein levels of Fpn1 (A), H-Ft (B), Tfr1 (C), Lcn2 (D), IFN-g (E) and Nos2 (F), normalized for total protein content, are depicted as means ± standard deviations. Statistically significant differences as calculated by unpaired, two-sided student t-test are indicated. n=10 for Hfe+/+, n=12 for Hfe-/-, n=20 for AlfpCre– Hfefl/fl, n=14 for AlfpCre+ Hfefl/fl, n=15 for LysMCre– Hfefl/fl, n=15 for LysMCre+ Hfefl/fl.

contributed to the insufficient induc mice is linked to increased cellular iron levels, we next maintained WT mice on an iron adequate (IA) or high iron diet for three weeks to induce iron overload (IO) prior to Salmonella infection. We observed an increased bacterial load in the serum (Figure 6D), spleen and liver (Online Supplementary Figure S5B and C) along with unaltered IL-6 (Figure 6E) but reduced IFN-g concentrations in the serum (Figure 6F). This finding suggests that reduced levels of IFN-g, the central cytokine orchestrator of immune responses against intracellular bacteria,28 are a direct consequence of increased serum iron.

Discussion The challenge of mice with the facultative intracellular bacterium S. Tm. uncovered an important extra-hepatic function of Hfe in macrophages and novel cell type-specific roles of Hfe in infection control and immune regulation: mice lacking Hfe either in all cell types or selectively in the myeloid compartment were more resistant to Salmonella infection and protected from early death compared to WT littermates expressing Hfe. Conversely, hepatocyte-specific Hfe deletion was deleterious to the host, triggering early death in response to Salmonella infection. These findings are somewhat unexpected because the primary iron overload patterns of Hfe-/- mice and hepatocyte-specific Hfe 3158

knockout mice are alike.25 Mice with myeloid-specific Hfe deletion by contrast, show no apparent iron-phenotype but are resistant to S. Tm. infection much like Hfe-/- mice.13 This suggests that the putative immune-regulatory roles of Hfe in macrophages are partially separated from its iron-regulatory functions or mediated via micro-environmental rather than systemic effects. Further studies using combinations of pathogens and exogenous iron sources will be required to ravel out underlying regulatory networks. However, the lack of Hfe in macrophages is sufficient to explain the improved survival of constitutive Hfe-/- mice infected with Salmonella. This fact may directly be related to the profound tropism of Salmonella for myeloid cells and points to an important Hfe function in macrophages.29,30 We noted that upon Salmonella infection, LysMCre+ Hfefl/fl and Hfe-/- mice have lower iron levels in the spleen. Indeed, it has been suggested that lower levels of iron in macrophages may be protective against pathogens such as Salmonella that propagate within macrophages.31 Surprisingly, we found accelerated death of AlfpCre+ Hfefl/fl mice due to impaired resistance to Salmonella infection, although macrophage iron content was unaffected, and bacterial loads in the spleen and liver were not different as compared to AlfpCre– Hfefl/fl mice. Our data rather indicate that enhanced extracellular bacterial proliferation in the iron-rich serum is deleterious to AlfpCre+ Hfefl/fl mice. Apart from iron-induced bacterial growth, the reduced levels of haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

Figure 6. Elevated iron levels correlate with reduced IFN-g production and increased bacterial numbers in the serum. Serum complement factor C3 (A), IL-6 (B) and IFN-γ (C) concentrations were analyzed in AlfpCre+ Hfefl/fl mice infected for 72 hours with Salmonella. n=19-20 for AlfpCre– Hfefl/fl, n=8-14 for AlfpCre+ Hfefl/fl. sIndependently, wild-type (WT) mice were fed an iron-adequate (IA) or iron-enriched diet (IO) 3 weeks prior to and during S. enterica Typhimurium (S. Tm.) infection. Serum bacterial load (D), IL-6 (E) and IFN-g (F) concentrations were determined. Statistically significant differences as calculated by Mann-Whitney test are indicated. n= 8-9 for IA, n=8 for IO.

IFN-g detected in the spleen and serum of infected AlfpCre+ Hfefl/fl mice and of WT mice maintained on a high iron diet may offer a partial explanation. Unlike AlfpCre+ Hfefl/fl mice in steady state,25 we herein exclusively report on the setting of Salmonella infection and observed that Salmonella-infected AlfpCre+ Hfefl/fl mice show normal iron content in the spleen and bone marrow macrophages, suggesting that high serum iron levels impair IFN-g production and its antimicrobial activity as shown in vitro and in vivo.32,33 To a large extent, host defense against intracellular microbes relies on direct antimicrobial effector functions of macrophages.28,34 Reactive nitrogen (RNS) and oxygen species, generated by Nos2 and phagocyte oxidase (phox), interfere with bacterial metabolism and exert toxic effects to limit Salmonella replication within macrophages and counteract systemic spread in infected mice.35-38 TNF and IFN-g have partly overlapping functions in the sense that both of them stimulate the expression of Nos2 and the assembly of phox subunits in Salmonella-infected macrophages.36 IL-6 in contrast, is the major cytokine inducer of the acute-phase reaction and centrally involved in the adaptation of iron homeostasis upon inflammatory stress.39 In the setting of Salmonella infection, IL-6 finetunes myeloid cell functions as it promotes bacterial killing but is also associated with alternative macrophage activation.40 haematologica | 2021; 106(12)

Given unaltered bacterial loads in the mononuclear phagocyte system, the high numbers of bacteria found in the serum of AlfpCre+ Hfefl/fl mice may result from enhanced extracellular proliferation rather than differential phagocytosis, which is supported by the enhanced bacterial growth we observed in AlfpCre+ Hfefl/fl serumspiked medium. Hfe-/- mice, in contrast, have reduced numbers of Salmonella in the serum, which may in part be attributable to increased serum Lcn2 concentrations (Figure 3), which are already present in the absence of infection.13,41 Our findings also support the concept that after invasion of myeloid cells, Salmonella has limited access to serum and hepatocellular iron pools. Rather, Salmonella may use intramacrophage iron sources such as ferritin.26 Iron metabolism and immune function have multiple interconnections including effects of iron availability on immune cell differentiation as well as direct effects of iron on cytokine formation and innate immune responses.42,43 In addition, iron genes and their products modulate the body’s response to inflammation.44 Here we extend these observations by demonstrating that the expression of the antimicrobial enzyme Nos2 is partially affected by Hfe: Nos2 expression in the spleen was highest in Hfe-/- mice, moderately increased in the setting of myeloid-specific Hfe deficiency and markedly reduced in mice with hepa3159

M. Nairz et al.

tocyte-specific Hfe deficiency. This suggests that the transcriptional induction of Nos2 in macrophages may be affected by a paracrine Hfe-dependent pathway. In contrast, when myeloid cells express Hfe and serum iron levels are high because of hepatocyte-specific Hfe deficiency, the induction of Nos2 was severely impaired in the spleen. Iron inhibits Nos2 transcription45 but this regulation fails to explain the reduced Nos2 mRNA and protein levels in the spleens of AlpfCre+ Hfefl/fl mice, which are relatively iron-poor in steady-state and iron-adequate in S. Tm. Infection.25 We propose that an iron-mediated immunederegulation secondary to the low levels of Ifn-g mRNA in spleens of these mice is a possible explanation because IFN-g is a major inducer of Nos2.46 In addition, RNS counteract Salmonella’s virulence and IFN- promotes Salmonella degradation in mouse macrophages.47 These mechanisms are also of central importance for immunity in human subjects because monogenetic defects in the IFN-g pathway result in increased susceptibility to non-typhoid Salmonella and atypical mycobacteria.48 However, additional studies are required to characterize the regulatory networks that appear to link Hfe to IFN-g and Nos2 levels. Additionally, the observed differences in the expression of immune genes between spleen and liver argue for the involvement of other types of immune, non-parenchymal or stromal cells. In this context, it will be particularly interesting to study the effects of Hfe depletion in lymphocyte subsets in the context of bacteremia because an effect of Hfe on T-cell differentiation has been proposed in different models.49 Salmonella can acquire iron via different pathways. Its major siderophores, enterobactin and salmochelins, bind ferric iron with extremely high affinity, thus initiating its uptake via siderophore receptors. Independently of siderophores, ionic iron is acquired via feo, sitABCD and a less well characterized low affinity iron uptake system.15 It is interesting to note that the entC sit feo triple mutant did grow better in medium spiked with sera of hepatocyte-specific Hfe-deficient mice than in sera of other mice. The growth of the triple mutant was significantly impaired as compared to Salmonella with only single deletions of iron uptake systems, entC, sit or feo, respectively (Figure 3). This suggests that the growth attenuation of this strain was partially conserved in high iron conditions. Moreover, the addition of high concentrations of recombinant murine Lcn2 to spiked sera of AlfpCre+ Hfefl/fl mice and corresponding controls inhibited bacterial growth in liquid cultures but did not abolish the differences between the two genotypes of mice whereas the iron chelator DFX did. These findings suggest that Salmonella is able to circumvent Lcn2’s growth inhibiting effects by iron uptake mechanisms that act independent of enterobactin and are thus resistant to Lcn2. Salmochelins, which are glycosylated

References 1. Pietrangelo A. Hereditary hemochromatosis--a new look at an old disease. N Engl J Med. 2004;350(23):2383-2397. 2. Weiss G. Genetic mechanisms and modifying factors in hereditary hemochromatosis. Nat Rev Gastroenterol Hepatol.;7(1):50-58. 3. Drakesmith H, Sweetland E, Schimanski L, et al. The hemochromatosis protein HFE inhibits iron export from macrophages. Proc

3160

enterobactin derivatives to which Lcn2 cannot bind, may one of the ways by which Salmonella resists the immune response.27 However, further studies are required to understand the mechanisms of Salmonella’s metabolic adaptation to iron withdrawal by the host’s immune response. In conclusion, our study exclusively reports on Hfe’s role in infection to demonstrate a pivotal extra-hepatic function of Hfe and to highlight the importance of intracellular iron levels within macrophages for the control of infections with Salmonella. We found that the lack of Hfe in macrophages increased host resistance to this particular pathogen secondary to reduced intracellular iron availability and increased Nos2 expression. Importantly, this effect was dominant over possible growth-promoting effects that increased serum iron levels may impose. In contrast, serum iron levels determined both, IFN-g levels and extracellular bacterial replication and increased bacteremia preceding early mortality in mice lacking Hfe exclusively in hepatocytes. Our data suggest that the high penetrance of HFE mutations may originate from the immune modulatory effects of HFE enabling a better control of infections with intracellular pathogens. Disclosures The authors declare that there is no conflict of interest. Contributions MN and CM planned and conducted experiments, acquired and analyzed data and drafted the manuscript; MV-S, A-MM, AS, DH, AH, LvR, RS, CWH, HT and PLM performed experiments; MM and GW conceived and designed the study, obtained funding and wrote the manuscript. Acknowledgments The authors would like to thank Sylvia Berger, Ines Brosch, Sabine Engl, Ines Glatz and Markus Seifert for excellent technical support. We also would like to thank Ferric C. Fang, Departments of Laboratory Medicine and Microbiology, University of Washington, for providing Salmonella mutants and intellectual input. Funding This work was supported by grants from the Austrian Research Fund (FWF sponsored doctoral programme W-1253 HOROS to GW and stand-alone project P 33062, to MN), the Christian Doppler Society (to GW), the Tyrolean Research Fund (TWF, to MN) and by the ‘Verein zur Förderung von Forschung und Weiterbildung in Infektiologie und Immunologie an der Medizinischen Universität Innsbruck’. CM was funded through a postdoctoral scholarship from the Medical Faculty of Heidelberg University, Germany and the Virtual Liver Network funding initiative (BMBF). MUM was supported by a grant from the Deutsche Forschungsgemeinschaft (SFB 1036).

Natl Acad Sci U S A. 2002; 99(24):1560215607. 4. Cairo G, Recalcati S, Montosi G, Castrusini E, Conte D, Pietrangelo A. Inappropriately high iron regulatory protein activity in monocytes of patients with genetic hemochromatosis. Blood. 1997;89(7):25462553. 5. Montosi G, Paglia P, Garuti C, et al. Wildtype HFE protein normalizes transferrin iron accumulation in macrophages from subjects with hereditary hemochromatosis. Blood.

2000;96(3):1125-1129. 6. Ramos P, Guy E, Chen N, et al. Enhanced erythropoiesis in Hfe-KO mice indicates a role for Hfe in the modulation of erythroid iron homeostasis. Blood. 2011; 117(4):13791389. 7. Hollerer I, Bachmann A, Muckenthaler MU. Pathophysiological consequences and benefits of HFE mutations: 20 years of research. Haematologica. 2017;102(5):809-817. 8. Bennett MJ, Lebron JA, Bjorkman PJ. Crystal structure of the hereditary haemochromato-

haematologica | 2021; 106(12)

Cell-specific role of Hfe in Salmonella infection

sis protein HFE complexed with transferrin receptor. Nature. 2000;403(6765):46-53. 9. Bridle KR, Frazer DM, Wilkins SJ, D et al. Disrupted hepcidin regulation in HFE-associated haemochromatosis and the liver as a regulator of body iron homoeostasis. Lancet. 2003;361(9358):669-673. 10. Muckenthaler MU, Rivella S, Hentze MW, Galy B. A Red Carpet for Iron Metabolism. Cell. 2017 Jan 26;168(3):344-361. 11. Zoller H, Theurl I, Koch RO, McKie AT, Vogel W, Weiss G. Duodenal cytochrome b and hephaestin expression in patients with iron deficiency and hemochromatosis. Gastroenterology. 2003;125(3):746-754. 12. Brissot P, Pietrangelo A, Adams PC, de Graaff B, McLaren CE, Loreal O. Haemochromatosis. Nat Rev Dis Primers. 2018;4:18016. 13. Nairz M, Theurl I, Schroll A, et al. Absence of functional Hfe protects mice from invasive Salmonella enterica serovar Typhimurium infection via induction of lipocalin-2. Blood. 2009;114(17):3642-3651. 14. Schaible UE, Kaufmann SH. Iron and microbial infection. Nat Rev Microbiol. 2004;2(12):946-953. 15. Andrews SC, Robinson AK, RodriguezQuinones F. Bacterial iron homeostasis. FEMS Microbiol Rev. 2003;27(2-3):215-237. 16. Gomes-Pereira S, Rodrigues PN, Appelberg R, Gomes MS. Increased susceptibility to Mycobacterium avium in hemochromatosis protein HFE-deficient mice. Infect Immun. 2008;76(10):4713-4719. 17. Olakanmi O, Schlesinger LS, Britigan BE. Hereditary hemochromatosis results in decreased iron acquisition and growth by Mycobacterium tuberculosis within human macrophages. J Leukoc Biol. 2007;81(1):195204. 18. Frank KM, Schneewind O, Shieh WJ. Investigation of a researcher's death due to septicemic plague. N Engl J Med. 2011; 364(26):2563-2564. 19. Miller HK, Schwiesow L, Au-Yeung W, Auerbuch V. Hereditary Hemochromatosis Predisposes Mice to Yersinia pseudotuberculosis Infection Even in the Absence of the Type III Secretion System. Front Cell Infect Microbiol. 2016;6:69. 20. Nairz M, Schroll A, Haschka D, et al. Genetic and Dietary Iron Overload Differentially Affect the Course of Salmonella Typhimurium Infection. Front Cell Infect Microbiol. 2017;7:110. 21. Nairz M, Fritsche G, Crouch ML, Barton HC, Fang FC, Weiss G. Slc11a1 limits intracellular growth of Salmonella enterica sv. Typhimurium by promoting macrophage immune effector functions and impairing bacterial iron acquisition. Cell Microbiol. 2009;11(9):1365-1381. 22. Sonnweber T, Ress C, Nairz M, et al. Highfat diet causes iron deficiency via hepcidin-

haematologica | 2021; 106(12)

independent reduction of duodenal iron absorption. J Nutr Biochem. 2012; 23(12):1600-1608. 23. Theurl I, Hilgendorf I, Nairz M, et al. Ondemand erythrocyte disposal and iron recycling requires transient macrophages in the liver. Nat Med. 2016;22(8):945-951. 24. Herrmann T, Muckenthaler M, van der Hoeven F, et al. Iron overload in adult Hfedeficient mice independent of changes in the steady-state expression of the duodenal iron transporters DMT1 and Ireg1/ferroportin. J Mol Med (Berl). 2004;82(1):39-48. 25. Vujic Spasic M, Kiss J, Herrmann T, et al. Hfe acts in hepatocytes to prevent hemochromatosis. Cell Metab. 2008; 7(2):173-178. 26. Nairz M, Ferring-Appel D, Casarrubea D, et al. Iron Regulatory Proteins Mediate Host Resistance to Salmonella Infection. Cell Host Microbe. 2015;18(2):254-261. 27. Crouch ML, Castor M, Karlinsey JE, Kalhorn T, Fang FC. Biosynthesis and IroC-dependent export of the siderophore salmochelin are essential for virulence of Salmonella enterica serovar Typhimurium. Mol Microbiol. 2008;67(5):971-983. 28. Weiss G, Schaible UE. Macrophage defense mechanisms against intracellular bacteria. Immunol Rev. 2015;264(1):182-203. 29. Vazquez-Torres A, Jones-Carson J, Baumler AJ, et al. Extraintestinal dissemination of Salmonella by CD18-expressing phagocytes. Nature. 1999;401(6755):804-808. 30. Vazquez-Torres A, Vallance BA, Bergman MA, et al. Toll-like receptor 4 dependence of innate and adaptive immunity to Salmonella: importance of the Kupffer cell network. J Immunol. 2004;172(10):6202-6208. 31. Weinberg ED. Iron availability and infection. Biochim Biophys Acta. 2009; 1790(7):600605. 32. Weiss G, Fuchs D, Hausen A, et al. Iron modulates interferon-gamma effects in the human myelomonocytic cell line THP-1. Exp Hematol. 1992;20(5):605-610. 33. Mencacci A, Cenci E, Boelaert JR, et al. Iron overload alters innate and T helper cell responses to Candida albicans in mice. J Infect Dis. 1997;175(6):1467-1476. 34. Fang FC. Antimicrobial reactive oxygen and nitrogen species: concepts and controversies. Nat Rev Microbiol. 2004;2(10):820-832. 35. Vazquez-Torres A, Jones-Carson J, Mastroeni P, Ischiropoulos H, Fang FC. Antimicrobial actions of the NADPH phagocyte oxidase and inducible nitric oxide synthase in experimental salmonellosis. I. Effects on microbial killing by activated peritoneal macrophages in vitro. J Exp Med. 2000;192(2):227-236. 36. Mastroeni P, Vazquez-Torres A, Fang FC, et al. Antimicrobial actions of the NADPH phagocyte oxidase and inducible nitric oxide synthase in experimental salmonellosis. II. Effects on microbial proliferation and host

survival in vivo. J Exp Med. 2000;192(2): 237-248. 37. Richardson AR, Payne EC, Younger N, et al. Multiple targets of nitric oxide in the tricarboxylic acid cycle of Salmonella enterica serovar typhimurium. Cell Host Microbe. 2011;10(1):33-43. 38. Rosenberger CM, Finlay BB. Macrophages inhibit Salmonella typhimurium replication through MEK/ERK kinase and phagocyte NADPH oxidase activities. J Biol Chem. 2002;277(21):18753-18762. 39. Kim DK, Jeong JH, Lee JM, et al. Inverse agonist of estrogen-related receptor gamma controls Salmonella typhimurium infection by modulating host iron homeostasis. Nat Med. 2014;20(4):419-424. 40. Fuster JJ, Walsh K. The good, the bad, and the ugly of interleukin-6 signaling. EMBO J. 2014;33(13):1425-1427. 41. Xiao X, Yeoh BS, Vijay-Kumar M. Lipocalin 2: An Emerging Player in Iron Homeostasis and Inflammation. Annu Rev Nutr. 2017;37: 103-130. 42. Ganz T, Nemeth E. Iron homeostasis in host defence and inflammation. Nat Rev Immunol. 2015;15(8):500-510. 43. Soares MP, Weiss G. The Iron age of hostmicrobe interactions. EMBO Rep.;16(11): 1482-1500. 44. Wu Q, Shen Y, Tao Y, Wei J, et al. Hemojuvelin regulates the innate immune response to peritoneal bacterial infection in mice. Cell Discov. 2017;3:17028. 45. Weiss G, Werner-Felmayer G, Werner ER, Grunewald K, Wachter H, Hentze MW. Iron regulates nitric oxide synthase activity by controlling nuclear transcription. J Exp Med. 1994;180(3):969-976. 46. Bogdan C. Nitric oxide synthase in innate and adaptive immunity: an update. Trends Immunol. 2015;36(3):161-178. 47. McCollister BD, Bourret TJ, Gill R, JonesCarson J, Vazquez-Torres A. Repression of SPI2 transcription by nitric oxide-producing, IFNgamma-activated macrophages promotes maturation of Salmonella phagosomes. J Exp Med. 2005;202(5):625-635. 48. Jouanguy E, Doffinger R, Dupuis S, Pallier A, Altare F, Casanova JL. IL-12 and IFN-gamma in host defense against mycobacteria and salmonella in mice and men. Curr Opin Immunol. 1999;11(3):346-351. 49. Cruz E, Melo G, Lacerda R, Almeida S, Porto G. The CD8+ T-lymphocyte profile as a modifier of iron overload in HFE hemochromatosis: an update of clinical and immunological data from 70 C282Y homozygous subjects. Blood Cells Mol Dis. 2006;37(1):3339. 50. Nairz M, Schleicher U, Schroll A, et al. Nitric oxide-mediated regulation of ferroportin-1 controls macrophage iron homeostasis and immune function in Salmonella infection. J Exp Med. 2013;210(5):855-873.

3161

ARTICLE Ferrata Storti Foundation

Myeloproliferative Disorders

Bone marrow megakaryocytic activation predicts fibrotic evolution of Philadelphia-negative myeloproliferative neoplasms Mattia Schino,1* Vincenzo Fiorentino,1* Elena Rossi,2,3 Silvia Betti,3 Monica Di Cecca,2 Valentina Ranucci,1 Patrizia Chiusolo,2,3 Maurizio Martini,1,3# Valerio De Stefano2,3# and Luigi Maria Larocca1,3#

Haematologica 2021 Volume 106(12):3162-3169

Department of Life Sciences and Public Health, Università Cattolica del Sacro Cuore; Department of Radiological and Hematological Sciences, Università Cattolica del Sacro Cuore and 3Fondazione Policlinico Universitario A. Gemelli IRCCS, Rome, Italy 2

*MF and VF contributed equally as co-first authors. # VDS, MM and LML contributed equally as co-senior authors.

ABSTRACT

Correspondence: LUIGI MARIA LAROCCA luigimaria.larocca@unicatt.it Received: June 24, 2020. Accepted: October 2, 2020. Pre-published: November 19, 2020. https://doi.org/10.3324/haematol.2020.264143

3162

hiladelphia-negative chronic myeloproliferative neoplasms (MPN) have been traditionally considered as indistinctly slowly progressing conditions; recent evidence proves that a subset of cases have a rapid evolution, so that MPN prognosis needs to be personalized. We identified a new morphological parameter, defined as megakaryocytic activation (M-ACT) based on the coexistence of megakaryocytic emperipolesis, megakaryocytes (MK) cluster formation and evidence of arrangement of collagen fibers around the perimeter of MK. We retrospectively analyzed the bone marrow biopsy of two MPN cohorts of patients with polycythemia (PV) (n=64) and non-PV patients (including essential thrombocythemia, and early/prefibrotic primary myelofibrosis [PMF]) (n=222). M-ACT showed a significant correlation with splenomegaly, white blood cell count, and lactate dehydrogenase serum levels in both groups, with JAK2 V617F allele burden in PV patients, and with CALR mutations, and platelet count in non-PV patients. Progression-free survival, defined as PV-to-secondary MF progression and non-PV-to-overt PMF, was worse in both PV and early/prefibrotic PMF patients with M-ACT in comparison to those without M-ACT (P<0.0001). Interestingly, M-ACT was not found in the subgroup of essential thrombocythemia patients. In conclusion, M-ACT can be helpful in the differential diagnosis of MPN and can represent a new morphologic parameter with a predictive value for progression of MPN.

Introduction Philadelphia-negative chronic myeloproliferative neoplasms (MPN) represent a group of hematological disorders that originates from the neoplastic transformation of a pluripotent stem cell and are characterized by clonal proliferation of one or more hematopoietic progenitors in the bone marrow (BM) and in extramedullary sites. According to the World Health Organization (WHO) 2017 classification, MPN can be divided into three main sets: polycythemia vera (PV), essential phrombocythemia (ET) and primary myelofibrosis (PMF), whose early stages’ differential diagnosis is often challenging.1 While MPN have been traditionally considered as indistinct slow progressing conditions,2,3 recent evidence, on the contrary, demonstrated that a subset of cases had a rapid evolution, leading different groups to develop several prognostic scores, mainly based on clinical and laboratory parameters with less emphasis on morphological, immunophenotypic and molecular data.4 The first prognostic score was the International Prognostic Scoring System (IPSS), edited in 2009 by an international

haematologica | 2021; 106(12)

Megakaryocytic activation and fibrotic evolution of MPN

study group, which enabled survival estimation at the time of diagnosis primarily employing five clinical and hematologic parameters;5 this model was further revised as the Dynamic International Prognostic Scoring System (DIPSS) and then as DIPSS-plus score.6,7 The above, nonetheless, applied to already-established myelofibrosis (both PMF and post-PV/ET MF) only, determining survival from the time of disease progression/transformation to death without considering the heterogeneous disease history before the appearance of BM changes.8 On the basis of advances in MPN molecular profiling, in order to improve the prognostic prediction in PMF patients, novel models included JAK2, CALR, and MPL mutation status in addition to the IPSS parameters.8 Moreover, novel insights were provided by in-depth analysis of genomic subsets with different clinical outcomes.9 Recent publications have introduced new risk models for PMF, namely MIPSS70 (mutation-enhanced international prognostic scoring system for transplant-age patients),10 MIPSS70+ version 2.0 (karyotype-enhanced MIPSS70) and GIPSS (genetically-inspired prognostic scoring system).11,12 Similar risk models have been recently introduced for both ET and PV under the name of MIPSS-ET and MIPSS-PV, highlighting the prognostic contribution of spliceosome gene mutations.13 However, all these predictive models do not consider morphological and phenotypical features, except BM fibrosis grade in the MIPPS70 model. In this study we evaluated a new morphological parameter, defined by the coexistence of emperipolesis of megakaryocytes (MK) (i.e., the presence of an intact cell within the cytoplasm of another cell), MK clustering and peri-MK fibrosis in BM biopsy, which was named megakaryocytic activation (M-ACT). Larocca et al. in 2015 demonstrated that extensive BM emperipolesis associated to BM fibrosis was present in patients affected by gray platelet syndrome, with up to 65% MK containing two of four leukocytes engulfed within the cytoplasm;14 a similar phenomenon has been described either in BM patients with PMF,15 and in the BM of animal models of myelofibrosis.16,17 We demonstrated that M-ACT is a useful morphological parameter in forecasting both PV and early/prefibrotic PMF to myelofibrosis progression and could also help in the differential diagnosis between ET and early/prefibrotic PMF.18

Methods Patients' features Formalin-fixed, paraffin-embedded BM biopsy specimens, obtained from the posterior superior iliac spine,19 were available in our Institute of Pathology for 460 patients clinically diagnosed with a MPN and followed at our Institute of Hematology (Fondazione Policlinico Universitario “A. Gemelli”, IRCCS) from January 2005 to October 2019. The study was carried out in accordance with the Declaration of Helsinki and the consent for retrospective analysis of all clinical data, according to the Ethical Committee of the Università Cattolica del Sacro Cuore School of Medicine, and obtained by all the patients at the hospital admission. Patients were clinically followed-up over the observation time by one single team physician (VDS and ER as senior members). All 286 cases were sorted until October 2019, according to three inclusion criteria: clinical diagnosis of either PV or non-PV MPN, first BM biopsy at diagnosis for non-PV cohort and within 0-24

haematologica | 2021; 106(12)

months from the clinical diagnosis for PV cohort and no grade 2-3 BM fibrosis. Accordingly, patients with diagnosis of overt PMF or secondary myelofibrosis were excluded. Furthermore, BM biopsies were revised by two skilled pathologists (LMR and MM) and categorized according to the WHO 2017 criteria (PV, ET, early/prefibrotic PMF). Clinical and hematological data (according to WHO 2017 criteria) were collected in order to trace lactose dehydrogenase (LDH) increase (i.e., LDH serum levels ≥250 UI/L), palpable splenomegaly, leukocytosis (i.e., white blood cell [WBC] count ≥ 11×109/L), high hemoglobin (Hg) level (i.e., Hgb >16 g/dL for women and Hgb > 16.5 g/dL for men) and thrombocytosis (i.e., PLT ≥ 450×109/L) for each patient at diagnosis. We also verified the occurrence of arterial/venous (A/V) thrombotic events and/or major bleeding events during the clinical course (until October 2019) for each case. Thrombotic and bleeding events were defined as previously described.20 JAK2 V617F mutation and allele burden analysis, CALR exon 9 mutations and MPL exon 10 mutations were performed as previously described.20 Progression to secondary myelofibrosis was defined from the patient chart review and based on the International Working Group for Myelofibrosis Research and Treatment (IWG-MRT) consensus criteria.21 The main clinical, hematological and molecular characteristics of the 286 patients are shown in Table 1 for the PV cohort (64 patients), in Table 2 and the Online Supplementary Table S1 for nonPV cohort (including 199 early/prefibrotic PMF patients [Table 2] and 23 ET patients [Online Supplementary Table S1]).

Bone marrow biopsy analysis and megakaryocytic activation histological parameters All biopsy specimens had a suitable length (at least 1.5-2 cm) in order to obtain at least ten partially preserved intertrabecular areas, since subcortical medullary lacunae are less cellular than deep ones (especially in the elderly) and since focal pathologies can have a deep localization.22 After collection, each biopsy specimen was kept in a properly-labeled clean container filled with 10% natural buffered formalin at pH 7.6 for 12 hours for fixation, was then decalcified with a Decalcifier II solution (Leica Biosystems, Milan, Italy) for 1 hour at room temperature, then fixed with 10% natural buffered formalin at pH 7.6 for 2 hours and finally embedded in paraffin. Sections (3-5 mm thick) were cut from each block for staining with hematoxylin and eosin (H&E) and Gordon&Sweet’s silver staining to evaluate morphological features and fibrosis.23,24 The specimens were concurrently examined and reviewed by two pathologists experienced in BM biopsy interpretation (LML and MM), who were blinded toward the patients’ characteristics and survival. Cases with disagreement were discussed using a multiheaded microscope until agreement was achieved. The agreement indices (Cohen’s K) between the two pathologists were very good: k=0.83 and k=0.85 for PV group and for non-PV group, respectively. In the definition of M-ACT the following parameters were examined in detail (as shown in Figure 1): (i) MK emperipolesis, (ii) MK clustering and (iii) peri-MK fibrosis: i) MK emperipolesis was defined as the presence of one or more leukocyte or a precursor of hematopoiesis within the cytoplasm of at least 30% MK in the specimen; ii) MK clustering was defined as an aggregation of three or more megakaryocytes in close contact with each other and at least 25% of MK distributed in clusters in the specimen; iii) periMK fibrosis was defined as the arrangement of collagen fibers around the perimeter of the vast majority of MK, underlining their primary role in the genesis of fibrosis. M-ACT positive patients showed the contemporary presence of all three parameters and M-ACT was evaluated only on the first BM biopsy at diagnosis and before any treatment.

3163

M. Schino et al.

Table 1. Correlation between megakaryocytic activation and the main polycythemia vera patient’s clinical and molecular features.

PV 1

Age Male Female JAK2-RT > 50% Secondary MF-progression Time-progression2 Hgb3 LDH4 Palpable splenomegaly WBC5 PLT6 A/V thrombosis Major bleeding

Overall (n=64)

M-ACT - (n=26)

M-ACT + (n=38)

HR [CI 95%]

59.1 (40-74) 38 (59.4%) 26 (40.6%) 43 (67.2%) 41 (64.1%) 78.3 (26-116) 18.7 (15.2-19.9) 351.7 (210-522) 36 (56.2%) 11.3 (7.2-20.5) 590.8 (380-851) 24 (37.5%) 12 (18.7%)

59.3 (42-74) 15 (23.4%) 11 (17.2%) 16 (25.0%) 6 (9.4%) 95.5 (72-116) 16.2 (15.2-18.4) 241.3 (210-380) 16 (25.0%) 9.8 (7.2-13.4) 507.2 (380-623) 8 (12.5%) 1 (1.6%)

58.8 (40-73) 23 (35.9%) 15 (23.4%) 27 (42.2%) 32 (50%) 44.7 (26-48) 17.1 (15.4-19.9) 410.5 (238-522) 20 (31.3%) 15.6 (10.3-20.5) 720.4 (430-851) 16 (25.0%) 11 (17.2%)

1.00 [0.821 - 1.036] 1.01 [0.791 – 1.059]

1.00 1.00

1.24 [0.058 - 2.589] 2.87 [0.332-1.591] 2.78 [0.036-2.589] 1.31 [0.228-2.746] 1.54 [0.025-3.532] 1.83 [0.179-3.795] 1.81 [0.054-3.273] 2.39 [0.082-2.629] 1.02 [0.882 – 1.069] 1.54 [0.332-1.591]

0.059 0.0001 0.0001 0.06 0.002 0.001 0.001 0.0001 0.43 0.06

JAK2-RT: JAK2 V617F allele burden; Hgb: hemoglobin, serum levels; LDH: lactate dehydrogenase, serum levels; WBC: white blood cell count; PLT: platelet count; A/V: arterial/venous; HR: hazard ratio; CI: Confidence Interval. 1In years; 2In months; 3(g/dL); 4(UI/L); 5(x109/L); 6(x109/L).

Table 2. Correlation between megakaryocytic activation and the main early/prefibrotic primary myelofibrosis patient’s clinical and molecular features.

Early/prefibrotic PMF 1

Age Male Female JAK2-RT > 50% CALR mut. CALR type 1 CALR type 2 MPL mut. Time-progression2 Hgb3 LDH serum levels4 Palpable splenomegaly WBC5 PLT6 A/V thrombosis Major bleeding

Overall (n=199)

M-ACT - (n=109)

M-ACT + (n=90)

HR [CI 95%]

66.2 (46-78) 91 (45.7%) 108 (54.3%) 37 (18.6%) 60 (30.1%) 46 (23.1%) 14 (7.0%) 3 (1.5%) 67.3 (15-109) 15.3 (12.8-16.4) 402.7 (205-612) 111 (55.8%) 11.8 (7.5-18.7) 620.1 (366-861) 74 (37.2%) 30 (15.1%)

64.2 (48-76) 57 (28.6%) 52 (26.7%) 26 (13.1%) 19 (9.5%) 11 (5.5%) 8 (4.0%) 2 (1.0%) 70.2 (53-109) 14.8 (12.8-15.5) 230.3 (205-390) 49 (24.6%) 10.3 (7.5-14.8) 520.6 (366-650) 30 (15.1%) 14 (7.0%)

66.3 (46-78) 34 (17.1%) 56 (28.1%) 11 (5.5%) 41 (20.6%) 35 (17.6%) 6 (3.0%) 1 (0.5%) 33.7 (15-56) 15.2 (13.6-16.4) 407.9 (288-612) 62 (31.1%) 14.4 (9.9-18.7) 710.3 (490-861) 44 (22.1%) 16 (8.0%)

1.01 [0.67078-1.036] 1.04 [0.872-1.055]

1.00 0.05

1.27 [0.073-2.879] 2.14 [0.301-1.902] 3.01 [1.527-3.812] 0.97 [0.833-1.056] 1.00 [0.663-1.044] 2.83 [0.328-1.913] 1.04 [0.923-1.088] 1.54 [0.045-3.235] 1.82 [0.1797-3.759] 1.80 [0.045-3.733] 2.46 [0.087-2.279] 1.03 [0.243-1.912] 1.00 [0.928-1.079]

0.04 0.001 0.0001 1.00 1.00 0.0001 0.43 0.003 0.001 0.002 0.0001 0.001 0.42

JAK2-RT: JAK2 V617F allele burden; CALR mut: CALR exon 9 mutations (type 1 + type 2); MPL mut: MPL exon 10 mutations; Hgb: hemoglobin: serum levels; LDH: lactate dehydrogenase, serum levels; WBC: white blood cell count; PLT: platelet count; A/V: arterial/venous; HR: hazard ratio; CI: Confidence Interval. 1In years; 2In months; 3(g/dL); 4(UI/L); 5 (x109/L); 6(x109/L).

Statistical analysis Statistical analysis was performed using GraphPad-Prism 5 software (Graph Pad Software, San Diego, CA) and MedCalc version 10.2.0.0 (MedCalc Software, Mariakerke, Belgium).25 Statistical comparison of continuous variables was performed by the MannWhitney U test (t-test), as appropriate. Comparison of categorical variables was performed by c2 statistic, using the Fisher’s exact test. In order to evaluate the agreement between the two pathologists about the presence or absence of M-ACT in BM biopsies, the interrater agreement (Kappa) using MedCalc software was calulated. The endpoint was progression-free survival (PFS), defined as the time between the first diagnosis and PV-to-secondary MF progression and early/prefibrotic PMF-to-overt PMF progression, respectively. We followed the WHO 2017 criteria to establish the progression for PV-to-secondary MF and for early/prefibrotic PMF to overt myelofibrosis progression.1

3164

Kaplan-Meier survival curves were plotted and differences in survival between groups of patients were compared using the logrank test. Multivariate analysis was performed using the Cox proportional hazards regression analysis including only those clinical and biological variables with a P-value of 0.10 or lower at the univariate analysis. P-values less than 0.05 were considered as statistically significant.

Results Megakaryocytic activation in the polycythemia vera cohort Twenty-six of the 64 PV did not meet histological criteria for M-ACT (40%), versus 38 who did (60%). In the PV cohort, M-ACT showed a significant correlation with one clinical parameter, i.e., palpable splenomegaly (P=0.001), haematologica | 2021; 106(12)

Megakaryocytic activation and fibrotic evolution of MPN

Figure 1. Representative images of megakaryocytic activation in polycythemia vera (A to C) and in early/prefibrotic primary myelofibrosis (D to F). (A and D) Emperipolesis, defined as the presence of at least one leukocyte (indicated by black arrows) or a precursor of hematopoiesis within the cytoplasm of a megakaryocyte (MK) (400X magnification); (B and E) MK clustering, defined as an aggregation of three or more MK in close contact with each other (200X and 400X magnification for panels B and E, respectively); (C and F) peri-MK fibrosis, defined as the arrangement of collagen fibers around the perimeter of activated MK (400X and 200X magnification for panel C and F, respectively).

Table 3. Polycythemia vera cohort: progression-free survival multivariate analysis.

Covariate Major bleeding JAK2 status M-ACT WBC

Wald

Exp(b)

95% CI of Exp(b)

-0.1562 1.0726 2.3435 -0.1520

0.3890 0.4191 0.4761 0.3604

0.1614 6.5500 24.2335 0.1780

0.6879 0.0105 <0.0001 0.6731

0.8554 2.9229 10.4180 0.8590

0.3991 to 1.8333 1.2855 to 6.6460 4.0978 to 26.4858 0.4239 to 1.7407

b: coefficient estimates; SE: standard error for coefficient estimates b; Exp(b): hazard ratio value; 95% CI of Exp(b)= 95% Confidence Interval of hazard ratio. Major bleeding vs. no major bleeding; JAK2 burden≥50% vs. JAK2 burden<50% (JAK2 status); megakaryocyte activation vs. no megakaryocyte activation (M-ACT); WBC≥11.0x109/L vs. WBC<11.0x109/L; WBC: whitel blood cell count.

Table 4. Early/prefibrotic primary myelofibrosis subset: progression-free survival multivariate analysis.

Covariate WBC M-ACT CALR status Sex LDH Palpable splenomegaly PLT

Wald

Exp(b)

95% CI of Exp(b)

0.4334 0.7659 0.3690 0.4071 0.1831 0.3084 0.2415

0.1879 0.1640 0.1685 0.1520 0.1749 0.1746 0.1638

5.3207 21.8235 4.7956 7.1702 1.0958 3.1217 2.1729

0.0211 <0.0001 0.0285 0.0074 0.2952 0.0773 0.1405

1.5425 2.1510 1.4463 1.5024 1.2009 1.3613 1.2731

1.0673 to 2.2294 1.5598 to 2.9661 1.0395 to 2.0124 1.1153 to 2.0240 0.8524 to 1.6919 0.9668 to 1.9167 0.9235 to 1.7551

b: coefficient estimates; SE: standard error for coefficient estimates b; Exp(b): hazard ratio value; 95% CI of Exp(b): 95% Confidence Interval of hazard ratio; WBC: white blood cell count; PLT: platelet count; LDH: lactose dehydrogenase; M-ACT: megakaryocyte activation. CALR mutations vs. no CARL wt (CARL); sex male vs. female (sex); megakaryocyte activation vs. no megakaryocyte activation (M-ACT); WBC count ≥11.0x109/L vs.: WBC<11.0x109/L; LDH≥250 UI/L vs. LDH<250 UI/L; palpable splenomegaly vs.no palpable splenomegaly; PLT≥450x109/L vs. PLT<450x109/L.

and with hematologic parameters, like platelet count (P=0.0001), LDH serum levels (P=0.002) and WBC count (P=0.001). On the other hand, no significant correlation was found between M-ACT and age (P=1.00), sex (P=1.00), A/V thrombosis (P=0.43), while major bleeding (P=0.06), Hgb level (P=0.06) and JAK2 V617F burden>50% (P=0.059) showed a certain associative trend (Table 1). We found that patients with M-ACT had a significant lower PFS than those without M-ACT (Table 1; Figure 2 panel A, for PFS: median PFS for M-ACT positive patients haematologica | 2021; 106(12)

58 months vs. median PFS for M-ACT negative patients 108 months, P<0.0001, hazard ratio [HR] 6.81, 95% Confidence Interval [CI]: 3.48-13.32). Moreover, JAK2 V617F allele burden ≥50% and history of major bleeding had a significant correlation with a worse PFS (P=0.0225 and P=0.0174, respectively, Online Supplementary Figure S1), while WBC count>11x109/L showed a certain trend toward significance (P=0.0823, Online Supplementary Figure S1). Conversely, age (P=0.3718), sex (P=0.3645), LDH serum level (P=0.1305), PLT count (P=0.5643), Hgb level 3165

M. Schino et al. A

Figure 2. Kaplan-Meier curves for progression-free survival of polycythemia vera and of early/prefibrotic primary myelofibrosis patients. (A) Kaplan-Meier curve for progression-free survival (PFS) of polycythemia vera (PV) patients stratified by megakaryocytic activation (M-ACT); (B) Kaplan-Meier curve for PFS of early/prefibrotic primary myelofibrosis (PMF) patients by M-ACT. M-ACT positive patients (red-line) was significantly associated to a worse PFS (P<0.0001) with respect to those without M-ACT (blue-line) in both groups.

(P=0.1024) and A/V thrombosis (P=0.4216) did not show significant correlation with PFS. Multivariate analysis of PFS, including M-ACT status, JAK2 status, WBC count and history of major bleeding, showed that the presence of M-ACT and the JAK2 V617F allele burden were the only significant predictors (for M-ACT status, P<0.0001, HR 10.4180, 95% CI: 4.097826.4858; for JAK2 V617F allele burden ≥50%, P=0.0105, HR 0.0105, 95% CI: 1.2855-6.6460; Table 3).

Megakaryocytic activation in non-polycythemia vera myeloproliferative neoplasm cohort One hundred and nine of 199 early/prefibrotic PMF patients did not meet histological criteria for M-ACT 3166

(55%) versus 90 who did (45%). In this cohort, M-ACT showed a strong correlation with clinical parameters, such as palpable splenomegaly (P=0.001) and history of major bleeding (P=0.001), and with hematologic parameters, like platelet count (P=0.0001), LDH serum levels (P=0.003), WBC count (P=0.002), presence of CALR mutations (P=0.001; Table 2). Notably, we found a significant association between M-ACT and CALR type 1 mutation (P=0.0001) while we did not find a significant correlation between M-ACT and CALR type 2 mutation (P=1.0). We found a significant yet milder correlation with sex, with M-ACT being more prevalent in females (P=0.05), and with JAK2 V617F allele burden≥50% (P=0.04). On the contrary, no significant correlation was found between haematologica | 2021; 106(12)

Megakaryocytic activation and fibrotic evolution of MPN

M-ACT and age (P=1.00), Hgb level (P=0.43), MPL mutations (P=1.00) and A/V thrombosis (P=0.42; Table 1). Similarly to what happened in the PV cohort, when we correlated M-ACT status with PFS, we found that patients with early/prefibrotic PMF and with M-ACT had a significant lower PFS than those without M-ACT (Table 1; Figure 2 panel B, for PFS: median PFS for M-ACT positive patients 44 months vs. median PFS for M-ACT negative patients 77 months, P<0.0001, HR 3.17, 95% CI: 2.274.44). Moreover, male sex, CALR type 1 mutations, WBC count >11x109/L, presence of palpable splenomegaly, PLT≥600x109/L and LDH ≥250 U/L had a significant correlation with a worse PFS (P=0.0187, P<0.0001, P<0.0001, P<0.0001, P<0.0001 and P=0.0025, respectively, Online Supplementary Figure 2S). Conversely, age (P=0.8831), major bleeding (P=0.7244), JAK2 V617F allele burden≥ 50% (P=0.3459), Hgb level (P=0.5234), MPL mutations (P=0.2268) and A/V thrombosis (P=0.2003) did not show significant correlation with PFS. Multivariate analysis of PFS, including M-ACT status, CALR status, WBC count, sex, LDH serous level, splenomegaly, and platelet count, showed that the presence of M-ACT and CALR type 1 mutation, WBC count >11x109/L and male sex were the significant predictors (for M-ACT status, P<0.0001, HR 2.1510, 95% CI: 1.55982.9661; for CALR status, P=0.0285, HR 1.446, 95% CI: 1.0395-2.0124; for WBC count, P=0.0211, HR 1.5425, 95% CI: 1.0673-2.2294; for sex, P=0.0074, HR 1.5024, 95% CI: 1.1153-2.0240; Table 4). In the non-PV MPN cohort, we also analyzed a small subgroup of 23 ET patients. We did not find M-ACT in any of the ET BM biopsies performed at the time of the diagnosis. The ET patients had a better PFS in comparison to patients with early/prefibrotic PMF either with M-ACT (Online Supplementary Figure S3; P<0.0001) and without MACT (Online Supplementary Figure S3; P<0.0001). Interestingly, we found that the incidence of M-ACT among triple-negative patients in the non-PV cohort was significantly lower in respect to patients with a driver gene mutation (21 of 85 [24.7%] triple negative patients vs. 90 of 222 [40.5%]; P=0.0115), and that triple-negative patients with M-ACT also had a significant lower PFS than those without M-ACT (median PFS for M-ACT positive triple-negative patients 56 months vs. median PFS for M-ACT negative triple-negative patients 79 months, P<0.0023, HR 2.76, 95% CI: 1.44-5.32; data not shown).

Discussion For about two decades, one of the most important problems in the treatment of patients with MPN has been the identification of biological and non-biological factors that could represent a determinant key to the prediction of prognosis. Accordingly, several prognostic scores have succeeded over time, mainly based on clinical, hematological and molecular parameters, in identifying the fraction of MPN patients that could have a high risk of developing a leukemic transformation or a bone marrow fibrotic failure. However, none of these models take the morphological parameters into factual consideration, while these parameters play an important role in the diagnostic phase. In this retrospective and single-center study, we propose a novel morphological parameter, defined as M-ACT, as a new possible predictive marker of fibrotic evolution haematologica | 2021; 106(12)

among Philadelphia-negative MPN. Furthermore, this new parameter seems to be useful to supplement WHO 2017 classification criteria in the differential diagnosis of the MPN subtype between ET and early/prefibrotic PMF. In our study, carried out on a large cohort of MPN BM biopsies at diagnosis, extensive evidence support this statement. In fact, in univariate analysis M-ACT correlates with relevant MPN clinical and hematologic parameters (see Table 1 and 2), such as palpable splenomegaly, WBC or PLT count, and LDH levels, but also with molecular profiles defined by the JAK2 V617F allele burden and CALR mutations (especially the CALR type 1 mutation). In PV patients the PFS was influenced at the multivariate analysis by the JAK2 V617F allele burden >50%, as already reported;26 in early/prefibrotic PMF patients’ PFS was influenced by the presence of the CALR type 1 mutation, WBC count >11x109/L and male sex, in agreement with previous reports.27 Moreover, patients with M-ACT had a significant correlation with a worse PFS and with an overt-myelofibrotic BM failure, in both PV and early/prefibrotic PMF (P<0.0001). This last result is also confirmed at multivariate analysis. Interestingly, PV and early/prefibrotic PMF patients with this parameter showed a rapid clinical progression before the end of the 5-year follow-up, suggesting that M-ACT could be an early predictive marker capable of precociously identifying patients that need a closer follow-up. Numerous scientific papers have highlighted that in the evolution towards myelofibrosis of MPN, a central role seems to be played by MK. Patients with MPN and fibrotic evolution showed a significantly increased number of MK with an abnormal nuclear/cytoplasmic ratio and a reduced polyploid state, often organized in clusters.27,28 Experiments using in vitro cultures of CD34+ hematopoietic stem cells of patients with fibrotic MPN have shown that MK expand excessively, are immature and show delayed apoptosis owing to increased expression of the anti-apoptotic factor BCL-XL.29 Moreover, mice with a MK-specific deficiency of the transcription factor–encoding gene GATA1 show elevated numbers of immature MK in the BM and an increased and pathologic neutrophil emperipolesis that may represent one of the mechanisms leading to myelofibrosis by releasing fibrogenic MK cytokines and neutrophil proteases in the microenvironment of in vivo experiments.14,16 Finally, MK from individuals with PMF secrete increased levels of the fibrotic cytokines such as TGF-b, compared to MK from healthy individuals, and the extracellular matrix (ECM) microenvironment, especially the fibronectin component, is able to sustain progenitor cell proliferation and megakaryopoiesis in a TPO-independent manner.16,30,31 These pro-fibrotic cytokines would presumably act mainly in the microenvironment near to those MK clusters which are, in turn, their main producers. Furthermore, the criteria defining the megakaryocytic activation could represent the morphological counterpart of what is postulated by in vitro and in vivo studies regarding the role of MK in the BM fibrotic evolution of patients with MPN. Recent evidence has suggested that treating patients with early-stage MF may lead to better outcomes with a less severe splenomegaly, a lower incidence of cytopenia, and less-severe BM fibrosis. However, the argument is debated, especially considerating the adverse events of the JAK2 inhibitor treatment (ruxolitinib). M-ACT parameter, 3167

M. Schino et al.

as an early predictive marker capable of precociously identifying patients with an overt-myelofibrotic BM failure, could also select those patients that would benefit from precocious treatment.32 Our analysis not only supports the role of the MK and their activation in the evolution of PV and early/prefibrotic PMF, but also seems to suggest that, for the treatment of this neoplasia, as well as the inhibition of specific mutations, which may partially alter the natural history of the disease, the blockage of the fibrotic evolution and therefore of its main key-player, the MK, should be a future therapeutic strategy to be investigated.30 M-ACT is also a very useful morphological parameter in the diagnostic phase of MPN. In fact, none of the ET patients showed M-ACT, which when present, identifies only an early/prefibrotic PMF. Interestingly the M-ACT showed a significant lower incidence in triple-negative patients in comparison to those with a driver gene mutation (24.7% vs. 40.5%) reinforcing the idea of a more indolent disease for this subgroup while maintaining its predictive role for the fibrotic evolution also in triple-negative patients. In addition, M-ACT parameter evaluation represents an easily executable analysis with a high agreement index between pathologists. Moreover, the search for MACT in BM biopsies in the diagnostic phase of MPN patients can be performed widely without the need for further analysis such as immunohistochemistry or molecular analysis. The main limitation of our study is the retrospective design, hence the time estimate of progression to overt myelofibrosis and the estimate of PFS based on the patient chart review can lack of accuracy. Moreover, the team physician remained the same over years and the cri-

References 1. Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA, Stein H, Thiele J. WHO classification of tumours of haematopoietic and lymphoid tissues (Revised 4th edition). IARC: Lyon 2017. 2. Murphy S. Diagnostic criteria and prognosis in polycythemia vera and essential thrombocythemia. Semin Hematol. 1999;36(1 Suppl 2):S9-13. 3. Georgii A, Buesche G, Kreft A. The histopathology of chronic myeloproliferative diseases. Baillieres Clin Haematol. 1998;11(4):721-749. 4. Barbui T, Thiele J, Gisslinger H, et al. The 2016 WHO classification and diagnostic criteria for myeloproliferative neoplasms: document summary and in-depth discussion. Blood Cancer J. 2018;8(2):15. 5. Cervantes F, Dupriez B, Pereira A, et al. New prognostic scoring system for primary myelofibrosis based on a study of the International Working Group for Myelofibrosis Research and Treatment. Blood. 2009;113(13):2895-2901. 6. Passamonti F, Cervantes F, Vannucchi AM, et al. A dynamic prognostic model to predict survival in primary myelofibrosis: a study by the IWG-MRT (International Working Group for Myeloproliferative Neoplasms Research and Treatment). Blood. 2010;115(9):1703-1708. 7. Gangat N, Caramazza D, Vaidya R, et al. DIPSS plus: a refined Dynamic International Prognostic Scoring System for

3168

teria for the diagnosing progression to overt myelofibrosis were aligned with those of the IWG-MRT consensus.21 Although we analyzed a large cohort, the results of this single-center study need confirmation in other independent MPN patient cohorts, and M-ACT should be validated as a prognostic tool. Disclosures The authors have no conflict of interest to disclose. The study was carried out in accordance with the Declaration of Helsinki and the consent for retrospective analysis of all clinical data, according to the Ethical Committee of the Catholic University School of Medicine, was obtained by all the patients at the hospital admission. The report does not present identifying images or other personal or clinical details of participants that compromise anonymity. Contributions LML, MS, VF and MM were the principal authors and the main contributors in writing the manuscript; ER, SB, VR, MDC and PC analyzed and interpreted the patient data; MM, LML, MS and VF performed the biopsies analysis; LML and VDS read and corrected the manuscript. All authors read and approved the final manuscript. Acknowledgments We thank Dr. Sara Capodimonti and Dr. Tonia Cenci for their technical support. Funding This study was supported by Università Cattolica del Sacro Cuore, Fondi d’Ateneo, Linea D1 (2018 and 2019; MM and LML).

primary myelofibrosis that incorporates prognostic information from karyotype, platelet count, and transfusion status. J Clin Oncol. 2011;29(4):392-397. 8. Rumi E, Cazzola M. Diagnosis, risk stratification, and response evaluation in classical myeloproliferative neoplasms. Blood. 2017;129(6):680-692. 9. Grinfeld J, Nangalia J, Baxter EJ et al. Classification and personalized prognosis in myeloproliferative neoplasms. N Engl J Med. 2018;379(15):1416-1430. 10. Guglielmelli P, Lasho TL, Rotunno G, et al. MIPSS70: mutation-enhanced international prognostic score system for transplantation-age patients with primary myelofibrosis. J Clin Oncol. 2018;36(4):310-318. 11. Tefferi A, Guglielmelli P, Lasho TL, et al. MIPSS70+ Version 2.0: mutation and karyotype-enhanced international prognostic scoring system for primary myelofibrosis. J Clin Oncol. 2018;36(17):1769-1770. 12. Tefferi A, Guglielmelli P, Nicolosi M, et al. GIPSS: genetically inspired prognostic scoring system for primary myelofibrosis. Leukemia. 2018;32(7):1631-1642. 13. Tefferi A, Guglielmelli P, Lasho TL, et al. Mutation-enhanced international prognostic systems for essential thrombocythaemia and polycythaemia vera. Br J Haematol. 2020;189(2):291-302. 14. Larocca LM, Heller PG, Podda G, et al. Megakaryocytic emperipolesis and platelet function abnormalities in five patients with gray platelet syndrome. Platelets. 2015;26 (8):751-757.

15. Schmitt A, Jouault H, Guichard J, Wendling F, Drouin A, Cramer EM. Pathologic interaction between megakaryocytes and polymorphonuclear leukocytes in myelofibrosis. Blood. 2000;96(4):1342-1347. 16. Centurione L, Di Baldassarre A, Zingariello M, et al. Increased and pathologic emperipolesis of neutrophils within megakaryocytes associated with marrow fibrosis in GATA-1(low) mice. Blood. 2004;104(12):3573-3580. 17. Yan XQ, Lacey D, Hill D, et al. A model of myelofibrosis and osteosclerosis in mice induced by overexpressing thrombopoietin (mpl ligand): reversal of disease by bone marrow transplantation. Blood. 1996;88(2): 402-409. 18. Gianelli U, Iurlo A, Vener C, et al. The significance of bone marrow biopsy and JAK2V617F mutation in the differential diagnosis between the "early" prepolycythemic phase of polycythemia vera and essential thrombocythemia. Am J Clin Pathol. 2008;130(3):336-342. 19. Ridgeway JA, Tinsley S, Kurtin SE. Practical guide to bone marrow sampling for suspected myelodysplastic syndromes. J Adv Pract Oncol. 2017;8(1):29-39. 20. De Stefano V, Ruggeri M, Cervantes F, et al. High rate of recurrent venous thromboembolism in patients with myeloproliferative neoplasms and effect of prophylaxis with vitamin K antagonists. Leukemia. 2016; 30(10):2032-2038. 21. Barosi G, Mesa RA, Thiele J, et al. Proposed criteria for the diagnosis of post-poly-

haematologica | 2021; 106(12)

Megakaryocytic activation and fibrotic evolution of MPN

cythemia vera and post-essential thrombocythemia myelofibrosis: a consensus statement from the International Working Group for Myelofibrosis Research and Treatment. Leukemia. 2008;22(2):437-438. 22.Lee SH, Erber WN, Porwit A, Tomonaga M, Peterson LC. International council for standardization in hematology. ICSH guidelines for the standardization of bone marrow specimens and reports. Int J Lab Hematol. 2008;30(5):349-364. 23.Vardiman JW, Thiele J, Arber DA, et al. The 2008 revision of the World Health Organization (WHO) classification of myeloid neoplasms and acute leukemia: rationale and important changes. Blood. 2009;114(5):937-951. 24.Afkhami M, Vergara-Lluri M, Brynes RK, Siddiqi IN. Peripheral blood smears, bone marrow aspiration, trephine and clot biopsies: methods and protocols. Methods Mol

haematologica | 2021; 106(12)

Biol. 2014;1180:257-269. 25.Martini M, Cenci T, D'Alessandris GQ, et al. Epigenetic silencing of Id4 identifies a glioblastoma subgroup with a better prognosis as a consequence of an inhibition of angiogenesis. Cancer. 2013;119(5):10041012. 26.Passamonti F, Rumi E, Pietra D, et al. A prospective study of 338 patients with polycythemia vera: the impact of JAK2 (V617F) allele burden and leukocytosis on fibrotic or leukemic disease transformation and vascular complications. Leukemia. 2010;24(9):1574-1579. 27.Cerquozzi S, Tefferi A. Blast transformation and fibrotic progression in polycythemia vera and essential thrombocythemia: a literature review of incidence and risk factors. Blood Cancer J. 2015;5(11): e366. 28.Ciurea SO, Merchant D, Mahmud N, et al.

Pivotal contributions of megakaryocytes to the biology of idiopathic myelofibrosis. Blood. 2007;110(3):986-993. 29.Villeval JL, Cohen-Solal K, Tulliez M, et al. High thrombopoietin production by hematopoietic cells induces a fatal myeloproliferative syndrome in mice. Blood. 1997;90(11):4369-4383. 30.Wen QJ, Yang Q, Goldenson B, et al. Targeting megakaryocytic-induced fibrosis in myeloproliferative neoplasms by AURKA inhibition. Nat Med. 2015;21(12):1473-1480. 31.Malara A, Gruppi C, Abbonante V, et al. EDA fibronectin-TLR4 axis sustains megakaryocyte expansion and inflammation in bone marrow fibrosis. J Exp Med. 2019;216(3):587-604. 32.Palandri F, Sabattini E, Maffioli M. Treating early-stage myelofibrosis. Ann Hematol. 2019;98(2):241-253.

3169

ARTICLE Ferrata Storti Foundation

Haematologica 2021 Volume 106(12):3170-3175

Non-Hodgkin Lymphoma

Phase I study of selinexor in combination with dexamethasone, ifosfamide, carboplatin, etoposide chemotherapy in patients with relapsed or refractory peripheral T-cell or natural-killer/T-cell lymphoma Tiffany Tang,1 Peter Martin,2 Nagavalli Somasundaram,1 Cindy Lim,3 Miriam Tao,1 Eileen Poon,1 Maica J.D. Yunon,3 Shu Q. Toh,3 Sean X. Yan,4 Mohamad Farid,1 Jason Y. Chan1 and Soon T. Lim1 1

Division of Medical Oncology, National Cancer Center Singapore, Singapore; Division of Medicine, Weill Cornell Medical College, New York, NY, USA; 3Division of Clinical Trials and Epidemiology, National Cancer Centre Singapore, Singapore and 4 Department of Nuclear Medicine and Molecular Imaging, Singapore General Hospital, Singapore 2

ABSTRACT

Correspondence: TIFFANY TANG Tiffany.Tang.PL@gmail.com Received: February 28, 2020. Accepted: October 7, 2020. Pre-published: November 5, 2020.

elinexor is a selective inhibitor of nuclear export with anti-cancer properties. We performed a phase I study to determine the safety and maximum tolerated dose of selinexor when combined with high-dose dexamethasone, ifosfamide, carboplatin and etoposide (DICE) in relapsed/refractory T-cell lymphoma (TCL) and natural-killer/T-cell lymphoma (NKTL). Patients with relapsed/refractory TCL and NKTL were treated with standard dose ICE, dexamethasone 20 mg on days 3 to 7, and escalating doses of oral selinexor on days 3, 5 and 7 in a 3+3 design. Dose levels (DL) 1, 2 and 3 were 40, 60 and 80 mg, respectively. Eleven patients with a median age of 60 years were enrolled; six at DL1 and five at DL2. Patients had received a median of two (range, 1-4) prior lines of treatment and seven had primary refractory disease at entry into the study. Patients received a median of three cycles (range, 1-6) of selinexor-DICE. The most common grade 1 or 2 toxicities included nausea (64%), fatigue (55%), and anorexia (45%) and the most common grade 3 or 4 toxicities included thrombocytopenia (82%), anemia (82%), neutropenia (73%), and hyponatremia (73%). Two patients developed dose-limiting toxicities at DL2 and one at DL1. Five patients discontinued treatment for reasons other than disease progression or lack of response. Of the ten evaluable patients, the overall and complete response rates were 91% and 82%, respectively. The maximum tolerated dose of selinexor was 40 mg when combined with DICE. The combination showed promising complete response rates in patients with relapsed/refractory TCL and NKTL but was poorly tolerated. (clinicaltrials.gov identifier: NCT03212937).

https://doi.org/10.3324/haematol.2020.251454

3170

Introduction T-cell lymphoma (TCL) is a heterogeneous group of non-Hodgkin lymphomas seen more commonly in Asia than in the West.1,2 The 5-year overall survival rates are approximately 30% for the most common subtypes of TCL, including peripheral-T cell lymphoma (PTCL)-not otherwise specified (NOS), angioimmunoblastic T-cell lymphoma (AITL) and natural-killer/T-cell lymphoma (NKTL).2 Patients with PTCL who relapse or progress after initial therapy have poor survival outcomes with median progression-free and overall survival of 3.1 and 5.5 months, respectively.3 However, patients who achieve a complete response to salvage therapy have better median progression-free and overall survival (12.2 and 18 months, respectively).3 Some patients who achieve a complete response to salvage therapy may be considered for high-dose chemotherapy (HDC) and autologous stem cell

haematologica | 2021; 106(12)

Phase I study of selinexor-DICE in relapsed PTCL and NKTL

transplantation (SCT) consolidation with curative intent.4 Thus there is a need to improve complete response rates for salvage regimens. Exportin 1 (XPO1/CRM1) is a nuclear export protein that is responsible for the nuclear to cytoplasmic translocation of tumor suppressor proteins (TSP) and growth regulator proteins (GRP) such as TP53, p21, p27, FOXO3 and nucleophosmin 1 (NPM1), leading to their inactivation.5 XPO1 is overexpressed in many malignancies including TCL and increased XPO1 expression is associated with poor survival.6-10 XPO1 also transports topoisomerase II enzymes to the cytoplasm and cytoplasmic localization of topoisomerase II enzymes has been identified as a mechanism of cancer resistance. Therefore, when topoisomerase IIα enzymes are not in contact with DNA, topoisomerase II inhibitors, such as doxorubicin, are unable to induce cell death.11 Selinexor® is an oral, first-in-class, potent selective inhibitor of nuclear export, which binds to XPO1, leading to nuclear retention of the TSP, GRP, and topoisomerase IIα enzymes, restoring their function. Selinexor has received Food and Drug Administration approval for relapsed or refractory multiple myeloma and diffuse large B-cell lymphoma, and has shown significant anticancer activity across a range of preclinical models of cancer, including T-cell acute lymphoblastic leukemia.12 There were also preclinical studies demonstrating the ability of selinexor to sensitize cancer cells to topoisomerase inhibitors,13 alkylating agents5 and steroids.14 A phase I study of selinexor in relapsed/refractory non-Hodgkin lymphomas showed overall response rates of about 30%.15 We hypothesized that selinexor could synergize with ifosfamide (an alkylating agent) and etoposide (a topoisomerase II inhibitor) in the ifosfamide, carboplatin and etoposide (ICE) regimen and we added high-dose dexamethasone to this regimen to improve the efficacy of ICE as a salvage regimen for TCL. We conducted a phase I study to identify the dose of selinexor that could be combined safely with standard-dose ICE and high-dose dexamethasone (DICE) in relapsed or refractory TCL (clinicaltrials.gov identifier: NCT03212937).

dose)15,16 and there was concern that Asian patients tolerated selinexor less well than Caucasian patients. Hence DL -1, 1, 2 and 3 were 20, 40, 60 and 80 mg, respectively. All patients received intravenous doses of ICE in a 21-day cycle: ifosfamide 5 g/m2 over days 1-3, carboplatin (area under the curve 5) on day 1 and etoposide 100 mg/m2 on days 1-3. Selinexor was administered on days 3, 5 and 7. Additionally, all patients received oral dexamethasone 20 mg/day for 5 days on days 3-7 for anticipated anticancer synergy of steroids with selinexor. Anti-emetics included oral aprepitant and granisetron 3 mg on days 1-3 and dexamethasone 8 mg on days 1-2. Oral olanzapine 5 mg was recommended with each dose of selinexor. Eligible patients could undergo HDC and SCT after at least two cycles of study treatment. Patients who were not eligible for SCT could receive up to six cycles of the study treatment. Patients could also receive maintenance selinexor (60 mg weekly) if they had not progressed upon completion of selinexor-DICE.

Assessment of adverse events and dose-limiting toxicities Dose-limiting toxicities were defined as any of the following treatment-related toxicities occurring during the first cycle of treatment: failure to resolve any grade 3 or higher non-hematologic toxicities, platelet count of less than 75x109/L or absolute neutrophil count of less than 1x109/L by day 29, a platelet count of less than 25x109/L or an absolute neutrophil count of less than 0.5x109/L lasting more than 14 days, a platelet nadir of 10x109/L or less, or any grade 5 toxicities.

Response assessment Responses were assessed using the revised International Working Group Criteria for non-Hodgkin lymphoma.17 Tumor measurements with positron emission tomography and computed tomography scans were performed at baseline, after two cycles of selinexor-DICE, and 6-8 weeks after the last cycles of selinexorDICE or after HDC/SCT.

Statistical analysis Any patient who received one dose of selinexor was included in the safety population and only the patients who completed two cycles of treatment and the first response assessment were included in the efficacy analysis.

Methods Patients Recruited patients had histologically confirmed relapsed or refractory PTCL or NKTL. Patients with CD30+ anaplastic large cell lymphoma (ALCL) had to have failed treatment with brentuximab vedotin. The study was conducted at the National Cancer Center Singapore and the Singapore General Hospital after approval by the Singhealth Institutional Review Board and in accordance with the Declaration of Helsinki and Good Clinical Practice Guidelines of the International Conference on Harmonization. Written consent was obtained from all patients prior to their entry into the study.

Study design This was an open-label, phase I study in which eligible patients were treated with DICE plus escalating doses of oral selinexor in a 3+3 design. The primary objective was to assess the safety and determine the maximum tolerated dose of selinexor that could be combined with DICE. The first dose level (DL) of selinexor was chosen as 40 mg because at the time of developing the study the recommended phase II dose of selinexor from phase I studies was 60 mg (fixed

haematologica | 2021; 106(12)

Results Eleven patients were recruited into the study. The median age was 60 years (range, 34-74), and nine were male. All patients were Asian; seven (64%) were Chinese, two (18%) were Malay, one (9%) was Indian and another (9%) was Myanmarese. The most common histological subtype in this study was AITL (n=5), followed by PTCL-NOS (n=2). There were one of each of the following histological subtypes; ALK-negative ALCL, ALK-positive ALCL, NKTL and PTCL with T-follicular helper phenotype. Two patients had stage II disease and the rest had stage IV disease at study entry. The patients had received a median of two prior lines of treatment, one had received prior HDC and autologous SCT, and one had been previously administered radiotherapy. Seven patients (64%) had primary refractory disease, defined as disease that had not responded to any prior chemotherapy, or disease that progressed within 8 weeks from the end of treatment response assessment. All patients had an Eastern Cooperative Oncology Group performance status of 0 or 1. (Table 1) 3171

T. Tang et al.

Patients received a median number of three cycles (range, 1-6) of selinexor-DICE. Three patients were eligible for HDC/SCT. (Table 2) Two patients underwent SCT, one autologous and the other allogeneic SCT. The patient who underwent autologous SCT had 8.52x106/L CD34 cells collected prior to the transplant: engraftment of neutrophils and platelets occurred on days 9 and 8, respectively. Engraftment of neutrophils and platelets in the patient who underwent allogeneic SCT occurred on days 10 and 14, respectively. The third patient eligible for HDC experienced disease progression just before autologous SCT and he was given alectinib before proceeding to allogeneic SCT. Eight patients were not eligible for HDC/autologous SCT and these patients received a median of 3.5 of the planned six cycles of study treatment (Table 2). Three patients completed all six cycles of study treatment, two patients discontinued treatment because of adverse events, another two refused to continue treatment, and one patient’s treatment was discontinued as a result of the investigator’s decision. Two patients received maintenance selinexor upon completing six cycles of selinexor-ICE. The most common grade 1 and 2 toxicities included nausea (64%), fatigue (55%), and anorexia (45%). Grade 3 or 4 toxicities occurring in at least one patient included thrombocytopenia (82%), anemia (82%), neutropenia (73%), hyponatremia (73%), leukopenia (64%), febrile neutropenia (45%), and one patient each who developed fatigue, anorexia, fever, hypokalemia, sepsis, and an upper respiratory tract infection (Table 3). There were no treatment-related deaths. All patients had hyponatremia (any grade), which took a median of 7 days (range, 1 to 21) to resolve. Considering all the cycles of treatment, dose reductions for selinexor, ifosfamide, carboplatin and etoposide were required in five (45%), ten (91%), eight (73%), and nine (82%) of patients, respectively. At DL1 and DL2, the median dose intensities (range) for selinexor were 94.2% (range, 66.7-100%) and 77.8% (range, 72.288.9%), respectively. The median dose intensities for ifosfamide, carboplatin and etoposide were 79.8% (range, 74.8-99.4%), 87.7% (range, 76.9-105.7%), and 73.0% (range, 26.0-98.7%), respectively. Six and five patients received selinexor at the dose of 40 mg and 60 mg, respectively. Patient 11 developed a doselimiting toxicity at DL1. He developed a platelet nadir of less than 10x109/L and failed to recover his platelet count to at least 75x109/L by day 28. Two patients developed doselimiting toxicities at DL2 (patients 5 and 8): one had a platelet nadir of less than 10x109/L on day 11 and in another the platelet count failed to recover to at least 75x109/L by day 28 but did recover by day 32. For these two patients, the platelet counts recovered to 75x109/L within 11 days of the dose-limiting toxicity and they both remained on study. All three patients who developed dose-limiting toxicities had a baseline platelet count of more than 100x109/L. Thus, 40 mg of selinexor was the maximum tolerated dose that could be combined with DICE in this study. Ten patients were evaluable for response after two cycles of treatment. One patient (patient 11) progressed before he could be evaluated by positron emission tomography. Of the ten evaluable patients, all responded; nine (82% [95% confidence interval: 48-98]) achieved a complete response and one (10%) achieved a partial response (Table 3). The median follow-up of the study was 32.3 months (range, 4.4-36.6). During this period, seven patients experienced disease progression and five patients 3172

Table 1. Baseline demographic data.

Characteristic

N. (%)

Age at trial entry, years median (range) Gender Female Male Histological subtype AITL PTCL-NOS ALCL NKTL Other (PTCL with T-follicular helper phenotype) Ann Arbor staging 2 4 N. of prior chemotherapy regimens Median (range) 1 2 3 4 Previous treatments CHOP/CHOPE/CEPP GDP/GEMOX/GIFOX Brentuximab Bendamustine DHAP Etoposide/cyclophosphamide Pembrolizumab PUVA Romidepsin SMILE Relapsed or refractory disease Relapsed Refractory Prior HDC/ASCT No Yes Prior radiotherapy No Yes ECOG performance status 0 1 Eligibility for HDC/ASCT on study entry No Yes

60 (34-74) 2 (18%) 9 (82%) 5 (45%) 2 (18%) 2 (18%) 1 (9%) 1 (9%) 2 (18%) 9 (82%) 2 (1-4) 3 (27%) 5 (45%) 2 (18%) 1 (9%) 9 (82%) 6 (55%) 2 (18%) 1 (9%) 1 (9%) 1 (9%) 1 (9%) 1 (9%) 1 (9%) 1 (9%) 4 (36%) 7 (64%) 10 (91%) 1 (9%) 10 (91%) 1 (9%) 3 (27%) 8 (73%) 8 (73%) 3 (27%)

AITL: angioimmunoblastic T-cell lymphoma;, PTCL: peripheral T-cell lymphoma; NOS: not otherwise specified; ALCL: anaplastic large cell lymphoma; NKTL: natural-killer/T-cell lymphoma, CHOP: cyclophosphamide, doxorubicin, vincristine, prednisolone; CHOPE: cyclophosphamide, doxorubicin, vincristine, prednisolone, etoposide; CEPP: cyclophosphamide, etoposide, procarbazine, prednisolone; GDP: gemcitabine, dexamethasone, cisplatin; GEMOX: gemcitabine, oxaliplatin; GIFOX: gemcitabine, ifosfamide, oxaliplatin; DHAP: dexamethasone, highdose cytarabine, cisplatin; PUVA: psoralen and ultraviolet A; SMILE: steroids, methotrexate, ifosfamide, l-asparaginase, etoposide; HDC/ASCT: high-dose chemotherapy and autologous stem cell transplantation); ECOG: Eastern Cooperative Oncology Group.

haematologica | 2021; 106(12)

Phase I study of selinexor-DICE in relapsed PTCL and NKTL

died. The median overall survival was not reached and the 1-year overall survival rate was 66.7% (95% confidence interval: 28.2-87.8%) (Figure 1).

Discussion In this study of Asian patients with relapsed or refractory TCL, we found that the maximum tolerated dose of selinexor that could be combined with high-dose dexamethasone and standard dose ICE i.e., DICE, in a 21-day cycle, was 40 mg on days 3, 5 and 7. The combination was highly active with a response rate of 100% among the

evaluable patients. However, toxicities were significant. Patients ineligible for HDC/autologous SCT underwent a median of 3.5 cycles of this treatment, and five of 11 (45.5%) discontinued treatment for reasons other than disease progression or a lack of response. Hyponatremia is a known and well-established adverse event associated with selinexor and was the most common non-hematologic adverse event that occurred in this study. All-grade hyponatremia occurred after selinexor was administered, was transient and resolved after a median of 7 days from the onset. Patients who developed hyponatremia were generally asymptomatic and managed with oral rehydration salts, sodium tablets, intravenous

Figure 1. Kaplan-Meier plot of overall survival in the efficacy population.

Table 2. Disposition of the patients.

Patient

Age (in years)/ gender

Dose level

Eligibility for HDC/ASCT

Reasons for HDC/ASCT ineligibility

Number of treatment cycles completed

Reasons for treatment discontinuation

Maintenance selinexor treatment

1 2 3 4

70/Male 38/Female 52/Male 61/Male

1 1 1 2

No Yes No No

2 3 4 6

Patient’s decision Proceeded to HDC/ASCT Adverse events Completed 6 cycles

No No Yes Yes

52/Male

Yes

Advanced age NA Prior HDC/ASCT Poor heart function EF 36% from IHD NA

6 7

60/Male 49/Male

2 2

No No

6 6

No No

8 9 10 11

34/Male 67/Male 74/Female 68/Male

2 1 1 1

Yes No No No

NKTL* Poor heart function EF 45% from non-ischemic cardiomyopathy NA Advanced age Advanced age Advanced age

Proceeded to allogeneic transplant Completed 6 cycles Completed 6 cycles

3 3 3 1

Proceeded to HDC/ASCT Investigator’s decision Patient’s decision Adverse events

No No No No

HDC/ASCT: high-dose chemotherapy and autologous stem cell transplantation; NA: not applicable; EF: ejection fraction; IHD: ischemic heart disease; NKTL: natural-killer/T-cell lymphoma. *HDC/ASCT in not performed for NKTL in our institution due to historical lack of efficacy and this patient declined an allogeneic stem cell transplant.

haematologica | 2021; 106(12)

3173

T. Tang et al. Table 3. Treatment-related adverse events.

Adverse event Non-hematologic Electrolyte imbalances Hyponatremia Hypokalemia Gastrointestinal Nausea Anorexia Vomiting Diarrhea Oral ulcer General Fatigue Edema Others Blurred vision Upper respiratory infection Dysgeusia Dizziness Encephalopathy Lethargy Rhinorrhea Rash Hematologic Thrombocytopenia Anemia Leukopenia Neutropenia Febrile neutropenia

Grade 1

Grade 2

Grade 3 Non-hematologic

Grade 4

Any grade

3 (27%) 1 (9%)

0 1 (9%)

8 (73%) 1 (1%)

0 0

11 (100%) 3 (27%)

2 (18%) 3 (27%) 2 (18%) 3 (27%) 2 (18%)

5 (45%) 2 (18%) 2 (18%) 1 (9%) 0

0 1 (9%) 0 0 0

0 0 0 0 0

7 (64%) 6 (55%) 4 (36%) 4 (36%) 2 (18%)

4 (36%) 2 (18%)

2 (18%) 1 (9%)

1 (9%) 0

0 0

7 (64%) 3 (27%)

4 (36%) 0 3 (27%) 2 (18%) 2 (18%) 2 (18%) 2 (18%) 2 (18%)

0 3 (27%) 0 0 0 0 0 0

0 1 (9%) 0 0 0 0 0 0

0 0 0 0 0 0 0 0

4 (36%) 4 (36%) 3 (27%) 2 (18%) 2 (18%) 2 (18%) 2 (18%) 2 (18%)

2 (18%) 0 0 1 (9%) 0

0 2 (18%) 0 2 (18%) 0

1 (9%) 9 (82%) 1 (9%) 1 (9%) 5 (45%)

8 (73%) 0 6 (55%) 7 (64% 0

11 (100%) 11 (100%) 7 (64%) 11 (100%) 5 (45%)

The table lists treatment-related adverse events experienced by at least 10% of patients. One patient (9%) developed grade 4 sepsis (not recorded in the table above).

saline infusions and, from mid-way through the study, patients were also encouraged to prophylactically hydrate with electrolyte-rich salt drinks during the period they were on selinexor. Our clinical trial together with currently available data suggest that there may be a higher incidence of grade 3 or more selinexor-induced hyponatremia among Asian patients16,18-20 and further studies will be required to substantiate and understand this phenomenon better. Although hyponatremia can occur with high-dose ifosfamide, the rates of hyponatremia that occurred in this study were much higher than those based on experiences with ICE therapy alone. The incidence of grade 3 or 4 thrombocytopenia in this study was high (82%) and both dose-limiting toxicities were related to low platelet counts. This can be expected given the overlapping toxicities of selinexor and ICE (especially carboplatin). However, there were no bleeding complications associated with the thrombocytopenia. In addition, the thrombocytopenia was transient and resolved within 11 days, with platelet counts reaching 75x109/L. In phase I studies of single-agent selinexor, the rates of grade 3 or 4 thrombocytopenia were between 14-50%,15,16,21 with the higher rates seen in the phase I study of selinexor in relapsed or refractory non-Hodgkin lymphoma.15 In the phase II study of selinexor in multiply relapsed multiple myeloma, the rates of grade 3 or 4 thrombocytopenia 3174

were about 58%.22 Preclinical studies suggest that the mechanism of selinexor-induced thrombocytopenia are related to selinexor inhibiting megakaryocyte maturation from progenitor cells.23 Thrombocytopenia, which is a well-established side effect of selinexor, appears dose- and schedule-dependent and can be managed with dose interruptions and modifications.24 It may be that patients who had received more myelotoxic chemotherapy, prior to receiving selinexor, were more prone to severe thrombocytopenia and this will be an important consideration for future clinical trial development. The most common grade 1 and 2 adverse events were nausea, anorexia and fatigue and the rates of these adverse events were not very different from those in phase I studies of single-agent selinexor in solid tumors and hematologic malignancies.15,16,21 It is likely that these adverse events, which overlap with the adverse events seen with ICE, were not much more frequent than those seen in the single-agent studies because selinexor was administered with an aggressive anti-emetic strategy and only in the first week of each cycle so that patients had 2 weeks to recover before the next cycle was due. Although not powered to assess response, we found the high complete response rate of 90% striking among this group of patients, the majority of whom had primary refractory disease on entering the study, including a haematologica | 2021; 106(12)

Phase I study of selinexor-DICE in relapsed PTCL and NKTL

patient with NKTL who progressed on SMILE. Although not directly comparable, in a retrospective multicenter study of 76 PTCL and NKTL patients treated with ICE chemotherapy in Singapore, we found that the overall and complete response rates were 57% and 34%, respectively, among this group of patients who were treated in the relapsed or refractory setting. The median progressionfree survival for this group of patients was 4.5 months.25 Zelenetz et al. previously also reported that PTCL patients treated with ICE had an overall response rate of about 54% and a complete response rate of 31%.26 The high response rates seen in this clinical trial may warrant further investigation of the role of selinexor with DICE in patients with relapsed and refractory PTCL and NKTL. Disclosures No conflicts of interests to disclose.

References 1. Anderson JR, Armitage JO, Weisenburger DD. Epidemiology of the non-Hodgkin's lymphomas: distributions of the major subtypes differ by geographic locations. NonHodgkin's Lymphoma Classification Project. Ann Oncol. 1998;9(7):717-720. 2. Vose J, Armitage J, Weisenburger D, International TCLP. International peripheral T-cell and natural killer/T-cell lymphoma study: pathology findings and clinical outcomes. J Clin Oncol. 2008;26(25):41244130. 3. Mak V, Hamm J, Chhanabhai M, et al. Survival of patients with peripheral T-cell lymphoma after first relapse or progression: spectrum of disease and rare longterm survivors. J Clin Oncol. 2013;31(16):1970-1976. 4. Hwang WY, Koh LP, Lim ST, et al. Multicenter study of comparative outcomes of hematopoietic stem cell transplant for peripheral T cell lymphoma and natural killer/T-cell lymphoma. Leuk Lymphoma. 2011;52(7):1382-1386. 5. Turner JG, Dawson J, Cubitt CL, Baz R, Sullivan DM. Inhibition of CRM1-dependent nuclear export sensitizes malignant cells to cytotoxic and targeted agents. Semin Cancer Biol. 2014;27:62-73. 6. Huang WY, Yue L, Qiu WS, Wang LW, Zhou XH, Sun YJ. Prognostic value of CRM1 in pancreas cancer. Clin Invest Med. 2009;32(6):E315. 7. Noske A, Weichert W, Niesporek S, et al. Expression of the nuclear export protein chromosomal region maintenance/exportin 1/Xpo1 is a prognostic factor in human ovarian cancer. Cancer. 2008;112(8):17331743. 8. Shen A, Wang Y, Zhao Y, Zou L, Sun L, Cheng C. Expression of CRM1 in human gliomas and its significance in p27 expression and clinical prognosis. Neurosurgery. 2009;65(1):153-159; discussion 159-160. 9. van der Watt PJ, Maske CP, Hendricks DT, et al. The karyopherin proteins, Crm1 and karyopherin beta1, are overexpressed in cervical cancer and are critical for cancer

haematologica | 2021; 106(12)

Contributions TT and PM developed the trial concept; TT, NS, MT, EP, MF, JYC, and STL were involved in recruiting and managing the patients on trial; CL performed the statistical analyses for the study; SXY evaluated all the scans to assess the responses; MJDY and SQT were the clinical trial coordinators involved in the data collection and running of the study; TT, PM and STL supervised the conduct of the study. All the authors reviewed the final manuscript. Funding The authors would like to thank the National Medical Research Council (Singapore) Clinical Trials Grants (NMRC/CTGICT/0001/2017) and Karyopharm Therapeutics for funding this study. The protocol development was supported by the American Society of Hematology, Clinical Research Training Institute (2015).

cell survival and proliferation. Int J Cancer. 2009;124(8):1829-1840. 10. Kojima K, Kornblau SM, Ruvolo V, et al. Prognostic impact and targeting of CRM1 in acute myeloid leukemia. Blood. 2013;121(20):4166-4174. 11. Engel R, Valkov NI, Gump JL, Hazlehurst L, Dalton WS, Sullivan DM. The cytoplasmic trafficking of DNA topoisomerase IIalpha correlates with etoposide resistance in human myeloma cells. Exp Cell Res. 2004;295(2):421-431. 12. Etchin J, Sanda T, Mansour MR, et al. KPT330 inhibitor of CRM1 (XPO1)-mediated nuclear export has selective anti-leukaemic activity in preclinical models of T-cell acute lymphoblastic leukaemia and acute myeloid leukaemia. Br J Haematol. 2013;161(1):117-127. 13. Turner JG, Marchion DC, Dawson JL, et al. Human multiple myeloma cells are sensitized to topoisomerase II inhibitors by CRM1 inhibition. Cancer Res. 2009; 69(17):6899-6905. 14. Muqbil I, Aboukameel A, Elloul S, et al. Anti-tumor activity of selective inhibitor of nuclear export (SINE) compounds, is enhanced in non-Hodgkin lymphoma through combination with mTOR inhibitor and dexamethasone. Cancer Lett. 2016;383(2):309-317. 15. Kuruvilla J, Savona M, Baz R, et al. Selective inhibition of nuclear export with selinexor in patients with non-Hodgkin's lymphoma. Blood. 2017;129(24):31753183. 16. Abdul Razak AR, Mau-Soerensen M, Gabrail NY, et al. First-in-class, first-inhuman phase I study of selinexor, a selective inhibitor of nuclear export, in patients with advanced solid tumors. J Clin Oncol. 2016;34(34):4142-4150. 17. Juweid ME, Wiseman GA, Vose JM, et al. Response assessment of aggressive nonHodgkin's lymphoma by integrated International Workshop Criteria and fluorine-18-fluorodeoxyglucose positron emission tomography. J Clin Oncol. 2005; 23(21):4652-4661. 18. Kuruvilla J, Gutierrez, M, Shah, BD, et al.

Preliminary evidence of anti tumor activity of selinexor (KPT-330) in a phase I trial of a first-in-class oral selective inhibitor of nuclear export (SINE) in patients with relapsed/refractory non-Hodgkin's lymphoma and chronic lymphocytic leukemia. Blood. 2013;122(21):90. 19. Tan DSP, Pang, MY, Yong, WP, et al Phase I study of the safety and tolerability of the Exportin 1 (XPO1) inhibitor selinexor (SXR) in Asian patients (pts) with advanced solid cancers. J Clin Oncol. 2015;22(15 suppl):2542. 20. Sweet K, Komrokji R, Padron E, et al. Phase I clinical trial of selinexor in combination with daunorubicin and cytarabine in previously untreated poor-risk acute myeloid leukemia. Clin Cancer Res. 2020;26(1):5460. 21. Garzon R, Savona M, Baz R, et al. A phase 1 clinical trial of single-agent selinexor in acute myeloid leukemia. Blood. 2017;129 (24):3165-3174. 22. Chari A, Vogl DT, Gavriatopoulou M, et al. Oral selinexor-dexamethasone for tripleclass refractory multiple myeloma. N Engl J Med. 2019;381(8):727-738. 23. Machlus KR, Wu SK, Vijey P, et al. Selinexor-induced thrombocytopenia results from inhibition of thrombopoietin signaling in early megakaryopoiesis. Blood. 2017;130(9):1132-1143. 24. Wang AY, Weiner H, Green M, et al. A phase I study of selinexor in combination with high-dose cytarabine and mitoxantrone for remission induction in patients with acute myeloid leukemia. J Hematol Oncol. 2018;11(1):4. 25. Tay T, Somasundaram N, Khoo LP, et al. Treatment outcomes of T and naturalkiller/T-cell lymphoma with ifosfamide, carboplatin, etoposide (ICE) chemotherapy. Br J Haematol. 2019;185(Suppl. 1):3-202. 26. Zelenetz AD, Hamlin P, Kewalramani T, Yahalom J, Nimer S, Moskowitz CH. Ifosfamide, carboplatin, etoposide (ICE)based second-line chemotherapy for the management of relapsed and refractory aggressive non-Hodgkin's lymphoma. Ann Oncol. 2003;14(Suppl 1):i5-10.

3175

ARTICLE Ferrata Storti Foundation

Plasma Cell Disorders

Expression of the chemokine receptor CCR1 promotes the dissemination of multiple myeloma plasma cells in vivo Mara N. Zeissig,1,2 Duncan R. Hewett,1,2 Vasilios Panagopoulos,1,2 Krzysztof M. Mrozik,1,2 L. Bik To,3 Peter I. Croucher,4,5 Andrew C.W. Zannettino1,2,6,7# and Kate Vandyke1,2# Myeloma Research Laboratory, Adelaide Medical School, Faculty of Health and Medical Sciences, University of Adelaide, Adelaide, South Australia; 2Precision Medicine Theme, South Australian Health and Medical Research Institute, Adelaide, South Australia; 3 Department of Hematology, Royal Adelaide Hospital, Adelaide, South Australia; 4Bone Biology Division, Garvan Institute of Medical Research, Sydney, New South Wales; 5St Vincent’s Clinical School, Faculty of Medicine, University of New South Wales, Sydney, New South Wales; 6Central Adelaide Local Health Network, Adelaide, New South Wales and 7Center for Cancer Biology, University of South Australia, Adelaide, New South Wales, Australia. 1

Haematologica 2021 Volume 106(12):3176-3187

ACWZ and KV contributed equally as co-senior authors.

ABSTRACT

https://doi.org/10.3324/haematol.2020.253526

ultiple myeloma (MM) disease progression is dependent on the ability of MM plasma cells (PC) to egress from the bone marrow (BM), enter the circulation and disseminate to distal BM sites. Expression of the chemokine CXCL12 by BM stromal cells is crucial for MM PC retention within the BM. However, the mechanisms which overcome CXCL12-mediated retention to enable dissemination are poorly understood. We have previously identified that treatment with the CCR1 ligand CCL3 inhibits the response to CXCL12 in MM cell lines, suggesting that CCL3/CCR1 signaling may enable egress of MM PC from the BM. Here, we demonstrated that CCR1 expression was an independent prognostic indicator in newly diagnosed MM patients. Furthermore, we showed that CCR1 is a crucial driver of dissemination in vivo, with CCR1 expression in the murine MM cell line 5TGM1 being associated with an increased incidence of bone and splenic disseminated tumors in C57BL/KaLwRij mice. Furthermore, we demonstrated that CCR1 knockout in the human myeloma cell line OPM2 resulted in a >95% reduction in circulating MM PC numbers and BM and splenic tumor dissemination following intratibial injection in NSG mice. Therapeutic targeting of CCR1 with the inhibitor CCX9588 significantly reduced OPM2 or RPMI-8226 dissemination in intratibial xenograft models. Collectively, our findings suggest a novel role for CCR1 as a critical driver of BM egress of MM PC during tumor dissemination. Furthermore, these data suggest that CCR1 may represent a potential therapeutic target for the prevention of MM tumor dissemination.

Introduction

Correspondence: KATE VANDYKE kate.vandyke@adelaide.edu.au Received: March 25, 2020. Accepted: October 23, 2020. Pre-published: November 5, 2020.

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.

3176

Multiple myeloma (MM) is an incurable hematological cancer characterized by the uncontrolled proliferation of clonal plasma cells (PC) within the bone marrow (BM).1 One of the key features of MM is the presence of MM PC at multiple sites throughout the BM, highlighting that dissemination of transformed PC is a critical process during disease development.1,2 In support of this, circulating MM PC are detectable by flow cytometry in approximately 75% of newly diagnosed MM patients.3 Importantly, the presence of elevated circulating MM PC predicts faster time to progression and poorer overall survival, independent of BM tumor burden.4-12 The dissemination of MM PC is a multi-step process requiring release from the supportive niche in the BM, intravasation into nearby blood vessels and subsequent

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

extravasation and homing to another BM site. Integrin mediated adhesion of MM PC to BM stromal cells (BMSC), and extracellular matrix components synthesized by BMSC, is well-established to mediate retention of MM PC within the niche.13 For example, MM PC express the integrin α4b1 (also known as very late antigen 4, VLA-4) that mediates adhesion to vascular cell–adhesion molecule 1 (VCAM-1) on BMSC and to the extracellular matrix component fibronectin.13 Importantly, the C-X-C chemokine ligand CXCL12 (also known as stromal cell-derived factor1; SDF-1), abundantly produced by BMSC,14 enhances adhesion to fibronectin and VCAM-1 through binding to its receptor CXCR4 on the surface of MM PC and inducing rapid conformational changes of the integrin α4b1 complex on MM PC.15 Notably, plerixafor-mediated inhibition of the CXCL12 receptor CXCR4 on MM PC results in mobilization of MM cells to the peripheral blood (PB) in a preclinical model of MM.15 These data suggest that CXCL12 is a critical BM retention signal for MM PC and that overcoming the CXCL12/CXCR4 signal may be required for release from the niche during dissemination. In a previous study by Azab and colleagues, increased hypoxia in the BM was shown to be associated with an increase in circulating MM PC in a preclinical model.16 Additionally, we have previously identified that overexpression of the hypoxia-inducible factor 2α (HIF-2α) in MM cell lines reduces their response to exogenous CXCL12 in vitro, suggesting that hypoxia may overcome CXCL12-mediated retention. Furthermore, we identified that hypoxia and HIF-2α increased expression of the C-C chemokine receptor CCR1 in human MM cell lines.17 CCR1 is a seven-transmembrane G-protein coupled receptor and its most potent activator is CCL3 (also known as macrophage inflammatory protein 1α; MIP-1α). Previous literature suggests that MM PC abundantly produce CCL318-21 which activates CCR1 expressed on osteoclasts leading to increased osteolysis,19 with CCR1 antagonists reducing osteolysis in a murine model of MM.22,23 In addition, CCL3 is a potent inducer of migration of patientderived MM PC and MM cell lines in vitro.17,19,20,24 In hematopoietic progenitors and natural killer cells, CCL3/CCR1 signaling drives mobilization from the BM, in part by inactivation of CXCL12/CXCR4.25,26 Similarly, our previous studies showed that either pre-treatment of MM cell lines with CCL3 or elevated CCR1 expression decreased tumor cell migration towards CXCL12 in vitro.17 Taken together, these data suggest that hypoxia-mediated increases in CCR1 expression may desensitize cells to CXCL12-mediated BM retention and thereby facilitate dissemination. In support of this, we have previously shown that expression of CCR1 in MM PC is associated with poorer prognosis and an increase in the number of circulating MM PC in newly diagnosed MM patients.17 Here, we further investigated the association between CCR1 expression and poor overall survival rates in MM patients. Furthermore, we investigated the role for CCR1 in the dissemination of MM PC in vivo. Initially, we determined whether CCR1 overexpression can promote tumor dissemination in the syngeneic 5TGM1/KaLwRij murine model of MM. Furthermore, using xenograft models of MM, we assessed whether CCR1 knockout limits the dissemination of MM PC in vivo. Lastly, we investigated whether pharmacological inhibition of CCR1 can be used as a viable therapeutic strategy to limit MM PC dissemination. haematologica | 2021; 106(12)

Methods Flow cytometry on patient samples Ethical approval for this study was obtained from the University of Freiburg Medical Center Ethics Review Committee and all patients provided written, informed consent, in accordance with the Declaration of Helsinki. CCR1 analysis was conducted on BM mononuclear cells from BM aspirates from 28 newly diagnosed MM (median age: 68 years [range, 49–84]; male:female ratio 1.15:1) and seven monoclonal gammopathy of undetermined significance (MGUS)1 (median age: 74 years [range, 53-88]; male:female ratio 1.7:1) patients. Cell surface CCR1 expression was assessed on viable CD38++/CD138+/CD45lo/CD19- malignant PC by multicolor flow cytometry (FACSARIA III; BD Biosciences, San Jose, CA) as previously described.17

Murine multiple myeloma models models C57BL.KaLwRijHsd (KaLwRij) mice were inoculated into the left tibia with 1x105 5TGM1-CCR1 or 5TGM1-EV cells. After 25 days, splenic tumor burden was assessed by bioluminescent imaging (Xenogen IVIS 100; Perkin Elmer), and tumor burden in the PB, injected tibiae, and pooled tibiae and femora from the contralateral leg was assessed by flow cytometry (LSRFortessa flow cytometer). NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were inoculated intratibially with 5x105 OPM2-CCR1-KO-1 or OPM2-EV-1 cells. For CCX9588 studies, mice were treated at 12-hour intervals via oral gavage with either the CCR1 antagonist CCX9588 (15 mg/kg; ChemoCentryx, CA) or polyethylene glycol (PEG) vehicle alone, commencing day 3 or day 14 following tumor cell injection. Primary and secondary BM and splenic tumor burden and PB tumour cells were assessed 28 days after tumor cell injection. Detailed methods can be found in the Online Supplementary Methods.

Results High CCR1 expression is associated with poorer prognosis in multiple myeloma patients We used flow cytometry to examine the expression of CCR1 on CD38++/CD138+/CD45lo/CD19- BM PC in a cohort of MM and MGUS patients who had not received previous treatment. BM PC expression of CCR1 was detectable by flow cytometry in 14.3% (one of seven) of MGUS patients and in 53.6% (15 of 28) of MM patients (Figure 1A). Furthermore, BM PC expression of CCR1 was significantly higher in MM patients than in MGUS patients (P<0.05, Figure 1A), consistent with our previous microarray analysis.17 Our previous analysis suggested that high levels of CCR1 expression in BM PC from newly diagnosed MM patients was associated with poorer overall survival in patients enrolled in the total therapy 3 trial.17 Here, we performed Cox regression analysis to determine if elevated (above median) CCR1 expression was an independent predictor of poor prognosis. In univariate analyses, elevated CCR1 expression, high-risk gene expression signature, elevated serum b2 microglobulin (≥5.5 mg/L), anemia (hemoglobin < 100 mg/L) and high plasma cell proliferative index (PI ≥10%) were associated with inferior overall survival (P<0.05, Table 1). Notably, multivariable analysis demonstrated that elevated CCR1 retained its association with poor prognosis (P<0.05, hazard ratio [HR]=2.5, 95% confidence interval [CI]: 1.0-5.9), when these other prognostic 3177

M.N. Zeissig et al.

factors were taken into account (Table 1). In order to further investigate if CCR1 expression was associated with poor outcomes for MM patients, we assessed CCR1 expression in BM PC using RNA-sequencing data from a cohort of MM patients who had a sample taken at diagnosis (baseline) and a sample taken following at least one line of therapy (subsequent). These data suggested that patients with relatively high CCR1 expression at baseline (n=7; P<0.05, HR=4.3, 95% CI: 1.0-18.1) or patients with elevated CCR1 expression following treatment (n=10; P=0.080, HR=3.0, 95% CI: 0.9-10.4) tended to have inferior survival compared with patients with low CCR1 expression both at baseline and following therapy (n=26; Figure 1 B and C). Taken together, these data suggest that CCR1 expression either at baseline or following treatment may be associated with poorer overall survival for MM patients.

Expression of CCR1 in the mouse multiple myeloma cell line 5TGM1 does not affect proliferation in vitro and increases incidence of splenic and bone dissemination in vivo As we have previously shown that CCR1 expression is associated with increased circulating MM PC numbers in MM patients,17 we hypothesized that the association between increased CCR1 expression and poor prognosis was due to a role for CCR1 in MM PC dissemination. In order to investigate this, we initially assessed whether constitutive expression of CCR1 affected the migration and dissemination of the mouse MM cell line 5TGM1, which does not express detectable CCR1 basally (Figure 2A), and exhibits low levels of spontaneous dissemination in vivo.29 Expression of functional HA-tagged CCR1 was confirmed by quantitative polymerase chain reaction (PCR) and by immunoprecipitation/western blotting (Figure 2A and B) and by the ability of the 5TGM1-CCR1 cells to migrate

3178

towards recombinant human CCL3 (rhCCL3) in a transwell assay (Figure 2C). Expression of CCR1 did not affect the proliferation of 5TGM1 cells, relative to 5TGM1-EV controls, either basally (P=0.63; Figure 2D) or following addition of rhCCL3 (P=0.99; Figure 2E). In order to investigate whether CCR1 expression affects 5TGM1 dissemination in vivo, 5TGM1-CCR1 or 5TGM1EV cells were intratibially injected into C57BL/KaLwRij mice. Primary tumor burden in the injected tibiae was not significantly different between animals injected with 5TGM1-CCR1 cells and controls (P=0.82; Figure 3A). Similarly, the numbers of circulating MM cells in the PB, or the tumor burden in the contralateral leg, were also not significantly different between groups (P=0.62 and P=0.41, respectively; Figure 3B and C). However, there was a significant increase in the incidence of tumor in the 5TGM1CCR1 group, with eight of 11 mice (73%) in this group having detectable green fluorescence protein positive (GFP+) cells in the contralateral leg, compared with four of 11 mice (36%) injected with 5TGM1-EV cells (P<0.0001; Figure 3D). Furthermore, an increase in the incidence of dissemination to the spleen was also observed in the 5TGM1-CCR1 group, with nine of 11 mice (82%) having tumors detectable in the spleen by bioluminescence imaging, compared with four of eight mice (50%) in the 5TGM1-EV cohort (P<0.0001; Figure 3E and F). Collectively, these data suggest that expression of CCR1 increases dissemination of MM PC, without affecting primary tumor growth.

Knockout of CCR1 in the human multiple myeloma cell line OPM2 does not affect proliferation in vitro and prevents dissemination in vivo In order to further investigate the role of CCR1 in tumor dissemination in MM, we generated CRISPR/Cas9-mediated CCR1 knockouts (KO) in the

Figure 1. CCR1 is expression is elevated in multiple myeloma patients and is associated with poor prognosis. (A) CCR1 expression (ΔMFI) on CD38++/CD138+/CD45lo/CD19- bone marrow (BM) plasma cells (PC) from newly diagnosed monoclonal gammopathy of undetermined significance (MGUS) (n=7) and multiple myeloma (MM) (n=28) patients was assessed by flow cytometry. Graph depicts median with interquartile range, showing all data points. (B) CCR1 expression is shown for CD138-selected BM MM PC from patients with a sample taken at diagnosis (baseline) and a sample taken following at least one line of therapy with bortezomib (subsequent) (CoMMpass RNA-sequencing dataset, n=43 patients). Patients were categorized as having low tumor expression of CCR1 (CCR1 <10 FPKM at both baseline and subsequent biopsy; n=26), high CCR1 (CCR1 ≥10 FPKM at baseline; n=7) or increased CCR1 (baseline CCR1 <10 FPKM and subsequent CCR1 ≥10 FPKM; n=10). (C) Kaplan-Meier plots of overall survival are shown for MM patients stratified based on BM MM PC expression of CCR1 at baseline and subsequent to therapy.

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

human MM cell line OPM2 (Online Supplementary Figure S1). Loss of CCR1 protein expression in OPM2-CCR1KO-1 and OPM2-CCR1-KO-2 cell lines was confirmed by flow cytometry (Figure 4A). Furthermore, migration towards rhCCL3 was not observed in the OPM2-CCR1KO cell lines, confirming loss of CCR1 function (Figure 4B). Proliferation of OPM2 cell lines was unaffected by

CCR1 KO, either basally (Figure 4C and D) or following addition of rhCCL3 (Figure 4E). In order to determine if CCR1 KO limited MM PC dissemination in vivo, NSG mice were injected with either OPM2-EV-1 or OPM2-CCR1-KO-1 cells. KO of CCR1 reduced primary tumor burden by 45.5%, compared with controls (Figure 5A). Circulating tumor cell numbers were

Table 1. Univariate and multivariable analysis of factors associated with overall survival in multiple myeloma patients.1

Univariate analysis P-value2 HR3 (95% CI)

n (%) CCR1 > median High-risk gene signature4 Age ≥65 years b2-microglobulin ≥5.5 mg/L Albumin <35 g/L Hemoglobin <100 g/L Proliferative index ≥10%

71 (50%) 38 (26.8%) 36 (25.4%) 33 (23.2%) 31 (21.8%) 40 (28.2%) 20 (14.1%)

0.026 0.044 0.62 0.021 0.54 0.001 0.018

2.46 (1.11-5.45) 2.16 (1.02-4.57) 0.77 (0.32-1.97) 2.45 (1.15-5.24) 1.31 (0.56-3.08) 3.52 (1.67-7.45) 2.69 (1.18-6.12)

Multivariable analysis P-value2 HR3 (95% CI) 0.039 0.50 0.60 0.009 0.61 (0.48-3.44)

2.48 (1.05-5.86) 1.36 (0.56-3.27) 1.26 (0.53-3.00) 3.18 (1.34-7.56) 1.29

1 Gene expression analysis and clinical data from n=142 newly diagnosed multiple myeloma patients in the total therapy 3 (TT3) trial (E-TABM-1138);28 2Cox proportional hazards models; 3Hazard ratio (HR); 4MS, MF or PR gene-expression profiling-defined subgroups;48 CI: confidence interval.

Figure 2. CCR1 expression in 5TGM1 murine multipe myeloma cell line increases migration towards CCL3 but does not affect proliferation. (A) Expression of murine Ccr1 mRNA was confirmed in 5TGM1-CCR1 cells. (B) CCR1-HA protein expression in 5TGM1-CCR1 cells was confirmed by immunoprecipitation using an anti-HA antibody followed by western blotting with anti-HA antibody. A representative of two independent experiments is shown. (C) Migration of 5TGM1-CCR1 and empty vector control (EV) cells towards 100 ng/mL rhCCL3 was assessed after 24 hours. (D) Relative number of 5TGM1-CCR1 and -EV cells was assessed over 72 hours. (E) Relative number of 5TGM1-CCR1 and -EV cells was assessed following 72 hours of culture with or without addition of 100 ng/mL rhCCL3. Graphs depict mean ± standard error of the mean of three biological replicates (A) or three or more independent experiments (C to E). **P<0.01, ****P<0.0001, unpaired t-test (A), twoway ANOVA with Sidak’s multiple comparison test (C).

haematologica | 2021; 106(12)

3179

M.N. Zeissig et al.

reduced by 97.8% in mice bearing CCR1 KO cells compared with controls (P<0.0001; Figure 5B). Additionally, dissemination of OPM2 cells to the contralateral leg was not observed in mice bearing OPM2-CCR1-KO-1 cells, with a 99.9% reduction in BM disseminated tumor cells compared with empty vector (EV) controls (P<0.0001; Figure 5C). Similar results were seen in the development of splenic dissemination, with mice inoculated with OPM2-EV-1 cells developing splenomegaly (Figure 5D) resulting from tumor cell infiltration, as confirmed by immunohistochemistry for GFP+ cells (Figure 5E), which was markedly reduced in mice inoculated with OPM2CCR1-KO-1 cells (P<0.0001; Figure 5D). We have previously demonstrated that CCL3 binding to CCR1 completely abrogates MM PC response to exogenous CXCL12 in vitro, without affecting CXCR4 expression,17 suggesting a mechanism whereby increased CCR1 expression may enable the dissemination of MM PC

from the BM. We therefore hypothesized that CCR1 KO cell lines may retain their response to BM CXCL12, thereby leading to retention within the BM niche. Consistent with our previous data, pre-treatment of OPM2-EV-1 cells with CCL3 prevented their migration towards CXCL12 (Online Supplementary Figure S2). In contrast, OPM2-CCR1-KO-1 cells retained their ability to migrate in response to exogenous CXCL12, even in the presence of CCL3 (Online Supplementary Figure S2). CCR1 KO had no effect on the expression of CXCR4 or CXCL12 in OPM2 cells (Online Supplementary Figure S3), consistent with our previous findings.17 In order to investigate the mechanism whereby CCR1 loss abrogates the dissemination of OPM2 cells in vivo, we assessed whether CCR1 KO had a compensatory effect on the expression of other factors that are known to play a role in MM PC adhesion and migration. CCR1 KO in OPM2 cells did not lead to a compensatory expression of the alternate CCL3 receptor CCR5, nor did it affect expres-

Figure 3. CCR1 expression in 5TGM1 multiple myeloma plasma cells increases incidence of bone and splenic dissemination in a C57BL/KaLwRij intratibial model of MM. (A) Primary tumor burden in injected tibiae after 3.5 weeks in C57BL/KaLwRij mice injected with 5TGM1-CCR1 or control 5TGM1-EV cells. Percentage of green fluorescence protein positive (GFP+) multiple myeloma (MM) cells of total mononuclear cells were quantitated using flow cytometry. (B) Number of circulating 5TGM1CCR1 or -EV cells in peripheral blood of mice. (C) Tumor burden disseminated to the non-injected contralateral leg in mice injected with 5TGM1-CCR1 or -EV cells. (D) Proportion of mice with detectable GFP+ MM cells in the contralateral long bones. (E) Spleens were collected from eight mice (5TGM1-EV) and 11 mice (5TGM1CCR1) and imaged using bioluminescence imaging, with representative spleens from each group shown. (F) Proportion of mice with detectable bioluminescence signal in the spleen. Box and whisker plots depict median and interquartile range for 11 mice/group (A to C). ****P<0.0001, Fisher’s exact test. EV: empty vector.

3180

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

sion of integrin α4 (ITGA4) and integrin b1 (ITGB1), critical adhesion molecules that play a role in MM PC BM retention (Online Supplementary Figure S3).

The CCR1 inhibitor CCX9588 inhibits migration towards CCL3 in vitro Next, the effects of a selective small molecule CCR1 inhibitor, CCX9588, on MM cells was assessed in vitro. In order to investigate whether the small molecule CCR1 inhibitor CCX9588 effects cell survival and/or proliferation, CCR1-expressing OPM2-EV-1 or RPMI-8226-luc17 cells were cultured with increasing concentrations of CCX9588 or vehicle alone. OPM2-EV-1 cell number (P=0.88, Figure 6A) and viability (P=0.70, Figure 6B) were not affected by treatment with up to 1 mM CCX9588. However, there was a 35% decrease in cell number in RPMI-8226-luc cells treated with 1 mM CCX9588 (P<0.01, Figure 6C), while cell survival was unaffected (P=0.50, Figure 6D), suggesting that high concentrations may decrease proliferation of these cells. Based on these results, concentrations up to 100 nM and 1 mM were used for further characterization in RPMI8226 and OPM2 cells, respectively. In order to confirm the anti-CCR1 function of CCX9588, OPM2-EV-1 or RPMI-8226-luc cells were treated with CCX9588 or vehicle alone and were then stimulated with rhCCL3. Western blot analysis revealed that, in OPM2-EV1 cells, CCL3 treatment induced AKT and ERK1/2 phosphorylation, which was inhibited by 10 nM CCX9588 or higher (Figure 6E). In RPMI-8226-luc cells, AKT phosphorylation was increased by CCL3 and this was inhibited by 10 nM CCX9588 or higher (Figure 6F). Furthermore, pre-treat-

ment of OPM2 or RPMI-8226 cells with CCX9588 resulted in a complete inhibition of migration towards rhCCL3 in a transwell assay (OPM2: P<0.001, Figure 6G; RPMI-8226: P<0.01, Figure 6H).

CCX9588 treatment reduces dissemination of multiple myeloma plasma cells in vivo In order to investigate the effectiveness of CCR1 inhibition in suppressing MM PC dissemination in vivo, the effects of the CCR1 inhibitor CCX9588 were assessed in mice bearing OPM2-EV-1 or RPMI-8226-luc tumors. CCX9588 treatment did not have appreciable adverse effects on the mice, as assessed by body weight (Online Supplementary Figure S4A) or analysis of PB cell counts (Online Supplementary Table S1). Mean trough serum concentration of CCX9588 achieved in vivo was 328 nM (range, 76.8-886 nM; Online Supplementary Figure S4B). In mice bearing OPM2-EV-1 or RPMI-8226-luc cells, primary tumor burden was unaffected by CCX9588 treatment (OPM2-EV-1: P=0.91, Figure 7A; RPMI-8226-luc: P=0.49, Figure 7B). Consistent with the effect of CCR1 KO in OPM2 cells, we observed a 66% decrease in the mean number of circulating tumor cells in the OPM2-EV-1 model (P<0.0001; Figure 7C); while the decrease in circulating tumor cells in the RPMI-8226-luc model did not reach statistical significance (P=0.09; Figure 7D). CCX9588 treatment significantly reduced dissemination to the bone, with a 22% and 70% reduction in mean tumor burden in the BM of the contralateral limb in the OPM2-EV-1 (Figure 7E) and the RPMI-8226-luc models, respectively, compared with controls (P<0.0001; Figure 7F). Furthermore, the degree of

Figure 4. Knockout of CCR1 in human OPM2 multiple myeloma plasma cells decreases migration towards CCL3 and does not affect proliferation. (A) CRISPR-Cas9mediated knockout (KO) of CCR1 was confirmed in OPM2-CCR1-KO-1 and OPM2-CCR1-KO-2 cells following staining with an anti-hCCR1 antibody or isotype control. (B) Migration of OPM2-CCR1-KO-1 and OPM2-CCR1-KO-2 cells and empty vector (EV) control cells towards 100 ng/mL rhCCL3, or media alone, was assessed after 18 hours. Migration is expressed relative to no chemoattractant controls. (C) Relative numbers of OPM2-CCR1-KO-1 or OPM2-EV-1 control cells were assessed over 72 hours. (D) Relative numbers of OPM2-CCR1-KO-2 or OPM2-EV-2 control cells were assessed over 72 hours. (E) The effect of 72 hours of treatment with 100 ng/mL rhCCL3 on relative numbers of OPM2-CCR1-KO-1 and OPM2-CCR1-KO-2, and EV-1 and EV-2 control cells. Graphs depict mean ± standard error of the mean of three or more independent experiments (A to E). **P<0.001, two-way ANOVA with Sidak’s multiple comparison test.

haematologica | 2021; 106(12)

3181

M.N. Zeissig et al.

splenomegaly in the OPM2-EV-1 model was significantly reduced compared with vehicle controls in CCX9588-treated mice (P<0.001; Figure 7G). Splenomegaly was not observed in the RPMI-8226-luc model, precluding assessment of the effect of CCX9588 on splenic dissemination (Figure 7G). When treatment was delayed until 2 weeks post OPM2-EV-1 tumor cell inoculation, CCX9588-treated mice showed significantly reduced numbers of circulating tumor cells (P<0.01; Online Supplementary Figure S5B) although delayed treatment did not significantly decrease tumor burden in the contralateral leg (P=0.08; Online Supplementary Figure S5C).

Discussion MM is characterized by the presence of multiple tumors throughout the skeleton, and in some patients, soft tissues. The dissemination of MM PC is central to the progression of disease and subsequent disease relapse, highlighting the therapeutic potential of targeting key factors that regulate dissemination to delay disease progression and prevent overt relapse. While the inhibition of several factors, including selectins,30 N-cadherin31,32 and CXCR433 have been demonstrated to slow BM homing of MM cells in vivo, very few genes have been demonstrated to play a role in the spontaneous dissemination of MM PC from the BM. For example, overexpression of heparanase, an enzyme that cleaves heparan sulphate chains, has been reported to increase the incidence of spontaneous dissemination of

MM cells in a mouse MM xenograft model.34 Additionally, recent data suggests that the transcription factor Twist-1 increases dissemination in an intratibial 5TGM1/KaLwRij model in vivo.35 Furthermore, as far as we are aware, no therapeutic interventions have been described that can inhibit spontaneous dissemination of MM PC in vivo. Here, our findings suggest a novel role for the chemokine receptor CCR1 in regulating the egress of MM PC from the BM to the circulation during dissemination. These findings are consistent with a role for CCR1 in metastasis in other cancer settings, with a study showing that short hairpin RNAknockdown of CCR1 decreased migration of hepatocellular carcinoma cells in vitro and reduced the incidence of lung metastasis in vivo.36 We have previously demonstrated that hypoxia, through induction of HIF-2α, increases the expression of CCR1 in human MM cell lines.17 This led us to hypothesize that tumor growth in the BM exacerbates BM hypoxia, leading to increased CCR1 expression and tumor dissemination.17 Consistent with this hypothesis, our flow cytometric analysis suggests that CCR1 expression on BM PC is increased in MM patients compared with MGUS patients. In addition, our analysis suggested that elevated CCR1 expression is an independent predictor of poor overall survival in MM patients. Mechanistically, we have previously demonstrated that CCL3 treatment of human MM cell lines reduces their capacity to migrate towards exogenous CXCL12 or undergo cytoskeletal remodeling in response to CXCL12 treatment.17 Furthermore, we found that the human MM cell line U266, which does not respond to exogenous

Figure 5. Dissemination of human multiple myeloma cell line OPM2 in NSG mice is abrogated by knockout of CCR1. (A) Primary tumor burden (percentage of green fluorescence protein positive [GFP+] multiple myeloma [MM] cells of total mononuclear cells) in injected tibiae after 4 weeks in NSG mice injected with OPM2-EV-1 or OPM2-KO-1 cells. (B) Number of circulating GFP+ OPM2-EV-1 or OPM2-KO-1 cells in peripheral blood of mice. (C) Tumor burden disseminated to the non-injected contralateral leg in mice injected with OPM2-EV-1 or OPM2-KO-1 cells. (D) Length of spleens collected from naïve NSG mice (n=7 mice) or mice bearing OPM2-EV-1 (n=3 mice) or OPM2-KO-1 (n=3 mice) cells were measured. Image of three representative spleens from OPM2-EV-1- and OPM2-CCR1-KO-1-bearing mice. Scale bar: 10 mm (E) Splenic tumor cell infiltration in mice bearing OPM2-EV-1 or OPM2-CCR1-KO-1 cells was confirmed by immunohistochemistry with an anti-GFP antibody (brown). Representative flow plots of percentage of GFP+ MM cells of total mononuclear cells from one mouse per group are shown (B to C). A representative of 5 mice/group is shown; scale bar: 10 mm (E). Box and whisker plots depict median and interquartile range, n=9-10 mice/group. **P<0.01, ****P<0.0001, MannWhitney U test (A to C), ****P<0.0001, one-way ANOVA with Tukey’s multiple comparisons test.

3182

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

Figure 6. CCX9588 treatment prevents activation of CCR1 signaling in multiple myeloma plasma cells and their migration towards CCL3 in vitro. (A) Relative number of OPM2-EV-1 cells was assessed after 72 hours of treatment with 10 nM to 1 mM CCX9588 (all containing 0.01% dimethyl sulfoxide [DMSO)], or 0.01% DMSO vehicle alone. (B) Viability of OPM2-EV-1 cells was assessed after 72 hours of treatment with 10 nM to 1 mM CCX9588 (all containing 0.01% DMSO), or 0.01% DMSO vehicle alone. (C) Relative number of RPMI-8226-luc cells was assessed after 72 hours of treatment with 10 nM-1 mM CCX9588 (all containing 0.01% DMSO), or 0.01% DMSO vehicle alone. (D) Viability of RPMI-8226-luc cells was assessed after 72 hours of treatment with 10 nM to 1 mM CCX9588, or 0.01% DMSO vehicle by WST-1 assay. (E) OPM2-EV-1 cells were treated with CCX9588 (10 nM to 1 μM) or media alone for 24 hours, and cells were stimulated with 100 ng/mL rhCCL3 for 5 minutes. Cells were lysed and western blotting was performed with antibodies against p-AKT, p-ERK1/2, total AKT and total ERK. Hsc70 was used as a loading control. A representative of three experiments is shown. (F) RPMI-8226-luc cells were treated with CCX9588 (10 nM to 1 μM) or media alone for 24 hours, and cells were stimulated with 100 ng/mL rhCCL3 for 5 minutes. Cells were lysed and western blotting was performed with antibodies against p-AKT, p-ERK1/2, total AKT and total ERK. Hsc70 was used as a loading control. A representative of two experiments is shown. (G) OPM2-EV-1 cells were treated with 1 mM CCX9588 or 0.01% DMSO vehicle control for 24 hours and migrated towards 100 ng/mL rhCCL3 or media alone. (H) RPMI-8226-luc cells were treated with 100 nM CCX9588 or 0.01% DMSO vehicle control for 24 hours and migrated towards 100 ng/mL rhCCL3 or media alone. Graphs depict mean ± standard error of the mean of three or more independent experiments (A to D, G to H). **P<0.01, *P<0.05, one-way ANOVA with Tukey’s multiple comparisons test (C) two-way ANOVA with Sidak’s multiple comparisons test (G-H).

haematologica | 2021; 106(12)

3183

M.N. Zeissig et al.

CXCL12, produces abundant CCL3, suggesting that endogenous CCL3 expression can suppress response to CXCL12. Notably, migration of U266 cells towards CXCL12 could be restored by either CCR1 KO or treatment with a CCR1 inhibitor.17 Here, we observed that treatment of OPM2-EV-1 control cells with CCL3 prevented migration towards CXCL12, in accordance with our previous study,17 whereas, the chemotactic response of OPM2CCR1-KO-1 cells to exogenous CXCL12 was maintained following CCL3 treatment. These data strongly suggest

3184

that CCL3/CCR1 signaling is responsible for blocking migration towards CXCL12 in these cell lines. Notably, CCL3/CCR1 signaling has been shown to drive mobilization of hematopoietic progenitors and natural killer cells from the BM in part through inactivation of CXCR4.25,26,37 Based on these studies, we postulated that CCL3/CCR1 signaling may inhibit the CXCL12 mediated retention of MM PC in the BM, enabling the egress of MM PC into the circulation and subsequent dissemination. Indeed, we demonstrate here that CCR1 expression increases the

Figure 7. CCR1 inhibition reduces circulating multiple myeloma plasma cell numbers and tumor dissemination in NSG mice bearing OPM2 or RPMI-8226 cells. (A) Primary tumor burden in injected tibiae after 4 weeks in NSG mice injected with OPM2-EV-1 cells and treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12hour intervals. (B) Primary tumor burden in injected tibiae after 4 weeks in NSG mice injected with RPMI-8226-luc cells and treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12-hour intervals. (C) Number of circulating OPM2-EV-1 cells in peripheral blood of mice treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12-hour intervals. (D) Number of circulating RPMI-8226-luc cells in peripheral blood of mice treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12-hour intervals. (E). Tumor burden disseminated to the non-injected contralateral leg in mice injected with OPM2-EV-1 cells treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12-hour intervals. (F) Tumor burden disseminated to the non-injected contralateral leg in mice injected with RPMI-8226-luc cells treated days 3-28 with CCX9588 (15 mg/kg) or vehicle control at 12-hour intervals. (G) Spleens collected from naïve NSG mice or vehicle- or CCX9588-treated mice bearing OPM2-EV-1 or RPMI-8226luc cells were measured to assess the degree of splenomegaly. Naïve mice splenic sizes are duplicated from Figure 5D for comparison. Box and whisker plots depict median and interquartile range, n=10-12 mice/group (A, C and E), n=17 mice/group (B, D and F), n=7-17 mice/group (G). **P<0.01, ***P<0.001, ****P<0.0001, Mann-Whitney test (C, E and F), ordinary one-way ANOVA with Tukey’s multiple

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

capacity for MM PC dissemination, while CCR1 inhibition or KO decreases mobilization of MM PC to the PB, and the subsequent formation of disseminated tumors in vivo. Our findings support our hypothesis that hypoxia-mediated upregulation of CCR1 may be critical for overcoming CXCL12-mediated BM retention and enabling mobilization. In addition to its role in counteracting CXCL12/CXCR4 signaling, CCL3 is known to act as a potent chemoattractant for murine and human MM cell lines and patientderived MM PC in vitro.17,20,38 In accordance with this, we demonstrated that CCL3 acts as a chemoattractant for OPM2 and RPMI-8226 cells, which could be blocked with CCR1 KO or inhibition. Furthermore, expression of CCR1 in 5TGM1 cells resulted in a chemotactic response to CCL3. However, while CCL3 has been shown to be produced by mesenchymal stem cells20 and osteoclasts19,20 in the BM, the most abundant source of CCL3 in the BM in MM patients is suggested to be the MM PC themselves.18-21,24 It is therefore likely that autocrine CCL3 production would interfere with the chemoattractant effect of exogenous CCL3. In further support of this, CCL3 is present at higher levels in the BM than in the PB of MM patients,21,39,40 suggesting that migration towards CCL3 in the PB would not play a significant role in the mobilization of MM PC from the BM. Instead, it is possible that autocrine CCL3 production may increase non-directional migration (chemokinesis), as has been described for chemokines CCL2 and IGF-1 in MM cell lines.41,42 In accordance with this, we found that inhibition of CCR1 in RPMI-8226-luc cells using CCX9588 resulted in a reduction in basal migration. This is consistent with a previous study, whereby the CCR1 inhibitor BX471 prevented basal migration of the human acute monocytic leukemia cell line THP-1.43 Alternatively, CCR1 has been suggested to signal without the presence of ligand and induce agonistindependent migration in some cell types.43 Decreased basal migration or chemokinesis of these cells in the presence of CCR1 inhibitor may, therefore, in part be contribute to the decrease in dissemination of RPMI-8226 and OPM2 cells observed in vivo. We observed no effect of CCR1 expression or KO on the proliferation of MM cell lines in vitro in either the presence or absence of exogenous CCL3. This was despite the ability of CCL3 to induce AKT and ERK phosphorylation, which are involved in survival/proliferation pathways in MM.24 This contrasts with a previous study suggesting that recombinant CCL3 increases human MM cell line proliferation in vitro.24 As such, the possibility that the relatively high serum concentration used here could be providing sufficient other growth factors to mask the effects of CCL3 cannot be excluded. Mice injected with OPM2-CCR1-KO cells had lower primary tumor burden compared with controls, suggesting that CCR1 KO may affect growth of OPM2 cells in vivo. However, we did not observe an effect of CCR1 overexpression or inhibition on primary MM tumor growth in our other in vivo models, suggesting this effect may be independent of CCR1. Additional studies are required to determine whether the retention of MM PC in the bone marrow with CCR1 KO may be causing environmental pressures, such as an increase in hypoxia,44 that is slowing their proliferation in vivo. In contrast, while CCX9588 treatment had no effect on the proliferation of OPM2 cells in vitro, we observed a decrease in the proliferation of RPMI-8226 cells with 1 mM CCX9588 treatment. This contrasts with a previous study which reported that treatment with the CCR1 haematologica | 2021; 106(12)

inhibitor CCX721 at high doses had no effect on the proliferation of RPMI-8226 cells in vitro,23 suggesting that the effects observed at high concentrations of CCX9588 here may be due to off-target effects. In support of this, CCX9588 treatment did not affect OPM2 or RPMI-8226 tumor growth in vivo. Inhibition of CCL3 or CCR1 in the murine 5T2MM and 5TGM1 models has previously been shown to decrease primary BM tumor growth, but not growth of subcutaneous tumors or cells in vitro.22,23 This suggests that CCL3/CCR1 inhibitors may affect growth factor production by cells of the BM microenvironment to indirectly affect 5TMM tumor growth.23 Similar effects were observed with osteoclast ablation using zoledronate, suggesting that these results may be secondary to decreased osteoclast activity/numbers in this model.23 The CCR1 inhibitor MLN3897 has previously been shown to decrease the pro-proliferative effects of osteoclast coculture on a CCR1-negative human MM cell line, at least in part through indirectly decreasing osteoclast IL-6 secretion, supporting the idea that effects of CCR1 inhibition on tumor growth in some in vivo models may be due to secondary effects on osteoclasts.19 However, we found no effect of CCR1 inhibition on primary tumor growth in vivo, suggesting that inhibition of microenvironmental CCR1 was not contributing to the effects observed here. Notably, we have previously demonstrated that treatment with the CXCR4 inhibitor T140 had no effect on intratibial RPMI-8226 tumor growth, despite dramatic effects on osteolysis and decreased osteoclast numbers, suggesting that inhibition of osteoclasts does not affect primary tumor growth in this model.45 Importantly, we are the first to assess the efficacy of the small molecule CCR1 inhibitor CCX9588 on dissemination in a pre-clinical model of MM. CCX9588 has been previously reported to decrease chemotaxis of T cells towards liver conditioned media in vitro.46 CCX9588 is an analogue of CCX354, which has previously been investigated as a therapeutic for rheumatoid arthritis in a clinical trial,47 and CCX721, which has been shown to have anti-osteolytic activity in an in vivo MM model.23 While we were not able to completely prevent dissemination of MM PC using CCX9588 in OPM2 and RPMI-8226 xenograft models at this dose, these studies suggest that impeding the egress of MM PC from the BM to the PB could slow the development of disease. Further studies are required to determine whether combination therapy with other anti-myeloma agents, or more intensive treatment regimens, could achieve an enhanced effect on tumor dissemination. Notably, while both of the human MM cell lines used here do not express the alternate CCL3 receptor CCR5, CCR5 is expressed at the mRNA level in up to one third of MM patients (data not shown). Therefore, additional studies are warranted to determine whether CCR1 inhibition alone is sufficient to block dissemination when both CCR1 and CCR5 are expressed. In summary, our studies have identified a novel role for the chemokine receptor CCR1 in the context of MM pathogenesis, demonstrating that CCR1 is a key driver of MM PC egress from the BM to the circulation during dissemination. Furthermore, we have shown that inhibition of CCR1 via therapeutic targeting or KO can slow MM PC dissemination. Together with previous studies demonstrating that targeting of CCR1 prevents the development of severe osteolytic lesions in vivo, and our data demonstrating that CCR1 is an independent prognostic factor in MM patients, our 3185

M.N. Zeissig et al.

study suggests that CCR1 is a potential attractive therapeutic target for MM. Future preclinical studies are warranted to investigate whether therapeutic inhibition of CCR1 has efficacy as a maintenance therapy, extending post-therapy remission and preventing overt relapse. Disclosure No conflicst of interest to disclose. Contributions MNZ performed experiments and wrote the manuscript; KV, DRH, KMM and VP assisted with experiments; KV, ACWZ, DRH and VP reviewed the manuscript; KV, ACWZ, PIC and LBT designed the study; KV, ACWZ and DRH supervised the study. All authors read and approved the final manuscript. Acknowledgments The authors thank ChemoCentryx for generously providing the CCR1 antagonist CCX9588 for these studies. We are grateful to Vicki Wilczek and Elyse Bell for their assistance with the animal studies.

References 1. Rajkumar SV, Dimopoulos MA, Palumbo A, et al. International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. Lancet Oncol. 2014;15 (12):e538-548. 2. Ghobrial IM. Myeloma as a model for the process of metastasis: implications for therapy. Blood. 2012;120(1):20-30. 3. Gonsalves WI, Rajkumar SV, Gupta V, et al. Quantification of clonal circulating plasma cells in newly diagnosed multiple myeloma: implications for redefining high-risk myeloma. Leukemia. 2014;28(10):20602065. 4. Chakraborty R, Muchtar E, Kumar SK, et al. Risk stratification in myeloma by detection of circulating plasma cells prior to autologous stem cell transplantation in the novel agent era. Blood Cancer J. 2016;6(12):e512. 5. Chakraborty R, Muchtar E, Kumar SK, et al. Serial measurements of circulating plasma cells before and after induction therapy have an independent prognostic impact in patients with multiple myeloma undergoing upfront autologous transplantation. Haematologica. 2017;102(8):1439-1445. 6. Peceliunas V, Janiulioniene A, Matuzeviciene R, Zvirblis T, Griskevicius L. Circulating plasma cells predict the outcome of relapsed or refractory multiple myeloma. Leuk Lymphoma. 2012;53(4): 641-647. 7. Dingli D, Nowakowski GS, Dispenzieri A, et al. Flow cytometric detection of circulating myeloma cells before transplantation in patients with multiple myeloma: a simple risk stratification system. Blood. 2006;107(8):3384-3388. 8. Nowakowski GS, Witzig TE, Dingli D, et al. Circulating plasma cells detected by flow cytometry as a predictor of survival in 302 patients with newly diagnosed multiple myeloma. Blood. 2005;106(7):22762279. 9. Witzig TE, Gertz MA, Lust JA, Kyle RA, O'Fallon WM, Greipp PR. Peripheral blood monoclonal plasma cells as a predictor of survival in patients with multiple myeloma. Blood. 1996;88(5):1780-1787.

3186

Funding This research was supported by grant 2002138 awarded through the 2020 Priority-driven Collaborative Cancer Research Scheme (PdCCRS), co-funded by Cancer Australia and Cure Cancer, and grant 1163245 awarded through the 2018 PdCCRS, co-funded by Cancer Australia, Cure Cancer, and Leukaemia Foundation of Australia, awarded to KV. The work was partially supported by a Hans-Jürgen and Marianne Ohff Research Grant from the University of Adelaide, awarded to MZ. MZ was supported by the Florey Medical Research Foundation Doctor Chun Chung Wong and Madam So Sau Lam Memorial Postgraduate Cancer Research Top-Up Scholarship and a Short-Term Research Grant from the German Academic Exchange Service (DAAD). VP was supported by a National Health & Medical Research Council Early Career Fellowship. KV and KMM were supported by Early Career Cancer Research Fellowships from the Cancer Council SA Beat Cancer Project on behalf of its donors and the State Government of South Australia through the Department of Health.

10. Bianchi G, Kyle RA, Larson DR, et al. High levels of peripheral blood circulating plasma cells as a specific risk factor for progression of smoldering multiple myeloma. Leukemia. 2013;27(3):680-685. 11. Kumar S, Rajkumar SV, Kyle RA, et al. Prognostic value of circulating plasma cells in monoclonal gammopathy of undetermined significance. J Clin Oncol. 2005;23(24):5668-5674. 12. Gonsalves WI, Rajkumar SV, Dispenzieri A, et al. Quantification of circulating clonal plasma cells via multiparametric flow cytometry identifies patients with smoldering multiple myeloma at high risk of progression. Leukemia. 2017;31(1):130-135. 13. Sanz-Rodríguez F, Ruiz-Velasco N, PascualSalcedo D, Teixidó J. Characterization of VLA-4-dependent myeloma cell adhesion to fibronectin and VCAM-1. Br J Haematol. 1999;107(4):825-834. 14. Nie Y, Waite J, Brewer F, Sunshine MJ, Littman DR, Zou YR. The role of CXCR4 in maintaining peripheral B cell compartments and humoral immunity. J Exp Med. 2004;200(9):1145-1156. 15. Azab AK, Runnels JM, Pitsillides C, et al. CXCR4 inhibitor AMD3100 disrupts the interaction of multiple myeloma cells with the bone marrow microenvironment and enhances their sensitivity to therapy. Blood. 2009;113(18):4341-4351. 16. Azab AK, Hu J, Quang P, et al. Hypoxia promotes dissemination of multiple myeloma through acquisition of epithelial to mesenchymal transition-like features. Blood. 2012;119(24):5782-5794. 17. Vandyke K, Zeissig MN, Hewett DR, et al. HIF-2a promotes dissemination of plasma cells in multiple myeloma by regulating CXCL12/CXCR4 and CCR1. Cancer Res. 2017;77(20):5452-5463. 18. Uneda S, Hata H, Matsuno F, et al. Macrophage inflammatory protein-1 alpha is produced by human multiple myeloma (MM) cells and its expression correlates with bone lesions in patients with MM. Br J Haematol. 2003;120(1):53-55. 19. Vallet S, Raje N, Ishitsuka K, et al. MLN3897, a novel CCR1 inhibitor, impairs osteoclastogenesis and inhibits the interaction of multiple myeloma cells and osteo-

clasts. Blood. 2007;110(10):3744-3752. 20. Moreaux J, Hose D, Kassambara A, et al. Osteoclast-gene expression profiling reveals osteoclast-derived CCR2 chemokines promoting myeloma cell migration. Blood. 2011;117(4):1280-1290. 21. Roussou M, Tasidou A, Dimopoulos MA, et al. Increased expression of macrophage inflammatory protein-1 on trephine biopsies correlates with extensive bone disease, increased angiogenesis and advanced stage in newly diagnosed patients with multiple myeloma. Leukemia. 2009;23(11):21772181. 22. Menu E, De Leenheer E, De Raeve H, et al. Role of CCR1 and CCR5 in homing and growth of multiple myeloma and in the development of osteolytic lesions: a study in the 5TMM model. Clin Exp Metastasis. 2006;23(5-6):291-300. 23. Dairaghi DJ, Oyajobi BO, Gupta A, et al. CCR1 blockade reduces tumor burden and osteolysis in vivo in a mouse model of myeloma bone disease. Blood. 2012;120(7): 1449-1457. 24. Lentzsch S, Gries M, Janz M, Bargou R, Dorken B, Mapara MY. Macrophage inflammatory protein 1-alpha (MIP-1 alpha) triggers migration and signaling cascades mediating survival and proliferation in multiple myeloma (MM) cells. Blood. 2003;101(9):3568-3573. 25. Broxmeyer HE, Hangoc G, Cooper S, Campbell T, Ito S, Mantel C. AMD3100 and CD26 modulate mobilization, engraftment, and survival of hematopoietic stem and progenitor cells mediated by the SDF1/CXCL12-CXCR4 axis. Ann N Y Acad Sci. 2007;1106:1-19. 26. Bernardini G, Sciume G, Bosisio D, Morrone S, Sozzani S, Santoni A. CCL3 and CXCL12 regulate trafficking of mouse bone marrow NK cell subsets. Blood. 2008;111 (7):3626-3634. 27. Zannettino ACW, Farrugia AN, Kortesidis A, et al. Elevated serum levels of stromalderived factor-1 are associated with increased osteoclast activity and osteolytic bone disease in multiple myeloma patients. Cancer Res. 2005;65(5):1700-1709. 28. Shaughnessy JD, Jr., Qu P, Usmani S, et al. Pharmacogenomics of bortezomib test-

haematologica | 2021; 106(12)

CCR1 drives dissemination of multiple myeloma plasma cells

dosing identifies hyperexpression of proteasome genes, especially PSMD4, as novel high-risk feature in myeloma treated with Total Therapy 3. Blood. 2011;118(13):35123524. 29. Hewett DR, Vandyke K, Lawrence DM, et al. DNA barcoding reveals habitual clonal dominance of myeloma plasma cells in the bone marrow microenvironment. Neoplasia. 2017;19(12):972-981. 30. Asosingh K, Günthert U, De Raeve H, Van Riet I, Van Camp B, Vanderkerken K. A unique pathway in the homing of murine multiple myeloma cells: CD44v10 mediates binding to bone marrow endothelium. Cancer Res. 2001;61(7):2862. 31. Mrozik KM, Cheong CM, Hewett D, et al. Therapeutic targeting of N-cadherin is an effective treatment for multiple myeloma. Br J Haematol. 2015;171(3):387-399. 32. Groen RWJ, de Rooij MFM, Kocemba KA, et al. N-cadherin-mediated interaction with multiple myeloma cells inhibits osteoblast differentiation. Haematologica. 2011;96 (11):1653. 33. Roccaro AM, Mishima Y, Sacco A, et al. CXCR4 regulates extra-medullary myeloma through epithelial-mesenchymal-transition-like transcriptional activation. Cell Rep. 2015;12(4):622-635. 34. Yang Y, Macleod V, Bendre M, et al. Heparanase promotes the spontaneous metastasis of myeloma cells to bone. Blood. 2005;105(3):1303-1309. 35. Cheong CM, Mrozik KM, Hewett DR, et al. Twist-1 is upregulated by NSD2 and

haematologica | 2021; 106(12)

contributes to tumour dissemination and an epithelial-mesenchymal transition-like gene expression signature in t(4;14)-positive multiple myeloma. Cancer Lett. 2020; 475:99-108. 36. Zhu Y, Gao X-M, Yang J, et al. C-C chemokine receptor type 1 mediates osteopontin-promoted metastasis in hepatocellular carcinoma. Cancer Sci. 2018;109(3): 710-723. 37. Lord BI, Woolford LB, Wood LM, et al. Mobilization of early hematopoietic progenitor cells with BB-10010: a genetically engineered variant of human macrophage inflammatory protein-1 . Blood. 1995;85 (12):3412-3415. 38. Moller C, Stromberg T, Juremalm M, Nilsson K, Nilsson G. Expression and function of chemokine receptors in human multiple myeloma. Leukemia. 2003;17(1):203210. 39. Choi SJ, Cruz JC, Craig F, et al. Macrophage inflammatory protein 1-alpha is a potential osteoclast stimulatory factor in multiple myeloma. Blood. 2000;96(2):671-675. 40. Wang X-T, He Y-C, Zhou S-Y, et al. Bone marrow plasma macrophage inflammatory protein protein-1 alpha(MIP-1 alpha) and sclerostin in multiple myeloma: relationship with bone disease and clinical characteristics. Leuk Res. 2014;38(5):525-531. 41. Vanderkerken K, Asosingh K, Braet F, Van Riet I, Van Camp B. Insulin-like growth factor-1 acts as a chemoattractant factor for 5T2 multiple myeloma cells. Blood. 1999;93(1):235-241.

42. Johrer K, Janke K, Krugmann J, Fiegl M, Greil R. Transendothelial migration of myeloma cells is increased by tumor necrosis factor (TNF)-alpha via TNF receptor 2 and autocrine up-regulation of MCP-1. Clin Cancer Res. 2004;10(6):1901-1910. 43. Gilliland CT, Salanga CL, Kawamura T, Trejo J, Handel TM. The chemokine receptor CCR1 is constitutively active, which leads to G protein-independent, -arrestinmediated internalization. J Biol Chem. 2013;288(45):32194-32210. 44. Muz B, de la Puente P, Azab F, Luderer M, Azab AK. Hypoxia promotes stem cell-like phenotype in multiple myeloma cells. Blood Cancer J. 2014;4(12):e262. 45. Diamond P, Labrinidis A, Martin SK, et al. Targeted disruption of the CXCL12/CXCR4 axis inhibits osteolysis in a murine model of myeloma associated bone loss. J Bone Miner Res. 2009;24(7):1150-1161. 46. Conroy MJ, Galvin KC, Kavanagh ME, et al. CCR1 antagonism attenuates T cell trafficking to omentum and liver in obesityassociated cancer. Immunol Cell Biol. 2016;94(6):531-537. 47. Tak PP, Balanescu A, Tseluyko V, et al. Chemokine receptor CCR1 antagonist CCX354-C treatment for rheumatoid arthritis: CARAT-2, a randomised, placebo controlled clinical trial. Ann Rheum Dis. 2013;72(3):337-344. 48. Zhan F, Huang Y, Colla S, et al. The molecular classification of multiple myeloma. Blood. 2006;108(6):2020-2028.

3187

ARTICLE Ferrata Storti Foundation

Red Cell Biology & its Disorders

Danicopan: an oral complement factor D inhibitor for paroxysmal nocturnal hemoglobinuria Antonio M. Risitano,1,2 Austin G. Kulasekararaj,3,4 Jong Wook Lee,5 Jaroslaw P. Maciejewski,6 Rosario Notaro,7,8 Robert Brodsky,9 Mingjun Huang,10 Michael Geffner11 and Peter Browett12 1

Federico II University of Naples, Naples, Italy; 2AORN Moscati, Avellino, Italy; King's College Hospital-NHS Foundation Trust, NIHR/Wellcome King’s Clinical Research Facility, London, UK; 4King’s College London, London, UK; 5Seoul St. Mary’s Hospital, The Catholic University of Korea, Seoul, Republic of Korea; 6Cleveland Clinic, Cleveland, OH, USA; 7Core Research Laboratory, Istituto per lo Studio, la Prevenzione e la Rete Oncologica (ISPRO), Firenze, Italy; 8Azienda Ospedaliera-Universitaria Careggi, Firenze, Italy; 9Johns Hopkins University School of Medicine, Baltimore, MD, USA; 10Achillion, Inc., A Subsidiary of Alexion Pharmaceuticals, Inc., New Haven, CT, USA and Alexion Pharmaceuticals, New Haven CT, USA; 11Achillion Inc., A Subsidiary of Pharmaceuticals, Inc., Blue Bell, PA, USA and 12University of Auckland, Auckland, New Zealand 3

Haematologica 2021 Volume 106(12):3188-3197

Presented as an oral presentation at the 24th European Hematology Association Congress, Amsterdam, NL, June 15, 2019. Presented as a poster presentation at the 61st annual meeting of the American Society of Hematology, Orlando, FL, USA December 8, 2019.

ABSTRACT

Correspondence: ANTONIO M. RISITANO amrisita@unina.it Received: June 3, 2020. Accepted: October 6, 2020. Pre-published: October 29, 2020. https://doi.org/10.3324/haematol.2020.261826

3188

aroxysmal nocturnal hemoglobinuria (PNH) is characterized by complement-mediated intravascular hemolysis due to the absence of complement regulators CD55 and CD59 on affected erythrocytes. Danicopan is a first-in-class oral proximal, complement alternative pathway factor D inhibitor. Therapeutic factor D inhibition was designed to control intravascular hemolysis and prevent C3-mediated extravascular hemolysis. In this open-label, phase II, dose-finding trial, ten untreated PNH patients with hemolysis received danicopan monotherapy (100-200 mg thrice daily). Endpoints included changes in the concentrations of lactate dehydrogenase (LDH) at day 28 (primary endpoint), of LDH at day 84, and of hemoglobin. Safety, pharmacokinetics/pharmacodynamics, and patient-reported outcomes were assessed. Ten patients reached the primary endpoint; two later discontinued treatment: one because of a serious adverse event (elevated aspartate aminotransferase/alanine aminotransferase coincident with breakthrough hemolysis, resolving without sequelae) and one for personal reasons unrelated to safety. Eight patients completed treatment. Intravascular hemolysis was inhibited, as demonstrated by a mean decrease of LDH (5.7 times upper limit of normal [ULN] at baseline vs. 1.8 times ULN at day 28 and 2.2 times ULN at day 84; both P<0.001). Mean baseline hemoglobin, 9.8 g/dL, increased by 1.1 (day 28) and 1.7 (day 84) g/dL (both P<0.005). No significant C3 fragment deposition occurred on glycosylphosphatidylinositol-deficient erythrocytes. Mean baseline Functional Assessment of Chronic Illness Therapy–Fatigue score, 34, increased by 9 (day 28) and 13 (day 84) points. The most common adverse events were headache and upper respiratory tract infection. These phase II, monotherapy data show that proximal inhibition with danicopan provides clinically meaningful inhibition of intravascular hemolysis and increases hemoglobin concentration in untreated PNH patients, without evidence of C3-mediated extravascular hemolysis. This trial was registered at www.clinicaltrials.gov (#NCT03053102).

haematologica | 2021; 106(12)

Danicopan, an oral factor D inhibitor for PNH

Introduction Paroxysmal nocturnal hemoglobinuria (PNH) is a rare hematologic disease characterized by chronic intravascular hemolysis, severe thrombophilia, and bone marrow failure.1 PNH is due to somatic mutations of the phosphatidylinositol N-acetylglucosaminyltransferase subunit A (PIGA) gene in hematopoietic stem cells that impair biosynthesis of glycosylphosphatidylinositol (GPI) anchors and surface expression of GPI-linked proteins.2-4 While bone marrow failure is secondary to T-cell-mediated immune attack that spares PIGA-mutated hematopoietic stem cells,5,6 both hemolysis and thromboembolism are complement-mediated. GPI-deficient erythrocytes (and platelets) lack GPI-linked complement regulators CD557 and CD598 and are exquisitely vulnerable to complement activation, which occurs continuously and spontaneously due to “C3 tick-over”9 and acutely with specific triggers. PNH treatment was revolutionized by introduction of the terminal complement C5 inhibitor eculizumab, which has proven effective in addressing intravascular hemolysis10,11 and thromboembolism,12 with a significant impact on long-term survival.13,14 Recently, a new long-acting C5 inhibitor dosed every 8 weeks, ravulizumab, has demonstrated non-inferiority to eculizumab in controlling intravascular hemolysis.15,16 Although the significant benefits of C5 inhibition in the treatment of PNH patients have been clearly demonstrated, the hematologic benefit may be hampered by the emergence of C3-mediated extravascular hemolysis from early phases of complement activation,17,18 which C5 inhibition cannot address.19-21 Thus, alternative strategies of complement inhibition are required to improve PNH treatment, and agents are in development to address this and other unmet patients’ needs, including improved convenience.22 The complement cascade has three activating pathways (alternative, classical, and lectin-mannose) that merge at the key complement component C3; from this level (which is amplified by alternative pathway [AP] proteins), the effector pathway starts, with generation of anaphylatoxins and the membrane attack complex (MAC).23 Novel strategies of therapeutic complement inhibition in development aim to intercept the complement cascade upstream of C5, some targeting upstream at pathway-specific initiating events.22 During AP initiation, the serine protease complement factor D cleaves factor B, leading to AP C3 convertase generation. Danicopan (ACH-4471, ACH-044471, ALXN2040) is a first-in-class, oral, smallmolecule factor D inhibitor that prevents new AP C3 convertase formation.24 Consequently, proximal inhibition at the level of factor D blocks AP-initiated upstream events and up to 80% of classical or lectin pathway-initiated downstream events via amplification-loop inhibition.25 In vitro, danicopan inhibited both AP-mediated hemolysis and C3 fragment deposition on red blood cells from PNH patients.24 Phase I data from healthy human volunteers in single and multiple ascending-dose trials showed that danicopan was generally well tolerated and could achieve inhibition of AP complement activity.26 This work identified danicopan 200 mg thrice daily (tid) as a safe and effective dose/regimen.26 For PNH, targeting factor D inhibition with a small molecule represents a potentially important treatment advancement because proximal AP inhibition may disable both terminal complement activation (inhibiting MAC–mediated intravascular hemolysis) and C3 fragment opsonization (preventing extravascular haematologica | 2021; 106(12)

hemolysis), with additional convenience of oral administration. We investigated the factor D inhibitor danicopan as single-agent treatment for untreated PNH, aiming to control intravascular hemolysis while preventing C3-mediated extravascular hemolysis.

Methods Study design This international, open-label, single-arm, dose-finding, phase II study investigated danicopan in patients with hemolytic PNH not receiving complement inhibitor treatment. This trial was approved by regulatory agencies/local ethics committees and conducted according to International Conference on Harmonisation and Good Clinical Practice Standards. Achillion, Inc., a subsidiary of Alexion Pharmaceuticals, Inc., designed and sponsored the study, with advice from the investigators. All participants provided written informed consent. The trial is registered at www.clinicaltrials.gov as #NCT03053102.

Patients This study was conducted from March 2017 to November 2018 and involved adults with untreated PNH. To be enrolled, patients had to have hemoglobin <12 g/dL (and adequate reticulocytosis according to the investigator), GPI-deficient granulocytes or type III erythrocyte clone size ≥10%, lactate dehydrogenase (LDH) ≥1.5 times upper limit of normal (ULN), platelet counts ≥50x109/L, and willingness to be vaccinated for N. meningitidis, H. influenzae, and S. pneumoniae. Investigators used their clinical judgement to assess whether patients had enough bone marrow capacity to derive potential benefit by comparing the level of pre-entry hemoglobin to the absolute reticulocyte count. None of the subjects was receiving eculizumab, because of lack of availability and/or the patients’ willingness.

Treatment Patients received oral danicopan at a starting dose of 100 mg or 150 mg tid. The starting dose was based on phase I data showing that danicopan doses of 200 to 600 mg reached peak plasma levels within 1 to 2 h and were well-tolerated, and that 200 mg tid was effective on PNH red blood cells.22,26 Dose escalations were permitted based on hemolysis control, assessed by LDH, per investigator assessment in stepwise increments up to 200 mg tid. Dose escalation criteria for the first 28 days were specified in the study protocol with potential dose escalation points occurring at days 7 and 14 (Online Supplementary Appendix). Dose escalation was permitted thereafter and was done in consultation between the investigator and sponsor based on the hemoglobin response; absolute reticulocyte count, LDH, and indirect bilirubin were reviewed to evaluate evidence for possible additional clinical benefit from dose escalation. Because this study was the first treatment experience with danicopan in PNH patients, dose escalations were approached cautiously, especially when moving to the higher doses in the study (from 150 to 175 mg and 175 to 200 mg) or if alanine aminotransferase levels had fluctuated relative to the baseline value. Patients were instructed to take their medication approximately every 8 h. On study center days, when blood for laboratory tests was drawn, the morning dose was to be administered in the study center by the site personnel following safety and pharmacokinetic assessments. Additionally, patients were instructed to take each dose with food, including prior to intensive sampling for pharmacokinetic studies on days 1, 13, and 28. The planned duration of treatment was 84 days; patients completing

3189

A.M. Risitano et al.

treatment with clinical benefit entered a long-term extension study (ClinicalTrials.gov; NCT03181633).

Endpoints The primary efficacy endpoint was change in LDH concentration from baseline at day 28. Secondary efficacy parameters were change of hemoglobin concentration from baseline at days 28 and 84 and LDH change from baseline at day 84. Safety, tolerability, pharmacokinetics, and pharmacodynamics were also investigated. Laboratory assessments comprised hematology, chemistry and urinalysis. Patient-reported outcomes were measured at baseline and during the study via the validated Functional Assessment of Chronic Illness Therapy (FACIT)–Fatigue scale. Haptoglobin, bilirubin, reticulocytes, GPI-deficient clone size of erythrocytes (type III) and granulocytes (type II and III unless type III values provided), and percentage of C3 fragment-coated erythrocytes were also monitored. Soluble C5b-9 was evaluated as an exploratory endpoint (non-GLP); C5a was not monitored. Transfusion history up to 3 years prior to and during the study was collected.

Assay methods Plasma danicopan concentrations were determined by liquid chromatography. Pharmacodynamics were determined by measuring serum AP activity with an AP Wieslab assay. Plasma Bb concentration, serum factor D, C3, and C4 concentrations were also monitored. All aforementioned complement tests were conducted in a central laboratory using commercial kits. C3 fragment deposition on erythrocytes was measured using flow cytometry with FITC-conjugated anti-human C3d antibody (see the Online Supplement for details).

baseline hemoglobin concentration was heterogeneous among patients (9.8±1.8 g/dL); two patients received transfusions in the 12 weeks preceding study entry, with one of these patients also having a medical history of aplastic anemia.

Study disposition and safety The disposition of patients in the study is shown in Online Supplementary Figure S1. Two patients started danicopan at 100 mg tid and increased to 150 mg tid, and eight started at 150 mg tid. Increases to 175 mg and 200 mg tid were performed in eight and four patients, respectively. All ten patients reached day 28 and are included in the primary endpoint evaluation. Two discontinued before day 84: one because of a serious adverse event, elevated aspartate aminotransferase/alanine aminotransferase coincident with breakthrough hemolysis, which resolved without sequelae; the other withdrew for personal reasons unrelated to safety. All patients were evaluated until they left the study or reached day 84 (n=8). Nine patients (90%) developed at least one adverse event during treatment; only one was serious (described above). In total, 38 unique treatment-emergent adverse events were recorded, of which four were considered possibly related and two probably related to danicopan. The most frequent events were PNH-related (hemolysis and its signs or symptoms) and infections, usually of the upper respiratory tract (Table 2). With few exceptions, adverse events were mild and resolved during the study. There were no clinically significant changes in other key laboratory parameters during treatment (Online Supplementary Table S2).

Statistical analysis This was a proof-of-concept, first-in-patients, exploratory, phase II study. The sample size was determined based on the very limited number of untreated PNH patients and the exploratory nature of this study to evaluate effectiveness of danicopan. Given the small sample size, only descriptive and exploratory statistics were utilized to present results for continuous biochemical and quality-of-life measurements. Patients who discontinued treatment during the trial were not replaced. Missing values were not imputed. To summarize categorical data, frequency counts and percentages are presented. The Pearson correlation coefficient (Pearson r) was used to examine the relationship between two variables. The quantitative analysis between pharmacokinetics (plasma danicopan concentration) and pharmacodynamics (AP inhibition) was conducted with nonlinear regression using the simple Emax dose-response equation (Prism 5.02, GraphPad Software, La Jolla, CA, USA).

Results Patients’ characteristics Eleven patients were screened. Ten untreated patients with hemolytic PNH were enrolled and received danicopan. These patients’ baseline characteristics are presented in Table 1 and Online Supplementary Table S1. Their median age was 33 years (range, 17-63 years) and median disease duration was 5.9 years (range, 0-14 years). The mean GPI-deficient clone size was 79% for granulocytes and 32% for erythrocytes. Overt hemolysis was demonstrated by elevated LDH levels (1416±540 U/L, corresponding to 5.7±2.17 times ULN), increased reticulocyte count (154±69×109/L), increased total bilirubin (1.3±0.74 mg/dL), and reduced haptoglobin (5.8±2.9 mg/dL). The 3190

Pharmacokinetics and pharmacodynamics Danicopan was bioavailable with dose-proportional exposure (peak serum concentration [Cmax] and area under the curve [AUC]) at 100, 150, and 175 mg tid doses, demonstrated by intensive pharmacokinetic and pharmacodynamic profiling on days 6, 13, and 20 (Figure 1A; Online Supplementary Table S3), whereas trough concentrations assessed biweekly at single time points from days 28 to 84 were variable (Online Supplementary Table S4). Trough drug concentration was much more variable than AUC and Cmax. For example, at the 175 mg tid dose, the concentrations varied from 62.6 to 223.1 ng/mL. There was appreciable inter-patient variability, as anticipated for a study with a small number of patients. One of two patients who received 200 mg tid was not included in day 56 analyses because the sample was not available (missed study visit). Per protocol, no patients were receiving 200 mg tid by day 20 (pharmacokinetic sampling was performed on days 6, 13, and 20). Plasma factor D concentration did not change during treatment (Online Supplementary Figure S2A). As anticipated from its mechanism of action, treatment resulted in selective AP inhibition (Figure 1B, upper panel; Online Supplementary Table S4) with no effect on classical pathway activity (Online Supplementary Figure S2B). Notably, AP activity ≤10% was observed at all time points except at 0 and 8 h when danicopan concentration was lowest (Figure 1B) indicating that the AP activity in patients may not be fully inhibited at the end of the 8 h dosing period. Pharmacokinetic/pharmacodynamic analysis showed a dose-response relationship between danicopan and AP inhibition (Online Supplementary Figure S2D). haematologica | 2021; 106(12)

Danicopan, an oral factor D inhibitor for PNH

Table 1. Baseline characteristics and clinical results. Patient 1* Patient 2 Patient 3§ Patient 4° Patient 5 BL D28 D84 BL D28 D84 BL D28 D84 BL D28D84 BL D28 D84 Danicopan (mg po tid) Hb (g/dL) LDH (xULN) Reticulocytes (109/mL) Total bilirubin (mg/dL) GPI-deficient RBC clone size (%) C3d+ RBC (%) FACIT-Fatigue††

100 150 7.5 9.3 3.76 2.64 84 76 0.53 0.70 20 31

175 8.7 2.36 69 0.41 36

100 11.7 7.39 150 2.34 11

150 14.1 2.22 71 0.88 41

175 12.3 3.27 85 0.47 43

150 12.0 3.60 80 2.40 36

0.1 -† 0.1 0.4 -† 0.1 0.1 22 33 31 32 47 49 42

175 11.8 0.77 42 0.94 49

200 13.7 0.94 45 1.05 58

-† 0.1 43 52

150 175 11.7 12.6 5.09 1.84 170 109 0.58 0.35 24 38

- 150 175 175 - 10.4 10.2 10.9 - 1.63 0.89 0.90 - 45 55 49 - 0.41 0.41 0.41 - 18 33 42

0.3 -† 23 36 -

0.1 0.1 0.1 22 23 40

Patient 6 BL D28 D84

Patient 7# Patient 8 BL D28 D84 BL D28 D84

Patient 9 BL D28 D84

Patient 10* BL D28 D84

150 150 9.2 11.2 7.93 0.90 236 38 1.17 0.23 78 82

150 8.7 6.68 137 0.83 20

150 175 10.0 11.4 6.08 2.29 171 78 1.32 0.54 13 35

150 6.9 6.02 249 1.64 75

150 11.4 2.50 105 0.53 78

0.1 -† 0.4 20 52 52

175 10.2 3.95 50 1.04 -

- 150 175 175 - 9.5 10.2 11.6 - 8.55 1.51 3.26 - 217 74 48 - 1.88 0.88 0.74 - 21 47 59

0.5 0.1 49 51 -

0.2 -† -⁋ 44 49 50

200 11.9 2.98 118 0.71 42

0.1 -$ -⁋ 39 46 46

150 8.4 0.80 109 0.35 37

150 11.1 1.02 130 0.41 92

0.1 0.3 0.1 47 52 52

BL: baseline; D: day; po: orally; tid: three times a day; Hb: hemoglobin; LDH: lactic acid dehydrogenase; ULN: upper limit of normal; GPI: glycosylphosphatidylinositol; RBC: red blood cells; FACIT: Functional Assessment of Chronic Illness Therapy. *Patient reported a history of aplastic anemia. †C3 fragment deposition was not tested for day 28; no data entry for this visit. ††Scores based on the FACIT-Fatigue Scale V4; score range 0-52 (a score <30 indicates severe fatigue). §Patient received a protocol waiver to enter the study despite the <12 g/dL Hb inclusion criterion; for the sponsor and site investigator, the Hb level did not represent a clinically meaningful difference relative to the threshold in the protocol and the patient had significant hemolysis as evidenced by LDH values. °Patient withdrew due to personal reasons not related to safety (day 40). #Patient withdrew due to a serious adverse event (day 51). ⁋C3 fragment deposition was not tested at day 84; no data entry for this visit. $Incorrect container type.

Efficacy Primary endpoint Change in LDH concentration from baseline to day 28 was the primary efficacy endpoint. A significant reduction was observed in all ten patients from a mean value of 5.7±2.17 times ULN at baseline to 1.8±1.03 times ULN at day 28 (P<0.001) (Figure 2A, Table 1), demonstrating achievement of the primary endpoint. The percentages of patients showing LDH <3 times ULN, <2 times ULN, and <1 time the ULN were 90%, 60%, and 40%, respectively.

Secondary endpoints Significant reductions in LDH from baseline were sustained throughout the study, with values being 2.3±1.41 times ULN at day 56 (P<0.005) and 2.2±1.04 times ULN (P<0.001) at day 84 (Figure 2A). The percentages of patients with LDH <3 times ULN, <2 times ULN, and <1 time ULN were 71%, 43%, and 43% at day 56, and 75%, 37.5%, and 25% at day 84, respectively. Danicopan is a highly permeable drug and even in patients with high body mass index, plasma concentrations do not appear to cause a clinically significant change in effect. Fluctuations in LDH indicated that low-level residual intravascular hemolysis remained in most patients, with possible transient exacerbations; this was due to transient weaker AP inhibition around predose periods, as described above, in addition to an increase of susceptible GPI-deficient erythrocytes during treatment (see below). Nevertheless, only two breakthrough hemolytic events were recorded by the investigator as adverse events (Table 2); a third patient experienced recurrent subclinical breakthrough episodes as a consequence of inadequate control of complement activation. Treatment with danicopan translated into an improvement of anemia: mean hemoglobin increased from 9.8 g/dL at baseline (range, 6.9 to 12.0 g/dL) to 10.9 g/dL at day 28 (range, 8.4 to 14.1 g/dL; P<0.005), 10.9 g/dL at day 56 (range, 8.5 to 13.1 g/dL; P<0.005), and 11.5 g/dL at day 84 (range, 8.7 to 13.7 g/dL; P<0.005) (Figure 2B, Table 1). The mean increase from baseline was 0.9 g/dL at day 28 and 1.7 g/dL at day 84. In the 12 weeks preceding study entry, two patients received transfusions (Figure 3A). One of these patients, who had aplastic anemia (not receiving immunosuppressive therapy), had received five transfusions totaling ten units. During treatment, this patient received three transfusions totaling seven units. The sechaematologica | 2021; 106(12)

ond patient had one transfusion (two units) during breakthrough hemolysis in the setting of a viral infection; in this case, danicopan treatment was not detrimental to the course of the infection, but (irrespective of its possible effect on other complement pathways) did not effectively counteract the transient acute complement activation (possibly due to pharmacokinetic/pharmacodynamic reasons). The remaining patients in the trial were transfusion independent through the 84 days of treatment. Accompanying the control of intravascular hemolysis and improvement of anemia, patient-reported outcomes were assessed. The mean FACIT–Fatigue score at baseline was 34 and increased by 9 and 13 points at days 28 and day 84, respectively (P<0.05) ( Figure 3B, Table 1). Control of intravascular hemolysis by danicopan was further confirmed by changes in other laboratory parameters. Danicopan significantly increased the percentage of GPI-deficient erythrocytes (56±19.9% at day 84 vs. 32±24.6% at baseline; P=0.001), whereas no change was observed for GPI-deficient granulocytes (Figure 4A). Total bilirubin decreased after danicopan treatment (0.6±0.23 mg/dL at day 84 vs. 1.3±0·74 mg/dL at baseline, P<0.05) (Figure 4B). A sustained but not statistically increase in haptoglobin was observed (15.3±16.08 mg/dL at day 84 vs. 5.8±2.89 mg/dL at baseline; P=0.15) (Online Supplementary Figure S3A), as was a transient decrease in free hemoglobin (83±51 mg/dL at day 28 vs. 138±132 mg/dL at baseline; P=0.26) (Online Supplementary Figure S3B). Absolute reticulocyte count decreased quickly and was sustained with treatment (81±33.6×109/L at day 84 vs. 154±69×109/L at baseline; P<0.05) (Figure 4C). Additional laboratory results can be found in Online Supplementary Table S6.

Exploratory endpoints To investigate the mechanistic effect of proximal complement inhibition by danicopan on PNH, complement biomarkers were monitored. Bb fragment, an activation product of factor B, tracks complement AP activation in vivo. Plasma Bb level was significantly elevated at baseline (2.24±0.77 mg/mL) compared with that in healthy volunteers (0.84±0.212 mg/mL; P<0.05). After danicopan, the Bb levels were significantly reduced: 0.84±0.84 mg/mL at day 28 (P<0.005); 0.47±0.09 mg/mL at day 56 (P<0.005); and 0.47±0.06 µg/mL at day 84 (P<0.005) (Figure 1C). In contrast to residual AP activity, Bb level remained consis3191

A.M. Risitano et al.

tently low irrespective of subtherapeutic plasma danicopan levels in predose periods (Figure 1B, middle panel). A strong positive correlation was found between Bb and LDH (Pearson r=0.80, P<0.0001) (Online Supplementary Figure S2E), supporting Bb as a reliable biomarker of in vivo AP activation in PNH and, therefore, its value for monitoring efficacy. Danicopan also showed strong linear correlations with Bb and LDH (negative), as did AP with Bb and LDH (positive); there was no correlation of classical pathway activity with any of these parameters (Online Supplementary Table S5), validating the role of danicopan in AP inhibition and subsequent in vivo changes of Bb and LDH. Additional laboratory results can be found in Online Supplementary Table S6. There was a slight increase in serum C3 (114.2±17.3 mg/dL at day 84 vs. 102.2±20.2 mg/dL at baseline, P=0.08) (Online Supplementary Figure S2C), likely from reduced C3 consumption because of upstream complement blockade. Importantly, C3 fragment deposition on erythrocytes was

very low (<0.5% of erythrocytes) throughout treatment (Figure 4A). sC5b-9 was normal at baseline and remained relatively constant over time (data not shown).

Discussion Current PNH treatments target C5 inhibition.10-14 Novel complement inhibitors in development aim to address unmet needs of PNH patients.22 Here, we describe a novel approach to PNH treatment, which aims to change the current paradigm of PNH therapy by improving hemoglobin levels in addition to reducing hemolysis, with the added advantage and convenience of oral administration. We investigated danicopan, a first-in-class oral factor D inhibitor, which blocks the proximal complement cascade upstream of C5 at the level of AP initiation and amplification. In untreated PNH patients, danicopan monotherapy resulted in inhibition of intravascular hemolysis, with sig-

Figure 1. Pharmacokinetic-pharmacodynamic evaluation of danicopan. (A) The mean plasma danicopan concentration by dose at hours 0 (predosing), 1, 1.5, 2, 2.5, 3, 4, 6, 8 (predosing) and 12 of days 6, 13, and 20. The number of patients is, respectively, two (100 mg tid) and eight (150 mg tid) at day 6 and day 13, and five (150 mg tid) and five (175 mg tid) at day 20. (B) The mean ± standard deviation of ex vivo serum alternative pathway (AP) activity, plasma Bb concentration, and plasma danicopan concentration, combining all dosing groups together. Serum AP activity and plasma Bb concentration were determined for a subset of the aforementioned time points by the AP Wieslab assay (Euro Diagnostica) and Bb enzyme-linked immunosorbent assay, respectively. (C) Plasma Bb concentration (mean ± standard deviation) at baseline (day 1) through the end of the study (day 84) with descriptive statistics. The dashed lines represent the upper and lower limits of normal, which were derived from phase I studies in healthy volunteers (see Assay Methods in the Online Supplement). N values of <10 for plasma Bb on days 42, 56, and 84 reflect the two early discontinuations and additional samples not collected. **P<0.005. tid: thrice daily; ULN: upper limit of normal; LLN: lower limit of normal; SD: standard deviation.

3192

haematologica | 2021; 106(12)

Danicopan, an oral factor D inhibitor for PNH

nificant reductions of LDH at day 28 (primary endpoint) and throughout treatment duration. In contrast with standard C5 inhibitor therapies, the inhibition of intravascular hemolysis observed during danicopan treatment was not associated with persistently increased bilirubin and reticulocyte count, or with the emergence of C3 deposition on surviving GPI-deficient erythrocytes, in agreement with its anticipated effect on extravascular hemolysis. Concomitant inhibition of intravascular hemolysis and prevention of C3-mediated extravascular hemolysis

resulted in improvement of anemia, with a mean hemoglobin gain of 1.7 g/dL after 12 weeks of treatment. Consistent with these findings, all patients exhibited significant increases in the percentage of GPI-deficient erythrocytes, confirming the extended half-life of these cells in vivo. Furthermore, all patients had improvements in FACIT-Fatigue quality-of-life measurements; FACITFatigue scores were used in this proof-of-concept study due to the lack of validated instruments for patient-reported outcomes in PNH. Additional patient-reported out-

Table 2. Adverse events.

Standard severity grade Moderate Severe Life-threatening

Primary system organ class preferred term*

% of total (N=10)

Mild

Number of subjects reporting Number of unique TEAE† Number of subjects with SAE Blood and lymphatic system disorders Hemolysis Gastrointestinal disorders Abdominal pain Mouth ulceration Nausea Vomiting General disorders and administration site conditions Fatigue Non-cardiac chest pain Edema, peripheral Vaccination site erythema Infections and infestations Pharyngitis Upper respiratory tract infection Viral upper respiratory tract infection Injury, poisoning, and procedural complications Contusion Investigations Alanine aminotransferase increased Aspartate aminotransferase increased Metabolism and nutrition disorders Iron deficiency Musculoskeletal and connective tissue disorders Back pain Myalgia Nervous system disorders Headache Psychiatric disorders Irritability Renal and urinary disorders Hemoglobinuria Paroxysmal nocturnal hemoglobinuria Reproductive system and breast disorders Dysmenorrhea Skin and subcutaneous tissue disorders Rash, papular

9 38 1 2 2 3 1 1 1 1 4 1 1 1 1 5 1 4 1 1 1 1 1 1 1 1 3 2 1 4 4 1 1 3 2 1 1 1 1 1

90 NA 10 20 20 30 10 10 10 10 40 10 10 10 10 50 10 40 10 10 10 10 10 10 10 10 30 20 10 40 40 10 10 10 20 10 10 10 10 10

1 1 1 1 1 1 1 1 3 1 1 1 2 1 4 1 2 1 1

1 1 1 1 -

1 1 .. 1 -

1 -

*MedDRA Version 18.1. †This row represents the number of events; all other rows represent the number of subjects. TEAE: treatment-emergent adverse events; NA: not available; SAE: serious adverse events.

haematologica | 2021; 106(12)

3193

A.M. Risitano et al.

come instruments will be utilized in a phase III trial. With the caveat of limited sample size and exposure, no safety concerns emerged during the study other than those described; in particular, infectious events were rare, clinically mild, and self-limiting, irrespective of unchanged danicopan treatment. Although the short study duration through the primary endpoint limits the ability to draw

robust conclusions, the data do not appear to support the postulated increased risk of infectious complications (due to upstream inhibition of the complement cascade),27 in agreement with in vitro data showing that killing of encapsulated and unencapsulated meningococci was nearly unaffected relative to that occurring with eculizumab.28,29 Danicopan is the first oral complement inhibitor treat-

Figure 2. Effect of danicopan on lactate dehydrogenase and hemoglobin levels. (A) Change in lactate dehydrogenase (LDH) concentration from baseline (day 1) to day 28 was the primary efficacy endpoint. LDH reduction per patient is shown here including the reduction from a mean value of 5.7±2.17 times upper limit of normal (ULN) at baseline to 1.8±1.03 times ULN at day 28 (P<0.001) demonstrating achievement of the primary endpoint. A significant mean LDH reduction from baseline was sustained throughout the study up to day 84 (2.2±1.04 times ULN; P<0.001). (B) Per patient effects on hemoglobin with a mean group increase from 9.8 g/dL at baseline (day 1) (range, 6.9 to 12.0 g/dL) to 10.9 g/dL at day 28 (range, 8.4 to 14.1 g/dL; P<0.005), and 11.5 g/dL at day 84 (range, 8.7 to 13.7 g/dL, P<0.005). Note, patient 3 received a protocol waiver to enter the study despite the <12 g/dL hemoglobin inclusion criterion. For the sponsor and site investigator, the hemoglobin level did not represent a clinically meaningful difference relative to the threshold in the protocol and the patient had significant hemolysis, as evidenced by LDH values. **P<0.005; SD:standard deviation.

Figure 3. Effect of danicopan on blood transfusions and Functional Assessment of Chronic Illness Therapy-Fatigue score. (A) Two patients required transfusions during the trial, for a total of nine units on four occasions over 84 days. The transfusion history (84 days prior to screening through to the end of the study; sum for all patients) is provided. (B) Mean Functional Assessment of Chronic Illness Therapy–Fatigue score values (± standard deviation) at baseline (day 1) through to the end of the study (day 84) with descriptive statistics. The range of scores was 0 to 52; a score of <30 indicates severe fatigue. *P<0.05; **P<0.005. Note, data were not obtained for one patient at day 56. FACIT: Functional Assessment of Chronic Illness Therapy; SD: standard deviation.

3194

haematologica | 2021; 106(12)

Danicopan, an oral factor D inhibitor for PNH

ment demonstrating efficacy and safety in PNH patients as monotherapy in a phase II study. The clinical effects observed were achieved irrespective of low-level, residual intravascular hemolysis, which remained detectable in some patients. This residual intravascular hemolysis is likely the consequence of the increase of GPI-deficient erythrocytes susceptible to complement-mediated hemolysis and the pharmacokinetic/pharmacodynamic characteristics of danicopan. Danicopan plasma level and AP activity also correlated with in vivo biomarkers (LDH and Bb), and Bb seems a reliable indicator of complement activation in vivo during danicopan treatment. Based on the results of the phase I single-dose and multiple ascendingdose trials in healthy human volunteers, 200 mg tid was shown to be both safe and efficacious.26 However, in this cohort of patients, full blockade of AP activity was not consistently achieved in all patients, irrespective of individual dose adjustment and the broad dose ranges used during the study. These observations suggest that residual intravascular hemolysis is due to low residual AP activity observed in some patients, which may be better inhibited by a more potent second-generation factor D inhibitor analog that will be assessed for safety and efficacy in a phase II trial (Clinicaltrials.gov, NCT04170023). In a concurrent phase II trial, patients (n=12) on a stable regimen

of eculizumab with hemoglobin <10 g/dL and who were transfusion dependent (≥1 red blood cell transfusion within 12 weeks of screening) received oral danicopan 100–150 mg tid, with possible response-based dose escalation to 200 mg tid at predefined time points. The addition of danicopan led to clinically and statistically significant reductions in frequency of red blood cell transfusions and in the number of transfused units in patients compared to those in patients with a history of eculizumab treatment alone.30 Clinical development of proximal complement inhibitors has been motivated by the description of C3mediated extravascular hemolysis as a mechanism driving significant anemia and limiting hematologic benefit in some PNH patients on eculizumab and other C5 inhibitors.17,31 Indeed, it is conceivable that, in combination with C5 inhibitors, proximal inhibitors can address C3-mediated extravascular hemolysis by preventing the generation of C3 convertase, eventually leading to better hematologic response in PNH patients.17,31 Moreover, it has been hypothesized that proximal inhibitors, by preventing generation of downstream C5 convertases, could be effective even in the absence of terminal inhibitors (i.e., eculizumab or other C5 inhibitors). Danicopan is the first compound demonstrating that proximal complement inhibitors can be used safely and effectively as monother-

Figure 4. Additional clinical efficacy evaluation of danicopan. (A) The mean (± standard deviation) clone size percentages displayed for GPI-deficient erythrocytes and granulocytes at baseline (day 1) through to the end of the study (day 84) and mean (± standard deviation) percentage of erythrocytes with C3 fragment deposition with descriptive statistics for GPI-deficient erythrocytes. Descriptive statistics for GPI-deficient granulocytes and erythrocytes with C3 fragment deposition are listed in Online Supplementary Table S6. Erythrocytes with C3 fragment deposition are defined as erythrocytes that stain positive for anti-C3d antibody. (B) Total bilirubin at baseline (day 1) through to the end of the study (day 84) with descriptive statistics. Data for one patient were not obtained at day 56. (C) Absolute reticulocyte count at baseline (day 1) through to the end of the study (day 84) with descriptive statistics. Data for one patient were not obtained at day 56. *P<0.05; **P<0.005. SD: standard deviation; GPI: glycosylphosphatidylinositol.

haematologica | 2021; 106(12)

3195

A.M. Risitano et al.

apy in PNH. By disabling the initiating event of complement activation, danicopan prevents generation of C5 convertases, obviating the need for downstream C5 inhibition. Additionally, specific targeting of AP can preserve classical and lectin pathway-mediated antimicrobial activity. The landscape of complement inhibition in the treatment of PNH will continue to evolve with the availability of proximal inhibitors. In addition to danicopan, other oral agents targeting the AP, such as factor B inhibitors, are under investigation for the treatment of PNH.22,31,32 Proximal complement inhibitors also include subcutaneously administered agents targeting C3; one of these agents was reported to be effective in a recent phase III trial in patients with PNH.33 All of these approaches look promising for the treatment of PNH and clinical data should tell us very soon what are the viable treatment options in terms of safety and efficacy, and how we can best utilize them in the appropriate patients (i.e., monotherapy vs. add-on treatment). Preclinical data seem to suggest that AP inhibitors may be as effective as terminal complement blockade with clinical differences mostly due to the specific pharmacokinetic/pharmacodynamic profile of the individual inhibitor, rather than to its target in the complement cascade.22,31 Indeed targeting the AP, as PNH is a disease due to AP dysregulation (i.e., continuous, spontaneous C3 tick-over, eventually exacerbated at times of additional complement activation), possible applications in other diseases will require an understanding of their pathogenic mechanisms.34 In conclusion, in this study danicopan appeared to be well-tolerated and showed clinically meaningful inhibition of intravascular hemolysis and hemoglobin improvement in untreated PNH patients. This is the first evidence that a proximal complement inhibitor, used as monotherapy, can have a clinical impact on PNH by inhibiting intravascular hemolysis and preventing extravascular hemolysis. While second-generation compounds with improved pharmacokinetics and pharmacodynamics are in development, this study paves the way to improved hematologic response and novel standards of care, with an easier mode of administration, for hemolytic PNH patients. Disclosures AMR has received research support from Alexion, Novartis, Alnylam and Ra Pharma, lecture fees from Alexion, Novartis, Pfizer and Apellis, served as a member of advisory/investigator boards for Achillion, Alexion, Apellis, Biocryst, Novartis, Roche, and Samsung, and served as a consultant for Amyndas. AGK has served on advisory boards for Alexion, Celgene, Novartis, Ra Pharma, and Regeneron and has received travel grants from Achillion, Celgene, and Ra Pharma. JWL has received grants from Alexion and Achillion, has served as a member of advisory

References 1. Luzzatto L, Notaro R. Paroxysmal nocturnal hemoglobinuria. In: Handin RI, Lux SE, Stossel TP, eds. Blood, Principles and Practice of Hematology. 2nd edition. Philadelphia: Lippincott Williams & Wilkins; 2003:319-334. 2. Takeda J, Miyata T, Kawagoe K, et al. Deficiency of the GPI anchor caused by a somatic mutation of the PIG-A gene in paroxysmal nocturnal hemoglobinuria.

3196

boards for Alexion and Apellis, has received honoraria from Alexion, and has served as a consultant for AlloVir. JPM has received lecture honoraria from Alexion, Novartis, and Celgene and served as a consultant for Novartis. RN has received lecture fees from Alexion. RB has served on scientific advisory boards for Achillion and Alexion. PB has served as a member of advisory boards for Merck, Janssen, Roche and AbbVie, has served as an associate editor for the Internal Medicine Journal, and received research funding from Roche, Beigene, and Amgen. MH is an employee of Alexion and has equity ownership in the same company. MG was an employee of Achillion, Inc., a subsidiary of Alexion Pharmaceuticals, Inc. and had equity ownership in the same company. Contributions MH and MG developed the study protocol, with the contribution of AMR, AGK, RN, RB and PB; AMR, AGK, JWL, and PB recruited and treated patients, and collected the data; AMR, MH and MG analyzed and interpreted the data, and wrote the manuscript; AGK, JWL, JPM, RN, RB and PB contributed to the manuscript and approved its final version. Acknowledgments We thank the patients and investigators, as well as their staff, who participated in this trial: Serena Marotta, Luana Marano and Fabiana Cacace (Naples), Petra Muus and Shreyans Gandhi (London), Federica Barone (Florence), Sung-Soo Park (Seoul), and Paul Hamilton (Auckland). Funding The study was sponsored and entirely supported by Achillion Inc., a subsidiary of Alexion Pharmaceuticals, Inc. We thank Heather Robison (an employee of Achillion Inc., a subsidiary of Alexion Pharmaceuticals, Inc.) and Steven Podos, Danny Shin and Julia Catini (Alexion employees and former employees of Achillion Inc., a subsidiary of Alexion Pharmaceuticals, Inc.) for their assistance in writing this manuscript and The Curry Rockefeller Group (funded by Achillion Inc., a subsidiary of Alexion Pharmaceuticals, Inc.) for their editorial assistance. Data sharing Alexion will consider requests for disclosure of clinical study participant-level data provided that participant privacy is assured through methods such as data de-identification, pseudonymization, or anonymization (as required by applicable law), and if such disclosure was included in the relevant study informed consent form or similar documentation. Qualified academic investigators may request participant-level clinical data and supporting documents (statistical analysis plan and protocol) pertaining to Alexion-sponsored studies. Further details regarding data availability and instructions for requesting information are available in the Alexion Clinical Trials Disclosure and Transparency Policy at http://alexion.com/research-development.

Cell. 1993;73(4):703-711. 3. Miyata T, Yamada N, Iida Y, et al. Abnormalities of PIG-A transcripts in granulocytes from patients with paroxysmal nocturnal hemoglobinuria. N Engl J Med. 1994;330(4):249-255. 4. Medof ME, Gottlieb A, Kinoshita T, et al. Relationship between decay accelerating factor deficiency, diminished acetylcholinesterase activity, and defective terminal complement pathway restriction in paroxysmal nocturnal hemoglobinuria ery-

throcytes. J Clin Invest. 1987;80(1):165-174. 5. Rotoli B, Luzzatto L. Paroxysmal nocturnal haemoglobinuria. Baillieres Clin Haematol. 1989;2(1):113-138. 6. Luzzatto L, Risitano AM. Advances in understanding the pathogenesis of acquired aplastic anaemia. Br J Haematol. 2018; 182(6):758-776. 7. Nicholson-Weller A, March JP, Rosenfeld SI, Austen KF. Affected erythrocytes of patients with paroxysmal nocturnal hemoglobinuria are deficient in the complement

haematologica | 2021; 106(12)

Danicopan, an oral factor D inhibitor for PNH

regulatory protein, decay accelerating factor. Proc Natl Acad Sci U S A. 1983;80(16):5066-5070. 8. Holguin MH, Fredrick LR, Bernshaw NJ, Wilcox LA, Parker CJ. Isolation and characterization of a membrane protein from normal human erythrocytes that inhibits reactive lysis of the erythrocytes of paroxysmal nocturnal hemoglobinuria. J Clin Invest. 1989;84(1):7-17. 9. Lachmann PJ, Halbwachs L. The influence of C3b inactivator (KAF) concentration on the ability of serum to support complement activation. Clin Exp Immunol. 1975; 21(1):109-114. 10. Hillmen P, Young NS, Schubert J, et al. The complement inhibitor eculizumab in paroxysmal nocturnal hemoglobinuria. N Engl J Med. 2006;355(12):1233-1243. 11. Brodsky RA, Young NS, Antonioli E, et al. Multicenter phase 3 study of the complement inhibitor eculizumab for the treatment of patients with paroxysmal nocturnal hemoglobinuria. Blood. 2008; 111(4): 1840-1847. 12. Hillmen P, Muus P, Dührsen U, et al. Effect of the complement inhibitor eculizumab on thromboembolism in patients with paroxysmal nocturnal hemoglobinuria. Blood. 2007;110(12):4123-4128. 13. Kelly RJ, Hill A, Arnold LM, et al. Longterm treatment with eculizumab in paroxysmal nocturnal hemoglobinuria: sustained efficacy and improved survival. Blood. 2011;117(25):6786-6792. 14. Loschi M, Porcher R, Barraco F, et al. Impact of eculizumab treatment on paroxysmal nocturnal hemoglobinuria: a treatment versus no-treatment study. Am J Hematol. 2016;91(4):366-370. 15. Lee JW, Sicre de Fontbrune F, Wong Lee Lee L, et al. Ravulizumab (ALXN1210) vs eculizumab in adult patients with PNH naive to complement inhibitors: the 301 study. Blood. 2019;133(6):530-539. 16. Kulasekararaj AG, Hill A, Rottinghaus ST, et al. Ravulizumab (ALXN1210) vs eculizumab in C5-inhibitor-experienced adult patients with PNH: the 302 study. Blood. 2019;133(6):540-549. 17. Risitano AM, Notaro R, Marando L, et al. Complement fraction 3 binding on erythro-

haematologica | 2021; 106(12)

cytes as additional mechanism of disease in paroxysmal nocturnal hemoglobinuria patients treated by eculizumab. Blood. 2009;113(17):4094-4100. 18. Hill A, Rother RP, Arnold L, et al. Eculizumab prevents intravascular hemolysis in patients with paroxysmal nocturnal hemoglobinuria and unmasks low-level extravascular hemolysis occurring through C3 opsonization. Haematologica. 2010; 95(4):567-573. 19. Luzzatto L, Risitano AM, Notaro R. Paroxysmal nocturnal hemoglobinuria and eculizumab. Haematologica. 2010; 95(4): 523-526. 20. Risitano AM, Notaro R, Luzzatto L, Hill A, Kelly R, Hillmen P. Paroxysmal nocturnal hemoglobinuria--hemolysis before and after eculizumab. N Engl J Med. 2010; 363(23):2270-2272. 21. Notaro R, Sica M. C3-mediated extravascular hemolysis in PNH on eculizumab: mechanism and clinical implications. Semin Hematol. 2018;55(3):130-135. 22. Risitano AM, Marotta S. Toward complement inhibition 2.0: Next generation anticomplement agents for paroxysmal nocturnal hemoglobinuria. Am J Hematol. 2018; 93(4):564-577. 23. Taylor RP, Lindorfer MA. Mechanisms of complement-mediated damage in hematological disorders. Semin Hematol. 2018; 55(3):118-123. 24. Yuan X, Gavriilaki E, Thanassi JA, et al. Small-molecule factor D inhibitors selectively block the alternative pathway of complement in paroxysmal nocturnal hemoglobinuria and atypical hemolytic uremic syndrome. Haematologica. 2017; 102(3):466-475. 25. Harboe M, Ulvund G, Vien L, Fung M, Mollnes TE. The quantitative role of alternative pathway amplification in classical pathway induced terminal complement activation. Clin Exp Immunol. 2004; 138(3): 439-446. 26. Wiles JA, Galvan MD, Podos SD, Geffner M, Huang M. Discovery and development of the oral complement factor D inhibitor danicopan (ACH-4471). Curr Med Chem. 2020;27(25):4165-4180. 27. Sprong T, Roos D, Weemaes C, et al.

Deficient alternative complement pathway activation due to factor D deficiency by 2 novel mutations in the complement factor D gene in a family with meningococcal infections. Blood. 2006;107(12):4865-4870. 28. Granoff DM, Kim H, Topaz N, MacNeil J, Wang X, McNamara LA. Differential effects of therapeutic complement inhibitors on serum bactericidal activity against non-groupable meningococcal isolates recovered from patients treated with eculizumab. Haematologica. 2019; 104(8): e340-e344. 29. Konar M, Granoff DM. Eculizumab treatment and impaired opsonophagocytic killing of meningococci by whole blood from immunized adults. Blood. 2017; 130(7):891-899. 30. Kulasekararaj A, Risitano AM, Maciejewski JP, et al. A phase 2 open-label study of danicopan (ACH-0144471) in patients with paroxysmal nocturnal hemoglobinuria (PNH) who have an inadequate response to eculizumab monotherapy. Blood. 2019; 134(Suppl_1):3514. 31. Risitano AM, Marotta S, Ricci P, et al. Anticomplement treatment for paroxysmal nocturnal hemoglobinuria: time for proximal complement inhibition? A position paper from the SAAWP of the EBMT. Front Immunol. 2019;10:1157. 32. Risitano AM, Roth A, Soret J, et al. LNP023 - a new oral complement factor-B inhibitor normalizes hemoglobin in paroxysmal nocturnal hemoglobinuria patients with poor response to eculizumab, both as add-on and monotherapy. The 46th Annual Meeting of the European Society for Blood and Marrow Transplantation; August 29 September 1, 2020; O019. 33. Hillmen P, Szer J, Weitz I, et al. Results of the Pegasus phase III randomized trial demonstrating superiority of the C3 inhibitor, pegcetacoplan, compared to eculizumab in patients with paroxysmal nocturnal hemoglobinuria. The 25th European Hematology Association Annual Congress; June 12, 2020; 295012; S192. 34. Holers VM. The complement system as a therapeutic target in autoimmunity. Clin Immunol. 2003;107(3):140-151.

3197

LETTERS TO THE EDITOR Daratumumab, an original approach for treating multi-refractory autoimmune cytopenia Immune thrombocytopenia (ITP) and warm autoimmune hemolytic anemia (AIHA) are antibody-mediated autoimmune diseases in which plasma cells secrete pathogenic antibodies directed against platelet and red blood cell antigens.1 Some patients show no response to first- or second-line treatments, including corticosteroids, rituximab, immunosuppressive drugs, splenectomy and, in the case of ITP, thrombopoietin receptor agonists, a situation Table 1. Characteristics and outcomes of patients. Patient Age/ Autoimmune Disease Active or # Sex cytopenia duration past underlying (months) disease

that increases the risk of morbidity and mortality.2,3 In these refractory cases, persistent autoreactive long-lived plasma cells in the bone marrow could explain treatment failure. Daratumumab, an anti-CD38 monoclonal antibody developed to target tumoral plasma cells in multiple myeloma,4 was recently found to be effective in antibody-mediated diseases, such as autoimmune cytopenia following hematopoietic stem cell transplantation5–11 and systemic lupus.12 Here we report the characteristics and outcome of patients who received daratumumab “off label” (compassionate use) for severe refractory ITP or warm AIHA.

Time Previous Other previous Number of Treatments Response Time to Duration of Relapse from last splenectomy therapies daratumumab given with response response rituximab infusions daratumumab (days) (months) infusion

34/M

ITP

Evans syndrome, antiphospholipid syndrome

Yes

CS (resistant), IVIg (response), HCQ (failure), eltrombopag (failure), MMF (failure), CSA (failure)

None

35/F

ITP

128

Evans syndrome

No (obesity)

CS (resistant), IVIg (response), romiplostim (failure), eltrombopag (failure), MMF (failure), AZA (failure), CSA (failure), CYC (failure)

None

Yes

70/M

ITP

103

Recurrent venous thrombosis

Yes

CS (dependant), MMF (failure), sirolimus (failure), CSA (failure)

None

Failure

20/F

ITP

174

Ischemic stroke with hemorrhagic transformation

Yes

CS (resistant), IVIg (response), romiplostim (failure), eltrombopag (failure), MMF (failure), disulone (failure)

High-dose CS Failure and IVIg

35/F

ITP AGS

Hodgkin disease (9 years before daratumumab)

Yes

For AGS: CS (failure), IVIg (failure), MMF (failure). For ITP: romiplostim (failure), eltrombopag (failure), disulone (failure)

Failure

Yes (at 6 months, ITP)

69/M

ITP

Evans syndrome

High-dose CS Failure and IVIg

55/F

Warm AIHA

Evans syndrome

CS (response)

Yes

55/F

Warm AIHA

Evans syndrome

Yes

CS (response), AZA (failure), CSA (failure), everolimus (failure), bortezomib (failure)

CS (response), 4 IVIg (response), romiplostim (failure), eltrombopag (response then adverse event)

None

High-dose CS with CR complete weaning at 6 weeks None

Failure

AGS: acquired Glanzmann syndrome; AIHA: autoimmune hemolytic anemia; AZA: azathioprine; CR: complete response; CS: corticosteroids; CSA: cyclosporine; CYC: cyclophosphamide; DAT: direct antiglobulin test; ITP : immune thrombocytopenia; IVIg: intravenous immunoglobulin; HCQ: hydroxycholoquine; MMF: mycophenolate mofetil; NA : not applicable;

3198

haematologica | 2021; 106(12)

Letters to the Editor

We conducted an observational, retrospective study including patients >18 years who were treated with daratumumab for ITP and/or warm AIHA (primary or secondary) according to international criteria.13,14 The patients were identified through the French National Reference Center for Adult Immune Cytopenia network from November 2019 to November 2021. Any patient who

opposed data collection was not included. All patients were informed and gave consent to ‘off-label’ use of daratumumab. The initial treatment regimen was extrapolated from the one commonly used in myeloma4 (i.e., at least 4 daratumumab infusions at a dose of 16 mg/kg per week combined with oral dexamethasone 20 mg before each infusion). For ITP, complete response was defined as a

Figure 1. Autoimmune cytopenia evolution after daratumumab treatment. (A) Evolution of platelet counts in patients with immune thrombocytopenia after daratumumab. Patient #5 had acquired Glanzmann syndrome and red stars indicate hemorrhagic symptoms. (B) Evolution of hemoglobin levels in patients with warm autoimmune hemolytic anemia after daratumumab. Week 0 corresponds to the first daratumumab infusion (orange “D”). CS: corticosteroids, CYC: cyclophosphamide; IVIG: intravenous immunoglobulin, MMF: mycophenolate mofetil, RBC: red blood cell transfusion, SPL: splenectomy.

haematologica | 2021; 106(12)

3199

Letters to the Editor

platelet count >100x109/L and response as a platelet count between 30 and 100x109/L with at least a 2-fold increase from baseline.13 For warm AHAI, complete response was defined as a hemoglobin level ≥12 g/dL in the absence of recent transfusion and response as a hemoglobin level ≥10 g/dL with an increase of at least 2 g/dL from the pre-treatment level in the absence of recent transfusion (<1 month).14 Patients who required any other treatment for autoimmune cytopenia including rescue therapy >6 weeks after the first daratumumab infusion were considered non-responders regardless of platelet or hemoglobin levels. The first daratumumab infusion was considered day 0 for subsequent time points. Clinical and biological data were collected with a standardized form. The study received institutional review board approval (00011558, UPEC University, AP-HP). Eight patients (5 females [62.5%]; median age 45 years [range, 34–70]) from six participating centers received daratumumab for refractory ITP (n=5), acquired Glanzmann syndrome with former ITP and normal platelet count (n=1) or warm AIHA (n=2). Six had secondary ITP or warm AIHA, five had Evans syndrome with no underlying immunodeficiency (including one patient with primary antiphospholipid syndrome) and one patient had a history of Hodgkin lymphoma that was cured (#5) 9 years before the onset of ITP/Glanzmann syndrome (Table 1). At the time of starting daratumumab treatment, the median disease duration was 84.5 months (range, 18– 174). The median number of previous therapies was 6 [range, 6-11]. No patients had shown a response to their last course of rituximab (except one patient who relapsed at 3 months) and five had undergone splenectomy (Table 1). For the five patients with ITP, the median platelet count was 11x109/L (range, 0–21x109/L), and all patients had skin and/or mucosal bleeding within the month before daratumumab. Two patients (#1, #2) achieved complete responses at 4 weeks (Figure 1). Patient #1 had a history of arterial and venous antiphospholipid syndrome, with strong positivity of lupus anticoagulant and IgM antib2GP1 antibodies without IgG or anticardiolipin antibodies. Nine months after daratumumab, the patient had no remaining IgM anti-b2GP1 antibodies, and lupus anticoagulant was barely detectable. He had no recurrence of thrombosis during follow-up with ongoing vitamin K antagonist treatment. Patient #3 was dependent on corticosteroids and had no response at 4 weeks, which resulted in a transient increase in corticosteroid doses. However, he achieved long-lasting complete response even after corticosteroid discontinuation 24 weeks after starting daratumumab, suggesting a delayed effect of anti-CD38 treatment. The two remaining patients (#4, #6) showed no response after daratumumab. One 35-year-old patient (#5) had chronic ITP and, after splenectomy, developed acquired Glanzmann thrombasthenia with bleeding despite normal platelet counts. She had anti-GPIIbIIIa antibodies, and platelet aggregation studies showed no aggregation with adenosine diphosphate, epinephrine, or arachidonic acid, with reverse ristocetin agglutination. She showed no response to intravenous immunoglobulins, corticosteroids or mycophenolate mofetil and was given daratumumab, with no response for hemorrhagic symptoms. She eventually had an ITP relapse 24 weeks after starting daratumumab. For the two patients with warm AIHA, the median baseline hemoglobin level was 9.4 g/dL (range, 8.2–10.7), median reticulocyte count 174x109/L (range, 124–225 x109/L), and median bilirubin level 27 mmol/L (range, 24– 3200

30). For both patients, the haptoglobin level was <0.1 mg/L, and the median lactate dehydrogenase level was 2.8 times the normal range (range, 1.34–4.2). Both had a history of ITP but normal platelet counts at the time of the first daratumumab infusion. One patient achieved a complete response after four cycles of daratumumab but relapsed after 9 months, and one had no response after 11 cycles, despite a progressive decrease in transfusion requirement after 3 months. After a median follow-up of 24 weeks (range, 24–36) from the first infusion of daratumumab, five patients experienced at least one moderate adverse event. Three (#3, #7 and #8) had a minor reaction at the first daratumumab infusion. No further infusion-related reactions occurred afterwards. Two patients had infectious events: patient #5 had bacterial pneumonia requiring hospitalization at 4 weeks and patient #8 had COVID-19 pneumonia requiring hospitalization and convalescent plasma therapy because of persistent symptoms at 20 weeks (chronic viremia without seroconversion). The median gammaglobulin level decreased from 7.1 g/L (range, 4.8–16.2) before treatment to 4.2 g/L (range, 3.5–7.6) at week 12 and 6.1 g/L (range, 6–15.5) at week 36 (Figure 2). Hypogammaglobulinemia (i.e., gammaglobulin level <6 g/L) was observed in five out of six patients at 12 weeks (after exclusion of 2 patients who had received intravenous immunoglobulins in the 3 weeks before dosage). Although this study has some limitations, including its uncontrolled design and the heterogeneity of the patients, our results suggest that daratumumab may be effective for some patients with refractory ITP and/or warm AIHA, two conditions associated with a high mortality rate.2,3 Three of eight patients achieved a complete response and one patient a delayed response despite previous failure to respond to several treatment lines including rituximab, splenectomy and at least one immunosuppressant. One patient showed complete disappearance of antiphospholipid antibodies. Whether depletion of CD38-positive cells correlates with response and/or relapses remains an open question that should be addressed in future studies. We observed rapid response (within 1 month) among responders, suggesting that four daratumumab infusions are sufficient to induce remission in responders. However, the optimal number of infusions remains to be determined. Importantly, two patients with an initial complete response eventually relapsed at 3 and 9 months, suggesting that plasma cell reconstitution had occurred. In these patients, the ongoing autoimmune B-cell response may not have been affected by daratumumab, which targets plasma cells and spares CD38-negative B cells. Thus, combining B-cell depletion with an anti-CD20 antibody and daratumumab may impair the generation of newly formed autoreactive plasma cells and prevent relapse.15 In this particular group of immunocompromised patients, the risk of severe infection was a main concern. All but one experienced profound although transient hypogammaglobulinemia, and two had a symptomatic infection. Therefore, the risk/benefit balance of such therapy should be carefully discussed according to the patient's clinical history. In conclusion, daratumumab may provide clinical benefit in a subset of patients with ITP or warm AIHA refractory to standard therapy. However, the efficacy seems relatively modest and hypogammaglobulinemia exposes patients to a risk of infection. Future prospective trials will evaluate the use of CD38-directed monoclonal antibodies for the management of patients with immune cytopenias, such as daratumumab in ITP (NCT04703621) or isatuximab in warm AIHA (NCT04661033). haematologica | 2021; 106(12)

Letters to the Editor

Etienne Crickx,1 Sylvain Audia,2 Ailsa Robbins,3 David Boutboul,4 Thibault Comont,5 Morgane Cheminant,6 Eric Oksenhendler,4 Bertrand Godeau,1 Marc Michel1 and Matthieu Mahevas1 1 Service de Médecine Interne, Centre National de Référence des Cytopénies Auto-immunes de l’Adulte, Hôpital Henri Mondor Assistance Publique Hôpitaux de Paris, Université Paris Est Créteil, Créteil; 2Service de Médecine Interne et d’Immunologie Clinique, Centre Constitutif de Référence des Cytopénies Auto-immunes de l’Adulte, Hôpital François Mitterrand, Dijon Cedex; 3Service de Médecine Interne,Maladies Infectieuses, Immunologie Clinique, CHU Robert Debré, Reims; 4Service d'Immunologie Clinique, Hôpital Saint Louis, Paris, and Paris Université; 5Service de Médecine Interne et Immunopathologie, Institut Universitaire du Cancer Toulouse – Oncopôle, Toulouse and 6Service d’Hématologie Clinique, Hôpital Necker, Paris, France Correspondence: MATTHIEU MAHÉVAS - matthieu.mahevas@aphp.fr doi:10.3324/haematol.2021.279232 Received: May 16, 2021. Accepted: July 1, 2021. Pre-published: August 5, 2021. Disclosures: no conflicts of interest to disclose. Contributions: EC and MMa designed the study and analyzed the data. EC, SA, AR, TC, DB, MC, EO, MMi, BG and MMa wrote the manuscript and included patients in the study. Acknowledgments: we thank L. Languille for logistic coordination.

References 1. Audia S, Mahévas M, Samson M, Godeau B, Bonnotte B. Pathogenesis of immune thrombocytopenia. Autoimmun Rev. 2017;16(6):620-632. 2. Mahévas M, Gerfaud-Valentin M, Moulis G, et al. Characteristics, outcome, and response to therapy of multirefractory chronic

haematologica | 2021; 106(12)

immune thrombocytopenia. Blood. 2016;128(12):1625-1630. 3. Barcellini W, Fattizzo B, Zaninoni A, et al. Clinical heterogeneity and predictors of outcome in primary autoimmune hemolytic anemia: a GIMEMA study of 308 patients. Blood. 2014;124(19):29302936. 4. Lonial S, Weiss BM, Usmani SZ, et al. Daratumumab monotherapy in patients with treatment-refractory multiple myeloma (SIRIUS): an open-label, randomised, phase 2 trial. Lancet. 2016;387(10027): 1551-1560. 5. Jain A, Gupta DK. Daratumumab for refractory warm autoimmune hemolytic anemia. Ann Hematol. 2021;100(5):1351-1353. 6. Cooling L, Hugan S. Daratumumab in combination with standard treatment for autoimmune hemolytic anemia in a pediatric patient. Transfusion. 2019;59(12):3801-3802. 7. Schuetz C, Hoenig M, Moshous D, et al. Daratumumab in lifethreatening autoimmune hemolytic anemia following hematopoietic stem cell transplantation. Blood Adv. 2018;2(19):2550-2553. 8. Blennerhassett R, Sudini L, Gottlieb D, Bhattacharyya A. Post-allogeneic transplant Evans syndrome successfully treated with daratumumab. Br J Haematol. 2019;187(2):e48-e51. 9. Even-Or E, Naser Eddin A, Shadur B, et al. Successful treatment with daratumumab for post-HSCT refractory hemolytic anemia. Pediatr Blood Cancer. 2020;67(1):e28010. 10. Driouk L, Schmitt R, Peters A, et al. Daratumumab therapy for post-HSCT immune-mediated cytopenia: experiences from two pediatric cases and review of literature. Mol Cell Pediatr. 2021;8(1):5. 11. Zaninoni A, Giannotta JA, Gallì A, et al. The immunomodulatory effect and clinical efficacy of daratumumab in a patient with cold agglutinin disease. Front Immunol. 2021;12:649441. 12. Ostendorf L, Burns M, Durek P, et al. Targeting CD38 with daratumumab in refractory systemic lupus erythematosus. N Engl J Med. 2020;383(12):1149-1155. 13. Rodeghiero F, Stasi R, Gernsheimer T, et al. Standardization of terminology, definitions and outcome criteria in immune thrombocytopenic purpura of adults and children: report from an international working group. Blood. 2009;113(11):2386-2393. 14. Michel M, Terriou L, Roudot-Thoraval F, et al. A randomized and double-blind controlled trial evaluating the safety and efficacy of rituximab for warm auto-immune hemolytic anemia in adults (the RAIHA study). Am J Hematol. 2017;92(1):23-27. 15. Dossier C, Prim B, Moreau C, et al. A global antiB cell strategy combining obinutuzumab and daratumumab in severe pediatric nephrotic syndrome. Pediatr Nephrol. 2021;36(5):1175-1182.

3201

Letters to the Editor

Biallelic mutations in the SARS2 gene presenting as congenital sideroblastic anemia Mitochondrial disorders (MID) are a clinically heterogeneous group of metabolic disorders characterized by impaired mitochondrial function. They can be caused by mutations in either mitochondrial DNA (mtDNA) or nuclear DNA encoding mitochondrial proteins. In addition to energy deficiency, consequences of mitochondrial dysfunction can include excessive reactive oxygen species (ROS) production, defect in heme synthesis, aberrant calcium handling, and apoptosis dysregulation, all of which contribute to the pathogenesis of MID.1 MID are usually multisystemic disorders, affecting various organs but only a few of them such as congenital sideroblastic anemias (CSA) display major erythroid abnormalities. CSA is a series of rare, heterogeneous disorders characterized by pathological iron accumulation in the mitochondria of erythroblasts and presence of ring sideroblasts in the bone marrow at varying degrees.2,3 So far, nearly two-thirds of CSA cases have been attributed to a mutation in a specific gene or genes.4 Here we report the original case of a patient presenting with CSA, in which we identified biallelic mutations of the seryl-tRNA synthetase 2 (SARS2) gene. The affected child was a first-born son to a non-consanguineous healthy couple with no neonatal antecedents and no familial medical history. His younger sister is healthy (Figure 1A). He presented at the age of 3 years with chronic

renal failure associated with cerebellar atrophy, hypertension, failure to thrive, polyuria and polydipsia. Kidney biopsy revealed a tubulointerstitial nephritis. Metabolic workup showed hyperlactatemia and abnormal acylcarnitine profile, findings suggesting of a MID. Blood tests concomitantly revealed a non-regenerative anemia associated with mild hemolysis, which was refractory to erythropoietin (EPO) treatment and required monthly blood transfusions. Negative direct antiglobulin test ruled out the presence of an immune disorder. No abnormal hemoglobin (Hb) was detected on electrophoresis profile and the levels of pyruvate kinase and G6PD enzymes were in normal values. Peripheral blood smear showed the presence of about 20% spherocytes (Online Supplementary Figure S1A). Accordingly, osmotic gradient ektacytometry revealed reduced erythrocyte deformability, supporting the diagnosis of hereditary spherocytosis (Online Supplementary Figure S1B). Bone marrow aspiration showed no signs of dyserythropoiesis but a small contingent of ring sideroblasts (5%) was observed (Figure 1B). Iron and transferrin serum levels were in normal ranges but ferritin level and transferrin saturation were elevated. Patient’s laboratory data are detailed in the Online Supplementary Table S1. Patient’s condition progressively deteriorated leading to terminal end-stage renal disease and very severe anemia requiring weekly blood transfusions. He ultimately died of pulmonary hypertension at the age of 5 years. Whole exome sequencing performed on the proband and

Figure 1. Clinical and molecular findings of the proband. (A) Pedigree and segregation of the pathogenic SARS2 variants in the reported family. (B) Bone marrow aspiration from the proband showing the presence of ring sideroblasts on iron stain. (C) Sanger sequencing validation of SARS2 c.1031 G>A (p.R344Q) and c.1205 G>A (p.R402H) mutations detected by next generation sequencing. Red arrows indicate the position of the nucleotide’s substitution.

3202

haematologica | 2021; 106(12)

Letters to the Editor

his parents identified a de novo heterozygous nonsense mutation (c. 6037C>T p.Q2013X) in the SPTB gene. This gene encodes a member of the spectrin family, involved in the stability of erythrocytes membranes. Mutations in this gene have been associated with hereditary spherocytosis and hemolytic anemia.5 While most symptoms of the proband were consistent with hereditary spherocytosis, the non-regenerative anemia, together with the presence of ring sideroblasts and extra-hematological signs remained unexplained, suggesting that additional mutations were involved in his phenotype. According to this hypothesis, we identified in the proband two compound heterozygous missense variants (c.1031 G>A p.R344Q and c.1205 G>A p.R402H) in the SARS2 gene (cDNA accession number: NM_017827.4, transcript ID: ENST00000221431), each inherited from one of two parents. SARS2 is an ubiquitously expressed nuclear gene encoding a mitochondrial aminoacyl-tRNA synthetase (mtaaRS) whose function is to catalyze the specific ligation of serine to two mitochondrial tRNA isoacceptors: tRNASer(AGY) and tRNASer(UCN).6 Both SARS2 variants were reported to be rare in public human genetic variant databases (GnomAD_exomes MAF: T=0.000012 and T=0.000008 respectively). As they affect highly conserved amino acid residues located in the core catalytic domain of the protein and are predicted to be damaging by in silico softwares, biallelic SARS2 mutations identified in the proband were very likely to result in loss-of function.

Sanger validation of the mutations in SPTB and SARS2 are shown in the Online Supplementary Figure S1C and Figure 1C. Missense mutations in the SARS2 gene (p.V223M, p.D390G and p.R402H) have been reported so far in six individuals from four distinct families who presented a specific MID characterized by hyperuricemia, pulmonary hypertension, renal failure in infancy and alkalosis and designated as HUPRA syndrome.7–9 Another SARS2 variant (homozygous splicing mutation in exon 14) was reported in a single case suffering from progressive spastic paresis instead of HUPRA syndrome,10 in agreement with the known phenotypic heterogeneity of mutations involving mitochondrial machinery. Interestingly, a careful review of the published cases of HUPRA syndrome revealed that five of six were anemic with a Hb level varying from 4.8 to 9 g/dL (Online Supplementary Table S2). However, it appears that anemia associated with HUPRA syndrome has never been investigated mechanistically. We therefore hypothesized that SARS2 plays an important role in erythropoiesis. As all SARS2 variants associated with HUPRA syndrome were expected to result in loss-of function,7–9 we used an in vitro SARS2 knockdown model to explore the consequences of SARS2 inactivation on erythropoiesis. In a widely used primary cell culture system that recapitulates erythropoiesis, CD34+ were isolated from unrelated cord blood and expanded for 5 days in a medium contain-

Figure 2. SARS2 depletion in early erythroblasts results in decreased proliferation rate and differentiation arrest due to increased apoptosis. (A) At day 5 (D5) of the first phase of culture, primary erythroid progenitors were transduced with scramble short hairpin RNA (shCTRL) or short hairpin RNA (shRNA) specifically targeting SARS2 (shSARS2). 48 hours later, cell proliferation of transduced cells was monitored for 150 hours by real-time videomicroscopy using the Incucyte® system. Results are represented as the fold increase of cell confluence normalized to T0. Error bars represent standard deviation (SD) from mean of 3 technical replicates. Data are representative of 3 independent experiments. (B) At the indicated days of the second phase of culture, shCTRL- and shSARS2-transduced cells were stained with an antibody directed against the GPA erythroid marker. The percentage of GPA positive cells was then assessed by flow cytometry. Error bars represent SD from mean of 3 independent experiments. (C) At day 6 (D6) of the second phase of culture, shCTRL- and shSARS2-transduced cells were stained with an antibody directed against Annexin V and with propidium iodide (PI). The percentage of apoptotic cells, defined as the percentage of the Annexin V positive cells amongst the PI negative population was assessed by flow cytometry. Error bars represent SD from mean of 3 independent experiments. P-values are determined by two-tailed t-test. ns: not significant; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

haematologica | 2021; 106(12)

3203

Letters to the Editor

ing 100 ng/mL interleukin 6 (IL-6), 10 ng/mL IL-3 and 50 ng/mL SCF in Iscove’s Modified Dulbecco’s medium supplemented with 15% BIT 9500. Cells were then transduced with green fluorescent protein (GFP) lentiviruses expressing scramble short hairpin RNA (shRNA) or shRNA directed against SARS2 (knockdown efficiency of 70%). Forty-eight hours later, GFP positive cells were sorted and cultured with the differentiation medium containing 10 ng/mL IL-3, 50 ng/mL SCF and 2 U/mL of EPO to undergo terminal erythroid differentiation. SARS2 depletion in erythroid progenitors resulted in drastic decrease of cell proliferation (Figure 2A) and delayed erythroid differentiation (Figure 2B). In addition, an increased apoptotic rate was observed in SARS2-depleted cells (Figure 2C). As SARS2 is involved in mitochondrial protein synthesis, we wanted to evaluate whether

increased apoptosis was related to mitochondrial dysfunction. Accordingly, we measured mitochondrial membrane potential by flow cytometry and assessed the expression levels of activated caspases 3 and 9 by western blot. We observed a profound mitochondrial depolarization in a subset of SARS2-depleted cells (Figure 3A) associated with a cleavage of both caspase 3 and caspase 9 (Figure 3B). Those results indicate that the increased apoptosis reported in SARS2-depleted erythroblasts resulted at least partly, from mitochondrial perturbations. While mutations in mt-aaRS are expected to affect mitochondrial translation system, levels of COXII and ATP8, two mtDNA-encoded subunits of the mitochondrial respiratory chain complex IV and V respectively were found not to be affected in SARS2depleted cells (Figure 3C). In addition, no excessive production of ROS was detected (Online Supplementary Figure

Figure 3. Increased apoptosis in SARS2-depleted cells is mitochondria-mediated. (A) At day 4 of the second phase of culture, mitochondrial membrane potential was assessed by flow cytometry in scramble short hairpin RNA (shCTRL) and shSARS2-transduced cells after staining with the potentiometric DilC1(5) fluorescent dye. The percentage of depolarizing cells, defined as the Dilc1(5)low population was then assessed by flow cytometry. Data are representative of 2 independent experiments. (B) Western blot showing SARS2, cleaved caspase 3 and cleaved caspase 9 expression levels in shCTRL- and shSARS2-transduced cell at day 4 of the second phase of culture. Proteins levels were compared to b-actin expression. Data are representative of 2 independent experiments. (C) Representative western blot showing SARS2, COXII and ATP8 expression levels in shCTRL- and shSARS2-transduced cells at day 4 of the second phase of culture. Proteins levels were compared to b-actin expression.

3204

haematologica | 2021; 106(12)

Letters to the Editor

S1D). Nonetheless, the absence of oxidative phosphorylation or mitochondrial protein synthesis defect observed in the context of SARS2 inactivation is in accordance with the lack of clear biochemical phenotypes previously reported in skin fibroblasts and myoblasts from most of mutant mtaaRS patients.6 Despite their common biochemical role in mitochondrial protein synthesis, mutations of the mt-aaRS family have been correlated to a wide range of clinical manifestations. To date, three mt-aaRS defects affecting the tyrosyl-tRNA synthetase 2 (YARS2),11 the leucyl-tRNA synthetase 2 (LARS2)12 and the isoleucyl-tRNA synthetase 2 (IARS2)3 genes have been associated with erythroid disorders corresponding to CSA. It is now well established that mitochondria are involved in erythroid cells homeostasis through multiple ways, such as heme synthesis, apoptosis, and cell differentiation following transient caspase activation. Alteration of these mitochondrial functions can therefore lead to a variety of erythroid disorders.13 For instance, it has been reported that alteration of physiological pathways involving caspases and mitochondria in myelodysplastic syndromes leads to abnormal activation of a mitochondria-mediated apoptotic pathway in erythroid cells and therefore to an ineffective erythropoiesis.14 Accordingly, our results suggest that the non-regenerative anemia observed in the SARS2-mutated patients may be related to an exacerbation of the mitochondria-mediated apoptotic pathway. However, the reason why only few defects in mt-aaRS including SARS2 result in perturbed erythropoiesis remains to be elucidated. With regard to tissue specificity of the different mt-aaRS mutations, one can hypothesize that non-translational functions only crucial in some tissues or cells are specific to each mt-aaRS. It is also possible that the remaining residual activity of some mt-aaRS is sufficient to maintain mitochondrial function in most cell types but not in specific tissues. In summary, our results strongly suggest that SARS2 is a new gene involved in CSA, although concomitant hereditary spherocytosis, through chronic stress erythropoiesis, has potentially exacerbated the hematological phenotype in the present case. We therefore consider that SARS2 should be added to the list of genes which have to be explored in patients presenting with sideroblastic anemia in early childhood. Our study also emphasizes the relevance of performing next generation sequencing in patients presenting with severe undiagnosed anemia associated with extra hematological signs to identify new key actors of physiological erythropoiesis. Elia Colin,1 Geneviève Courtois,1 Chantal Brouzes,2 Juliette Pulman,3 Marion Rabant,4 Agnès Rötig,3 Hélène Taffin,5 Mathilde Lion-Lambert,5 Sylvie Fabrega,6 Lydie Da Costa,7 Mariane De Montalembert,8 Rémi Salomon,5 Olivier Hermine9 and Lucile Couronné10 1 Laboratory of Cellular and Molecular Mechanisms of Hematological Disorders and Therapeutic Implications, INSERM U1163, Imagine Institute, University of Paris, Laboratory of Excellence GR-Ex; 2Hematology Laboratory, Hôpital Necker-Enfants Malades, Assistance publique-Hôpitaux de Paris (AP-HP); 3Laboratory for Genetics of Mitochondrial Disorders, INSERM U1163, Imagine Institute, University of Paris; 4Department of Pathology, Hôpital Necker - Enfants Malades, Assistance Publique-Hôpitaux de Paris, University of Paris; 5Department of Pediatric Nephrology, MARHEA, Hôpital Necker - Enfants Malades, Assistance Publique-Hôpitaux de Paris (AP-HP); 6VVTG platform, SFR Necker; 7Hematology Laboratory, Robert Debré Hospital, Assistance Publique-Hôpitaux de Paris (AP-HP), University of Paris, Laboratory of Excellence GR-Ex; haematologica | 2021; 106(12)

Department of General Pediatrics and Pediatric Infectious Diseases, Reference Center for Sickle Cell Disease, Hôpital Necker - Enfants Malades, Assistance Publique-Hôpitaux de Paris (AP-HP), Université de Paris; 9Hematology Department, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris (APHP), Laboratory of Cellular and Molecular Mechanisms of Hematological Disorders and Therapeutical Implications, INSERM U1163, Imagine Institute, University of Paris, Laboratory of Excellence GR-Ex and 10 Laboratory of Onco-Hematology, Hôpital Necker - Enfants Malades, Assistance Publique - Hôpitaux de Paris (APHP), Laboratory of Cellular and Molecular Mechanisms of Hematological Disorders and Therapeutical Implications, INSERM U1163, Imagine Institute, University of Paris, Laboratory of Excellence GR-Ex, Paris, France Correspondence: ELIA COLIN - elia.colin@hotmail.fr doi:10.3324/haematol.2021.279138 Received: May 11, 2021. Accepted: July 26, 2021. Pre-published: August 19, 2021. Disclosures: no conflicts of interest to disclose. Contributions: EC, OH and LC designed the study; HT, ML-L, MdM, LDC and RS were responsible for patient management and collection of clinical and laboratory data; MR conducted histopathological analyses; EC, GC and JP designed experiments; EC performed experiments; EC and LC analyzed data and co-wrote the manuscript. All the authors reviewed the manuscript and approved the final version. Funding: this study was supported by grants from Laboratory of Excellence GR-Ex, reference ANR-11-LABX-0051. The labex GREx is funded by the IdEx program “Investissements d’avenir” of the French National Research Agency, reference ANR-18-IDEX-0001.

References 1. El-Hattab AW, Scaglia F. Mitochondrial cytopathies. Cell Calcium. 2016; 60(3):199-206. 2. Long Z, Li H, Du Y, Han B. Congenital sideroblastic anemia: advances in gene mutations and pathophysiology. Gene. 2018;668:182-189. 3. Barcia G, Pandithan D, Ruzzenente B, et al. Biallelic IARS2 mutations presenting as sideroblastic anemia. Haematologica. 2020;106(4):1220-1225. 4. Ducamp S, Fleming MD. The molecular genetics of sideroblastic anemia. Blood. 2019;133(1):59-69. 5. He B-J, Liao L, Deng Z-F, et al. Molecular genetic mechanisms of hereditary spherocytosis: current perspectives. Acta Haematol. 2018;139(1):6066. 6. Diodato D, Ghezzi D, Tiranti V. The mitochondrial aminoacyl tRNA synthetases: genes and syndromes. Int J Cell Biol. 2014;2014:787956. 7. Belostotsky R, Ben-Shalom E, Rinat C, et al. Mutations in the mitochondrial Seryl-tRNA synthetase cause hyperuricemia, pulmonary hypertension, renal failure in infancy and alkalosis, HUPRA syndrome. Am J Hum Genet. 2011;88(2):193-200. 8. Rivera H, Martín-Hernández E, Delmiro A, et al. A new mutation in the gene encoding mitochondrial seryl-tRNA synthetase as a cause of HUPRA syndrome. BMC Nephrol. 2013;14:195. 9. Zhou Y, Zhong C, Yang Q, et al. Novel SARS2 variants identified in a Chinese girl with HUPRA syndrome. Mol Genet Genomic Med. 2021; 9(4):e1650. 10. Linnankivi T, Neupane N, Richter U, Isohanni P, Tynismaa H. Splicing defect in mitochondrial seryl-tRNA synthetase gene causes progressive spastic paresis instead of HUPRA syndrome: Hum Mutat. 2016;37(9):884888. 11. Riley LG, Cooper S, Hickey P, et al. Mutation of the mitochondrial tyrosyl-tRNA synthetase gene, YARS2, causes myopathy, lactic acidosis, and sideroblastic anemia - MLASA syndrome. Am J Hum Genet. 2010; 87(1):52-59. 12. Riley LG, Rudinger-Thirion J, Schmitz-Abe K, et al. LARS2 variants associated with hydrops, lactic acidosis, sideroblastic anemia, and multisystem failure. JIMD. 2015;28:49-57. 13. Fontenay M, Cathelin S, Amiot M, Gyan E, Solary E. Mitochondria in hematopoiesis and hematological diseases. Oncogene. 2006;25(34):47574767. 14. Testa U. Apoptotic mechanisms in the control of erythropoiesis. Leukemia. 2004;18(7):1176-1199.

3205

Letters to the Editor

Combined transcriptome and proteome profiling of SRC kinase activity in healthy and E527K defective megakaryocytes Megakaryocytes (MK) and platelets express high levels of different SRC family kinases including the proto-oncogene SRC, a tyrosine kinase that has been extensively studied in platelets.1 The germline heterozygous missense variant E527K in SRC was detected in patients from three unrelated families with thrombocytopenia accompanied with bleeding symptoms, a paucity of α granules, and a variety of other variable symptoms including myelofibrosis, osteoporosis, facial dysmorphism and behavioral problems.2-4 The E527K-SRC variant resulted in increased kinase activity.2 These observations suggested an important role of SRC in megakaryopoiesis, but defects in downstream pathways because of increased kinase activity remain unknown. In this study, we further explored the effect of SRC kinase signaling in megakaryopoiesis using omics approaches. For RNA sequencing (RNAseq), CD34+ hematopoietic stem cells (HSC) were isolated from healthy controls before transduction with wild-type SRC (WT-SRC) and E527K-SRC lentiviral vectors in triplicate and differentiation to MK as described2 (Figure 1A). Day 12 MK were analyzed by flow cytometry for expression markers CD41 (ITGA2B) and CD42 (GP9) (Figure 1B). Though E527K-SRC cultures contained less CD41/CD42 doublepositive MK, there was no difference in CD41 positive immature MK when compared to the WT condition. RNA was extracted from the complete cell population as the aim was to detect differences in genes regulating MK maturation and differentiation. A total of six WT-SRC and six E527K-SRC RNA samples were used for RNAseq to identify differentially expressed genes (DEG). Principal component analysis performed on the RNAseq datasets showed obvious clustering of WT-SRC and E527K-SRC MK samples according to condition (Figure 1C). The SRC gene for WT-SRC samples was covered by 8,661±901 read counts while only 2,988±410 read counts were detected for E527K-SRC MK samples but 80,83±8,06% of these reads contained the E527K variant. Relative expression levels of all significant DEG in E527K-SRC versus WT-SRC samples were visualized in a heatmap (Figure 1D). A total of 852 significant (false discovery rate [FDR]<0.05 and |log2FC|>1) DEG were detected (see the Online Supplementary Table S1 for the full dataset), of which 369 upregulated genes and 483 downregulated genes as visualized in the volcano plot (Figure 1E). Reactome pathway analysis showed that the downregulated DEG were enriched in pathways like platelet activation, signaling and aggregation, interferon type I or α/β signaling, platelet degranulation, response to elevated platelet cytosolic Ca2+, and RUNX1 regulates genes involved in megakaryocytic differentiation and platelet function as top five (Figure 1F). On the other hand, the pathway analysis for upregulated DEG in E527K-SRC MK included interleukin-10 signaling, degradation of the extracellular matrix, chemokine receptors bind chemokines, collagen degradation and signaling by interleukins as top five pathways (Figure 1F). Interestingly, Barozzi et al. recently described that overactive SRC kinase may affect proplatelet formation by affecting the interaction of MK with the extracellular matrix.4 Shotgun proteomics was next used to analyze protein expression in transduced WT-SRC and E527K-SRC differentiated MK at day 12 (Figure 1A). Principal component analysis performed on the proteomics dataset showed a 3206

clear separation between WT-SRC and E527K-SRC MK samples, with 66% of total variance explained by the first principal component and 13% explained by the second principal component (Figure 2A). Statistical analysis of WT-SRC and E527K-SRC MK proteomes detected 142 significant (FDR<0.01) differentially expressed proteins (DEP) (see the Online Supplementary Table S2 for the full dataset), of which 43 were upregulated and 99 downregulated as shown in the volcano plot (Figure 2B). Similar to the RNAseq results, SRC was significantly downregulated in E527K-SRC MK (Figure 2B). Interestingly, the Reactome pathway analysis again identified an enrichment of downregulated DEP in the interferon pathways, but no platelet-related pathways were detected. By comparing the set of DEG and DEP, 20 upregulated and 24 downregulated joined DEG/DEP sets were detected in E527K-SRC MK. Reactome pathway analysis of these joined DEG/DEP identified an enrichment of interferon α/b signaling, interferon signaling and cytokine signaling in the immune system as significant (Figure 2C). This unbiased approach to identify pathways related to hyperactive SRC signaling in MK proposed decreased interferon α/b signaling as an unexpected player. Interferon signaling is neither well studied in megakaryopoiesis nor in SRC kinase signaling, encouraging our further studies in the immortalized megakaryocyte cell line (imMKCL) and in MK derived from a patient with the E527K-SRC variant. The top five downregulated interferon-stimulated genes (ISG) IFIT1, MX1, OAS2, IFIT3 and ISG15 (labeled in Figures 1E, 2B and 2C) were selected for these validation studies. Interestingly, SRC protein expression was significantly increased from day 0 (non-differentiated) to days 2 and 4 differentiated imMKCL (Figure 3A; Online Supplementary Figure S1A and B). This increase was confirmed by quantitative real-time ploymerase chain reaction (qRT-PCR) analysis (Figure 3B). The expression of ISG showed increased levels of IFIT1, MX1, and OAS2 during MK differentiation while no significant differences were detected for IFIT3 and ISG15 (Figure 3B). The expression of RUNX1 in contrast to that of ITGA2B remained unchanged during megakaryopoiesis (Figure 3B). Genetically modified imMKCL transduced with WTSRC and E527K-SRC lentiviral vectors showed similar results as detected in the RNAseq study. A significant downregulation of all ISG was present in E527K-SRC MK compared to WT-SRC MK (Online Supplementary Figure S1C). Adding interferon α (IFNα) to the medium resulted in higher expression of ISG but no difference in reactivity was detected between E527K-SRC and WT-SRC MK (Online Supplementary Figure S1C). In order to exclude the possibility that the difference in ISG expression between WT-SRC and E527K-SRC would be due to the difference in SRC expression between those conditions or the lentiviral transduction, transfection experiments were performed. In addition, a comparison with MK transfected with an empty vector was missing from previous experiments. Therefore, imMKCL were transfected with pSecTag2 empty, WT-SRC and E527K-SRC expression vectors and the expression of SRC and the different ISG was quantified in day 2 MK (Figure 3C). Compared to the condition with the empty vector, MK with WT-SRC and E527K-SRC now expressed similarly elevated SRC levels. Interestingly, both WT-SRC and E527K-SRC MK expressed significantly lower ISG levels compared to MK transfected with the empty vector (Figure 3C). Overexpression of SRC in the undifferentiated imMKCL affected ISG expression during megakaryopoiesis but this experiment also pointed out that the detection of the ISG haematologica | 2021; 106(12)

Letters to the Editor

1. Legend on the following page.

haematologica | 2021; 106(12)

3207

Letters to the Editor

Figure 1. Bulk RNA sequencing of WT-SRC and E527K-SRC megakaryocytes. (A) A total of three hematopoietic stem cell (HSC) pools were isolated from peripheral blood from 15 healthy controls and cultured in HSC amplification medium (with cytokine cocktail 100). On day 1 after isolation, HSC were transduced in triplicate with either wild-type SRC (WT-SRC) or E527K-SRC lentiviral vectors. On day 2, differentiation to megakaryocytes (MK) was initiated upon addition of the cytokines TPO, SCF and IL1b. Day 12 mature MK were collected for bulk RNA sequencing and shotgun proteomics. (B) Flow cytometry of day 12 MK stained for CD41 (forward scatter) and CD42 (side scatter). The % of CD41+ and CD41/CD42+ MK are shown for WT-SRC and E527K-SRC cells as mean and standard deviation for 3 cultures. (C) Principal component analysis of 6 WT-SRC and 6 E527K-SRC MK samples used for RNA sequencing. (D) Heatmap visualizing the 852 differentially expressed genes (DEG) identified in E527K-SRC vs. WT-SRC MK samples (red, upregulated genes; blue, downregulated genes). (E) Volcano plot showing the 852 DEG detected in E527K-SRC MK organized according to log2(fold-change [FC]) and -log10(false discovery rate [FDR]). (red, upregulated genes with FDR<0.05 and log2FC>1; blue, downregulated genes with FDR<0.05 and log2FC<-1). (F) The 852 DEG were used for Reactome pathway analysis. Top 5 downregulated (blue) and upregulated (red) pathways are shown as arranged by their -log(FDR) value.

Figure 2. Shotgun Proteomics of WT-SRC and E527K-SRC megakaryocytes. (A) Principal component analysis of three wild-type SRC (WT-SRC) and three E527KSRC megakaryocyte (MK) samples used for shotgun proteomics analysis. (B) Volcano plot showing the 141 differentially expressed proteins (DEP) detected in E527K-SRC MK organized according to log2(fold-change [FC]) and -log10(false discovery rate [FDR]). (red, upregulated proteins with FDR<0.01 and log2FC>1; blue, downregulated proteins with FDR<0.01 and log2FC<-1) (C) Top 3 enriched downregulated pathways in E527K-SRC MK from Reactome analyses using differentially expressed genes (DEG), DEP and the combined dataset (left panel). Genes selected for validation are depicted in blue. FDR values for these five most significant interferon-stimulated genes (ISG) are shown as found in the RNA sequencing and proteomics analyses (right panel).

3208

haematologica | 2021; 106(12)

Letters to the Editor

Figure 3. Legend on following page.

haematologica | 2021; 106(12)

3209

Letters to the Editor

Figure 3. Validation of interferon-stimulated genes in immortalized megakaryocyte cell line (imMKCL). (A) Immunoblot analysis of total SRC expression on day 0, 2 and 4 megakaryocytes (MK) (left). GAPDH was used as loading control. Quantification of SRC expression in triplicate immunoblots and statistical analysis using one-way ANOVA with multiple comparisons (right). (B) Quantitative real-time polymerase chain reaction (qRT-PCR) validation showing relative expression of SRC, IFIT1, MX1, OAS2, IFIT3 and ISG15 on day 0, 2 and 4. Statistical analysis was performed using one-way ANOVA with multiple comparisons. (C) qRT-PCR validation showing relative expression of SRC, IFIT1, MX1, OAS2, IFIT3 and ISG15 on day 2 MK transfected with empty vector, WT-SRC or E527K-SRC in pSecTagHygro expression vector (triplicated transfection experiment). Statistical analysis was performed using one-way ANOVA with multiple comparisons. (D) qRT-PCR validation showing relative expression of SRC, RUNX1, IFIT1, MX1, OAS2, IFIT3 and ISG15 in MK from a control and the patient carrying the E527K SRC variant (3 technical repeats). Statistical analysis was performed using one-way ANOVA with multiple comparisons.

in the RNAseq dataset could not be explained by the lower SRC expression in E527K-SRC compared to WTSRC MK. Though the difference between WT-SRC and E527K-SRC MK in the transfection experiment was only significantly different for IFIT3 expression, the difference for the other ISG was always more pronounced between E527K-SRC and the condition with the empty vector than when comparing WT-SRC with the empty vector. This difference could be due to the fact that these expression studies were performed on day 2 immature imMKCL cells while the omics was performed on day 12 mature MK. Finally, expression studies were validated in HSC-derived day 12 mature MK from a patient carrying E527K-SRC3 and a control. No significant difference in SRC and RUNX1 expression could be detected while patient MK showed significantly decreased expression of all ISG except OAS2 (Figure 3D). The downregulation was most pronounced for IFIT1 and MX1, the two genes that were the most significantly downregulated in the RNAseq dataset (Figure 2C). The combined omics approach detected IFNα/b signaling as most significantly downregulated pathway associated with E527K-SRC hyperactivity. This was an unexpected finding because interferons are generally considered to be negative regulators of cellular proliferation and maturation.5 As qRT-PCR data showed no evidence for an altered response towards IFNα, the difference in ISG expression in E527K-SRC is due to a downstream effect. The proteomics (but not the RNAseq) data showed that STAT1 and STAT2 were significantly downregulated in E527K-SRC MK (Online Supplementary Table S2). Further studies must be undertaken to pinpoint the exact place of SRC in this pathway. In chronic myeloid leukemia, the fusion gene BCR-ABL is the result from the translocation between chromosomes 9 and 22. Similar to E527K-SRC, BCR-ABL is an overactive tyrosine kinase. Studies showed that BCR-ABL in hematopoietic cells caused the transcriptional suppression of ISG such as ISG15, IRF1, IRF9 and IFIT1,7 which resulted in impaired IFNα-mediated protection against viral infection and reversal of IFNα-dependent growth suppression, thereby promoting malignant transformation.7 As downregulated ISG were also present in E527K-SRC MK, this points to a clear similarity between both overactive kinases, SRC and BCR-ABL. The suppression of ISG in BCR-ABL cells is reflected by the suppression of JAK-STAT pathway components, including STAT1.7 It is also interesting that previous proteomic studies using inducible pluripotent stem cell-derived MK from a patient with the GFI1B Q287* variant showed low expression levels of STAT1, MX1, IFIT1, IFIT3 and OAS2, similar to our findings.8 This study hypothesizes that the underlying pathway would be a failure of IFNg to activate its target genes via STAT1 although this was not experimentally studied. Of note, similar to SRC deficiency, GFI1B defects result in thrombocytopenia, α granule deficiency and myelofibrosis. In conclusion, we here describe that interferon signaling plays a role during megakaryopoiesis by acting downstream of SRC signaling. However, the exact mechanisms 3210

of how SRC can change ISG to influence megakaryopoiesis still remain unknown. Lore De Kock,1 Fabienne Ver Donck,1 Chantal Thys,1 Anouck Wijgaerts,1 Koji Eto,2,3 Chris Van Geet1 and Kathleen Freson1 1 Center for Molecular and Vascular Biology, Department of Cardiovascular Sciences, University of Leuven, Leuven, Belgium; 2 Department of Clinical Application, Center for iPS Cell Research and Application, Kyoto University, Kyoto, Japan and 3Department of Regenerative Medicine, Chiba University Graduate School of Medicine, Chiba, Japan Correspondence: KATHLEEN FRESON - kathleen.freson@kuleuven.be doi:10.3324/haematol.2021.279248 Received: May 17, 2021. Accepted: July 27, 2021. Pre-published: August 5, 2021. Disclosures: no conflicts of interest to disclose. Contributions: LDK analyzed the results, performed the experiments, and wrote the manuscript; LDK and FVD performed statistical analyses; AW contributed to the sample preparation for RNAseq; FVD analyzed the RNA sequencing and shotgun proteomics data and performed the pathway analysis; CT performed imMKCL experiments; KE provided differentiation protocol and imMKCL cells; CVG studied E527K defective patients; KF designed the study and analysis plan and co-wrote the manuscript. Funding: this work was supported by KULeuven BOF grant C14/19/096, FWO grant G072921N and research grants from Novo Nordisk and Swedish Orphan Biovitrum AB (SOBI). Data sharing statement: RNA sequencing data will be released in European Genome-phenome Archive (EGA).

References 1. De Kock L, Freson K. The (Patho)biology of src kinase in platelets and megakaryocytes. Med. 2020;56(12):1-11. 2. Turro E, Greene D, Wijgaerts A, et al. A dominant gain-of-function mutation in universal tyrosine kinase SRC causes thrombocytopenia, myelofibrosis, bleeding, and bone pathologies. Sci Transl Med. 2018; 8(328):328ra30. 3. De Kock L, Thys C, Downes K, et al. De novo variant in tyrosine kinase SRC causes thrombocytopenia: case report of a second family. Platelets. 2019;30(7):931-934. 4. Barozzi S, Di Buduo CA, Marconi C, et al. Pathogenetic and clinical study of a patient with thrombocytopenia due to the p.E527K gainof-function variant of SRC. Haematologica. 2020;106(3):918-922. 5. Pestka S, Langer JA, Zoon KC, Samuel CE. Interferons and their actions. Annu Rev Biochem. 1987;(57):727-777. 6. Couldwell G, Machlus KR. Modulation of megakaryopoiesis and platelet production during inflammation. Thromb Res. 2019; 179:114-120. 7. Katsoulidis E, Sassano A, Majchrzak-Kita B, et al. Suppression of interferon (IFN)-inducible genes and IFN-mediated functional responses in BCR-ABL-expressing cells. J Biol Chem. 2008; 283(16): 10793-10803. 8. Van Oorschot R, Hansen M, Koornneef JM, et al. Molecular mechanisms of bleeding disorderassociated GFI1BQ287* mutation and its affected pathways in megakaryocytes and platelets. Haematologica. 2019;104(7):1460-1472.

haematologica | 2021; 106(12)

Letters to the Editor

Melphalan dose intensity for autologous stem cell transplantation in multiple myeloma High-dose chemotherapy (HDC) and autologous stem cell transplantation (ASCT) remains the standard of care upfront treatment for transplant-eligible multiple myeloma (MM) patients.1,2 High-dose melphalan 200 mg/m2 (Mel200) is the standard preparative regimen for MM.1 Over the past two decades, several prospective clinical trials compared Mel200 to more intense preparative regimens in an attempt to improve efficacy.3 Generally, none of these regimens showed an overall survival (OS) advantage over Mel200. However, some reported superior progression-free survival (PFS),4 albeit with higher rates of regimen-related toxicity. Furthermore, older and frail patients are not candidates for more intense regimens. Reduced-dose melphalan is frequently used as an alternative for older patients and those considered unfit for Mel200 because of frailty or medical comorbidities. Melphalan 100 mg/m2 (Mel100) was shown to be less effective compared to Mel200, albeit better tolerated.5,6 An intermediate dose of melphalan 140 mg/m2 (Mel140) has been widely used in older patients and those with significant comorbidities. Several groups reported feasibility of Mel140, particularly for older patients and those with renal impairment,7,8 however with conflicting outcomes when compared to Mel200.9-11 A recent report by the European Society for Blood and Marrow Transplantation (EBMT) showed similar outcomes between Mel140 and Mel200, except for a subset of patients with suboptimal responses to pretransplant induction therapy.12 In this study, we compared the safety and efficacy of Mel140 versus Mel200 in a recent cohort of MM patients who received an ASCT at our institution. We included all consecutive adult patients, who were 18 years or older, with newly diagnosed MM who underwent upfront ASCT consolidation and received single agent HDC of Mel140 or Mel200. Primary endpoints

were PFS and OS, computed from date of transplant. Secondary endpoints included cumulative incidence of relapse (CIR), non-relapse mortality (NRM), and toxicities. OS and PFS estimates for each melphalan group were obtained from Cox regression models, adjusting for the selected variables. We identified 911 eligible patients between January 2010 and December 2015, with a median age of 62 (range, 31-82) years. Ninety-seven (11%) received Mel140 and 814 (89%) received Mel200. Patient and disease characteristics are summarized in Table 1. Patients in the Mel140 group had significantly higher rates of hematopoietic cell transplantation-specific comorbidity index (HCT-CI) >3 and were significantly older. Furthermore, a higher proportion of patients in the Mel140 group had Karnofsky performance status (KPS) <90, International Staging System (ISS) stages II-III, renal impairment, and less frequently had received triplet induction therapy (Table 1). In order to correct for potential bias for the impact of disease status at transplant (i.e., partial response [PR] or worse and very good partial response [VGPR] or better), we performed 1:3 (Mel140 vs. Mel200) propensity score matching within disease status using the variables listed in Table 2. Matching was done separately for patients with ≤PR and those with ≥VGPR at ASCT. After matching, only HCT-CI scores for patients with ≥VGPR 1st transplant remained significantly different between the two melphalan groups (Table 2). The median age of the matched patients was 69 (range, 43-81) years. With a median follow-up of 54.6 (range, 0.3-112.2) months of all study patients, the median PFS and OS were 39.6 (95% confidence interval [CI]: 36.7-43.8) and 92.0 (95% CI: 85.0-101.1) months, respectively. At the time of transplant, 52% had ≥VGPR in the Mel140 group compared to 50% in the Mel200 group (P=0.75). At 3 months after transplant, the response rates were similar between the two melphalan groups (P=0.23). The complete response (CR)/stringent CR (sCR), VGPR, and PR rates were 26%, 49%, and 20%, respectively, in the Mel140 group, compared to 29%, 46%, and 22%,

Figure 1. Kaplan-Meier survival curves of multiple myeloma patients treated with either Mel140 (dashed lines) or Mel200 (solid lines) conditioning prior to autologous stem cell transplantation. (A) Progression-free survival; (B) overall survival. Mel140: intermediate dose of melphalan 140 mg/m2; Mel200: high dose of melphalan 200 mg/m2.

haematologica | 2021; 106(12)

3211

Letters to the Editor

respectively, in the Mel200 group. The median PFS and OS in the Mel140 group were 36.6 (95% CI: 26.3-43.7) and 83.0 (95% CI: 55.3-not reached) months, respectively, compared to 40.6 (95% CI: 37.0-45.1) and 92.0 (95% CI: 86.2-101.7) months, respectively, in the Mel200 group. Before adjusting for the known risk factors, Mel200 was associated with comparable PFS (hazard ratio [HR]: 0.81, 95% CI: 0.63-1.05; P=0.12), but improved OS (HR: 0.62, 95% CI: 0.44-0.87; P=0.005). After adjusting for age at ASCT, ISS, KPS, serum creatinine, disease status at ASCT, HCT-CI, and induction treatment, there were no statistically significant differences in PFS (HR 0.91, 95% CI: 0.66-1.26; P=0.58) or OS (HR: 0.79, 95% CI: 0.52-1.19; P=0.26) between the Mel140 (reference) and Mel200 groups (Figure 1).

The 1:3 matching produced a dataset of 304 patients, 76 in the Mel140 group and 228 in the Mel200 group. The response rates at 3 months after ASCT were similar to the estimates before matching (Mel140: 96%; Mel200: 96%). The CR/sCR, VGPR, and PR rates were 24%, 53%, and 20%, respectively, in the Mel140 group, compared to 30%, 49%, and 18%, respectively, in the Mel200 group. At a median follow-up of 51.8 (range: 0.3-105.1) months, the median PFS and OS for all matched patients were 38.3 (95% CI: 33.5-43.5) and 82.2 (95% CI: 73.0-97.9) months, respectively. Using fitted conditional regression models, the overall PFS (HR: 0.90, 95% CI: 0.63-1.29; P=0.56) and OS (HR: 0.72, 95% CI: 0.46-1.13; P=0.15) were not significantly different between the two melphalan groups. Likewise, there were no significant differ-

Table 1. Patient and disease characteristics by melphalan dose for all patients.

Melphalan Dose Measure, n (%) Age at ASCT < 65 years ≥ 65 years Sex Male Female ISS stage I II III Unknown Cytogenetic risk* Standard High Unknown Disease status at ASCT PR or worse VGPR or better LDH Normal High Unknown Creatinine < 2 mg/dL ≥ 2 mg/dL Unknown Karnofsky performance status < 90 ≥ 90 Unknown HCT-CI ≤3 >3 Unknown Induction treatment Conventional IMiD PI IMiD + PI Maintenance therapy No Yes

140 mg (N=97)

200 mg (N=814)

20 (21) 77 (79)

562 (69) 252 (31)

< 0.001

56 (58) 41 (42)

465 (57) 349 (43)

1.00

15 (19) 34 (43) 31 (39) 17

299 (44) 197 (29) 182 (27) 136

< 0.001

67 (74) 24 (26) 6

599 (77) 176 (23) 39

0.43

47 (48) 50 (52)

411 (50) 403 (50)

0.75

50 (85) 9 (15) 38

459 (88) 61 (12) 294

0.40

61 (68) 29 (32) 7

653 (85) 116 (15) 45

< 0.001

38 (40) 58 (60) 1

180 (23) 597 (77) 37

0.001

53 (55) 43 (45) 1

645 (81) 156 (19) 13

< 0.001

1 (1) 11 (11) 47 (48) 38 (39)

19 (2) 73 (9) 272 (33) 450 (55)

0.012

24 (25) 73 (75)

183 (22) 631 (78)

0.61

ASCT: autologous stem cell transplantation; HCT-CI: hematopoietic cell transplantation-specific comorbidity index; IMiD: immunomodulatory imide drug; ISS: International Staging System; LDH: lactate dehydrogenase; PR: partial response; PI: proteasome inhibitor; VGPR: very good partial response. *High-risk cytogenetic category was defined as patients with any of the following genetic abnormalities at diagnosis: 17p deletion, t(4;14), t(14;16), 1q gain, and del13 (only if by cytogenetics).

3212

haematologica | 2021; 106(12)

Letters to the Editor

ences in CIR (HR: 0.88, 95% CI: 0.65-1.20; P=0.43) or NRM (HR 1.39, 95% CI: 0.61-3.16; P=0.43) between the two groups. In the matched cohort with ≤PR at ASCT (n=140), there was a trend for improved OS (HR: 0.53, 95% CI: 0.28-1.02; P=0.06) for the Mel200 group, but not in PFS (HR: 1.01, 95% CI: 0.60-1.71; P=0.97), CIR (HR: 0.97, 95% CI: 0.63-1.49; P=0.90), or NRM (HR: 0.84, 95% CI: 0.30-2.37; P=0.74). In patients with ≥VGPR at ASCT (n=164), there were no statistical differences between the two melphalan groups in OS (HR: 0.93, 95% CI: 0.491.75; P=0.82), PFS (HR: 0.82, 95% CI: 0.49-1.36; P=0.43), CIR (HR: 0.81, 95% CI: 0.52-1.27; P=0.36), or NRM (HR: 2.41, 95% CI: 0.60-9.64; P=0.21). Hematological toxicities were universal as expected and there were no significant differences in either neutrophil or platelet engraftments between the two groups. Gastrointestinal toxicity of any grade was the most commonly reported side effect, observed in 91% and 96% of Mel140 and Mel200 patients, respectively (P=0.13). Mucositis rates were significantly higher with Mel200 (49%) compared to Mel140 (29%; P=0.002). The majority were grades 1-2, 26% in the Mel140 group and 48% in the Mel200 group. A higher percentage of patients in the Mel200 group had febrile neutropenia (34% vs. 25%;

P=0.20), but the documented infection rates were comparable (25% in the Mel200 group vs. 26% in the Mel140 group; P=0.88). Renal toxicity was observed more frequently in the Mel140 group (5% vs. 2%; P=0.11); this was expected because even after matching, higher proportion of patients in this group had renal impairment. Other non-hematological toxicities were less frequent. Higher rate of cardiac toxicity in the Mel140 group (8% in the Mel140 group vs. 3% in the Mel200 group; P=0.10). The rates of grades 3-4 non-hematological toxicities for the 304 matched patients were 33% for Mel140 and 22% for Mel200 (P=0.09). During the study period, 33 of the 76 patients in the Mel140 group and 89 of the 228 patients in the Mel200 group died. The most common cause of death was relapse/recurrent disease (81% for both Mel140 and Mel200; P=1.00). The 1-year NRM was 1% in the Mel140 group, compared to 3% in the Mel200 group (P=0.64). In this study, we found that a reduced dose of Mel140 is feasible for use in older MM patients and those considered ineligible for Mel200. In patients with VGPR or better at transplant, use of Mel140 had comparable response rates, PFS, and OS to Mel200. In patients with ≤PR at ASCT, there was a trend towards improved OS with Mel200.

Table 2. Patient and disease characteristics by melphalan dose for matched patients.

Measure, n (%) Age at ASCT < 65 years ≥ 65 years Sex Male Female ISS stage I II III Cytogenetic risk Standard High LDH Normal High Creatinine < 2 mg/dL ≥ 2 mg/dL Karnofsky performance status < 90 ≥ 90 HCT-CI ≤3 >3 Induction treatment Conventional IMiD PI IMiD + PI Maintenance therapy No Yes

PR or worse Mel140 Mel200 (N=35) (N=105)

VGPR or better Mel140 Mel200 (N=41) (N=123)

5 (14) 30 (86)

20 (19) 85 (81)

0.62

7 (17) 34 (83)

26 (21) 97 (79)

0.66

22 (63) 13 (37)

69 (66) 36 (34)

0.84

25 (61) 16 (39)

71 (58) 52 (42)

0.85

6 (17) 12 (34) 17 (49)

35 (33) 33 (31) 37 (35)

0.16

8 (20) 19 (46) 14 (34)

33 (27) 44 (36) 46 (37)

0.45

23 (70) 10 (30)

77 (75) 25 (25)

0.50

26 (67) 13 (33)

81 (71) 33 (29)

0.69

20 (87) 3 (13)

68 (92) 6 (8)

0.44

27 (87) 4 (13)

80 (89) 10 (11)

0.75

22 (63) 13 (37)

80 (76) 25 (24)

0.13

29 (71) 12 (29)

93 (76) 30 (24)

0.54

14 (40) 21 (60)

27 (26) 78 (74)

0.13

17 (41) 24 (59)

38 (31) 85 (69)

0.25

21 (60) 14 (40)

71 (68) 34 (32)

0.42

20 (49) 21 (51)

87 (71) 36 (29)

0.014

1 (3) 7 (20) 17 (49) 10 (29)

3 (3) 12 (11) 48 (46) 42 (40)

0.43

0 1 (2 ) 20 (49) 20 (49)

2 (2) 7 (6) 42 (34) 72 (59 )

0.40

6 (17) 29 (83)

23 (22) 82 (78)

0.64

11 (27) 30 (73)

37 (30) 86 (70)

0.84

haematologica | 2021; 106(12)

3213

Letters to the Editor

These two regimens have not been compared in randomized clinical trials, but few studies reported conflicting results.9-12 In a small non-randomized prospective study for patients older than 65 years, there was a trend for inferior PFS and OS for patients who received Mel140.9 In contrast, three recent studies reported comparable outcomes between the two regimens. The Mayo group reported no significant differences in survival estimates by melphalan dose intensity in older patients, despite observing improved post-transplant response rates with Mel200.11 The study by Katragadda et al. showed comparable response and survival outcomes for Mel140 versus Mel200.10 In a recent study by the EBMT group that provides the largest analysis of outcomes between the two melphalan doses,12 Mel200 was found to be associated with better outcomes only in patients with <PR at the time of ASCT. In contrast, Mel140 was associated with a better OS in patients transplanted in >VGPR. Our results from this large single center study further indicate that Mel140 has comparable efficacy to Mel200. Consistent with the EBMT results, we found that Mel140 can be particularly beneficial for patients who have achieved VGPR or better at ASCT. Of importance to note, the majority of patients in our study were (after matching) older (median age, 69 years) and/or with comorbidities. Hence, our findings might not be generalizable for younger fit patients. Furthermore, 46% of patients in the Mel140 group had an HCT-CI >3. Despite older age and medical comorbidities, the use of Mel140 was generally well-tolerated in this high-risk population. Mel140 was associated with less gastrointestinal toxicities and significantly less mucositis. Patients in the Mel140 group, despite matching, had higher rates of renal insufficiency and worse KPS, in addition to higher HCT-CI. All of these characteristics may have contributed to higher grades 3-4 non-hematological toxicities, when compared to Mel200. However, even with these adverse baseline characteristics, there was no increase in treatment-related mortality in patients who received Mel140. Besides the impact of dose intensity on the increased toxicity, the melphalan dosing in our study was fixed based on the body surface area, as opposed to pharmacokinetic-directed dosing, which could have potentially lead to inter-patient differences in their exposure to mephalan and hence a variability in the associated safety and efficacy profile. Our study has the predictable limitations of a retrospective analysis, including patient selection and missing data, such as revised ISS (R-ISS) stage and minimal residual disease status, the use of which were limited during the study period. In order to overcome some of these limitations, we limited the study period to patients transplanted after 2010 and used a propensity matched scoring model to account for most of the known risk factors that may influence the outcomes. Furthermore, the results were reproduced by a second analytic method using multivariable regression models fit on all patients and adjusting for the same variables that were used in the matched cohort. In conclusion, in this large, single-center matched analysis of a homogeneous patient population with MM we showed that Mel140 has comparable efficacy to Mel200, particularly in older patients and those with VGPR or better at the time of transplant.

3214

Samer A. Srour,1 Denái R. Milton,2 Qaiser Bashir,1 Yago Nieto,1 Neeraj Saini,1 May Daher,1 Jeremy Ramdial,1 Jin Im,1 Chitra Hosing,1 Ruby Delgado,1 Elisabet Manasanch,3 Hans C. Lee,3 Sheeba Thomas,3 Gregory Kaufman,3 Krina Patel,3 Uday Popat,1 Donna Weber,3 Robert Orlowski,3 Elizabeth Shpall,1 Richard E. Champlin1 and Muzaffar H. Qazilbash1 1 Department of Stem Cell Transplantation and Cellular Therapy, The University of Texas MD Anderson Cancer Center; 2Department of Biostatistics, The University of Texas MD Anderson Cancer Center, and 3Department of Lymphoma and Myeloma, The University of Texas MD Anderson Cancer Center, Houston, TX, USA Correspondence: SAMER A. SROUR - ssrour@mdanderson.org doi:10.3324/haematol.2021.279179 Received: May 8, 2021. Accepted: August 3, 2021. Pre-published: August 19, 2021. Disclosures: no conflicts of interest to disclose. Contributions: SAS and MQ conceived and designed the research; RD helped in data collection and monitoring; DRM performed statistical analysis; SAS, DRM and MQ interpreted data; SAS wrote the manuscript; and SAS, DRM, QB, YN, NS, MD, JR, JI, CH, RD, EM, HCL, ST, GK, KP, UP, DW, RO, ES, REC and MHQ critically reviewed and edited the manuscript for important intellectual content.

References 1. Mikhael J, Ismaila N, Cheung MC, et al. Treatment of multiple myeloma: ASCO and CCO joint clinical practice guideline. J Clin Oncol. 2019;37(14):1228-1263. 2. Kumar SK, Callander NS, Hillengass J, et al. NCCN guidelines insights: multiple myeloma, version 1.2020. J Natl Compr Canc Netw. 2019; 17(10):1154-1165 3. Martino M, Olivieri A, Offidani M, et al. Addressing the questions of tomorrow: melphalan and new combinations as conditioning regimens before autologous hematopoietic progenitor cell transplantation in multiple myeloma. Expert Opin Investig Drugs. 2013;22(5):619-634. 4. Bashir Q, Thall PF, Milton DR, et al. Conditioning with busulfan plus melphalan versus melphalan alone before autologous haemopoietic cell transplantation for multiple myeloma: an open-label, randomised, phase 3 trial. Lancet Haematol. 2019;6(5):e266-e275. 5. Palumbo A, Bringhen S, Bertola A, et al. Multiple myeloma: comparison of two dose-intensive melphalan regimens (100 vs. 200 mg/m2). Leukemia. 2004;18(1):133-138. 6. Palumbo A, Bringhen S, Bruno B, et al. Melphalan 200 mg/m2 versus melphalan 100 mg/m2 in newly diagnosed myeloma patients: a prospective, multicenter phase 3 study. Blood. 2010;115(10):1873-1879. 7. Badros A, Barlogie B, Siegel E, et al. Autologous stem cell transplantation in elderly multiple myeloma patients over the age of 70 years. Br J Haematol. 2001;114(3):600-607. 8. Badros A, Barlogie B, Siegel E, et al. Results of autologous stem cell transplant in multiple myeloma patients with renal failure. Br J Haematol. 2001;114(4):822-829. 9. Garderet L, Beohou E, Caillot D, et al. Upfront autologous stem cell transplantation for newly diagnosed elderly multiple myeloma patients: a prospective multicenter study. Haematologica. 2016;101(11):13901397. 10. Katragadda L, McCullough LM, Dai Y, et al. Effect of melphalan 140 mg/m2 vs 200 mg/m2 on toxicities and outcomes in multiple myeloma patients undergoing single autologous stem cell transplantation-a single center experience. Clin Transplant. 2016;30(8):894-900. 11. Muchtar E, Dingli D, Kumar S, et al. Autologous stem cell transplant for multiple myeloma patients 70 years or older. Bone Marrow Transplant. 2016;51(11):1449-1455. 12. Auner HW, Iacobelli S, Sbianchi G, et al. Melphalan 140 mg/m(2) or 200 mg/m2 for autologous transplantation in myeloma: results from the Collaboration to Collect Autologous Transplant Outcomes in Lymphoma and Myeloma (CALM) study. A report by the EBMT Chronic Malignancies Working Party. Haematologica. 2018;103(3):514521.

haematologica | 2021; 106(12)

Letters to the Editor

Predicting risk of progression in relapsed multiple myeloma using traditional risk models, focal lesion assessment with PET-CT and minimal residual disease status Novel therapeutic strategies have dramatically increased the depth of response and survival rates in multiple myeloma (MM), but the disease remains incurable in most patients because of eventual relapse.1 The timing and disease course of relapsed MM can be highly variable, and most often the presentation of the first relapse can give more information on disease biology and overall prognosis than parameters identified at diagnosis.2 The dynamic change of clinical parameters during the disease course has recently shown to significantly impact survival in MM patients,3 underscoring that prognostic models, such as fluorescence in situ hybridization (FISH), International Staging System (ISS), revised-ISS (RISS) and gene expression profiling (GEP) in addition to focal lesion (FL) assessment can be useful prognostic tools at initial diagnosis,4,5 even though they have not been fully validated in the relapse setting. Furthermore, in contrast to newly diagnosed myeloma, it is unknown whether the depth of response after salvage therapy also affects long term outcome in relapsed disease.6 This is particularly true for the achievement of minimal residual disease (MRD) negativity, a powerful prognostic tool in newly diagnosed MM,7 even though its importance in relapsed disease has started to be elucidated only recently.8 In order to explore whether reassessment of initial prognostic markers at relapse increases accuracy in predicting outcome after relapse and to determine whether MRD achievement after the first relapse improves outcome, we investigated 120 patients who relapsed after MM diagnosis and initial treatment on our total therapy (TT) 2-6 protocols between 2000-2016. All patients achieved a complete response (CR) and subsequently relapsed, with the first relapse occurring after January 2014, the time point at which MRD assessment by eightcolor flow cytometry was established systematically at the University of Arkansas for Medical Sciences.9 The majority of patients received an immunomodulatory imide drug (IMiD)-based triplet as second line, either in combination with a proteasome inhibitor (64%) or a CD38 targeting monoclonal antibody (27%). The median time to first relapse after initial diagnosis and treatment was 5 years (range, 0.9-18 years) with a median follow up after first relapse of 1.57 years (range, 0.18-6.0 years). Patients’ characteristics are presented in Table 1. GEP70 classified 17% as high-risk (HR) patients at diagnosis. The proportion increased significantly to 35% at relapse (P<0.05). While the diagnostic GEP70 classification retained significant prognostic value at relapse with HR patients having significantly worse PFS (median: 0.93 years vs. 1.86 years; P=0.03), and OS (median: 2.12 years vs. 5.01 years; P<0.01) (Online Supplementary Figure S1A and B), reassessment of GEP70 at relapse improved prognostic accuracy with a median PFS of 0.76 years for HR versus 2.15 years for low-risk (LR) patients, P<0.01, and a median OS of 1.87 years for HR, while LR patients had not reached their median OS, P<0.01, Figure 1A and B. Similarly, we saw that reassessment of FISH and RISS at relapse improved accuracy in outcome prediction over initial assessment at diagnosis. HR FISH alterations were characterized by translocations t(4;14) and t(14;16) and deletion 17p. In particular the proportion of patients with del17p increased significantly from 12.5% (n=11/88) at diagnosis to 28% (n=17/59) at relapse. Despite the relahaematologica | 2021; 106(12)

tively small number of patients with FISH at relapse, reassessment of FISH improved the predictive accuracy for PFS (median: 1.16 years vs. 1.75 years; P=0.1) and OS (median: 2.86 years vs. 4.38 years; P<0.05) (Figure 1C and D) compared to assessment at diagnosis (Online Supplementary Figure S1C and D). Furthermore, reassessment of RISS at relapse, was a more accurate tool in predicting PFS (median PFS RISS I: 1.8 years vs. median PFS RISS II/III: 1.15 years; P<0.05) and OS (median OS RISS I: not reached vs. median OS RISS II/III: 2.9 years, P<0.01) (Figure 1E and F), compared to RISS evaluation at diagnosis (Online Supplementary Figure S1E and F). Of note is that very few patients presented with RISS III at relapse, which is likely due to early detection of relapsing disease in most patients, and hence still a relatively small tumor burden with low b-2-microglobulin values and normal albumin and lactate dehydrogenase (LDH). In contrast to GEP70, FISH and RISS, we only saw a modest prognostic impact of ISS evaluation at diagnosis or even Table 1. Patient characteristics at diagnosis and relapse.

At diagnosis Age in yrs, (range) 59 (32-75) GEP high risk 20/115 (17.4%) ISS stage 1 47/120 (39%) 2 39/120 (33%) 3 34/120 (28%) FISH Translocation t4;14 11/84 (13%) Translocation t14;16 3/84 (3.75%) Deletion 17p 10/84 (12%) R-ISS stage 1 16/84 (19%) 2 54/84 (64.3%) 3 14/84 (16.7%) Focal lesions by PET 0 39/114 (34%) 1-3 26/114 (23%) >3 49/114 (43%) Treatment at relapse IMiD+PI CD38 ab + IMiD CD38 ab + PI other# Salvage ASCT at relapse+ Best response First line therapy sCR/CR 120/120 (100%) VGPR PR SD PD

At relapse 64 (37-81) 27/77 (35%)** 95/112 (84.8%)** 12/112 (10.7%)** 4/112 (3.8%)** 11/59 (18.6%) 3/59 (5%) 17/59 (28.8%)** 22/57 (38.6) * 32/57 (56%) 3/57 (5%)* 62/112 (55%)* 36/112 (32%)* 14/112 (12.5%)** 77/120 (64%) 32/120 (27%) 4/120 (4%) 6/120 (5%) 30/120 (25%) 2nd line therapy 60/119 (50.4%) 22/119 (18.5%) 17/119 (14.3%) 16/119 (13.4%) 4/119 (3.4%)

*P<0.05, **P<0.001 comparing presentation at relapse to diagnosis, McNemar’s test; #: regimen including intravenous chemotherapy such as cytoxan, adriamycin, etoposide, cisplatin (PACE, Metronomic, PACMED); +: salvage autologous stem cell transplant (ASCT) was performed in selected patients after brief reinduction with novel agents or intrvenous chemotherapy. GEP: gene expression profiling; ISS: International Staging System; FISH: fluorescence in situ hybridization; RISS: revisedISS; PET: positron emission tomography; IMiD: immunomodulatory imide drug; PI: protease inhibitor; CR: complete remission; VGPR: very good partial remission, PR: partial response; SD: stable disease; PD: progressive disease.

3215

Letters to the Editor

Figure 1. Legend on following page.

3216

haematologica | 2021; 106(12)

Letters to the Editor

Figure 1. Reassessment of traditional risk factors in first multiple myeloma relapse shows improved prognostic accuracy compared to their evaluation at diagnosis (shown in the Online Supplementary Appendix). GEP70 high-risk (HR) patients at relapse had significant worse progression-free survival (PFS) (A), and OS (B) compared to low-risk (LR) patients, P<0.01. HR fluorescence in situ hybridization (FISH) included translocation t14;16, t4;14 and del 17p and showed worse PFS (P=0.1) (C) and OS (P<0.05) (D), compared to patients with LR FISH. Assessment of revised International Staging System (RISS) at relapse showed significant worse PFS, E, and OS, F, for patients with RISS 2+3 compared to patients with RISS stage I. The presence of >3 focal lesions by positron emission tomography and computed tomography (PET CT) at relapse was associated with worse PFS (H) and OS (I) in first relapse. The results were not significant, likely due to the relative small patient number. Achievement of minimal residual disease (MRD) negativity after first relapse was a powerful marker for significantly improved PFS (J) and OS (K).

at first relapse with a mild non-significant trend to improved clinical outcome in earlier stages after relapsed disease (Online Supplementary Figure S2A and D). Imaging with positron emission tomography and computed tomography (PET CT) was performed at diagnosis (n=120) and at relapse (n=111). Of the 120 patients in our study cohort, 69% (n=75) had at least one 18 F-fluorodeoxyglucose (FDG) avid lesion at diagnosis. Sequential PET CT studies during first line treatment confirmed resolution of PET avid lesions during initial treatment. At first relapse, 44.5% (n=50/112) had at least one lesion. Of these, 68% (n=34/50) had also presented with a FL at diagnosis and 46% (n=23/50) had at least one FL at the same site as at initial diagnosis. The presence of >3 FL by PET at diagnosis, a previously identified adverse risk factor10 only had a small and non-significant adverse prognostic impact on outcome after first relapse with a median PFS of 1.4 years compared to 1.8 years for patients with 0-3 FL and a median OS of 3.9 years compared to 4.8 years (Online Supplementary Figure S3A and B). Reassessment of focal lesions by PET CT at relapse improved the prognostic value of this test, albeit not significantly (median PFS for 0-3 FL: 1.8 years vs. 1.0 year for >3 FL and median OS: 4.4 years for 0-3 FL vs. 2.1 years for >3 FL) (Figure 1H and I). In a further step we evaluated the prognostic significance of MRD achievement after the first relapse. In total, 116 patients had sequential MRD assessment by flow cytometry, as previously described,9 after initiation of a second line therapy of which 47 (40.5%) achieved MRD negativity. Nearly all of the MRD-negative patients also achieved a CR (n=45/47), while the remaining two patients had achieved a VGPR. Achievement of a deep response with MRD negativity during second line treatment was a strong predictor of outcome with a median PFS to second relapse of 1.3 years for patients who did not achieve MRD negativity compared to not reached for patients who achieved MRD negativity (P<0.01) (Figure 1J). Median OS was equally significantly better and not reached for patients who achieved MRD negativity compared to 3.7 years for patients who remained MRD-positive (Figure 1K). The time to achievement of MRD negativity varied greatly, reaching from 0.6 to 3 years with a median of 1.02 years. Intriguingly, a slower response to treatment and later achievement of MRD negativity (>1.02 years, n=24), was associated with significant better PFS (median PFS not reached vs. 1.6 years) and OS (median OS not reached vs. 2.8 years) compared to patients who achieved rapid MRD negativity(<1.02 years, n=23) (Online Supplementary Figure S3C and D). In a final step, we evaluated the association between aforementioned independent risk factors and the hazards of experiencing death as well as progression using a multivariable Cox proportional hazards model (Online Supplementary Table S1A and B). FISH and RISS were excluded from the analysis due to the overall small patient number that was assessed at relapse. Age at progression and time from initial MM diagnosis to first progression were included, as they previously had shown to be of prognostic significance in relapsed MM dishaematologica | 2021; 106(12)

ease.11,12 High risk by GEP70 and the presence of >3 FL were significantly and independently associated with worse PFS and OS as was older age at first progression. Similarly, achievement of MRD negativity after first relapse was a significant and independent prognostic marker for improved outcome. Though prolonged time to first relapse (TT1P) was suggestive of improved PFS and OS, the results were not quite significant in this cohort, suggesting that TT1P as a prognostic marker is determined by other more significant variables. Our study provides a strong rationale to incorporate reassessment of GEP, FISH and RISS at first relapse to improve clinical prognostication. We further underscore the importance of FL assessment at relapse, a practice that is currently inconsistently performed. The importance of identifying focal lesions can be vital to clinical management, as we show that patients with an increased number of FL tend to have worse outcome and furthermore previous reports indicate that bone marrow content and peripheral myeloma markers do not always correlate with the presence of focal lesions.13 Lastly, we show that MRD negativity after first relapse is associated with significantly better PFS and OS, which is in line with a previous report.8 Currently available therapies are increasingly effective, making the achievement of deep responses in relapsed disease a realistic goal.14 While the present study is limited by a relatively small patient size and differences in treatment at relapse, the investigated prognostic factors have been shown to be valid independent of the treatment modality.15 Furthermore, key characteristics of our study are that all patients were uniformly treated during upfront therapy and had been exposed to a protease inhibitor, an IMiD and stem cell transplantation, which are all currently part of standard first line treatment in newly diagnosed MM. Taken together our study provides important insight into prognostic features at first relapse that could help clinicians to reclassify patients and also suggests to treat patients – if performance status permits - to a deep clinical response. David Baker,1* Milan Bimali,2,3* Luis Carrillo,4 Archana Sachedina,5 Daisy Alapat,4 Md Shadiqul Hoque,1 Mathew Kottarathara,1 Richa Parikh,1 Amani Erra,1 Angel A. Mitma,1 Pankaj Mathur,1 Yetunde Ogunsesan,1 Lakshmi Yarlagadda,1 Sravani Gundarlapalli,1 Sharmilan Thanendrarajan,1 Maurizio Zangari,1 Frits van Rhee,1 Guido Tricot1 and Carolina Schinke1 1 Myeloma Center, Division of Hematology/Oncology, Winthrop P. Rockefeller Cancer Institute, University of Arkansas for Medical Sciences; 2Arkansas Children’s Nutrition Center, Arkansas Children's Hospital; 3Department of Biostatistics, University of Arkansas for Medical Sciences; 4Department of Pathology, University of Arkansas for Medical Sciences and 5Department of Radiology, University of Arkansas for Medical Sciences, Little Rock, AR, USA *DB and MB contributed equally as co-first authors. Correspondence: CAROLNA SCHINKE - cdschinke@uams.edu doi:10.3324/haematol.2021.278779 3217

Letters to the Editor

Received: April 12, 2021. Accepted: August 4, 2021. Pre-published: August 12, 2021. Disclosures: no conflicts of interest to disclose. Contributions: DB, MS and CS conceived and developed the study; AS performed imaging studies; LC and DA performed pathology and MRD assessment; MSH, MK, RP, AE, AAM, PM, YO, LY, SG, ST, MZ, FvR, GT and CS provided study material or patients; DB, MB, GT and CS wrote the paper. All authors reviewed and approved the paper. Funding: work completed by CS was funded the National Institutes of Health grants P20GM109005.

References 1. Kristinsson SY, Anderson WF, Landgren O. Improved long-term survival in multiple myeloma up to the age of 80 years. Leukemia. 2014; 28(6):1346-1348. 2. Continued improvement in survival in multiple myeloma: changes in early mortality and outcomes in older patients. Leukemia. 2014; 28(5):1122-1128. 3. Schinke M, Ihorst G, Duyster J, Wasch R, Schumacher M, Engelhardt M. Risk of disease recurrence and survival in patients with multiple myeloma: a German Study Group analysis using a conditional survival approach with long-term follow-up of 815 patients. Cancer. 2020; 126(15):3504-3515. 4. Fonseca R, Monge J, Dimopoulos MA. Staging and prognostication of multiple myeloma. Expert Rev Hematol. 2014;7(1):21-31. 5. Hillengass J, Usmani S, Rajkumar SV, et al. International myeloma work-

3218

ing group consensus recommendations on imaging in monoclonal plasma cell disorders. Lancet Oncol. 2019;20(6):e302-e312. 6. Sonneveld P, Broijl A. Treatment of relapsed and refractory multiple myeloma. Haematologica. 2016;101(8):995. 7. Anderson KC, Auclair D, Kelloff GJ, et al. The role of minimal residual disease testing in myeloma treatment selection and drug development: current value and future applications. Clin Cancer Res. 2017;23(15):39803993. 8. Avet-Loiseau H, San-Miguel J, Casneuf T, et al. Evaluation of sustained minimal residual disease negativity with Daratumumab-combination regimens in relapsed and/or refractory multiple myeloma: analysis of POLLUX and CASTOR. J Clin Oncol. 2021;39(10):1139-1149. 9. Schinke C, Hoering A, Wang H, et al. The prognostic value of the depth of response in multiple myeloma depends on the time of assessment, risk status and molecular subtype. Haematologica. 2017;102(8):e313-e316. 10. Jamet B, Bailly C, Carlier T, et al. Interest of Pet imaging in multiple myeloma. Front Med. 2019;6:69. 11. Chretien ML, Hebraud B, Cances-Lauwers V, et al. Age is a prognostic factor even among patients with multiple myeloma younger than 66 years treated with high-dose melphalan: the IFM experience on 2316 patients. Haematologica. 2014;99(7):1236-1238. 12. Palumbo A, Bringhen S, Falco P, et al. Time to first disease progression, but not beta2-microglobulin, predicts outcome in myeloma patients who receive thalidomide as salvage therapy. Cancer. 2007;110(4):824-829. 13. Rasche L, Alapat D, Kumar M, et al. Combination of flow cytometry and functional imaging for monitoring of residual disease in myeloma. Leukemia. 2019;33(7):1713-1722. 14. Chim CS, Kumar SK, Orlowski RZ, et al. Management of relapsed and refractory multiple myeloma: novel agents, antibodies, immunotherapies and beyond. Leukemia. 2018;32(2):252-262. 15. Fulciniti M, Munshi NC, Martinez-Lopez J. Deep response in multiple myeloma: a critical review. Biomed Res Int. 2015;2015:832049.

haematologica | 2021; 106(12)

Letters to the Editor

Allogeneic hematopoietic cell transplantation outcomes in patients with Richter’s transformation Approximately 2-10% of chronic lymphocytic leukemia (CLL) cases develop into Richter’s transformation (RT), a more aggressive disease typically manifesting as diffuse large B-cell lymphoma (DLBCL).1 Targeted therapies such as ibrutinib are now commonly used to treat CLL but the transformation rate remains comparable to the chemoimmunotherapy era.2 Moreover, these targeted therapies are often used to treat RT despite limited efficacy,1,3 and prognosis for these patients is poor.4-6 The treatment of RT therefore remains challenging in the current era of targeted therapy. Graft-versus-leukemia activity after allogeneic hematopoietic cell transplantation (alloHCT) is evident in patients with CLL where durable remissions can be achieved in all genetically defined high-risk subsets.7,8 Indeed, several small studies have reported benefit from alloHCT in RT.9-11 In order to better understand the therapeutic value of alloHCT in the modern era, we report alloHCT outcomes for 28 consecutive patients with RT who received chemoimmunotherapy and/or targeted therapy prior to alloHCT. The Blood and Marrow Transplant data repository of the Dana-Farber Cancer Institute was queried to identify all patients aged ≥18 years who underwent alloHCT for RT between January 1, 2010 and May 31, 2019. After obtaining Institutional Review Board approval in accordance with the Declaration of Helsinki, a retrospective chart review was performed to confirm the diagnosis of CLL and transformation to RT and 28 patients were identified. Clinical characteristics of these patients are summarized in the Online Supplementary Table S1. Median age was 61 years (range: 41-73 years) and 24 (85.7%) were male. Twenty-six patients received reduced intensity conditioning (RIC) HCT. The histologic diagnosis at alloHCT was DLBCL (n=27) and Hodgkin lymphoma (n=1). Median time from CLL diagnosis to RT was 4.5 years (range: 0- 24.4 years). Median time from RT to alloHCT was 0.6 years (range: 0.2-3.8 years). Twenty-six patients (92.8%) were in complete remission (CR) or partial remission (PR) at the time of alloHCT. Positron emis-

sion tomography (PET) scan was available for 23 patients and seven (30%) of these 23 patients were PET positive. Of note, since RT is a high risk disease, our current practice is to offer alloHCT only to those patients in at least PR. Median number of total therapies for CLL and RT combined prior to alloHCT was three (range: 1-7): one (range: 0-4) for CLL and two (range: 1-7) for RT. Nine patients received targeted therapies (4 for CLL and 5 for RT) in addition to chemoimmunotherapies before alloHCT. No patient received CAR-T cell therapy. All prior and post-transplant therapies are listed in Online Supplementary Table S2. Time from CLL diagnosis to RT and alloHCT, relapse, post HCT therapy, and duration of overall survival (OS) for the entire cohort are depicted in Figure 1 along with selected clinical features such as age, prior targeted therapy, total number of prior therapies, complex karyotype (defined as ≥5 abnormalities),12 HCT comorbidity index, disease status, donor type, bulky disease, high lactate dehydrogenase (LDH) and/or low platelet counts (<100x109/L), and a PET scan result at transplant and occurrence of grade 2-4 acute graft-versus-host disease (GvHD). Strikingly, the cohort is dichotomized into a group of long survivors and a group that experienced early deaths. In the first group (subjects 15-28), all patients remain alive (4-year overall survival [OS] 100%) with median follow-up 4.9 years (range: 2.2-7.7 years). In the second group (subjects 1-14), 11 of 14 died within 1 year (1-year OS 21%). Remarkably, two of three patients aged >70 years survived over 5 years. Subject 27 was 73 years old at the time of alloHCT, relapsed 11 months after alloHCT, and subsequently received post-transplant therapy (CHOP) and donor lymphocyte infusion from his brother. This patient remains alive 7.3 years after alloHCT. Subject 22 was also 73 years old at the time of alloHCT, had del(17p) and developed RT while on ibrutinib. This subject subsequently responded to R-EPOCH prior to alloHCT and remains alive in remission 5.2 years after alloHCT. For the entire cohort, eight relapses (7 RT and 1 CLL) and 13 deaths have occurred: five from disease progression, six from infection and two from GvHD. Of the eight non-relapse deaths, six died within 1 year and two within 2 years of alloHCT. Median follow-up among sur-

Table 1. Kaplan-Meier estimates for overall and progression-free survival and estimates of cumulative incidences of non-relapse mortality, relapse, acute graft-versus-host disease and chronic graft-versus-host disease in the competing risks framework.

All High Risk Standard Risk (N=28) (95% CI) (N=9) (95% CI) (N=19) (95% CI) 4-yr OS 4-yr PFS 4-yr NRM 4-yr Relapse 6 mo. Grade 2-4 aGvHD 6 mo. Grade 2-4 aGvHD 2 yr cGvHD

Grade II-IV aGvHD P

Age ≥65 Age<65 (N=10) (95% CI) (N=18) (95% CI)

53% (33-70) 39% (21-56) 29% (13-47) 32% (16-50) 36% (19-54) 18% (6-34) 52% (30-70)

11%* (0.6-39) 0% 33%* (5-67) 56% (16-83) 56% (17-82) 37% (6-71) 25% (2.5-60)

74% (48-88) 58% (33-76) 21% (6-42) 21% (6-42) 21% (6-42) 11% (1.7-29) 61% (33-80)

<0.0001 <0.0001 0.21 0.054 0.013 0.12 0.43

OS HR (95% CI)

PFS HR (95% CI)

NRM sHR (95% CI)

Relapse sHR (95% CI)

3.94 (1.36-12.4) 0.016

2.05(0.8-5.09) 0.13

7.36 (1.59-34) 0.01

0.53 (0.1-2.81) 0.45

40% (12-67) 10% (0.6-36) 20% (2-50) 70% (25-91)

61% (35-79) 55% (30-74) 34% (13-56) 11% (2-30)

P 0.16 0.006 0.58 0.007

Log-rank was used for comparisons of overall survival (OS) and progression-free survival (PFS). Gray test was used for comparisons of non-relapse mortality (NRM), relapse and graft-versus-host disease (GvHD). The table presents results of univariable analysis for the effect of grade 2-4 acute GvHD (aGvHD) on outcomes. Cox model was used for OS and PFS and cause-specific Cox model was used for NRM and relapse. Occurrence of grade 2-4 aGvHD was treated as a time dependent variable. HR: hazard ratio; CI: confidence interval; mo: months; yr: years; cGvHD: chronic GvHD. *3-year estimate as the last patient in this cohort was censored at 36.3 months.

haematologica | 2021; 106(12)

3219

Letters to the Editor

Figure 1. Swimmer plot. Right panel: time from allogeneic hematopoietic cell transplantation (alloHCT) to post transplant events. Left panel: time from chronic lymphocytic leukemia (CLL) diagnosis to Richter’s transformation (RT) and alloHCT. Middle panel: selected risk features. HCT-CI: HCT comorbidity index; Complex 5: complex karyotype defined as ≥5 abnormalities.12 High lactate dehydrogenase (LDH) is defined as LDH >205 U/L. Low PLT: platelet count <100x109/L.

vivors was 54 months (range: 16-92 month); median OS was not reached and median PFS was 11.2 months. Fouryear OS, progression-free survival (PFS), cumulative incidence of non-relapse mortality (NRM) and relapse were 53%, 39%, 29% and 32%, respectively. (Figure 2A and B). The cumulative incidence of grade 2-4 and grade 3-4 acute GvHD at 6 months were 36% and 18%, respectively (Table 1). As for risk factors, all four patients with low platelet counts (subjects 1, 2, 4 and 5) and six of seven patients with high LDH died within 17 months of alloHCT (subjects 3, 4, 5, 7, 9 and 14) (Figure 1). Due to the small number of patients with high LDH and low platelet counts, these two factors were combined and considered ‘high risk’. Four-year OS was 11% in this high risk group and 74% in the standard risk group (P<0.0001) (Table 1; Figure 2C). In addition, patients who developed grade 24 acute GvHD did poorly, with nine of 11 dying within 18 months (hazard ratio [HR] for OS: 3.94, P=0.016) (Figure 1; Table 1). High risk was also associated with poor PFS (4-year PFS 0% vs. 58%, P<0.0001) (Table 1; Figure 2D). Age was not a significant risk factor for OS but was significant for PFS (4-year PFS 10% for age ≥65 vs. 55% for age <65 years, P=0.006) (Table 1; Online Supplementary Figure S1A). Risk factors for NRM included the occurrence of grade 2-4 acute GvHD (HR: 7.08, P=0.017) (Table 1). Risk factors for relapse included age ≥65 years (4-year cumulative incidence 70% vs. 11%, P=0.007) and high risk (4-year cumulative incidence 56% vs. 21%, P=0.05). (Table 1; Online Supplementary Figure S1B and D). Other factors did not affect outcomes. In particular, remission status (CR vs. PR), Eastern Cooperative Oncology Group performance status, HCT comorbidity index, use of targeted therapy prior to alloHCT, number of prior therapies, year of HCT, PET positivity, bulky dis3220

ease, fluorescence in situ hybridization (FISH) abnormalities and complex karyotype did not affect outcomes. To our knowledge, this is the largest study reporting outcomes of patients with RT who underwent alloHCT in recent years. We report favorable outcomes for these previously treated patients. Importantly, half of these patients have extended OS, reaching a plateau after 1.5 years post transplant. This suggests that some RT patients could be cured with alloHCT. For factors that are associated with poor outcome, high risk disease (i.e., low platelet counts and/or high LDH) was significantly associated with shorter OS and PFS. Outcome for patients with standard risk at transplant was excellent (4-year OS and PFS: 74% and 58%, respectively) despite the fact that these patients had failed multiple therapies. In contrast, few patients with high risk showed benefit from alloHCT suggesting that LDH and platelet counts together could be a sensitive marker of residual disease, since radiologic remission status based on PET/CT imaging at transplant was not predictive of outcome. In addition to these factors, advanced age was associated with poor outcome. Interestingly, use of prior targeted therapy was not associated with improved outcome. Similarly, year of transplant and number of prior therapies for CLL or for RT did not affect clinical outcome. These findings are very different from CLL patients who undergo alloHCT in the modern era13 but resemble observations made in alloHCT of de novo DLBCL,14 suggesting that disease control and sensitivity to alloHCT may be most critical for an aggressive disease like RT. The survival outcome reported in the current study compares favorably to previously published alloHCT series in RT. The European Society for Blood and Marrow Transplantation10 (n=25, 72% RIC) reported 3-year OS haematologica | 2021; 106(12)

Letters to the Editor

41% among 16 patients who received alloHCT in CR/PR, and 17% for nine patients with progressive disease. In a series of single-center studies, Kharfan-Dabaja et al.11 (n=10, all patients were in CR/PR) reported 4-year OS and PFS 50%; Tsimberidou et al.9 (n=17) reported 3-year OS 75% for seven patients who received alloHCT as postremission therapy and 21% for 13 patients who received allo- or autoHCT as salvage therapy. For patients with RT who do not achieve remission, CAR-T cell therapy is a newer option, with recently reported results in a small series (n=9) with limited follow-up by Kittai et al.15 Further and larger studies with longer follow-up are warranted to evaluate the efficacy of this therapy on its own or as a bridge to alloHCT. This study has some limitations owing to its singlecenter retrospective design with a small sample size of 28 patients, which nonetheless is the largest study to date. Another limitation is the absence of data on clonal relationship between RT and CLL. Published literature,16 however, shows that the majority (~80%) of RT is clonally related to the preceding CLL, particularly in heavily pretreated patients like these, suggesting that most RT patients in this study were clonally related. With availability of less toxic/reduced induced intensity conditioning regimens, improved human leukocyte antigen typing, and better GvHD prophylaxis strategies, alloHCT has become a viable and safe treatment option for patients with high risk hematologic cancers, even with advanced age. Our study results show that a sizeable proportion of patients with RT in remission can achieve durable remissions, and that alloHCT should be considered as a treatment option for patients with RT who are fit and have controlled disease. haematologica | 2021; 106(12)

Figure 2. Clinical outcomes. (A) Overall survival (OS) and progression-free survival (PFS) and (B) cumulative incidence of non-relapse mortality (NRM) and relapse for the entire cohort. (C) OS and (D) PFS according to the risk group.

Haesook T. Kim,1 Peter O. Baker,2 Erin Parry,2 Matthew Davids,2 Edwin P. Alyea,3 Vincent T. Ho,2 Corey Cutler,2 John Koreth,2 Mahasweta Gooptu,2 Rizwan Romee,2 Sarah Nikiforow,2 Joseph H. Antin,2 Jerome Ritz,2 Robert J. Soiffer,2 Catherine J. Wu2# and Jennifer R. Brown2# 1 Department of Data Science, Dana Farber Cancer Institute, Harvard School of Public Health, Boston, MA; 2Department of Medical Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA and 3Duke Cancer Institute, Duke Medical School, Durham, NC, USA # CJW and JRB contributed equally as co-senior authors. Correspondence: JENNIFER R. BROWN - jennifer_brown@dfci.harvard.edu HAESOOK T. KIM - htkimc@jimmy.harvard.edu doi:10.3324/haematol.2021.279033 Received: April 19, 2021. Accepted: August 17, 2021. Pre-published: August 26, 2021. Disclosures: JRB has served as a consultant for Abbvie, Acerta, Astra-Zeneca, Beigene, Catapult, Dynamo Therapeutics, Juno/Celgene, Kite, MEI Pharma, Nextcea, Novartis, Octapharma, Pfizer, Sunesis, TG Therapeutics, Verastem; received honoraria from Janssen and Teva; received research funding from Gilead, Loxo, Sun and Verastem; and served on data safety monitoring committees for Morphosys and Invectys, outside the submitted work. CJW holds equity in BioNTech, Inc, and receives research funding from Pharmacyclics, Inc, outside the submitted work. JR has received research funding from Amgen, Equillium, Kite Pharma and Novartis; and served as a consultant for Avrobio, Celgene, Clade Therapeutics, Garuda 3221

Letters to the Editor

Therapeutics, Immunitas Therapeutics, LifeVault Bio, Novartis, Rheos Medicines, Talaris Therapeutics and TScan Therapeutics, outside the submitted work. JK has served as a consultant for Amgen, Equillium, Cugene, and Moderna and advisory board for therakos; and received funding from BMS, Miltenyi, Clinigen, and Regeneron, outside the submitted work. PA has served as a consultant for Merck, BMS, Pfizer, Affimed, Adaptive, Infinity, ADC Therapeutics, Celgene, Morphosys, Daiichi Sankyo, Miltenyi, Tessa, GenMab, C4, and Enterome; received research funding from Merck, BMS, Affimed, Adaptive, Roche, Tensha, Otsuka, Sigma Tau, Genentech, IGM and received honoraria from Merck and BMS, outside the submitted work. RJS has served as a consultant for Gilead, Rheos Therapeutics, Jazz, Cugene, Mana Therapeutics, VOR, and Novartis; served on data safety monitoring committees for Juno; served on board of directors for Kiadis and Be the Match/ NMDP, outside the submitted work. HTK has served as a consultant for Miltenyi, outside the submitted work. The remaining authors declare no competing financial interests. Contributions: HTK conceived and designed the study, performed statistical analysis, interpreted the data and wrote the manuscript; JRB, CJW and MD conceived the study; POB and JRB compiled the outcome data, provided FISH data and annotated the cytogenetic data; JRB and JR edited the manuscript. All authors contributed to the manuscript review and approved the final version for submission. Acknowledgments: we thank the patients and their families, the research coordinators, research nurses, and advanced practice providers. Funding: this work was supported by research funding from the National Cancer Institute (P01CA229092) and the Ted and Eileen Pasquarello Tissue Bank in Hematologic Malignancies.

References 1. Wang Y, Tschautscher MA, Rabe KG, et al. Clinical characteristics

and outcomes of Richter transformation: experience of 204 patients from a single center. Haematologica. 2020;105(3):765-773. 2. Rossi D, Spina V, Gaidano G. Biology and treatment of Richter syndrome. Blood. 2018;131(25):2761-2772. 3. Allan JN, Furman RR. Current trends in the management of Richter's syndrome. Int J Hematol Oncol. 2019;7(4):IJH09. 4. Kadri S, Lee J, Fitzpatrick C, et al.Clonal evolution underlying

3222

leukemia progression and Richter transformation in patients with ibrutinib-relapsed CLL. Blood Adv. 2017;1(12):715-727. 5. Maddocks KJ, Ruppert AS, Lozanski G, et al. Etiology of ibrutinib therapy discontinuation and outcomes in patients with chronic lymphocytic leukemia. JAMA Oncol. 2015;1(1):80-87. 6. Jain P, Thompson PA, Keating M, et al. Long-term outcomes for patients with chronic lymphocytic leukemia who discontinue ibrutinib. Cancer. 2017;123(12):2268-2273. 7. Dreger P, Schnaiter A, Zenz T, et al. TP53, SF3B1, and NOTCH1 mutations and outcome of allotransplantation for chronic lymphocytic leukemia: six-year follow-up of the GCLLSG CLL3X trial. Blood. 2013;121(16):3284-3288. 8. Dreger P, Montserrat E. European Society for Blood and Marrow Transplantation (EBMT); European Research Initiative on CLL (ERIC). Where does allogeneic stem cell transplantation fit in the treatment of chronic lymphocytic leukemia? Curr Hematol Malig Rep. 2015;10(1):59-64. 9. Tsimberidou AM, O’Brien S, Khouri I, et al. Clinical outcomes and prognostic factors in patients with Richter’s syndrome treated with chemotherapy or chemoimmunotherapy with or without stem-cell transplantation. J Clin Oncol. 2006;24(15):2343-2351. 10. Cwynarski K, Van Biezen A, De Wreede L, et al. Autologous and allogeneic stem-cell transplantation for transformed chroniclymphocytic leukemia (Richter’s syndrome): a retrospective analysis from the chronic lymphocytic leukemia subcommittee of the chronicleukemia working party and lymphoma working party of the European Group for Blood and Marrow Transplantation. J Clin Oncol. 2012;30(18):2211-2217. 11. Kharfan-Dabaja MA, Kumar A, Stingo F, et al. Allogeneic hematopoietic cell transplantation for richter syndrome: a single-centerexperience.Clin. Lymphoma. Myeloma Leuk. 2018:18(1):e35-e39. 12. Kim HT, Ahn KW, Hu ZH, et al Prognostic score and cytogenetic risk classification for reduced intensity conditioning allogeneic HCT in CLL patients: CIBMTR report. Clin Cancer Res. 2019;25(16):51435155. 13. Kim HT, Shaughnessy CJ, Rai SC, et al. Outcome of high-risk chronic lymphocytic leukemia patients undergoing allogeneic hematopoietic cell transplant after prior targeted therapy. Blood Adv. 2020; 4(17):4113-4123. 14. Fenske T, Ahn K, Graff T, et al. Allogeneic transplantation provides durable remission in a subset of DLBCL patients relapsing after autologous transplantation. Br J Haematol. 2016;174(2):235-248. 15. Kittai AS, Bond DA, William B, et al. Clinical activity of axicabtagene ciloleucel in adult patients with Richter syndrome. Blood Adv. 2020; 4(19):4648-4652. 16. Rossi D, Spina V, Deambrogi C, et al. The genetics of Richter syndrome reveals disease heterogeneity and predicts survival after transformation. Blood. 2011;117(12):3391-3401.

haematologica | 2021; 106(12)

Letters to the Editor

studies have established that BET inhibitors (BETi) disrupt adult b-globin expression mediated by GATA1 in mouse G1E cells7 and induce hemoglobin production in the UT7 human erythroid cell line.8 However, the effects of BETi on the transcription regulation of genes in the b-globin locus have not been evaluated comprehensively. Here, we demonstrate the ability of BETi to reactivate embryonic and fetal e/g-globins in erythroleukemia cell lines. We tested the ability of BETi to induce erythroid differentiation in TF-1 cells, an erythroleukemia cell line that expresses virtually no embryonic e-globin at baseline (Figure 1A). Quantitative polymerase chain reaction (qPCR) of b-globins after treatment with JQ1 and/or erythropoietin (EPO) demonstrated that although EPO upregulated all three types of b-globins, JQ1 specifically reactivated HBE1, embryonic e-globin (Figure 1). Although JQ1 alone did not affect the expression of HBB, JQ1 antagonized the EPO-induced HBB upregulation, reducing the expression of adult b-globin by 50% (Figure 1C). In contrast, we did not observe expression changes in HBG1/2 with or without EPO (Figure 1D). Strikingly, JQ1 upregulated the e-globin by 15-fold alone, and by more than 200-fold when combined with EPO (Figure 1E). Taken together, these expression changes account for the dramatic increase in HBE1 transcripts from <1%

BET inhibitors enhance embryonic and fetal globin expression in erythroleukemia cell lines Five genes encoding the human b-globins are located in a gene cluster on chromosome 11p15.4 (b-globin locus): 5’-HBE1 (e)-HBG2 (Gg)-HBG1 (Ag)-HBD (d)-HBB (b)-3’. These genes are expressed in distinct developmental stages, with transitions controlled by a series of transcriptional switches regulated by the interplay between local chromatin structure and erythroid-specific transcription factors (TF).1,2 Locally, expression of each b-globin is influenced by the spatial proximity of its promoter to the enhancer-rich locus control region (LCR).1,2 In addition to chromatin looping, a number of TF, including GATA1, BCL11A, COUP-TF2, NuRD, and MYB, specifically repress the expression of embryonic and fetal e/g-globins.1,3,4 Furthermore, MYB mRNA is targeted for degradation by microRNA (miR) miR-15A and miR-16-1.5 In mammals, the bromodomain and extra-terminal domain (BET) family consists of histone readers with two acetyl-lysine binding bromodomains (BD1 and BD2), including the ubiquitously expressed BRD2, BRD3, BRD4, and the germ cell-specific BRDT, that are crucial for epigenetic regulation of gene expression through recruitment of the transcription machinery.6 Previous

Figure 1. JQ1 specifically induces the ε-globin gene in TF-1 cells. (A) Percentage of HBB, HBG1/2, and HBE1 transcripts relative to total b-globin transcripts in TF-1 cells after treatment with JQ1 (red), erythropoietin (EPO) (blue), EPO+JQ1 (green) vs. dimethyl sulfoxide (DMSO) (black) for 5 days (n=3). (B) RNA sequencing quantification of α- and b-globin genes in TF-1 cells after treatment with JQ1 (red), EPO (blue), EPO+JQ1 (green) vs. DMSO (black) for 3 days. (C-E) Quantitative polymerase chain reaction (qPCR) quantification of (C) HBB, (D) HBG1/2, and (E) HBE1 expression in TF-1 cells after treatment with JQ1 (red), EPO (blue), EPO+JQ1 (green) vs. DMSO (black) for a total of 5 days (D1-5) (n=3). (F) Western blots of g- and e-globins in TF-1 cells treated for 3 days. The day at which the protein extracts were made is indicated by a black arrow labeled “WB” in panels (D and E). Quantification of the target normalized to loading control is shown to the right of the gel images (n=3). In (A, C to E), *P<0.05, **P<0.01, ***P<0.001 (Student’s t-test). In (B), * false discovery rate (FDR) <0.05, **FDR <0.01, ***FDR <0.001.

haematologica | 2021; 106(12)

3223

Letters to the Editor

Figure 2. JQ1 downregulates known inhibitors of fetal hemoglobin. (A to C) qPCR quantification of (A) MYB, (B) miR-15A, (C) miR-16-1, in TF-1 cells after treatment with JQ1 (red), erythropoietin (EPO) (blue), EPO+JQ1 (green) vs. dimethyl sulfoxide (DMSO) (control, black) for 3 days. (D) Heatmap representing RNA sequencing quantification of expression changes (log2 fold-changes in FPKM) of known miR-15A and miR-16-1 targets in JQ1 treated TF-1 cells vs. control, with or without EPO. Log2 fold-change ranges from -4 (red, suppressed by JQ1) to 4 (green, induced by JQ1). Genes with low expression (FPKM<1) were excluded or represented with gray box. (E to I) qPCR quantification of (E) IKZF1 (encodes IKAROS), (F) NR2F2 (encodes COUP-TF2), (G) GATA1, (H) BCL11A, and (I) KLF1 in TF-1 cells after treatment with JQ1 (red), EPO (blue), EPO+JQ1 (green) vs. DMSO (control, black) for 3 days. (J) Western blots for GATA1 and BCL11A with corresponding loading controls. Quantifications of signal intensity, normalized to loading controls, are shown as bar graphs to the right of the western blots. N=3 for panels (B), (C), and (J). N=2 for panel (D). N=4 for all other panels. *P<0.05, **P<0.01, ***P<0.001, n.s.: not statistically significant (Student’s t-test).

to nearly 20% as well as the decrease in HBB transcripts from over 30% to less than 20% of all b-globin transcripts in JQ1 treated cells versus control cells (Figure 1A), indicating that JQ1 is reactivating HBE1 in this cell line. RNA sequencing (RNA-seq) on TF-1 cells treated 3 days with dimethyl sulfoxide (DMSO), JQ1, EPO, and EPO+JQ1 confirmed qPCR quantification of b-globin expression (Figure 1B). Moreover, JQ1 treatment also induced the adult α-globin genes HBA1/2, suggesting that BETi treatment is unlikely to lead to an α-thalassemia-like condition (Figure 1B). We performed western blots for g- and e-globins using TF-1 cells treated for 3 days, and observed 3224

that HBE1 protein was increased by the combined treatment compared to EPO alone, as expected given the gene expression changes (Figure 1F). In order to test whether this effect was specific to this cell line and/or was a class effect of BETi, we first treated two other erythroleukemia cell lines, K562 and HEL, and one myeloid cell line (HL-60), in a similar fashion. We found that JQ1 induced e/g-globin genes only in K562 and HEL cells, without any effect in HL-60 cells (Online Supplementary Figure S1A to C) and without EPO induction of differentiation. Two other BETi currently tested in clinical trials (CPI-0610 and PLX51107) had similar haematologica | 2021; 106(12)

Letters to the Editor

Figure 3. JQ1 changes interaction frequencies between the locus control region and the b-globin genes. (A) Chromosome conformation capture quantitative polymerase chain reaction (3C-qPCR) quantification of interaction frequencies between the locus control region (LCR) and segments of the b-globin locus in TF-1 cells after treatment with JQ1 (red), erythropoietin (EPO) (blue), EPO+JQ1 (green) vs. dimethyl sulfoxide (DMSO) (control, black) for 3 days, (n=3). (B) Log2 (fold-change) and statistical significance comparing JQ1 vs. DMSO control (red) and EPO vs. DMSO control (blue). (C) Log2 (fold-change) and statistical significance comparing EPO+JQ1 vs. EPO (green) and EPO+JQ1 vs. JQ1 (purple). Track on top indicates chromosomal gene positions and the DNA fragments generated from an EcoRI digestion. Replicates included have EcoRI digestion efficiency >60% at all control sites. HS: DNase hypersensitivity site. The EcoRI-generated fragment containing the bait primer is shown as a yellow bar. *P<0.05, *P<0.01, ***P<0.001 (Student’s t-test). A genomic region containing no EcoRI site (chr7:117,293,465-117,293,547) was used as internal control for the amount of input DNA. All chimeric fragment quantifications was normalized to the control region quantification of the respective experimental conditions. Since BET inhibition results in global changes in genomic structure, we were unable to identify a genomic region as a reliable control region for this study.

effects to JQ1 in both K562 and TF-1 cells (Online Supplementary Figure S1D and E), demonstrating that the ability to reactivate e/g-globins is a class effect. In order to determine the effects of JQ1 on erythroid differentiation, we examined the expression of erythroidand myeloid-specific genes in TF-1 cells with or without EPO stimulation (Online Supplementary Figure S2A). As expected, EPO treatment resulted in erythroid maturation as shown by the upregulation of erythroid-lineage genes and the downregulation of HSPC or myeloid-lineage genes (Online Supplementary Figure S2A). However, JQ1 only mirrored the effects of EPO in a subset of these genes, including most of those that encode cell surface markers, transporters, and receptors (Online Supplementary Figure S2A). JQ1 exhibited opposite effects to those of EPO in many of the genes encoding components of the erythroid cytoskeleton (Online Supplementary haematologica | 2021; 106(12)

Figure S2A). These observations indicate that JQ1 and EPO have overlapping but distinct effects on the expression of erythroid-lineage genes, and the reactivation of e/g-globins by BETi is not purely due to erythroid differentiation and maturation. Transcriptome analyses further revealed the distinct effects of JQ1 and EPO (Online Supplementary Table S1; Online Supplementary Figure S2). Whereas EPO induces genes involved in erythropoiesisrelated pathways (e.g., heme-biosynthesis and iron ion homeostasis), JQ1 suppresses immune-activation pathways (Online Supplementary Figure S2B to E). Among the genes regulated by EPO or JQ1, only a small subset is regulated by both (Online Supplementary Figure S2F to G). Since e/g-globins are suppressed by a number of TF, we examined the expression of corresponding genes in our RNA-seq data and further validated the results with qPCR (Figure 2). We found that MYB expression was 3225

Letters to the Editor

decreased by JQ1 and EPO, and the combination treatment showed an additive effect (Figure 2A). Since MYB is a target of miR-15A and miR-16-1, we measured the levels of these miR and found that both were upregulated by either JQ1 and EPO in an additive fashion (Figure 2B and C), showing an inverse correlation compared to MYB expression as expected. RNA-seq quantification of known miR-15A/16-1 targets indicated that most were downregulated by treatment with JQ1 and/or EPO (Figure 2D), further suggesting that these miR play a central role in mediating the effects of JQ1. Next, we examined the genes encoding several well-established HBE1 and HBG1/2 inhibitors or HBB activators: IKZF1 (IKAROS), NR2F2 (COUP-TF2), BCL11A, and GATA1. Each of these genes was downregulated by JQ1 with or without EPO (Figure 2E to G). KLF1 promotes the expression of BCL11A,9 and therefore as expected, we observed downregulation of BCL11A only when we also saw KLF1 repression in the setting of joint JQ1 and EPO treatment (Figure 2H and I). GATA1 western blots confirmed a stable level of protein as expected from the qPCR data, whereas western blotting for BCL11A demonstrated almost no detectable protein after combination treatment (Figure 2J). In order to examine how the spatial structure of the locus changes in response to JQ1 treatment, we performed chromatin conformation capture followed by qPCR (3C-qPCR) to quantify the LCR interactions with the b-globin genes under each treatment (Figure 3A). Interestingly, differentiation alone via EPO did not change the interaction frequency in this locus (Figure 3B). In contrast, JQ1 decreased the interaction at multiple loci near or in the fetal and adult g/b-globin genes (Figure 3B). Similarly, the LCR interaction in EPO+JQ1 double-treated cells is more similar to that of JQ1-treated cells than that of EPO-treated cells, again showing that JQ1, but not EPO, decreases interactions between the LCR and the g/b-globin genes (Figure 3C). Notably, JQ1 treatment did not decrease the interaction between the LCR and the HBE1 promoter (Figure 3B and C), suggesting that JQ1 treatment relaxes the looping between the LCR and the g/b-globin genes, which biases LCR contacts in favor of interactions with the promoter of embryonic HBE1. Thus, the b-globin gene expression changes under BET inhibition are likely the result of both shifts in local chromatin looping and expression changes of b-globin inhibitors and activators. The shifts in LCR interactions favor HBE1 expression over HBB or HBG1/2. Although the overall interaction between LCR and the b-globin genes decreases under JQ1 treatment, e/g-globin expression still increased likely due to reduced inhibition and upregulation of genes involved in erythroid maturation. Together, these factors result in decreased HBB transcription and increased embryonic HBE1 transcription (Figure 1A). Our data in TF-1 cells show that BETi induces partial erythroid maturation and reactivate the embryonic e-globin HBE1 even without EPO-mediated erythroid maturation. An important clinical question is whether the HBE1 encoded e-globin could function as a reasonable substitute for the adult b-globin chain. Biochemical analysis of the embryonic hemoglobin Hb-Gower 2 (α2e2) shows that its P50 for oxygen, affinity to 2,3-BPG, Bohr coefficient, and Hill coefficient are comparable to those of adult hemoglobin A (HbA).10 Hb-Gower 2 also has a comparable tetramer-dimer dissociation constant to that of HbA.11 A study in transgenic α/b-thalassemia mice found that human embryonic hemoglobins consist of ζ-globin and e-globin rescue the lethal phenotype of 3226

α/b-thalassemia.12 Similarly, a study in sickle cell mice found that the presence of human Hb-Gower 2 (α2e2) greatly alleviated sickle cell phenotypes, and Hb-Gower 2 inhibits sickle cell hemoglobin (HbS) polymerization.13 Our findings suggest that specific inhibition of bromodomain-containing proteins could provide a reasonable alternative and/or adjuncts to the treatment of b-globinopathies like sickle cell anemia (SCA). SCA patients generally have elevated EPO levels,14 and thus treatment with JQ1 alone might be sufficient to induce e-globin in these patients. Currently, the most common treatment for SCA is administration of hydroxyurea, which elevates fetal hemoglobin production to alleviate symptoms.15 In recent years, multiple gene therapy strategies for SCA have emerged, including b-globin gene addition and nuclease-assisted b-globin gene modification/repair.16 Although these recent advances in gene therapy offer the potential for cure, they will not be accessible or affordable for all patients, especially those from less affluent countries or health systems. We hope that our work motivates future studies that focus on the effect of BET inhibition on globin expression in other erythroid cell lines and primary cells and how BET inhibition could be harnessed for therapeutic purposes. John Z. Cao,1 Kristina Bigelow,2 Amittha Wickrema1,2 and Lucy A. Godley1,2 1 Committee on Cancer Biology, Biological Sciences Division, The University of Chicago and 2Section of Hematology/Oncology, Department of Medicine, The University of Chicago, Chicago IL, USA Correspondence: LUCY A. GODLEY - lgodley@medicine.bsd.uchicago.edu doi:10.3324/haematol.2021.278791 Received: March 16, 2021. Accepted: August 20, 2021. Pre-published: August 26, 2021. Disclosures: LAG, AW and JZC are owners of the patent PCT/US20/52842 titled “Methods and compositions for treating sickle cell disease and thalassemia”, filed on September 25, 2020. Contributions: JZC designed and performed the experiments, analyzed the data, and wrote the manuscript; KB performed additional experiments; LAG conceived of the study and provided insights in experimental design and data interpretation; AW provided additional input for experimental design and data interpretation. Acknowledgments: we thank Dr. Alex Ruthenburg (University of Chicago) for providing the K562 cells used in this work and experimental advice; and Dr. Julie-Aurore Losman (Dana-Farber Cancer Institute) who provided the TF-1 cells. Funding: this work was supported by a grant by the Edwards P. Evans Foundation/EvansMDS to AW and LAG. Funding for JZC was supported by the University of Chicago Biological Sciences Division Dean’s Office, the University of Chicago Comprehensive Cancer Center Women’s Board, and the Goldblatt Scholarship.

References 1. Sankaran VG, Xu J, Orkin SH. Advances in the understanding of haemoglobin switching. Br J Haematol. 2010;149(2):181-194. 2. Wilber A, Nienhuis AW, Persons DA. Transcriptional regulation of fetal to adult hemoglobin switching: new therapeutic opportunities. Blood. 2011;117(15):3945-3953. 3. Aerbajinai W, Zhu J, Kumkhaek C, Chin K, Rodgers GP. SCF induces gamma-globin gene expression by regulating downstream transcription factor COUP-TFII. Blood. 2009;114(1):187-194. 4. Xu J, Bauer DE, Kerenyi MA, et al. Corepressor-dependent silencing of fetal hemoglobin expression by BCL11A. Proc Natl Acad Sci U S A. 2013;110(16):6518-6523. 5. Sankaran VG, Menne TF, Šćepanović D, et al. MicroRNA-15a and -16-1

haematologica | 2021; 106(12)

Letters to the Editor

act via MYB to elevate fetal hemoglobin expression in human trisomy 13. Proc Natl Acad Sci U S A. 2011;108(4):1519-1524. 6. Doroshow DB, Eder JP, LoRusso PM. BET inhibitors: a novel epigenetic approach. Ann Oncol. 2017;28(8):1776-1787. 7. Stonestrom AJ, Hsu SC, Jahn KS, et al. Functions of BET proteins in erythroid gene expression. Blood. 2015;125(18):2825-2834. 8. Goupille O, Penglong T, Lefèvre C, et al. BET bromodomain inhibition rescues erythropoietin differentiation of human erythroleukemia cell line UT7. Biochem Biophys Res Commun. 2012;429(1-2):1-5. 9. Zhou D, Liu K, Sun C-W, Pawlik KM, Townes TM. KLF1 regulates BCL11A expression and gamma- to beta-globin gene switching. Nat Genet. 2010;42(9):742-744. 10. He Z, Russell JE. Expression, purification, and characterization of human hemoglobins Gower-1 (zeta(2)epsilon(2)), Gower-2 (alpha(2)epsilon(2)), and Portland-2 (zeta(2)beta(2)) assembled in complex transgenic-knockout mice. Blood. 2001;97(4):1099-1105. 11. Manning LR, Russell JE, Padovan JC, et al. Human embryonic, fetal, and adult hemoglobins have different subunit interface strengths. Correlation

haematologica | 2021; 106(12)

with lifespan in the red cell. Protein Sci. 2007;16(8):1641-1658. 12. Russell JE, Liebhaber SA. Reversal of lethal alpha- and beta-thalassemias in mice by expression of human embryonic globins. Blood. 1998; 92(9):3057-3063. 13. He Z, Russell JE. A human embryonic hemoglobin inhibits Hb S polymerization in vitro and restores a normal phenotype to mouse models of sickle cell disease. Proc Natl Acad Sci U S A. 2002;99(16):10635-10640. 14. Gordeuk VR, Campbell A, Rana S, et al. Relationship of erythropoietin, fetal hemoglobin, and hydroxyurea treatment to tricuspid regurgitation velocity in children with sickle cell disease. Blood. 2009;114(21):46394644. 15. Ali MA, Ahmad A, Chaudry H, et al. Efficacy and safety of recently approved drugs for sickle cell disease: a review of clinical trials. Exp Hematol. 2020;92:11-18.e11. 16. Orkin SH, Bauer DE. Emerging genetic therapy for sickle cell disease. Annu Rev Med. 2019;70:257-271.

3227

Letters to the Editor

increased the platelet count and blunted the pro-thrombotic effect of sera and plasma from two patients with TTS, whereas IVIg and anti-platelets prevented in vitro TTS sera/plasma-supported thrombogenicity, platelet reactivity and markers of platelet activation. Patient 1 is a 47 years old man who had an episode of syncope on March 15th 2021, 7 days after the first ChAdOx1 nCov-19 injection. He had thalassemia trait and had never been previously exposed to heparin. His platelet count was 92x109/L at presentation and decreased to a nadir of 27x109/L on day 4. A computed tomography angiography (CTA) detected pulmonary embolism, which was hemodynamically stable. Patient 2 is a 36 years old woman who experienced severe abdominal pain on March 17th 2021, 18 days after the first ChAdOx1 nCov-19 injection. She had never been previously exposed to heparin and never used oral contraceptives. Platelet count at presentation was 133x109/L and decreased to a nadir of 106x109/L on day 4. An abdomi-

Platelet activation and modulation in thrombosis with thrombocytopenia syndrome associated with ChAdOx1 nCov-19 vaccine A severe clinical syndrome has been observed in some recipients of the ChAdOx1 nCov-19 or Ad26.COV2.S vaccine, characterized by the presence of antibodies against platelet factor 4 (PF4)/polyanions complexes, thrombocytopenia and thrombosis,1-6 thus resembling heparin-induced thrombocytopenia (HIT).1 The syndrome has been termed “thrombosis with thrombocytopenia syndrome (TTS)”, or “vaccine-induced immune thrombotic thrombocytopenia (VITT)”.7,8 Intravenous immunoglobulin (IVIg) has been successfully used to increase the platelet count in patients with TTS.3,4 Here we report on the management of two patients with TTS, the effect of their serum or plasma on normal platelets and its modulation by IVIg and anti-platelet agents. IVIg

Figure 1. Immunologic tests and platelet parameters in patients before and after intravenous immunoglobulin administration and healthy subjects. Immunologic tests and platelet parameters in patients before and after intravenous immunoglobulin administration and healthy subjects. Blood withdrawal for all after intravenous immunoglobulin (IVIg) experiments was performed on day 15 for patient 1 and on day 13 for patient 2. Open squares: healthy subjects; open circles: patient 1; closed triangles: patient 2; open diamonds: patient 3 (post-vaccine thrombocytopenia without thrombosis). (A) Detection of anti-platelet factor 4 (PF4)/polyanions immunoglobulins by enzyme-linked immunosorbent assay (ELISA) in patients’ sera in absence or presence of high concentrations of heparin (100 U/mL). The horizontal dotted line indicates the cut-off value of 0.4 optical density (O.D.) for normal values. (B) Platelet activation test (PAT), measured by light transmission aggregometry (LTA) in normal washed platelet suspensions (WPS). Serum samples (60 mL) from 7 healthy subjects and from patients 1 and 2 were added to 222 mL of normal WPS in a LTA aggregometer and platelet aggregation was measured as increase in light transmission for 30 minutes (min), in the absence and presence of low (0.2 U/mL) and high concentrations (100 U/mL) of heparin in 2 different experimental sessions, and in the presence of PF4 10 mg/mL in 2 (patient 1) and 3 (patient 2) experimental sessions. Individual results obtained in patients’ sera and mean values obtained in sera from 7 healthy subjects are displayed. The horizontal dotted line indicates the cut-off value of 3.2% for normal values, which was calculated as mean + 2 standard deviations of results obtained in healthy subjects. (C) PAT, measured by impedance aggregometry (HIMEA) in normal whole blood (WB) samples. Serum samples (200 mL) from 1 healthy subject and from patients 1 and 2 were added to 300 mL of normal WB in a multiplate aggregometer and platelet aggregation was measured as area under the curve (AUC) for 15 min in the absence and presence of low (1.0 U/mL) and high concentrations (200 U/mL) of heparin. Sera from patients 1 and 2 were tested only before IVIg infusions. (D) Effects of IVIg infusion (2 gr/Kg body weight over 5 days) on platelet count in patient 1 and patient 2. (E) Percent of platelet/monocyte hetero-aggregates before and after IVIg infusion in patients 1 and 2. The horizontal dotted line indicates the cut-off value of 13.44% for normal values, which was calculated as mean + 2 standard deviations of results obtained with normal sera from 5 healthy subjects. Hep 0.2: heparin 0.2 U/mL; Hep 1: heparin 1 U/mL; Hep 100: heparin 100 U/mL; Hep 200: heparin 200 U/mL; PF4: platelet factor 4.

3228

haematologica | 2021; 106(12)

Letters to the Editor

Table 1. In vitro effects of plasma or sera from patients or healthy subjects on parameters of platelet function in whole blood or washed platelet suspensions from healthy subjects.

Plasma* or serum* donor subjects

Thrombus formation in microchannels Surface coverage (%)

Healthy subjects (n. of subjects) HS + Ig (n. of subjects) HS + aspirin (n. of subjects) HS + cangrelor (n. of subjects)

5.35 ± 1.5 (n=18) 3.66 ± 1.5 (n=4) 5.20 ± 2.2 (n=4) 4.61 ± 1.36 (n=4)

23.14 ± 5.3 (n=18) 17.75 ± 4.2 (n=4) 16.75 ± 3.8 (n=4) 18.75 ± 2.2 (n=4)

1.86 ± 0.7 (n=7) 2.0 ± 0 (n=2) ND

11.30 ± 0.1 (n=6) ND

0.70 ± 0.0 (n=6) ND

Patient 1 before IVIg 7.06 ± 3.7 (n. of experiments) (n=3) Patient 1 before IVIg + Ig 2.97 ± 0.77 (n. of experiments) (n=3) Patient 1 before IVIg + aspirin 7.25 ± 2.2 (n. of experiments) (n=3) Patient 1 before IVIg + cangrelor 5.28 ± 2.4 (n. of experiments) (n=3) Patient 1 after IVIg 5.40 ± 2.1 (n. of experiments) (n=3)

24.67 ± 11.6 (n=3) 18.00 ± 6.1 (n=3) 19.67 ± 4.7 (n=3) 18.67 ± 6.5 (n=3) 25.33 ± 11.7 (n=3)

35 ± 13.8 (n=7) 1±0 (n=2) 1±0 (n=2) 1±0 (n=2) 1±0 (n=2)

22.06 ± 0.0 (n=2) ND

0.90 ± 0.0 (n=2) ND

12.02 ± 0.1 (n=2)

0.77 ± 0.0 (n=2)

Patient 2 before IVIg 10.86 ± 2.1 (n. of experiments) (n=3) Patient 2 before IVIg + Ig 5.45 ± 0.5 (n. of experiments) (n=3) Patient 2 before IVIg + aspirin 9.48 ± 1.2 (n. of experiments) (n=3) Patient 2 before IVIg + cangrelor 6.86 ± 2.7 (n. of experiments) (n=3) Patient 2 after IVIg 7.17 ± 1.0 (n. of experiments) (n=3)

34.33 ± 8.3 (n=3) 18.67 ± 1.5 (n=3) 23.00 ± 1.7 (n=3) 21.33 ± 6.8 (n=3) 25.50 ± 6.1 (n=3)

0±0 (n=2) ND

86.68 ± 0.04 (n=2) ND

16.01 ± 0.1 (n=2) ND

1±0 (n=2)

8.86 ± 0.0 (n=2)

0.55 ± 0.0 (n=2)

19.0 ±1.4 (n=2)

6.29 ± 0.0 (n=2)

0.51 ± 0.0 (n=2)

Patient 3 (n. of experiments)

4.63 ± 0.4 (n=2)

Thrombus formation PAT Flow cytometry Flow cytometry in microchannels Light transmission Platelets/monocytes Annexin V binding Mean thrombus area (%) heteroaggregates (%) (%) (mm2)

*Citrate plasma samples were used in experiments of thrombus formation in microchannels and of platelet activation test (PAT), while serum samples were used in flow cytometry experiments. Blood withdrawal for after-IVIg experiments was performed on day 15 for patient 1 and on day 13 for patient 2. Aspirin= 100 µmol/L; cangrelor= 1 µmol/L; Ig= 5 mg/mL. HS: healthy subjects; Ig:immunoglobulin; IVIg: intravenous immunoglobulin; pt: patient.

nal CT scan showed thrombosis of the portal, superior mesenteric and splenic veins, not associated with liver cirrhosis, occult malignancy or JAK2 V617F. Both patients had normal platelet counts before vaccination. Experiments for the confirmation of TTS diagnosis, the evaluation of platelet activation in such patients and its modulation by IVIg and anti-platelets were performed as follows. Anti-PF4/polyanions antibodies were measured by an enzyme-linked immunosorbent assay (ELISA, PF4 Enhanced Test, Immucor), which contains immunoglobulin G (IgG), IgA and IgM antibodies and is more sensitive than non-ELISA rapid immunoassays.9 The platelet activation test (PAT) was measured (i) by light transmission aggregometry (LTA) using normal washed platelet suspensions (WPS) prepared by the method described by Mustard et al.10 in the Platelet Aggregation Profiler-8E (Bio/Data, Milan, Italy), and (ii) by whole blood impedance aggregometry (HIMEA)11 using normal whole blood haematologica | 2021; 106(12)

(WB) in a Multiplate ECC (F. Hoffmann-La Roche). Platelets in WPS and WB were normally reactive to physiological agonists; patients’ sera were tested in parallel in the same experimental sessions. For flow cytometry experiments, normal citrate-anticoagulated WB was incubated with anti-CD14-PE or annexin V-PE and antiCD42b-FITC at room temperature (RT) for 20 minutes. Subsequently, samples for platelet-monocyte heteroaggregates were fixed, and red cells lysed. A total of 2,000 events of CD14+ or 10,000 events of CD42b for annexin V were acquired at medium flow rate by FACS Verse Cytometer (BD Biosciences, San Jose, CA, USA). In some experiments, patients’ sera were incubated with normal WB at RT for 20 minutes before staining. Experiments of in vitro thrombus formation were performed as previously described,12 perfusing normal WB anticoagulated with lepirudin (450 ATU/mL) (Refludan, Pharmion) on collagen-coated (100 mg/mL) microchan3229

Letters to the Editor

nels at constant blood flow of 950/s shear rate for 4 minutes. Six images were then captured and the surface coverage and area of thrombi (ATh) were calculated. Ig (5 mg/mL) (Venital, Kedrion Biopharma), aspirin (100 mmol/L) (Sanofi SPA) or the P2Y12 antagonist cangrelor (1 mmol/L) (The Medicines Company, Parsippany-Troy Hills, NJ, USA) were added in vitro in some experiments. The suspicion of TTS, based on the co-presence of thrombosis and thrombocytopenia, was supported by the positivity of the ELISA for anti-PF4/polyanions antibodies (Figure 1A), which was normalized by heparin at high concentration (100 U/mL). PAT was tested both by LTA and HIMEA after the addition of patients’ sera to normal WPS and normal WB. Different results were obtained in the two patients: only serum from patient 1 induced aggregation of WPS, which was inhibited by heparin at low (0.2 U/mL) and high (100 U/mL) concentrations (Figure 1B); in contrast, both patients’ sera induced platelet aggregation in normal WB, which was not inhibited by 1 U/mL heparin and was inhibited by 200 U/mL heparin only when induced by patient 2 serum (Figure 1C). The observed discrepant results obtained with WPS and WB might suggest a major role in patient 2 of leukocytes interaction with platelets and antiPF4/polyanions autoantibodies in the pathogenesis of platelet activation and thrombosis.13 As it has been demonstrated that the in vitro addition of PF4 increases the sensitivity of the PAT test in some patients, experiments were repeated in the presence of 10 µg/mL PF4 (Chromatec, Germany): under these conditions, serum from patient 1 induced platelet activation similarly in two separate experiments, while serum from patient 2 induced platelet activation in one experiment, but was still ineffective in two separate experiments (Figure 1B). Following the diagnosis of TTS, anticoagulant treatment, which was initially based on heparin preparations, was switched to alternative anticoagulants: fondaparinux during hospitalization and edoxaban at discharge for patient 1, argatroban and fondaparinux during hospitalization and apixaban at discharge for patient 2. Both patients were also treated with IVIg, 2 g/Kg body weight over 5 days, which normalized their platelet count (Figure 1D). The time needed to increase the platelet count was similar to that observed in other studies in which the same dose of IVIg was infused over 2 days.3,4,14 No steroids were given to patients. Patient 2 also underwent transjugular intrahepatic portosystemic shunt (TIPS), thrombo-aspiration and loco-regional fibrinolysis in the angiography room on day 2. The clinical courses were uneventful for both patients, who were discharged on days 9 and 16. Platelet counts of both patients were normal up to 7 weeks after completion of IVIg treatment (not shown). IVIg infusion had additional potentially protective effects: it i) reduced (patient 1) or normalized (patient 2) the serum reactivity detected by the ELISA test (Figure 1A), compatible with inhibition of antibody production;15 ii) reduced or abolished the activation of normal WPS by patients’ sera (Figure 1B); iii) normalized the percentage of circulating platelet/monocyte hetero-aggregates in both patients, a marker of platelet activation and interaction with leukocytes, which were increased at baseline (Figure 1E): similar findings were recently reported in other patients;13 iv) blunted the amplifying effect of patients’ sera on in vitro thrombus formation by normal blood (see below). Considering that markers of platelet hyper-reactivity could be secondary to the patients’ ongoing thrombotic process in vivo and that their improvement after IVIg could also be due to concomitant treatment with antico3230

agulants, we elected to evaluate the effects of patients’ sera or plasma on markers of activation and reactivity of platelets from healthy subjects and the inhibitory effects of Ig added in vitro. To this end, the effects of patients’ sera/plasma were compared not only to those of sera/plasma from six to 18 healthy subjects, but also to those of serum/plasma from a 76 years old man (patient 3) who developed thrombocytopenia (69x109/L) and epistaxis 6 days after the first ChAdOx1 nCov-19 injection, but had no thrombotic events and negative ELISA test results for anti-PF4/polyanions antibodies (Figure 1A). Compared to 18 plasma samples from healthy subjects, plasma from patients 2 and 1 (albeit less markedly), but not plasma from patient 3, increased thrombus formation by normal WB perfused over collagen-coated microchannels at 950/s shear rate (Table 1). Increased thrombus formation was not observed or was less marked with patients’ plasma obtained after IVIg, and was completely prevented by Ig added in vitro (Table 1). Similar effects of patients’ sera/plasma were observed on aggregation of normal WPS, formation of monocyte/platelet hetero-aggregates and binding of annexin V to procoagulant phosphatidylserine on the platelet membrane in normal WB, which were dramatically increased, especially by serum from patient 2. These effects were prevented by both the in vivo administration of IVIg and the in vitro addition of Ig, suggesting that Ig mostly inhibit platelet activation through FcgRIIa receptors, although a partial contribution by in vivo inhibition of antibody production cannot be ruled out.15 Even though serum/plasma from patient 2 did not activate normal WPS, it was more pro-thrombogenic than serum/plasma from patient 1 in all other tests in which normal WB was used, thus reproducing the discrepant results obtained with WPS and WB in the PAT test (Figure 1B and C). Finally, we also tested the in vitro effects of the antiplatelet drugs aspirin and cangrelor on these parameters of platelet reactivity. Both drugs prevented the potentiation of platelet reactivity induced by patients’ sera/plasma, although cangrelor tended to be more effective than aspirin. These results suggest that the thromboxane A2 and ADP/P2Y12 pathways of platelet activation might play a role in platelet activation in TTS. Whether or not these antiplatelet drugs could benefit TTS patients should only be determined by the results of ad hoc control studies. In conclusion, we found that IVIg curbed the plateletactivating properties of our patients’ sera and produced a lasting increase in platelet count even in absence of concomitant corticosteroid treatment. Mariangela Scavone,1* Bianca Clerici,1* Simone Birocchi,1 Tatiana Mencarini,2 Mariagrazia Calogiuri,1 Claudia Ghali,1 Daniele Prati,3 Silvia Bozzi,2 Paolo Villa,4 Marco Cattaneo1 and Gian Marco Podda1 1 Divisione di Medicina Generale II, ASST Santi Paolo e Carlo, Dipartimento di Scienze della Salute, Università degli Studi di Milano; 2 Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano; 3Dipartimento di Medicina Trasfusionale e di Ematologia, IRCCS Fondazione Ca' Granda, Ospedale Maggiore Policlinico and 4 U.O Medicina d'urgenza Sacco - Medico dell'ASST Fatebenefratelli Sacco, Milano, Italy. *MS and BC contributed equally as co-first authors

Correspondence: GIAN MARCO PODDA - gmpodda@gmail.com doi:10.3324/haematol.2021.279345 Received: July 1, 2021. haematologica | 2021; 106(12)

Letters to the Editor

Accepted: August 25, 2021. Pre-published: September 2, 2021. Disclosures: no conflicts of interest to disclose. Contributions: MS and BC contributed to the design of the study, analyzed the data, contributed to writing the manuscript and critically reviewed it; MS, MC and CG performed laboratory analyses; TM and SB contributed to microfluidic device production and analysed thrombus formation data; BC and SB consulted on patient management; PV provided information regarding the negative control; MC and GMP designed the study, coordinated the group, contributed to data analysis and interpretation and wrote and edited the manuscript. All authors read and approved the final manuscript. Acknowledgments: the authors would like to thank the nursing staff of the Hematology Day Hospital of Presidio San Paolo, ASST Santi Paolo e Carlo, Milan, whose co-operation was essential for the collection of samples for this study; the medical and nursing personnel involved in patient management operating at the Emergency Department of Presidio San Paolo and Presidio San Carlo, Internal Medicine II Division of Presidio San Paolo, and Emergency Medicine of Presidio San Carlo, ASST Santi Paolo e Carlo, Milan, Italy. Data sharing statement: the raw data that support the findings of this study will be made available by the authors, without undue reservation.

References 1. Greinacher A, Thiele T, Warkentin TE, Weisser K, Kyrle PA,

Eichinger S. Thrombotic thrombocytopenia after ChAdOx1 nCov19 Vaccination. N Engl J Med. 2021;384(22):2092-2101. 2. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021; 384(23):2202-2211. 3. Tiede A, Sachs UJ, Czwalinna A, et al. Prothrombotic immune thrombocytopenia after COVID-19 vaccine. Blood. 2021;138(4):350353.

haematologica | 2021; 106(12)

4. Thaler J, Ay C, Gleixner KV, et al. Successful treatment of vaccineinduced prothrombotic immune thrombocytopenia (VIPIT). J Thromb Haemost. 2021;19(7):1819-1822. 5. Vayne C, Rollin J, Gruel Y, et al. PF4 immunoassays in vaccineinduced thrombotic thrombocytopenia. N Engl J Med. 2021; 385(4):376-378. 6. See I, Su JR, Lale A, et al. US case reports of cerebral venous sinus thrombosis with thrombocytopenia after Ad26.COV2.S vaccination, March 2 to April 21, 2021. JAMA. 2021;325(24):2448-2456. 7. Cattaneo M. Thrombosis with thrombocytopenia syndrome associated with viral vector COVID-19 vaccines. Eur J Intern Med. 2021; 89:22-24. 8. American Society of Hematology. Thrombosis with thrombocytopenia syndrome (also termed vaccine-induced thrombotic thrombocytopenia). Version 1.4; last updated April 29, 2021. https://www.hematology.org/covid-19/vaccine-induced-immunethrombotic-thrombocytopenia. Accessed on May 1 , 2021. 9. Sachs UJ, Cooper N, Czwalinna A, et al. PF4-dependent immunoassays in patients with vaccine-induced immune thrombotic thrombocytopenia (VITT): results of an inter-laboratory comparison. Thromb Haemost. 2021 Jun 24. [Epub ahead of print] 10. Mustard JF, Perry DW, Ardlie NG, Packham MA. Preparation of suspensions of washed platelets from humans. Br J Haematol. 1972; 22(2):193-204. 11. Morel-Kopp MC, Mullier F, Gkalea V, et al; subcommittee on platelet immunology. Heparin-induced multi-electrode aggregometry method for heparin-induced thrombocytopenia testing: communication from the SSC of the ISTH. J Thromb Haemost. 2016 Dec;14(12):2548-2552. 12. Scavone M, Bozzi S, Mencarini T, Podda G, Cattaneo M, Redaelli A. Platelet adhesion and thrombus formation in microchannels: the effect of assay-dependent variables. Int J Mol Sci. 2020;21(3):750. 13. Greinacher A, Selleng K, Wesche J, et al. Towards understanding ChAdOx1 nCov-19 vaccine-induced immune thrombotic thrombocytopenia (VITT). https://www.researchsquare.com/article/rs440461/v1. Accessed on April 30, 2021. 14. Uzun G, Althaus K, Singh A, et al. The use of intravenous immunoglobulin in the treatment of vaccine-induced immune thrombotic thrombocytopenia. Blood. 2021;138(11):992-996. 15. Galeotti C, Kaveri SV, Bayry J. IVIG-mediated effector functions in autoimmune and inflammatory diseases. Int Immunol. 2017; 29(11):491-498.

3231

Letters to the Editor

Dose-adjusted EPOCH-rituximab or intensified B-non-Hodgkin lymphoma therapy for pediatric primary mediastinal large B-cell lymphoma. Results from the study B-NHL-BFM-04 and the NHL-BFM registry 2012 Treatment outcomes for children and adolescents with primary mediastinal large B-cell lymphoma (PMBCL) with chemotherapy designed for childhood mature Bnon-Hodgkin lymphoma (B-NHL) are inferior to those of children with other B-NHL-subtypes.1-3 Consequently, B-NHL-type chemotherapy was first intensified and subsequently replaced by dose-adjusted chemoimmunotherapy with etoposide, prednisone, cyclophosphamide, doxorubicin, and rituximab (DA-EPOCH-R) in the NHL-Berlin-Frankfurt-Münster (BFM)-study group. DA-EPOCH-R resulted in superior event-free (EFS) and

overall survival (OS) compared to the previous B-NHL chemotherapy, however, in four patients central nervous system (CNS)-relapses occurred. Treatment of children with PMBCL by chemotherapy protocols without rituximab including high-dose methotrexate, etoposide, ifosfamide, cyclophosphamide, cytarabine, vincristine, and corticosteroids, combined with intrathecal chemotherapy resulted in EFS rates at 5 years of 53–70%.1-3 In order to improve outcome, treatment was intensified for patients with PMBCL in the trial B-NHL-BFM-04 (B04) by adding two courses of chemotherapy and prolonging the infusion time of highdose methotrexate. In 2010, a modified DA-EPOCH-R regimen was recommended for PMBCL by the NHL-BFM study committee on the basis of a 5-year EFS of 93% in adults with PMBCL in a phase II study.4 The modifications included the addition of a least one dose of intrathecal triple therapy (ITT), and a cumulative doxorubicin

Table 1. Clinical characteristics of the children and adolescents with primary mediastinal large B cell lymphoma.

Study B04 N95 REG12 Sex f m Stage** III IV not evaluable* unknown CNS involvement not analyzed no Bone marrow involvement not analyzed no yes Age at diagnosis (years) mean range LDH at diagnosis (U/L) mean range above normal range < 500 500 – <1,000 ≥ 1,000 Duration of follow-up (months) mean range

All eligible Patients (N=116)

N95 (N=20)

Treatment B04 (N=29)

DA EPOCH R (N=67)

45 19 52

19 1

29 -

16 51

62 54

7 13

17 12

38 29

94 1 19 2

20 -

19 1 9 -

55 10 2

16 100

8 21

8 59

11 104 1

20 -

4 24 1

7 6 -

15.8 1.4–21.7

14.7 1.4–17.9

15.7 10.3–18.6

16.2 8.4–21.7

562 187–1,698 89/96 56 47 13

445 187–1,267 unknown*** 14 5 1

608 252–1,322 25/29 (86%) 13 12 4

578 188–1,698 64/67 (96%) 29 30 8

59 2–211.8

77 2–211.8

73 12.5–144.2

48 7.6–123.2

*no initial assessment of central nervous system (CNS) or bone marrow involvement; **St. Jude staging system;15 ***upper limit of normal not reported in the study N95.NHL: N95: study NHL-BFM 95; B04: study B-NHL BFM 04; REG12: NHL-BFM Registry 2012; f: female; m: male; LDH: lactate dehydrogenase; DA-EPOCH-R: dose-adjusted etoposide, prednisone, cyclophosphamide, doxorubicin, and rituximab.

3232

haematologica | 2021; 106(12)

Letters to the Editor

Figure 1. Patient allocation and treatment assignment. Patients with diagnosis of primary mediastinal large B-cell lymphoma (PMBCL) were identified from 3 trials. Two patients received treatment not according to protocol and were excluded from the analysis. One patient from the non-Hodgkin lymphoma BerlinFrankfurt-Münster (NHL-BFM) Registry 2012 (REG12) received treatment according to the NHL-BFM 95 treatment strategy (R2).

dose limit at 360 mg/m2 of body-surface area (BSA). A first analysis by our group of 15 patients treated with DA-EPOCH-R showed an EFS and OS of 92±8% after 2 years.5 A retrospective analysis of 156 adults and children with PMBCL treated with DA-EPOCH-R reports an EFS of 86% at 3 years.6 However, in the prospective Intergroup trial testing DA-EPOCH-R, the 2-year EFS of children and adolescents with PMBCL was 72%, not different from the historical control.7 We analyzed children and adolescents with PMBCL confirmed by central histopathological review, excluding mediastinal grey zone lymphoma, enrolled in the B04 trial (German clinical trial registry: DRKS00009436) or the NHL-BFM Registry 2012 between 2004 and 2019 to i) assess the efficacy of intensified B-NHL-BFM chemotherapy (n=29 patients) and modified DA-EPOCH-R (n=67 patients), ii) compare it retrospectively to the treatment regimen in the NHL-BFM 95 trial (N95, n=20 patients) and iii) identify risk factors for treatment failure with DA-EPOCH-R. Treatment details for N95, B04 and DA-EPOCH-R are summarized in the Online Supplementary Table S1. N95 treatment was stratified by lactate dehydrogenase (LDH) and stage to risk groups R2–R4, as previously reported.8 In B04, treatment was intensified by adding two courses: patients with LDH <500 U/L received six (PMBCL6), those with LDH ≥500 U/L seven (PMBCL7) 5-day courses of chemotherapy including high-dose methotrexate infused over 24 hours (h) and ITT. Outside the protocol, three patients received one or two doses of rituximab and one patient received initial emergency-mediastinal radiotherapy. One patient each after B04 and DA-EPOCH-R received radiotherapy for a persisting mediastinal mass. From September 2010, DA-EPOCH-R was recommended with the described modifications. Erroneously, 60 mg/m² prednisone was used instead of 120 mg/m² as protocolspecified for 26 patients. The primary endpoint was the EFS at 5 years, defined as time from diagnosis to death, relapse, progressive disease, or secondary malignancy, estimated using the haematologica | 2021; 106(12)

Kaplan-Meier method. OS was defined as time from diagnosis to death. Survival and competing risk comparisons were performed by log-rank analysis and Gray´s test.9 Data were updated as of January 3, 2021. For this analysis, 116 of 118 registered patients were included (Figure 1). Their median age was 16.2 years, 53% were female. Patient characteristics are summarized in Table 1. Fifteen patients in the trial N95 and 15 patients treated by DA-EPOCH-R enrolled in B04 have been reported previously.2,5,8 Of 20 patients treated according to N95, six patients with LDH levels at diagnosis ≥500 U/L received the intended treatment (R3/R4). Of 14 patients with LDH <500 U/L, eight received the protocol-intended treatment with four courses (R2) and six were treated more intensively (R3/R4). B04 therapy was used for 29 patients, of whom 12 with LDH <500 U/L were scheduled for six courses, one for seven. All 16 patients with LDH ≥500 U/L received the intended treatment with seven courses. Among 67 patients treated by DA-EPOCH-R, 15 received pretreatment other than one dose of rituximab or a BFM-type prephase (B04 chemotherapy in 13 patients - 1 course A24 in 2, 1 course AA24 in 10 patients, 2 courses AA24 and BB24 in 1 patient, 2 courses of OEPA and one course of R-CHOEP in 1 patient each). Fifty-two patients without pretreatment received six (n=50) or eight (n=2) courses of DA-EPOCH-R. The mean cumulative doxorubicin dose was 310 mg/m2 of BSA (range, 200–415 mg/m2). The median number of ITT was 2.5 (range, 0–8) in 50 patients with available data. The maximal dose levels reached were 1, 2, 3, 4, and 5 in 5 (10%), 6 (12%), 17 (34%), 17 (34%) and 5 (10%) patients, respectively, and unknown in two patients. The levels reached were slightly lower than reported by Dunleavy and colleagues.4 Dose decisions were at the discretion of the treating physicians, and reasons for nonescalation might include concerns for sequelae, overestimation of hematological toxicity due to frequent blood counts for dose decisions, or the fact that G-CSF was not 3233

Letters to the Editor

Figure 2. Event-free survival and overall survival at 5 years for patients with primary mediastinal large B-cell lymphoma treated with the treatment regimen NHL-BFM 95, B-NHL-BFM 04 or DA-EPOCH-R. (A) Event-free survival (EFS) and (B) overall survival for patients with primary mediastinal large B-cell lymphoma (PMBCL) according to the type of treatment. EFS was significantly different between DA-EPOCH-R and B-non-Hodgkin lymphoma Berlin-Frankfurt-Münster (B-NHL-BFM) B04 (P=0.024) and DA-EPOCH-R and NHL-BFM 95 (N95) (P<0.001). The difference between 04 and N95 was not significant (P=0.142). DA-EPOCHR: dose-adjusted chemo-immunotherapy etoposide, prednisone, cyclophosphamide, doxorubicin, and rituximab.

administered to all patients (only 39 of 46 (85%) of patients with available data). In 15 pretreated patients, the median number of DAEPOCH-R-courses was five, the median number of ITT was five, and the mean cumulative dose of doxorubicin was 260 mg/m2 BSA. For treatment by DA-EPOCH-R, B04 and N95, estimates for EFS at 5 years were 84% (95% confidence interval [CI]: 72–91), 59% (95% CI: 39–74), and 39% (95% CI: 19–60), respectively (Figure 2). OS 90% (95% CI: 79–95), 72% (95% CI: 51–85) and 70% (95% CI: 45– 85), respectively (Figure 2). EFS and OS with DA-EPOCH-R were significant superior to treatment with B04 (P=0.016 for EFS, P=0.039 for OS) and N95 P<0.001 for EFS and P=0.026 for OS). The observed EFS with DA-EPOCH-R was comparable to that of other trials ranging from 72% to 93%.4,6,7,10,11 To what extent rituximab alone contributed to the superior outcome cannot be answered by our data. The addition of rituximab to CHOP improved outcomes in adult patients with PMBCL.12 Recent preliminary data from the non-randomized, prospective IELSG37 trial suggest similar efficacy for DA-EPOCH-R and R-CHOP14.13 The AEIOP reported 13 pediatric PMBCL patients treated with a modified MTX-based BFM-type backbone combined with rituximab resulting in an EFS of 84%.14 These data indicate that addition of rituximab contributed substantially to the improved outcome. Estimated EFS at 5 years for patients with LDH <500 U/L receiving PMBCL6, R3/R4, and R2 in B04/N95 were 67% (95% CI: 34–86), 67% (95% CI: 19–90), and 19% (95% CI: 1–54), respectively (Online Supplementary Figure S1A), with a significant difference between PMBCL6 and R2 (P=0.047). PMBCL7 was given to 16 patients with LDH ≥500 U/L in B04, R3/R4 in N95 to eight patients. The estimated EFS was 50% (95% CI: 25–71) and 33% (95% CI: 5–68), respectively (P=0.45, Online 3234

Supplementary Figure S1B). The improvement with intensified B-NHL therapy in patients with LDH levels <500 U/L but not among those with LDH levels ≥500 U/L indicates a possible limit for further improvements by modifying standard B-NHL chemotherapy for PMBCL. In patients treated by DA-EPOCH-R without pretreatment, EFS and OS at 5 years were 87% (95% CI: 74–93) and 91% (95% CI: 78–97), not significantly different from the outcome for 15 patients receiving DA-EPOCHR after pretreatment (with an EFS and OS of 73% (95% CI: 44–89) and 86% (95% CI: 55–96), respectively (P=0.2 for EFS, P=0.54 for OS, Online Supplementary Figure S2). The heterogeneity in treatment with pretreatment in about 20% of patients is a limitation of our analysis, but likely reflects real-world diagnostic uncertainties, with a final diagnosis of PMBCL only made by central histopathological review in conjunction with the typical location. There was no significant difference in EFS according to sex, initial LDH, extra-thoracic involvement, prednisone dose or the maximal dose-level reached in DA-EPOCH-R. Mean age was lower in patients experiencing relapse (hazard ratio [HR]: 0.74, P=0.012), resulting in an EFS of 90% (95% CI: 76–96) for 41 patients ≥16 years, compared with 73% (95% CI: 52–86) for 26 patients <16 years (P=0.07). The limited number of patients might explain that we could not identify risk factors for treatment failure with DA-EPOCH-R except for younger age. At relapse four of 11 (37%) patients treated by DA-EPOCH-R had parenchymal CNS involvement compared to zero of 22 after B04 chemotherapy (Gray´s test, P=0.08). Three of these patients had received only 60 mg/m2 prednisolone, two reached only dose level 1 or 2 and one received only one ITT for CNS prophylaxis. Further explanations for a possibly higher risk of CNSrelapse after DA-EPOCH-R include the use of prednisone haematologica | 2021; 106(12)

Letters to the Editor

instead of dexamethasone and the omission of high-dose methotrexate, both part of the B-NHL therapy. In conclusion, our prospective data confirmed DA-EPOCH-R as effective treatment for children and adolescents with PMBCL with only one patient receiving consolidation radiotherapy. Further trials on PMBCL should address the risk of CNS relapse and identify prognostic markers. Low patient numbers in this orphan disease call for collaborative, international trials including patients of the whole age spectrum. Fabian Knörr,1 Martin Zimmermann,2 Andishe Attarbaschi,3 Edita Kabíčková,4 Britta Maecker-Kolhoff,2 Stephanie Ruf,5 Ingrid Kühnle,6 Martin Ebinger,7 Anne-Kathrin Garthe,8 Ingrid Simonitsch-Klupp,9 Ilske Oschlies,10 Wolfram Klapper,10 Birgit Burkhardt8 and Wilhelm Woessmann1 1 University Medical Center Hamburg-Eppendorf, Pediatric Hematology and Oncology, Hamburg, Germany; 2Hannover Medical School, Clinic for Pediatric Hematology and Oncology, Hannover, Germany; 3St. Anna Children’s Hospital, Medical University of Vienna, Department of Pediatric Hematology and Oncology, Vienna, Austria; 4Charles University and University Hospital Motol, Department of Pediatric Hematology and Oncology, Prague, Czech Republic; 5Justus-Liebig-University of Gießen, Pediatric Hematology and Oncology, Gießen, Germany; 6University Medical Center Göttingen, Division of Pediatric Hematology and Oncology, Göttingen, Germany; 7Children's University Hospital Tübingen, Department of General Pediatrics and Pediatric Hematology/Oncology, Tübingen, Germany; 8University-Hospital of Münster, Pediatric Hematology and Oncology, Münster, Germany; 9Clinical Institute of Pathology, Medical University of Vienna, Vienna, Austria; 10University Hospital SchleswigHolstein, Campus Kiel, Institute of Pathology, Hematopathology Section and Lymph Node Registry, Kiel, Germany Correspondence: WILHELM WOESSMANN- w.woessmann@uke.de doi:10.3324/haematol.2021.278971 Received: April 28, 2021. Accepted: August 27, 2021. Pre-published: September 9, 2021. Disclosures: AA received honoraria from Jazz Pharmaceuticals, Amgen, MSD, Novartis, travel grants from Jazz Pharmaceuticals, and has consulting or advisory roles for Jazz Pharmaceuticals, Amgen, MSD, Gilead and Novartis; ME has received travel grants and honoraria from Amgen, MSD and Jazz Pharmaceuticals; WK received research funding by Roche, Amgen, Regeneron and Takeda. The remaining authors have no conflicts of interest to disclose. Contributions: WW, BB, AA, EK designed and supervised treatment in the NHL-BFM registry 2012. WW and FK designed the concept and the analysis; FK, MZ, SR and AG collected and assembled data; AA, EK, BMK, IK, ME provided patient data; ISK, IO, WK performed reference pathology review; FK and WW wrote the first draft

haematologica | 2021; 106(12)

of the manuscript; MZ and FK carried out the statistical analysis; WW and BB supervised the analysis; all authors critically revised the manuscript; all authors gave their approval of the final manuscript. Acknowledgments: we thank all patients and their families for participating in the studies. We thank our colleagues in the hospitals and reference institutions, who contributed to this study, for their care for the children and families, and the supplied data.

References 1. Gerrard M, Waxman IM, Sposto R, et al. Outcome and pathologic classification of children and adolescents with mediastinal large B-cell lymphoma treated with FAB/LMB96 mature B-NHL therapy. Blood. 2013;121(2):278-285. 2. Seidemann K, Tiemann M, Lauterbach I, et al. Primary mediastinal large B-cell lymphoma with sclerosis in pediatric and adolescent patients: treatment and results from three therapeutic studies of the BerlinFrankfurt-Münster Group. J Clin Oncol. 2003;21(9):1782-1789. 3. Cairo MS, Sposto R, Gerrard M, et al. Advanced stage, increased lactate dehydrogenase, and primary site, but not adolescent age (≥ 15 years), are associated with an increased risk of treatment failure in children and adolescents with mature B-cell non-Hodgkin's lymphoma: results of the FAB LMB 96 study. J Clin Oncol. 2012;30(4):387-393. 4. Dunleavy K, Pittaluga S, Maeda LS, et al. Dose-adjusted EPOCH-rituximab therapy in primary mediastinal B-cell lymphoma. N Engl J Med. 2013;368(15):1408-1416. 5. Woessmann W, Lisfeld J, Burkhardt B. Therapy in primary mediastinal B-cell lymphoma. N Engl J Med. 2013;369(3):282-284. 6. Giulino-Roth L, O'Donohue T, Chen Z, et al. Outcomes of adults and children with primary mediastinal B-cell lymphoma treated with doseadjusted EPOCH-R. Br J Haematol. 2017;179(5):739-747. 7. Burke GAA, Gross TG, Pillon M, et al. Results of Inter-B-NHL Ritux 2010 - phase II study of DA-EPOCH-R for children and adolescents with primary mediastinal large B-cell lymphoma (PMLBL) on behalf of European Intergroup for Childhood Non Hodgkin's Lymphoma (EICNHL) and Children's Oncology Group (COG). Blood. 2017;130(Suppl 1):S4124. 8. Woessmann W, Seidemann K, Mann G, et al. The impact of the methotrexate administration schedule and dose in the treatment of children and adolescents with B-cell neoplasms: a report of the BFM Group Study NHL-BFM95. Blood. 2005;105(3):948-958. 9. Gray RJ. A Class of K-Sample Tests for comparing the cumulative incidence of a competing Risk. Ann Stat. 1988;16(3):1141-1154. 10. Shah NN, Szabo A, Huntington SF, et al. R-CHOP versus dose-adjusted R-EPOCH in frontline management of primary mediastinal B-cell lymphoma: a multi-centre analysis. Br J Haematol. 2018;180(4):534-544. 11. Melani C, Advani R, Roschewski M, et al. End-of-treatment and serial PET imaging in primary mediastinal B-cell lymphoma following doseadjusted EPOCH-R: a paradigm shift in clinical decision making. Haematologica. 2018;103(8):1337-1344. 12. Rieger M, Österborg A, Pettengell R, et al. Primary mediastinal B-cell lymphoma treated with CHOP-like chemotherapy with or without rituximab: results of the Mabthera International Trial Group study. Ann Oncol. 2011;22(3):664-670. 13. Martelli M, Zucca E, Botto B, et al. Impact of different induction regimens on the outcome of primary mediastinal B cell lymphoma in the Prospective Ielsg 37 Trial. Hematol Oncol. 2021;39(S2):S90-92. 14. Pillon M, Carraro E, Mussolin L, et al. Primary mediastinal large B-cell lymphoma: outcome of a series of pediatric patients treated with highdose methotrexate and cytarabine plus anti-CD20. Pediatr Blood Cancer. 2018;65(2). 15. Murphy SB. Classification, staging and end results of treatment of childhood non-Hodgkin's lymphomas: dissimilarities from lymphomas in adults. Semin Oncol. 1980;7(3):332-339.

3235

Letters to the Editor

Lower respiratory tract infection with Staphylococcus aureus in sickle-cell adult patients with severe acute chest syndrome - the STAPHACS Study Acute chest syndrome (ACS) is the most common acute pulmonary complication of sickle cell disease (SCD).1 It may progress to a life-threatening event requiring the use of mechanical ventilation, with a mortality rate ranging from 3%1 to 50% when acute respiratory distress syndrome develops.2 Lung infection may account for 30% of the aetiologies of ACS.1 The causal relationship between Staphylococcus aureus (S. aureus) and ACS has been described for many years.3 However, this microorganism has been identified in only 4% of cases in the largest series published to date.1 To our knowledge, there is no specific study focusing on acute lower respiratory tract infection (LRTI) associated with S. aureus in SCD adult patients with ACS, in the pneumococcal vaccine era. The objectives of this pilot study were to describe S. aureus LRTI in SCD patients with severe ACS, in terms of prevalence, clinical and laboratory findings and outcomes. We conducted a retrospective observational study from April 2015 to December 2017 in SCD patients with ACS admitted to the intensive care unit (ICU) of Tenon Hospital, Paris, France, a tertiary university hospital and referral center for SCD. The definition of ACS combined fever or chest pain and a new pulmonary infiltrate of at least one segment on thoracic imaging. ACS was considered to be associated with S. aureus (alone or associated with another microorganism) when respiratory tract

samples or blood cultures yielded S. aureus, in the absence of any identifiable clinical source other than the lung. Patients with ACS associated with S. aureus (S. aureus group) were compared to patients with ACS in whom another microorganism was identified or in whom no microbiological documentation was obtained despite a comprehensive microbiological workup (control group). The workup included i) respiratory tract samplings (sputum, tracheal aspirate [TA], or bronchoalveolar lavage [BAL]) with Gram staining and quantitative culture for bacterial microorganisms; ii) blood cultures; iii) urinary antigen testing for Streptococcus pneumoniae and Legionella pneumophila. Additionally, a respiratory multiplex polymerase chain reaction (mPCR) test (FilmArrayTM Respiratory Panel system) has been available in our unit since 2016. This study was conducted in accordance with the French law, and was approved by the Ethical Review Board of the Société de Pneumologie de Langue Française (CEPRO 2019-021). During the study period, 119 episodes of ACS were recorded in 114 patients. Forty-two episodes (40 patients) were excluded from the analysis because of incomplete microbiological investigation (Figure 1). Overall, S. aureus was identified in 29 of 119 episodes (24%), including respiratory tract samples cultures in 28 episodes, and blood culture in one episode (Table 1). Bacterial and viral co-infections were respectively diagnosed in four (14 %) and three (10%) episodes in the S. aureus group. The overall distribution of the ACS episodes associated with S. aureus was sporadic throughout the year (Online Supplementary Figure S1). More

Table 1. Microbiological investigations. Microbiological investigations performed, n (%) Sputum Tracheal aspirate Broncho-alveolar lavage Blood culture Streptococcus pneumoniae urinary antigen test¤ Legionella pneumophila urinary antigen test¤ Chlamydophila pneumoniae serology Mycoplasma pneumoniae serology Nasopharyngeal swab (multiplex PCR)† Parvovirus B19 serology Microbial identification, n (%) Respiratory tract MSSA£ MRSA£ Other bacterial microorganism Respiratory virus† Blood

S. aureus ACS group, N=29

Control ACS group, N=48

26 (90) 1 (3) 1 (3) 28 (97) 24 (83) 26 (90) 7 (24) 8 (28) 19 (66) 11 (38)

47 (98) 7 (15) 3 (6) 47 (98) 47 (98) 47 (98) 8 (17) 9 (19) 31 (65) 14 (29)

0.15* 0.25* 0.99* 0.99* 0.03 0.15* 0.42 0.37* 0.93 0.43

27 (93) 2 (7) 4 (14)§ 3 (10)¶ 1 (3)

N/A N/A 10 (21)# 5 (10)‡ 3 (7)$

0.44 0.99* 0.99*

Finally, all the patients had at least one respiratory tract sample, except 1 patient in the Staphylococcus aureus ( group (in whom S. aureus was identified in blood culture). ¤Urinary antigen testing for Streptococcus pneumoniae and Legionella pneumophila was the BinaxNOW kits (Alere, Jouy en Josas, France). £MSSA: Methicillin-sensitive S. aureus; MRSA: Methicillin-resistant S. aureus. §Streptococcus pneumoniae (n=2); Klebsiella pneumoniae (n=1); Mycoplasma pneumoniae (n=1). #Chlamydophila pneumoniae (n=2); Streptococcus pneumoniae (n=2); Klebsiella pneumonia (n=1); Streptococcus alpha-hemolyticus (n=1); Streptococcus agalactiae (n=1); Enterobacter aerogenes (n=1); Citrobacter freundii (n=1); Salmonella typhimirium (n=1). †Respiratory viruses were detected using the respiratory multiplex polymerase chain recation (mPCR) panel (FilmArrayTM Respiratory Panel system, BioFire®, Salt Lake City, UT) including 17 respiratory viruses (coronaviruses, adenovirus, human metapneumovirus, human enterovirus/rhinovirus, respiratory syncytial virus, parainfluenza viruses and influenza viruses A and B) and 3 bacteria (Chlamydophila pneumoniae, Mycoplasma pneumoniae, and Bordetella pertussis). ¶ Rhinovirus (n=1); Coronavirus 229E (n=1), Influenza B (n=1). ‡Rhinovirus (n=3); Enterovirus (n=1); Influenza A virus (n=1); Parvovirus B19 (n=1); one patient had 2 respiratory viruses (Enterovirus + Influenza A virus). $Salmonella typhimirium (n=1); Streptococcus pneumoniae (n=1); Citrobacter freundii (n=1). Data are presented as median [first through third quartiles] or number (%). Continuous variables are compared using a Wilcoxon method; categorical variables are compared either using a c2 test or Fisher’s exact test when followed by (*). ACS: acute chest syndrome.

3236

haematologica | 2021; 106(12)

Letters to the Editor

specifically, S. aureus was identified during the flu season in 24% of episodes (Online Supplementary Table S1), a rate that did not differ from that of the other episodes of ACS (35%; P=0.3). The baseline characteristics were similar between the two groups, except the presence of a more frequent prior history of S. aureus infections in the S. aureus group, as compared with the control group (28% vs. 6%; P=0.01) (Online Supplementary Table S1). Despite similar rates of influenza and pneumococcal vaccination, the pneumococcal vaccination strategies differed (13-valent pneumococcal vaccine only: none in the S. aureus group vs. 17% in the control group; P=0.02). The characteristics, management and outcomes of ACS were also similar between the two groups. Post-hospital outcomes marginally differed, in terms of number of and time to hospital readmission for vaso-occlusive crisis or ACS (Online Supplementary Figure S2). Among the 29 ACS episodes with S. aureus, 21 isolates (72%) were sent to the National Staphylococcus Reference Center for genotyping analysis. S. aureus strains were genetically diverse, covering the four accessory gene regulator groups and assigned to 12 clonal complexes (CC) (Table 2). At least one toxin gene was found in 13 isolates (62%); two methicillin-sensitive (MSSA) isolates (10%) had a gene coding for pantonvalentine leukocidin (PVL). This study underlines the importance and the clonal

diversity of S. aureus during severe episodes of ACS. Although unrelated to the influenza epidemic, ACS associated with S. aureus appeared inversely related to the pneumococcal vaccination strategy, raising the question of how the pneumococcal vaccination may affect the nasopharyngeal colonization of those patients. A history of S. aureus infection was associated with the subsequent development of a documented ACS episode with S. aureus. In our study, one quarter of the episodes of ACS were associated with S. aureus, although this microorganism is infrequently involved in community-acquired LRTI4 and in ACS.1 Current guidelines advise using antibiotics targeting S. pneumoniae and intracellular bacteria.5 In our series, S. pneumoniae was identified in only four episodes of ACS (accounting for 5% of episodes with a complete microbiological investigation) including two co-infections with S. aureus. This low proportion of S. pneumoniae is probably due to the good pneumococcal vaccination coverage in our cohort. One hypothesis explaining the increase in the prevalence of S. aureus could be a decrease in the carriage of S. pneumoniae related to the extensive vaccination in this fragile population. The effectiveness of pneumococcal vaccination has led to the decrease in invasive pneumococcal infection in SCD patients, and may have changed the prevalence of S. aureus infection in this population.6 An inverse relationship between the oropharyngeal carriage of S. pneumoniae and S. aureus

Figure 1. Selection of the episodes of acute chest syndrome. §The microbiological investigation was incomplete in 42 episodes, including 26 episodes (25 patients) with no respiratory tract samples, 14 episodes (14 patients) with no blood culture, and 16 episodes (14 patients) with no pneumococcal and legionella urinary antigen testing. Altogether, 77 episodes (in 74 patients) with a complete microbiological investigation were analyzed, including 29 episodes (24%) associated with Staphylococcus aureus (S. aureus) in 28 patients (1 patient had two episodes of acute chest syndrome [ACS] associated with S. aureus), and 48 episodes (in 49 patients) associated either with another microorganism or with no microorganism.

haematologica | 2021; 106(12)

3237

Letters to the Editor

has already been suggested, and could be related to an inhibitory effect of S. pneumoniae on S. aureus via the production of hydrogen peroxide.7 These findings raise the question of whether oropharyngeal carriage may be associated with a risk for developing ACS with S. aureus. In the general population, an increase in the prevalence of S. aureus carriage over a prolonged period has been suggested after pneumococcal vaccination.8 Moreover, S. aureus infection might be more common in patients who are colonized.9 Some series have highlighted the risk of S. aureus oropharyngeal carriage and the risk of S. aureus infection in the general population9 as well as in the critical care setting.10 A sickle-cell pediatric series suggested that S. aureus colonization was also associated with a subsequent risk of S. aureus infection, without specifying the site of infection.11 In our study, a history of S. aureus infection was associated with the subsequent occurrence of ACS associated with this microorganism. This finding may also suggest a chronic carriage of S. aureus in this population, and the subsequent risk for developing another S. aureus infection. S. aureus has been implicated in influenza LRTI.12 In our study, 24% of the ACS episodes associated with S. aureus occurred during the flu season, a rate that did not differ from that of the other ACS episodes. The viral co-infection rate was 10%, but influenza was identified in only one episode, despite a broad search using mPCR (66% of patients with S. aureus documentation). In addiTable 2. Genotypic markers of Staphylococcus aureus.

Variables, n (%) Toxin TSST-1 Enterotoxins§ PVL Exfoliative toxin¶ agr allele I II III IV Clonal Complex CC15 MSSA CC12 MSSA CC8 MSSA CC8 MRSA CC398 MSSA CC188 MSSA CC152 MSSA CC121 MSSA CC97 MSSA CC96 MSSA CC88 MSSA CC45 MSSA CC30 MSSA

Strains, n=21 13 (62) 2 (10) 11 (52) 2 (10) 1 (5) 10 (48) 7 (33) 3 (14) 1 (5) 5 (24) 2 (10) 2 (10) 2 (10) 2 (10) 1 (5) 1 (5) 1 (5) 1 (5) 1 (5) 1 (5) 1 (5) 1 (5)

TSST: toxic shock syndrome toxin; PVL: Panton-Valentine leucocidin; agr: accessory gene regulator; MSSA: methicillin-sensitive Staphylococcus aureus; MRSA: methicillin-resistant Staphylococcus aureus; CC: clonal complexes. §SEA (n=2); SEB (n=3); SEC (n=2); SED (n=1); SEG (n=3); SEI (n=3); SEJ (n=1); SEL (n=2); SEM (n=1); SEN (n=1); SEO (n=1); SEP (n=4); SEJ (n=2); SEQ (n=1); SER (n=2); SEU (n=2); ¶ETA (n=1).

3238

tion, the ACS episodes associated with S. aureus were not preponderant during the flu season. We identified two MRSA, accounting for 3% of the S. aureus strains, a rate higher than that usually reported in community-acquired pneumonia,4 but lower than the 16% rate in a recent series concerning community acquired pneumoniae.13 SCD patients are regularly hospitalized and exposed to repeat antimicrobial therapies, which could partly explain this proportion. Nevertheless, a pediatric series suggested a similar rate of MRSA oropharyngeal carriage between SCD and non-SCD populations.14 Despite a higher initial clinical severity, the S. aureus group had similar short-term outcomes to the control group, a finding that may be related to the fact that early antimicrobial treatments were administered to all patients, as recommended.5 Genotypic analysis demonstrated a significant clonal diversity, with a 62% rate of toxin genes, similarly to colonizing and pathogenic S. aureus strains.15 While the PVL strains may be involved in necrotizing communityacquired pneumonia,13 the two ACS episodes were not associated with necrotizing pulmonary lesions, and had favorable outcomes. Our study has limitations inherent to all retrospective monocentric studies, and our results should be extrapolated with caution. The selection criteria may have underestimated the prevalence of S. aureus, but also the prevalence of other microorganisms, in particular intracellular bacteria. In some patients, the vaccination program may have been just initiated before the ACS episode, precluding any formal conclusion about the relationship between S. aureus ACS and the pneumococcal vaccination strategy. Last, the distinction between S. aureus colonization and S. aureus infection may have been difficult in some cases. Whether the identification of S. aureus is associated with colonization in ACS, and whether this colonization is associated with a subsequent risk for developing a new infection is uncertain. In order to answer this, it would be necessary to take sequential samples at baseline and during episodes of ACS. In this context, and due to the limitations of the current diagnostic tests, the value of quantitative PCR may help to distinguish between colonization and deep pulmonary infection. Moreover, decontamination of the S. aureus carriage sites could be useful. Finally, antibiotic therapy targeting MSSA on admission of patients with ACS having a prior S. aureus infection may be warranted. Alexandre Elabbadi,1 Guillaume Voiriot,1,2 Anne Tristan,3,4 Aude Gibelin,1 Charlotte Verdet,5 Michel Djibré,1 Aline Santin,6 Etienne-Marie Jutant,1 Julien Lopinto,1 François Vandenesch,3,4 François Lionnet6 and Muriel Fartoukh1,2 1 Sorbonne Université, Assistance Publique - Hôpitaux de Paris, Service de Médecine Intensive Réanimation, Hôpital Tenon, Paris; 2 Groupe de Recherche Clinique CARMAS, Collegium Galilée, Créteil; 3 Centre National de Référence des Staphylocoques, Institut des Agents Infectieux, Hospices Civils de Lyon, Lyon; 4Centre International de Recherche en Infectiologie, INSERM U1111, Université Lyon 1, École Normale Supérieure de Lyon, Lyon; 5Sorbonne Université, Assistance Publique - Hôpitaux de Paris, Service de Bactériologie, Hôpital SaintAntoine, Paris and 6Sorbonne Université, Assistance Publique Hôpitaux de Paris, Service de Médecine Interne, Centre de Référence de la Drépanocytose, Hôpital Tenon, Paris, France Correspondence: ALEXANDRE ELABBADI - alexandre.elabbadi@aphp.fr haematologica | 2021; 106(12)

Letters to the Editor

doi:10.3324/haematol.2021.278827 Received: March 23, 2021. Accepted: September 8, 2021. Pre-published: September 16, 2021. Disclosures: no conflicts of interest to disclose. Contributions: MF and AE collected, analyzed, and interpreted the data; AE and MF drafted the manuscript; MF, AE and GV contributed to the study conception and design; FV, AT and CV performed investigation of S. aureus strains; MF, GV, FL, FV, CV and AT critically revised the manuscript. All the authors read and approved the final manuscript.

References 1. Vichinsky EP, Neumayr LD, Earles AN, et al. Causes and outcomes of the acute chest syndrome in sickle cell disease. National Acute Chest Syndrome Study Group. N Engl J Med. 2000;342(25):1855-1865. 2. Cecchini J, Boissier F, Gibelin A, et al. Pulmonary vascular dysfunction and cor pulmonale during acute respiratory distress syndrome in sicklers. Shock. 2016;46(4):358-364. 3. Charache S, Scott JC, Charache P. “Acute chest syndrome” in adults with sickle cell anemia. Microbiology, treatment, and prevention. Arch Intern Med. 1979;139(1):67-69. 4. Self WH, Wunderink RG, Williams DJ, et al. Staphylococcus aureus community-acquired pneumonia: prevalence, clinical characteristics, and outcomes. Clin Infect Dis. 2016;63(3):300-309. 5. Habibi A, Arlet J-B, Stankovic K, et al. [French guidelines for the management of adult sickle cell disease: 2015 update]. Rev Med Interne. 2015; 36(5 Suppl 1):S5S3-84.

haematologica | 2021; 106(12)

6. Soothill G, Darboe S, Bah G, Bolarinde L, Cunnington A, Anderson ST. Invasive bacterial infections in Gambians with sickle cell anemia in an era of widespread pneumococcal and hemophilus influenzae type b vaccination. Medicine (Baltimore). 2016;95(49):e5512. 7. Shak JR, Vidal JE, Klugman KP. Influence of bacterial interactions on pneumococcal colonization of the nasopharynx. Trends Microbiol. 2013;21(3):129-135. 8. Spijkerman J, Prevaes SMPJ, van Gils EJM, et al. Long-term effects of pneumococcal conjugate vaccine on nasopharyngeal carriage of S. pneumoniae, S. aureus, H. influenzae and M. catarrhalis. PLoS One. 2012;7(6):e39730. 9. von Eiff C, Becker K, Machka K, Stammer H, Peters G. Nasal carriage as a source of Staphylococcus aureus bacteremia. Study Group. N Engl J Med. 2001;344(1):11-16. 10. Paling FP, Hazard D, Bonten MJM, et al. Association of Staphylococcus aureus colonization and pneumonia in the intensive care unit. JAMA Netw Open. 2020;3(9):e2012741. 11. Rocha LC, Carvalho MOS, Nascimento VML, et al. Nasopharyngeal and oropharyngeal colonization by Staphylococcus aureus and Streptococcus pneumoniae and prognostic markers in children with sickle cell disease from the Northeast of Brazil. Front Microbiol. 2017;8:217. 12. Kallen AJ, Brunkard J, Moore Z, et al. Staphylococcus aureus community-acquired pneumonia during the 2006 to 2007 influenza season. Ann Emerg Med. 2009;53(3):358-365. 13. Gillet Y, Tristan A, Rasigade J-P, et al. Risk factors of severity in community-acquired staphylococcal pneumonia. MedRxiv. 2020:20162875. 14. Schaumburg F, Biallas B, Alabi AS, et al. Clonal structure of Staphylococcus aureus colonizing children with sickle cell anaemia and healthy controls. Epidemiol Infect. 2013;141(8):1717-1720. 15. Becker K, Friedrich AW, Lubritz G, Weilert M, Peters G, Von Eiff C. Prevalence of genes encoding pyrogenic toxin superantigens and exfoliative toxins among strains of Staphylococcus aureus isolated from blood and nasal specimens. J Clin Microbiol. 2003;41(4):1434-1439.

3239

Letters to the Editor

Oral azacitidine preserves favorable level of fatigue and health-related quality of life for patients with acute myeloid leukemia in remission: results from the phase III, placebo-controlled QUAZAR AML-001 trial Despite relatively high remission rates with intensive chemotherapy (IC), most patients with acute myeloid leukemia (AML) will relapse, and overall survival (OS) in relapsed AML is dismal.1 In the phase III, placebo-controlled QUAZAR AML-001 trial, oral azacitidine (OralAZA [CC-486]) significantly prolonged OS versus placebo (P=0.0009; median 24.7 vs. 14.8 months from randomization) and relapse-free survival (RFS) (P=0.0001; 10.2 vs. 4.8 months) as maintenance therapy for patients with AML in first remission after intensive chemotherapy (IC), and was associated with a manageable safety profile.2 Health-related quality of life (HRQoL) and fatigue generally improve over time for patients with AML in remission; an ideal maintenance treatment should prolong survival without compromising HRQoL.3,4 The impact of Oral-AZA on patient-reported fatigue and HRQoL, a key secondary endpoint in QUAZAR AML-001, was assessed using the self-administered Functional Assessment of Chronic Illness Therapy (FACIT)-Fatigue Scale and EuroQoL EQ-5D-3L instruments. We hypothesized that Oral-AZA treatment would not meaningfully worsen fatigue or overall HRQoL from baseline, and that mean changes from baseline in fatigue and HRQoL scores in the Oral-AZA arm would be comparable (i.e., not inferior) to those in the placebo arm. Topline HRQoL outcomes of this study are described briefly elsewhere.2 At study entry, patients reported generally favorable levels of fatigue and overall HRQoL. Mean FACIT-Fatigue and EQ-5D-3L health utility index (HUI) scores remained similar to baseline over time during Oral-AZA treatment, with similar changes between the Oral-AZA and placebo arms.2 We describe previously unreported HRQoL results from QUAZAR AML-001, including longitudinal analyses using linear mixed-effect models for repeated measures (MMRM), outcomes in patient subgroups defined by prognostic baseline characteristics, and rates of clinically meaningful deterioration in HRQoL scores. QUAZAR AML-001 was a randomized, double-blind, placebo-controlled phase III trial. Study design and endpoints are reported in detail elsewhere.2 Briefly, patients aged ≥55 years, with intermediate- or poor-risk cytogenetics at diagnosis, ECOG PS ≤3, and ineligible for transplant, were randomized to Oral-AZA 300 mg or placebo once-daily for 14 days/28-day cycle within 4 months after achieving first CR or CR with incomplete hematologic recovery (CRi) with IC (induction ± consolidation). Patients who relapsed on-study with 5-15% blasts could

receive an escalated 21-days/cycle dosing schedule at the discretion of the treating investigator. The FACIT-Fatigue Scale is a 13-item questionnaire that measures an individual’s level of fatigue during daily activities over the previous week. The EQ-5D-3L is a generic instrument that includes a descriptive questionnaire that assesses impairment across five dimensions (mobility, self-care, pain/discomfort, usual activities, anxiety/depression) at three severity levels (none, moderate, severe), and a visual analogue scale (VAS) that asks patients to rate their perceived HRQoL from 0-100. Higher scores indicate lower fatigue (FACIT-Fatigue) and better health state (EQ-5D-3L). Both instruments were completed on day 1 of each cycle and end-of-treatment (EOT). HRQoL-evaluable patients had non-missing assessments at baseline and ≥1 post-baseline visit. In order to interpret changes from baseline, we used predefined thresholds for clinically meaningful changes within/between treatment arms (i.e., minimally important differences [MID]) and at the individual level (i.e., responder definitions [RD]).5 Thresholds used to define clinically meaningful improvement and deterioration from baseline, respectively, were score changes of +3/–3 on the FACIT-Fatigue Scale;6 +0.08/–0.10 on the EQ-5D3L HUI;7,8 and +11/–11 on the EQ-5D VAS.8 MMRM models were performed to confirm the hypothesized non-inferiority of Oral-AZA and placebo;9 these models used an unstructured covariance matrix and included the intercept and visit as random effects, and treatment arm, randomization stratification factors,2 baseline HRQoL score, visit, baseline-by-visit interaction, and treatment-group-by-visit interaction as fixed effects. The dependent variable was change in HRQoL score from baseline. Non-inferiority of Oral-AZA versus placebo was demonstrated if the lower bound of the twosided 95% confidence interval (CI) of the between-group difference in the overall least-squares (LS) mean change from baseline was greater than the MID for deterioration at each assessment.5,10 Empirical cumulative distribution frequency (eCDF) curves were generated showing FACIT-Fatigue score changes from baseline for individual patients within each treatment arm at cycles 3, 6, 12, and 24, using the predefined RD for clinically meaningful improvement and deterioration (+3/–3 points). Time to confirmed deterioration was assessed for each patient from the time of randomization until the first of ≥ 2 consecutive visits with a change from baseline surpassing the RD for clinically meaningful deterioration, or until death. Time to confirmed deterioration was estimated using Kaplan-Meier product-limit methods and compared between treatment arms using a stratified Cox proportional hazards regression model with treatment group and baseline score as covariates.

Table 1. Mixed-effect models for repeated measures analyses: overall least-squares mean changes from baseline within in each arm, between-group differences in overall least-squares mean changes, and prespecified minimally important differences for each assessment.

Assessment

FACIT-Fatigue scale EQ-5D-3L health utility index EQ-5D visual analogue scale

Overall LS mean [95%CI] change from baseline Oral-AZA Placebo –0.60 [–2.19, 0.99] –0.01 [–0.03, 0.01] 2.64 [–0.59, 5.86]

0.29 [–1.44, 2.02] 0.00 [–0.02, 0.02] 3.59 [0.01, 7.17]

Difference in overall LS mean change, Oral-AZA vs. placebo, mean [95%CI]*

Prespecified MID for clinically meaningful worsening

–0.89 [–2.37, 0.59] –0.01 [–0.03, 0.01] –0.95 [–4.38, 2.47]

–3 –0.10 –11

*Mixed-effect models for repeated measures (MMRM) analyses confirmed the noninferiority of Oral-AZA effects on fatigue and overall HRQoL vs. placebo, as the lower bounds of the 95% confidence interval (CI) for between-group differences in least-square (LS) mean changes from baseline did not exceed the predefined minimally important difference (MID) for worsening on any assessment. AZA: azacitidine; FACIT: functional assessment of chronic illness therapy.

3240

haematologica | 2021; 106(12)

Letters to the Editor

Figure 1. Empirical cumulative distribution frequency curves of observed changes from baseline on FACIT-Fatigue scores for individual patients in the oral azacitidine and placebo arms at cycles 3, 6, 12 and 24. A positive change score includes an improvement from baseline. A change from baseline ≥3 was used to define clinically meaningful improvement and worsening. Odds ratio, 95% confidence interval (CI), and P values were estimated using Cochran-MantelHaenszel test, stratified by randomization stratification factors. ECDF: empirical cumulative distribution frequency; AZA: azacitidine; FACIT: functional assessment of chronic illness therapy; RD: responder definition.

haematologica | 2021; 106(12)

3241

Letters to the Editor

The FACIT-Fatigue-evaluable population comprised 225 of 238 patients (94.5%) randomized to Oral-AZA and 219 of 234 patients (93.6%) randomized to placebo, and the EQ-5D-3L–evaluable population included 225 and 217 patients, respectively. Baseline demographic and disease characteristics of HRQoL-evaluable patients were balanced between treatment arms (Online Supplementary Table S1). FACIT-Fatigue and EQ-5D-3L compliance rates

were >95% in both treatment arms at baseline and remained high (>85%) across postbaseline visits except at EOT (~65%), suggesting that HRQoL endpoints were unlikely to be confounded by missing data. Patientreported FACIT-Fatigue, EQ-5D-3L HUI, and EQ-5D VAS scores were comparable between treatment groups at baseline and similar to reference values from general populations in the United States (FACIT-Fatigue) and

Figure 2. Kaplan-Meier estimated times to confirmed deterioration from baseline. (A) FACIT-Fatigue scale. (B) EQ-5D-3L health utility index. (C) EQ-5D visual analogue scale scores. Time to definitive deterioration was defined as time from randomization to clinically meaningful deterioration sustained from ≥2 consecutive assessment visits. cAZA: azacitidine; FACIT: functional assessment of chronic illness therapy; CI: confidence interval; HR: hazard ratio.

3242

haematologica | 2021; 106(12)

Letters to the Editor

Germany (EQ-5D-3L) (Online Supplementary Table S2).2,11,12 Median treatment durations for HRQoL-evaluable patients were 12 cycles and 7 cycles in the Oral-AZA and placebo arms, respectively. As reported previously, there were no clinically meaningful differences in observed mean changes from baseline FACIT-Fatigue or EQ-5D-3L HUI scores within treatment arms, or between the Oral-AZA and placebo arms, at any postbaseline visit.2 Longitudinal MMRM analyses confirmed the non-inferiority of Oral-AZA effects on fatigue and overall HRQoL relative to placebo, as the lower bounds of the 95% CI for between-group differences in LS mean changes from baseline did not exceed the predefined MID for worsening on any instrument (Table 1). In subgroup analyses, observed mean HRQoL scores generally remained similar to baseline over time within each arm. Mean changes in FACIT-Fatigue, EQ-5D-3L HUI, and EQ-5D VAS scores were comparable between treatment arms within patient subgroups defined by cytogenetic risk at diagnosis (intermediate/poor), response after induction (CR/CRi), receipt of consolidation chemotherapy (yes/no), ECOG PS score (0-1/2-3), age (<65/65-74/≥75 years), and HRQoL domain score (<25th/25th-74th/≥75th percentile). Overall, 45 HRQoLevaluable patients experienced relapse with 5-15% blasts and received Oral-AZA for 21 days/cycle. Escalated OralAZA dosing was not associated with clinically meaningful differences in changes from baseline in mean FACITFatigue, EQ-5D-3L HUI, or EQ-5D VAS scores at any visit compared with 14-day Oral-AZA dosing. eCDF curves detailing individual FACIT-Fatigue changes from baseline in the Oral-AZA and placebo arms at cycles 3, 6, 12, and 24 generally overlapped, with similar proportions of patients reporting clinically meaningful improvement or deterioration at each visit (Figure 1). Proportions of patients with clinically meaningful deterioration for each measure were low in both treatment arms, and rates were similar between arms on each instrument at almost all post-baseline visits (Online Supplementary Figure S1); deterioration rates were significantly higher in the Oral-AZA arm at cycle 19 (EQ-5D VAS) and cycle 29 (FACIT-Fatigue), but these may have occurred by chance as these analyses did not include any adjustments for multiple testing. Times to confirmed deterioration were similar between the Oral-AZA and placebo arms on each instrument (Figure 2). Estimated median times to confirmed deterioration were 41 weeks for Oral-AZA and 44 weeks for placebo on the FACITFatigue (hazard ratio [HR]: 1.06; 95% CI: 0.80-1.40); 200 and 164 weeks, respectively, on the EQ-5D-3L HUI (HR: 0.91; 95% CI: 0.62-1.34); and not reached versus 136 weeks on the EQ-5D VAS (HR: 0.86; 95% CI: 0.61-1.22). Similar findings were observed when censoring patients at the time of death. While improving survival is the primary goal of AML treatment, systematic evaluation of the impact of treatment on HRQoL is essential because prolonged survival may be less meaningful if accompanied by drug-related HRQoL decrements. To our knowledge, QUAZAR AML001 is the first placebo-controlled study to prospectively investigate the impact of long-term maintenance therapy on HRQoL for patients with AML in remission post-IC. At study entry, these older patients (median age 68 years2) reported generally favorable levels of fatigue and overall HRQoL that were comparable to levels in general populations.11,12 Mean FACIT-Fatigue and EQ-5D-3L scores during Oral-AZA treatment remained at or above baseline levels at almost all post-baseline assessments, haematologica | 2021; 106(12)

and longitudinal MMRM analyses confirmed the noninferiority of Oral-AZA relative to placebo for preserving HRQoL. These HRQoL data are also consistent with the reported manageable safety profile and acceptable tolerability of Oral-AZA in QUAZAR AML-001.2 A potential limitation of this study was that HRQoL assessments were conducted on day 1 of each 28-day treatment cycle, allowing for 14 days of recovery after each 14-day dosing period. Additionally, patients in both arms had to undergo routine hospital visits, testing, and marrow collections, which could potentially negatively affect HRQoL outcomes compared with an “observationonly” approach during AML remission. Oral-AZA administration offers a number of potential benefits, including optimal convenience for patients, no injection-site reactions, fewer clinic visits and lower associated costs, and treatment flexibility for long-term use. Findings from QUAZAR AML-001 show that Oral-AZA significantly improves OS and RFS without compromising fatigue or overall HRQoL for patients with AML in remission. Gail J. Roboz,1,2 Hartmut Döhner,3 Christopher Pocock,4 Hervé Dombret,5 Farhad Ravandi,6 Jun Ho Jang,7 Dominik Selleslag,8 Jiří Mayer,9 Uwe M. Martens,10 Jane Liesveld,11 Teresa Bernal,12 Ming Chung Wang,13 Peiwen Yu,14 Ling Shi,14 Shien Guo,14 Ignazia La Torre,15 Barry Skikne,16,17 Qian Dong,16 Julia Braverman,16 Salem Abi Nehme,15 C. L. Beach16 and Andrew H. Wei18 1 Weill Cornell Medical College, New York, NY, USA; 2New York Presbyterian Hospital, New York, NY, USA; 3Ulm University Hospital, Ulm, Germany; 4Kent & Canterbury Hospital, Canterbury, UK; 5Hôpital Saint-Louis, Assistance Publique - Hôpitaux de Paris (AP-HP) and Institut de Recherche Saint-Louis, Université de Paris, Paris, France; 6The University of Texas MD Anderson Cancer Center, Houston, TX, USA; 7Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea; 8 AZ Sint-Jan Brugge-Oostende AV, Bruges, Belgium; 9 University Hospital Brno, Brno, Czech Republic; 10SLK-Kliniken GmbH, MOLIT Institute for Personalized Medicine, Heilbronn, Germany; 11Wilmot Cancer Institute, University of Rochester, New York, NY, USA; 12Hospital Universitario Central de Asturias, Oviedo, Spain; 13Chang Gung Medical Foundation, Kaohsiung, Taiwan; 14 Evidera, Waltham, MA, USA; 15Celgene, a Bristol-Myers Squibb Company, Boudry, Switzerland; 16Bristol Myers Squibb, Princeton, NJ, USA; 17University of Kansas Medical Center, Kansas City, KS, USA and 18The Alfred Hospital and Monash University, Melbourne, Vicoria, Australia Correspondence: GAIL J. ROBOZ - gar2001@med.cornell.edu doi:10.3324/haematol.2021.279174 Received: May 7, 2021. Accepted: September 15, 2021. Pre-published: September 23, 2021. Disclosures: GJR reports consultancy or advisory board or data and safety monitoring committee of AbbVie, Actinium, Agios, Amphivena, Amgen, Argenx, Array Biopharma, Astex, Astellas, AstraZeneca, Bayer, Bristol Myers Squibb, Celgene, Celltrion, Daiichi Sankyo, Eisai, Epizyme, GlaxoSmithKline, Helsinn, Janssen, Jasper Therapeutics, Jazz, Mesoblast, MEI Pharma (IDMC Chair), Novartis, Orsenix, Otsuka, Pfizer, Roche/Genentech, Sandoz, Takeda (IRC Chair), Trovagene; and research support from Cellectis. HD reports personal fees from Abbvie, Agios, Astellas, Astex Pharmaceuticals, Helsinn, Janssen, Oxford Biomedicals, and Roche; grants and personal fees from Amgen, Celgene, Jazz Pharmaceuticals, and Novartis; and grants from AROG Pharmaceuticals, Bristol Myers 3243

Letters to the Editor

Squibb, Pfizer, and Sunesis. HD reports grants and personal fees from Celgene, Amgen, Incyte, Novartis, Jazz Pharmaceuticals, Daiichi Sankyo, Servier, and Astellas; and personal fees from Pfizer, Cellectis, Menarini, Otsuka, AbbVie, Janssen, Shire-Baxalta, Celyad, Agios, and Immunogen. FR reports honoraria and consulting fees from Bristol Myers Squibb and Celgene; and research funding from Bristol Myers Squibb. DS reports honoraria from Novartis, Celgene, Amgen, Janssen-Cilag, AbbVie, Alexion, GSK, MSD, Pfizer, Sanofi , Takeda, Incyte, and Teva; consultancy for Novartis, Celgene, Amgen, Janssen-Cilag, AbbVie, Alexion, GSK, MSD, Pfizer, Sanofi, Takeda, Incyte, and Teva; and speakers’ bureau participation for Novartis, Celgene, Amgen, MSD, Takeda, and Teva. JM reports research funding from Celgene. UMM reports consultancy for Bristol Myers Squibb, Merck, Amgen, Roche, and Celgene; and travel accommodations/expenses from Bristol Myers Squibb, Amgen, Pierre-Fabre, and Celgene. JL reports participation in DSMB for Onconova. PY, LS, and SG are employed by Evidera. BS is employed by Bristol Myers Squibb. ILT, QD, JB, SAN, and CLB are employed at and have equity ownership in Bristol Myers Squibb. AHW reports study-related fees and personal fees from Celgene; royalties from Walter and Eliza Hall Institute of Medical Research; grants from the Medical Research Future Fund; grants and personal fees from Servier, AbbVie, Novartis, Celgene, Astra Zeneca, and Janssen; and personal fees from Astellas, Pfizer, Macrogenics, and Amgen. CP, JHJ, TB, and MCW report no conflicts of interest. Contributions: the sponsors collected and analyzed data in conjunction with all authors. The lead author wrote the initial draft of the manuscript. All authors revised the manuscript and reviewed and approved the final version for submission. Acknowledgments: additional support on an early draft of the manuscript was provided by Sheila Truten and Brian Kaiser from Medical Communication Company, Inc. (Wynnewood, PA, USA), funded by Bristol Myers Squibb and in accordance with Good Publication Practice guidelines Funding: this study was sponsored and funded by Celgene, a Bristol-Myers Squibb Company. Data sharing statement: BMS policy on data sharing may be found at https://www.bms.com/researchers-and-partners/independentresearch/data-sharing-request-process.html

3244

References 1. Brandwein JM, Saini L, Geddes MN, et al. Outcomes of patients with

relapsed or refractory acute myeloid leukemia: a population-based real-world study. Am J Blood Res. 2020;10(4):124-133. 2. Wei A, Dohner H, Pocock C, et al. Oral azacitidine maintenance for acute myeloid leukemia in first remission. N Engl J Med. 2020; 383(26):2526-2537. 3. Alibhai SM, Breunis H, Timilshina N, et al. Quality of life and physical function in adults treated with intensive chemotherapy for acute myeloid leukemia improve over time independent of age. J Geriatr Oncol. 2015;6(4):262-271. 4. Timilshina N, Breunis H, Tomlinson GA, et al. Long-term recovery of quality of life and physical function over three years in adult survivors of acute myeloid leukemia after intensive chemotherapy. Leukemia. 2019;33(1):15-25. 5. Gerlinger C, Schmelter T. Determining the non-inferiority margin for patient reported outcomes. Pharm Stat. 2011;10(5):410-413. 6. Patrick DL, Gagnon DD, Zagari MJ, Mathijs R, Sweetenham J, Epoetin Alfa Study G. Assessing the clinical significance of healthrelated quality of life (HrQOL) improvements in anaemic cancer patients receiving epoetin alfa. Eur J Cancer. 2003;39(3):335-345. 7. Kvam AK, Fayers PM, Wisloff F. Responsiveness and minimal important score differences in quality-of-life questionnaires: a comparison of the EORTC QLQ-C30 cancer-specific questionnaire to the generic utility questionnaires EQ-5D and 15D in patients with multiple myeloma. Eur J Haematol. 2011;87(4):330-337. 8. Pickard AS, Neary MP, Cella D. Estimation of minimally important differences in EQ-5D utility and VAS scores in cancer. Health Qual Life Outcomes. 2007;5:70. 9. Ashbeck EL, Bell ML. Single time point comparisons in longitudinal randomized controlled trials: power and bias in the presence of missing data. BMC Med Res Methodol. 2016;16:43. 10. U.S. Department of Health and Human Services - Food and Drug Administration. Non-inferiority clinical trials to establish effectiveness: guidance for industry. In: Services UDoHaH, ed, 2016. 11. Montan I, Lowe B, Cella D, Mehnert A, Hinz A. General population norms for the functional assessment of chronic illness therapy (FACIT)-fatigue scale. Value Health. 2018;21(11):1313-1321. 12. Szende A, Janssen B, Cabases J, editors. Self-reported population health: an international perspective based on EQ-5D [Internet]. Dordrecht (NL): Springer, 2014. 13. National Cancer Institute (NCI). Surveillance, Epidemiology, and End Results (SEER) Program. Cancer Stat Facts: Leukemia - Acute Myeloid Leukemia (AML). [cited March 1, 2021]; Available from: https://seer.cancer.gov/statfacts/html/amyl.html.

haematologica | 2021; 106(12)

CASE REPORTS Dominance of an UBA1 mutant clone over a CALR mutant clone: from essential thrombocytemia to VEXAS VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic) is a recently described syndrome caused by somatic mutations in the UBA1 gene that lead to a multiorgan inflammatory disease and anemia.1 The clinical spectrum of UBA1-related diseases seems to be broader than the initial description,2,3 and the potential co-mutations occurring in UBA1-mutated patients have not been described yet. Here, we report the case of a patient with an initial CALR-mutated essential thrombocythemia (ET) who developed myelodysplasia cutis associated with an UBA1 mutation, with a concomitant extinction of both the CALR clone and the ET phenotype. A 65-year-old man without major previous disease was diagnosed with ET and was treated with hydroxyurea

and low dose aspirin. A CALR mutation was retrospectively identified in the diagnostic blood sample. After 10 years of treatment, at the age of 75, he was referred for chronic edematous erythematous plaques on the chest, upper and lower limbs, suggestive of Sweet’s syndrome, and a mild biological inflammatory syndrome (C-reactive protein at 30-50 mg/L) (Figure 1A). Skin biopsies evidenced infiltration of myeloid cells with a histiocytoid morphology associated with incompletely segmented nuclei suggesting immature polymorphonuclear cells expressing CD163 and myeloperoxidase, which was consistent with the diagnosis of myelodysplasia cutis4 (Figure 1A). A targeted next generation sequencing (NGS) analysis covering 41 myeloid disease associated genes (ASXL1, ASXL2, ATM, BCOR, CALR, CBL, CEBPA, CSF3R, DDX41, DNMT3A, EZH2, FLT3, GATA2, HRAS, IDH1, IDH2, IKZF1, JAK2, KIT, KMT2A, KRAS, MPL, NF1, NPM1, NRAS, PHF6, PTPN11, RAD21, RUNX1, SETBP1,

Figure 1. Skin and sequential bone marrow morphology. (A) Skin lesions and histopathological findings of skin biopsy, showing infiltration of histiocytoid cells staining positively for myeloperoxidase (MPO). (B) May-Grünwald-Giemsa stained bone marrow (BM) aspirate, exhibiting vacuolated myeloid precursors suggestive for a VEXAS syndrome (upper, BM1 after 10 years; middle, BM2 after 11 years; and lower, BM3 after 11.5 years of evolution after essential thrombocythemia diagnosis).

haematologica | 2021; 106(12)

3245

Case Report

Figure 2. Clonal history of the two diseases. (A) Table of frequencies in bone marrow (BM) and skin biopsy: according to fluorescence in situ hybridization and next generation sequencing data. ND: not detected; y: years. (B) Sanger sequencing electrophoregram, showing mosaic UBA1 c.121A>C variant [NM_153280.3] in skin and successive three BM evaluations (BM1, 2, and 3). (C) Fish plot, representing clonal evolution upon BM evaluations (TET2 mutation is not displayed because of insufficient data in clonal evolution).

SF3B1, SH2B3, SMC1A, SMC3, SRSF2, STAG2, TET2, TP53, U2AF1, WT1, ZRSR2) was performed on the skin biopsy. Sequence analysis identified a DNMT3A p.R882H mutation (variant allele frequency [VAF] 22%) and a CALR type 1 mutation (VAF 6%). A monosomy of chromosome 7 was also identified by fluorescence in situ hybridization (FISH) in 30% of the cells of the biopsy (Figure 2A). A concomitant bone marrow (BM) examination (BM1 after 10 years of evolution) showed evidence of megakaryocyte hyperplasia but neither signs of myelodysplasia nor blast excess. Using the same NGS panel, we found the same mutations of CALR (VAF 29 %) and DNMT3A (VAF 38 %), while FISH analysis detected only 2 % of cells with monosomy 7, suggestive of a clonal relation between granulocytes in the BM and in the 3246

skin lesions (Figure 2A). A thalidomide-based treatment was introduced, with an initial clinical improvement on skin lesions and a persistent normalization of platelet count, in spite of hydroxyurea progressive tapering and arrest. During this treatment, the patient progressively developed severe anemia, without efficacy of erythropoietin support, and needed repeated blood transfusion. Serial BM evaluations (BM2 after 11 years, and BM3 after 11.5 years of evolution) were performed, which still showed no evidence of myelodysplastic changes, but a decreased number of typical ET-related megakaryocytes and a progressive profound erythroblastopenia. Interestingly, DNMT3A mutation persisted at similar frequency, whereas CALR allele burden dropped to 1%, and monosomy 7 became undetectable (Figure 2A). After discontinuation of all therapies except aspirin, haematologica | 2021; 106(12)

Case Report

macrocytic anemia persisted, in association with a biological inflammatory syndrome with elevated C-reactive protein, and cutaneous lesions that led to a VEXAS hypothesis. Sanger sequencing identified an UBA1 variant with a high allele burden (p.Met41Leu, c.121A>C [NM_153280.3]) (Figure 2B). Cytoplasmic vacuoles in erythroid and granulocytic precursors were also observed (Figure 1B). A retrospectively UBA1 mutation screening of the initial BM sample and skin biopsy detected the same variant at a lower allele burden, revealing the clonal sweep from the CALR mutated “ET clone” to the UBA1 “VEXAS” clone (Figure 2C). A treatment by ruxolitinib, the efficacy of which has been reported in VEXAS putatively due to its anti-inflammatory properties rather than a direct effect on the UBA1 mutated clone, was started and is currently ongoing with good improvement of cutaneous lesions.2 Hematopoietic stem cell transplantation was not considered due to the age of over 75 years at the time of VEXAS diagnosis. A recent publication reports the newly described VEXAS syndrome, linked to somatic mutations in UBA1.1 Our case illustrates similar clinical manifestations with regard to the described entity, but the patient did not exhibit all the key clinical features like fever, pulmonary involvement, or polychondritis. This case suggests that UBA1 mutations can be associated to a diversity of clinical presentations and that the clinical spectrum of UBA1 related disease has to be refined. Other teams already described the frequent incomplete clinical presentation, and other not initially described involvement (vasculitis-like aspect or other) has also been reported.2,3,5,6 Sweet’s syndrome was present in eight of 25 (32%) participants from the first series of VEXAS syndrome.1 There are two forms of Sweet’s syndrome: neutrophilic and histiocytoid.7,8 The histiocytoid subset, characterized by a dermal infiltration of immature myeloid cells with histiocytoid morphology, is associated with myelodysplastic syndromes in 24-31% of the cases.7 Some authors think that myelodysplasia cutis can be confirmed when the same cytogenetic abnormalities are found in the skin and BM of patients with histiocytoid Sweet’s syndrome without leukemia, as in our patient.4 To our knowledge, this is the first report of myelodysplasia cutis associated with UBA1 mutations. In our patient, we describe a progressive clonal replacement of a preexisting CALR-mutated ET clone, and illustrate the parallel clonal and phenotypic dynamics of two distinct driver-associated diseases. These results, thus demonstrate a clonal implication of different clones with a shared DNMT3A-mutated founder clone (Figure 2C). We assumed the association of DNMT3A and CALR mutations in the BM to be responsible for the initial ET. Then the association of the DNMT3A mutation, UBA1 mutation, and monosomy 7, occurring in a subclone distinct from the DNMT3A and CALR mutant subclone, was associated with a skin tropism, with a minority of cells harboring monosomy 7 in the BM. Finally, the CALR clone was outcompeted by the UBA1 clone, which resulted in the clinical remission of the ET, and in the development of the VEXAS phenotype associating anemia, vacuolated myeloid precursors and systemic inflammation. CALR mutations are strong driver mutations in myeloproliferative neoplasms. A decrease in allele burden can be observed during therapy such as interferon-α,8 or in around half of the cases of secondary acute myeloid leukemia related to myeloproliferative neoplasms.9 In the present observation, clonal dominance conferred by the UBA1 mutation seems to be a major event during clonal haematologica | 2021; 106(12)

evolution. This suggests that UBA1 mutations may provide a powerful selective advantage, able to overcome a CALR mutated clone. This might be due to a putative effect of UBA1 on the mutated hematopoietic stem cell proliferation and self-renewal, or might be in link with major changes induced by the autoinflammatory microenvironment that probably plays a role in shaping the clonal architecture, and might result in a selective advantage of the UBA1 clone over others. In this observation, the role of hydroxyurea or thalidomide in clonal evolution cannot be excluded. It remains however unlikely, as the effects of hydroxyurea on CALR clones are often very limited, and the effects of thalidomide are known to be limited in thromboembolism cases. Moreover, expansion of preexisting clones after lenalinomide therapy, which is closely related to thalidomide or incidence of second malignancy after lenalinomide have not been clearly demonstrated.10–12 In conclusion, we describe the clonal dynamics of a CALR mutated subclone associated with ET, crushed by an UBA1 mutated subclone associated with myelodysplasia cutis and VEXAS, developing on a preexisting DNMT3A clonal hematopoiesis. This suggests somatic mutations of UBA1 can be associated with other mutations, and can be a secondary major driver event in clonal evolution. Common clinical features between VEXAS and hematological neoplasms with inflammatory manifestations could lead to a misdiagnosis. Extensive screening for UBA1 mutation is required to determine the real prevalence of VEXAS among patients with atypical presentation of myeloid neoplasms. Mehdi Hage-Sleiman,1 Sophie Lalevée,2 Hélène Guermouche,1 Fabrizia Favale,1 Michael Chaquin,1 Maxime Battistella,3,4 Jean-David Bouaziz,1,4 Martine Bagot,2,4 François Delhommeau,1 Florence Cordoliani2 and Pierre Hirsch1 1 Sorbonne Université, INSERM, Centre de Recherche SaintAntoine, AP-HP, Hôpital Saint-Antoine, Service d’Hématologie Biologique; 2Dermatology Department, Saint-Louis Hospital, AP-HP; 3 Pathology Department, Saint-Louis Hospital, AP-HP and 4Université de Paris, Paris, France Correspondence: PIERRE HIRSCH - pierre.hirsch@aphp.fr doi:10.3324/haematol.2021.279418 Received: June 11, 2021. Accepted: July 21, 2021. Pre-published: July 29, 2021. Disclosures: no conflicts of interest to disclose. Contributions: MHS and PH performed molecular analyses and wrote the manuscript; SL, FC MB and JDB provided dermatologic data; MB performed histological analyses; FF performed molecular analyses; MC and FD performed cytological analyses; HG performed cytogenetic and FISH analyses. All authors contributed to manuscript writing. Funding: this work was supported by SiRIC CURAMUS (INCA-DGOS-Inserm_12560).

References 1. Beck DB, Ferrada MA, Sikora KA, et al. Somatic mutations in UBA1 and severe adult-onset autoinflammatory disease. N Engl J Med. 2020;383(27):2628-2638 2. Bourbon E, Heiblig M, Gerfaud-Valentin M, et al. Therapeutic options in Vexas syndrome: insights from a retrospective series. Blood. 2021;137(26):3682-3684. 3. Oganesyan A, Jachiet V, Chasset F, et al. VEXAS syndrome: still

3247

Case Report

expanding the clinical phenotype. Rheumatol Oxf Engl. 2021;60(9): e321-e323. 4. Osio A, Battistella M, Feugeas J-P, et al. Myelodysplasia cutis versus leukemia cutis. J Invest Dermatol. 2015;135(9):2321-2324. 5. Lee SMS, Fan BE, Lim JH-L, Goh LL, Lee JSS, Koh LW. A case of VEXAS syndrome manifesting as Kikuchi-Fujimoto disease, relapsing polychondritis, venous thromboembolism and macrocytic anaemia. Rheumatol Oxf Engl. 2021;60(9):e304-e306. 6. Poulter JA, Collins JC, Cargo C, et al. Novel somatic mutations in UBA1 as a cause of VEXAS syndrome. Blood. 2021;137(26):36763681. 7. Ghoufi L, Ortonne N, Ingen-Housz-Oro S, et al. Histiocytoid Sweet syndrome is more frequently associated with myelodysplastic syndromes than the classical neutrophilic variant: a comparative series of 62 patients. Medicine (Baltimore). 2016;95(15):e3033. 8. Verger E, Cassinat B, Chauveau A, et al. Clinical and molecular response to interferon-α therapy in essential thrombocythemia

3248

patients with CALR mutations. Blood. 2015;126(24):2585-2591. 9. Rampal R, Ahn J, Abdel-Wahab O, et al. Genomic and functional analysis of leukemic transformation of myeloproliferative neoplasms. Proc Natl Acad Sci U S A. 2014;111(50):E5401-E5410. 10. Palumbo A, Bringhen S, Kumar SK, et al. Second primary malignancies with lenalidomide therapy for newly diagnosed myeloma: a meta-analysis of individual patient data. Lancet Oncol. 2014; 15(3):333-342. 11. Jones JR, Cairns DA, Gregory WM, et al. Second malignancies in the context of lenalidomide treatment: an analysis of 2732 myeloma patients enrolled to the Myeloma XI trial. Blood Cancer J. 2016; 6(12):e506. 12. Kiladjian J-J, Rain J-D, Bernard J-F, Briere J, Chomienne C, Fenaux P. Long-term incidence of hematological evolution in three French prospective studies of hydroxyurea and pipobroman in polycythemia vera and essential thrombocythemia. Semin Thromb Hemost. 2006;32(4):417-421.

haematologica | 2021; 106(12)

Case Report

Fatal exacerbation of ChadOx1-nCoV-19-induced thrombotic thrombocytopenia syndrome after initial successful therapy with intravenous immunoglobulins - a rational for monitoring immunoglobulin G levels The present report describes a vaccine-induced thrombotic thrombocytopenia (VITT) case with fatal exacerbation after initial improvement following initial intravenous immunoglobulin (IVIg) administration and anticoagulation. An 83-year-old woman presented at the emergency room with an alteration of her general condition. She presented with symptoms of weakness, nausea, vomiting, weight loss and spontaneous bruises without any obvious reason, 14 days after having received her first dose of ChadOx1 nCov-19. According to our medical records, she did not receive heparin or derivative during the previous 4 months. Clinical examination unraveled bruising on the upper limbs. Computer tomography (CT) of thorax and abdomen was normal. Oxygen saturation was 98% at admission and the patient was tested negative for SARS-CoV-2 infection as assessed by reverse-transcriptase polymerase chain reaction (RT-PCR). The initial laboratory investigations on the day of admission revealed that the patient was suffering from marked thrombocytopenia (i.e., platelet count of 10,000 per mm3), dramatically increased D-dimers plasma levels (i.e., > 20,000 ng/mL) and slightly low plasma fibrinogen (i.e., 179 mg/dL) (Figure 1). She was transfused with a platelet concentrate (roughly 3.5x1011 platelets) on the day of admission. During the night, she suffered from dyspnea grade NYHA 4. Pulmonary ventilation and perfusion (V/Q) scan was performed and disclosed bilateral pulmonary

embolism. Anti-PF4 immunoglobulin G (IgG) antibodies (i.e. 1.80 AU/mL, Figure 1) were detected on day 1 postadmission using a PF4/polyvinylsulfonate rapid assay (HemosIL® AcuStar HIT IgG assay, Instrumentation Laboratory Belgium NV, Zaventem, Belgium). The diagnosis of VITT was confirmed using a heparin-induced multi-electrode aggregometry method.1 In face of the clinical picture, i.e., thrombocytopenia and thrombosis, with the presence of anti-PF4 antibodies and positive platelet activation tests within 30 days after vaccination with ChadOx1 nCov-19, VITT was diagnosed.2 The patient therefore promptly received 15 grams of IVIg (Privigen®, CSL Behring Gmbh, Marburg, Germany) and methylprednisolone 1 mg/kg. A second platelet concentrate (roughly 4.5x1011 platelets) was administered to allow initiation of anticoagulation as the platelet count was still below 30,000 per mm3. The platelet count rapidly improved, i.e., 53,000 per mm3, and anticoagulation was started with fondaparinux 5 mg once a day (od) subcutaneously from day 1 to day 3 (taking into account renal failure, i.e., Cockcroft-Gault creatinine clearance <50 mL/min). She received additional IVIg on day 2 and 3, at the dose of 60 grams per day for a total IVIg dose of 135 grams corresponding to 1.7 grams of IVIg per kg administered over a period of 48 hours. Fondaparinux dose was increased to 7.5 mg od from day 4 to day 11 since renal function improved. On day 6, the patient was stabilized, and her global health status was improved as witnessed by normalized platelet count, decrease in Ddimers (i.e., from > 20,000 ng/mL at admission to 14,380 ng/mL) and C-reactive protein (CRP) (from 145 mg/dL at admission to 23 mg/dL). Later that day, however, oxygen saturation dropped below 80%. Cough with sputum production was noted and exacerbation of COPD with Moraxella catarrhalis infection was diagnosed. Oxygen

Figure 1. Clinical and laboratory data of the case. CRP: C-reactive protein; CODP: chronic obstructive pulmonary disease; HIT: heparin-induced thrombocytopenia; od: once daily.

haematologica | 2021; 106(12)

3249

Case Report

Figure 2. Dilution experiments on samples collected at day 1 post-admission (diagnosis), days 4 post-admission (after IVIg administration), day 7 post-admission (marked clinical improvement with platelet count normalization) and day 15 post-admission (deterioration of patient’s status and death). Dilutions were made in normal heated plasma (containing normal immunoglobulin G [IgG] level) or in modified Tyrode’s buffer at 1/10, 1/20, 1/40 dilution ratios. Platelet activation was assessed without heparin in platelet-activating anti-platelet factor 4-serotonin-release assay (PF4-SRA) and results are expressed in percentage of serotonin release. IVIG: intravenous immunoglobulin.

supplementation was then started (2 liters per minute) combined with oral moxifloxacin 400 mg od for 5 days. On day 12 post-admission, anticoagulation was switched from fondaparinux to apixaban 5 mg twice a day (bid). Unfortunately, the clinical status worsened on day 12 post-admission with a de novo reduction of platelet count. Abdominal CT scan showed right adrenal hematoma with left adrenal infiltrate, which is a usual presentation of adrenal infarction, as described in autoimmune heparin-induced thrombocytopenia and also recently in VITT.4,5 Four units of 500 IU/mL of prothrombin complex concentrate were administered on day 14 as an attempt to control the adrenal hematoma, but she died later that day from hypovolemic shock probably secondary to adrenal hemorrhage. A causal adrenal infarction may have existed but could not be confirmed as neither an injected CT scan nor an autopsy was performed. Nevertheless, adrenal insufficiency was not documented and cortisol levels on day 14 was still in the upper range (i.e., 21 mg/dL, normal range: 6.2-18 mg/dL). Moreover, it must be noted that although apixaban was last administered on day 13 in the morning and was never reintroduced, its plasma level on day 14 was 353 ng/mL (usual Ctrough range: 22-177 ng/mL) and was still 132 ng/mL on day 15. The accumulation of apixaban may thus have contributed to, or even triggered, this bleeding event. The initial infusion of platelet concentrates may also have contributed to disease progression, as well as to the pulmonary embolism observed at diagnosis. An immediate treatment with IVIg, as now recommended by the American Society of Hematology,6 could have been beneficial but the presence of active bruising, the absence of documented thrombosis and the marked thrombocytopenia guided our therapeutic choice at that time. This case further highlights how VITT is a dynamical condition, which should not be discounted in a recently vaccinated patient with only thrombocytopenia and increased D-dimer levels, even in the absence of documented thrombosis, and that IVIg should be considered promptly, along with anti-PF4 testing and thrombosis screening. 3250

The rapid fatal outcome, occurring 2 weeks after VITT diagnosis while an improvement was noticed, raised the question of an early relapse or an exacerbation of the initial event. In an attempt to understand the possible cause(s), additional laboratory investigations were performed, as reported in Figure 1. Blood samples collected post-admission on day 1 (diagnosis), day 4 (after IVIg administration), day 7 (marked clinical improvement with platelet count normalization) and day 15 (deterioration of patient’s status, leading to death) were selected to assess time-related changes of anti-PF4 antibodies, platelet activation with PF4-SRA7 and total IgG levels. Result of the enzyme-linked immunosorbent assay (ELISA) with immobilized PF4/PVS complexes (LIFECODES PF4 IgG, Immucor Lifecodes, Jette, Belgium) remained strongly positive during the whole hospital stay with OD >3.00 measured with all samples collected from day 1 to day 15. Positive results were also obtained using a modified in-house ELISA in which the wells are only coated by PF4,8 demonstrating the presence of IgG antibodies that bound PF4 alone (data not shown). Such characteristics of VITT antibodies, shared with those of highly pathogenic auto-immune heparin-induced thrombocytopenia (HIT) antibodies,9 indicate a different specificity and affinity towards PF4 compared to classical HIT antibodies, and explain why their detection by HIT-dedicated immunoassays may be inadequate.7 The fact that anti-PF4 IgG antibodies were detected using a PF4/polyvinylsulfonate rapid assay (HemosIL® AcuStar HIT IgG assay) is particular since this assay failed to detect anti-PF4 antibodies in most of the reported VITT cases. Nevertheless, the HemosIL® Acustar HIT IgG assay rarely gives weak positive results in patients with likely VITT diagnosis, as stated by Platton et al., who reported two positive results out of 31 patients, all of whom had anti-PF4 IgG detected by ELISA.10 PF4 serotonin-release assay (PF4-SRA) was performed with the same samples, as previously described.7 Platelet activation, measured through maximal serotonin release, was 90% in the absence of heparin on day 1, decreased by more than 50% after IVIg administration (day 4, haematologica | 2021; 106(12)

Case Report

35%) and remained low when the patient was getting better (day 7, 36%). In contrast, platelet activation returned to a high level at the time of clinical deterioration, day 15 (79%). Time related changes of total IgG plasma levels mirrored those of PF4-SRA (Figure 1). While the total IgG level was rather low at diagnosis, it rose above 30 g/L after IVIg administration and then rapidly decreased, as observed on day 15, with a return to normal values for a healthy adult. These data support the hypothesis that the rebound of platelet activation observed in PF4-SRA at the time of clinical deterioration could be due to a rapid elimination of the infused immunoglobulins and the loss of their competing effect with platelet activating anti-PF4 antibodies on platelet FcɣRIIa. Additional experiments were performed with PF4-SRA to consolidate our hypothesis that the levels of IgG can explain the exacerbation The same samples were diluted either in normal plasma (containing normal IgG level) or in modified Tyrode’s buffer at 1/10, 1/20, 1/40 dilution ratios. When samples were diluted in normal plasma, platelet activation decreased proportionally with the dilution, and was completely abolished at 1/40, except for the day 1 sample, which still slightly activate platelets under these conditions (Figure 2). In contrast, when the samples were diluted in Tyrode’s buffer, platelet activation remained high (near 100%) for the samples collected on day 1 and day 15 but increased for those obtained at days 5 and 7, approaching 100% (Figure 2). These results strongly support that platelet activation by VITT antibodies was inhibited by normal IgG, and that lowering the concentrations of normal IgG led to the reappearance of platelet activation by loss of competition between the IVIg and anti-PF4 IgG. This case supports the concept that proper monitoring using an appropriate functional assay could help in the clinical decision making since PF4-SRA mirrored with the clinical evolution of the patient. Such an observation has also been made in a recent study, which demonstrated that platelet activation by VITT antibodies was inhibited in patients treated with IVIg.11 Nevertheless, this needs to be confirmed. Interestingly, the inhibitory effect of normal polyclonal IgG on the platelet activation induced by PF4-specific antibodies could also vary from one patient to another, as previously demonstrated in HIT patients.12 It is also important to note that administration of IVIg reduces the activation of platelets as assessed by the PF4-SRA. This has major consequences when collecting samples for confirmation of VITT diagnosis and dilution in appropriate buffer, as we did in our experiments, could be recommended to assess the competitive interaction between anti-PF4 IgG and IVIg. However, this test lacks worldwide availability, and cannot easily be used for emergency patient monitoring. Therefore, as total IgG concentration measured in the patient inversely correlated with platelet activation in PF4-SRA, quantitatively assaying anti-PF4 IgG antibodies levels and total IgG concentration in the patient’s plasma could help to identify situations where the competition between normal polyclonal IgG and anti-PF4 IgG on FcgRIIa may switch in favor of the platelet activating antibodies.13 Even though data on IVIg clearance parameters and target concentrations are lacking for such a very peculiar condition, a rapid decrease in total IgG concentrations within the normal range (i.e., 7-16 g/L)14 could alert to possible therapeutic escape, and the need for re-administration of IVIg, especially in a situation where anti-PF4 IgG remains high. In the patient, total IgG concentration haematologica | 2021; 106(12)

was reduced by half within 8 days, which is substantially faster than the median half-life of 30 days generally reported in the literature for IVIg.15 Although further studies are needed to understand the accelerated clearance and to assess the clinical relevance of total IgG measurement to monitor the efficacy of IVIg, it appears a very affordable tool in medical practice in combination with anti-PF4 IgG antibodies testing. Jonathan Douxfils,1,2* Caroline Vayne,3* Claire Pouplard,3 Thomas Lecompte,4 Julien Favresse,1,5 Florence Potier,6 Emy Gasser,7 Valérie Mathieux,8 Jean-Michel Dogné,1 Yves Gruel,3 Jérôme Rollin3# and François Mullier9# 1 University of Namur, Department of Pharmacy, Namur Research for Life Sciences, Namur Thrombosis and Hemostasis Center, Namur, Belgium; 2QUALIblood s.a., Namur, Belgium; 3University of Tours, EA7501 GICC, CHRU de Tours, Department of Hemostasis, Tours, France; 4Département de Médecine, Hôpitaux Universitaires de Genève, Service d’Angiologie et d’Hémostase et Faculté de Médecine, Geneva Platelet Group (GpG), Université de Genève, Geneva, Switzerland; 5Clinique Saint-Luc Bouge, Department of Laboratory Medicine, Bouge, Belgium; 6Service de Gériatrie, CHU UCL Namur site Sainte-Elisabeth, Namur, Belgium; 7Université Catholique de Louvain, Service de Gériatrie, CHU UCL Namur site SainteElisabeth, Namur, Belgium; 8CHU UCL Namur |site SainteElizabeth, Université Catholique de Louvain, Department of Hematology, Namur Research Institute for Life Sciences, Namur Thrombosis and Hemostasis Center, Yvoir, Belgium and 9 Université Catholique de Louvain, CHU UCL Namur, Namur Thrombosis and Hemostasis Center, Hematology Laboratory, Namur Research Institute for Life Sciences, Yvoir, Belgium. *JD and CV contributed equally as co-first authors # JR and FM contributed equally as co-senior authors Correspondence: JONATHAN DOUXFILS jonathan.douxfils@unamur.be doi:10.3324/haematol.2021.279509 Received: June 28, 2021. Accepted: August 3, 2021. Pre-published: August 12, 2021. Disclosures: JD is the CEO and founder of QUALIblood S.A., a contract research organization manufacturing the DP-Filter, is a co-inventor of the DP-Filter (patent application number: PCT/ET2019/ 052903) and reports personal fees from Daiichi-Sankyo, DOASense Gmbh, Gedeon Richter, Mithra Pharmaceuticals, Norgine, Portola, Stago, Roche and Roche Diagnostics outside the submitted work; TL reports non-personal fees from IRIS and Stago; FM reports institutional fees from Stago, Werfen, Nodia, Roche Sysmex and Bayer as well as speaker fees from Boehringer Ingelheim, Bayer Healthcare, Bristol- Myers Squibb-Pfizer, Stago, Sysmex and Aspen all outside the submitted work. The other authors have no conflicts of interest to disclose. Contributions: JD and CV analyzed the results, wrote the first draft of the manuscript and designed the figures; JR, YG, CP, CV provided and analyzed the results and revised the manuscript; TL analyzed the results and thoroughly revised the manuscript; JF provided and analyzed the results and revised the manuscript; FP, EG and VM managed the patient and thoroughly revised the manuscript; J-MD analyzed the results; FM designed and supervised the experiments, provided and analyzed the data, and interpreted the results. Acknowledgments: the authors would like to thank the technical staff of the CHU UCL Namur, Mrs Justine Baudar, Mrs Maïté Guldenpfennig and the technical staff of the CHRU Tours, Mrs Séverine Augereau. Informed consent and ethical committee approval: as the patient died, an independent Review Board of the Clinique Sainte-Elisabeth 3251

Case Report

CHU UCL Namur (OM070) was consulted to decide if reporting this case is ethically acceptable. This Ethical Committee (decision number: 27-21) was in favor of the publication of this case and did not see any ethical issue in publishing this case report. All data has been anonymized.

References 1. Morel-Kopp MC, Mullier F, Gkalea V, et al. Heparin-induced multielectrode aggregometry method for heparin-induced thrombocytopenia testing: communication from the SSC of the ISTH. J Thromb Haemost. 2016;14(12):2548-2552. 2. 2. Robert T. Proposed Brighton Collaboration process for developing a standard case definition for study of new clinical syndrome X, as applied to thrombosis with thrombocytopenia syndrome (TTS) V10.16.3. Accessed on May 30, 2021. Available from: https://brightoncollaboration.us/wpcontent/uploads/2021/04/TTS-Case-Finding-and-DefinitionProcess.v9.0-April-16-202115853.pdf. 3. 3. Guidance produced from the Expert Hematology Panel (EHP) focused on Covid-19 vaccine induced thrombosis and thrombocytopenia (VITT) updated guidance on management. Version 1.3. Accessed on Apr 7, 2021. Available from: https://b-sh.org.uk/media/19530/guidance-version-13-on-mngmt-of-thromb o s i s - w i t h - t h r o m b o c y t o p e n i a - o c c u r r i n g - a f t e r- c - 1 9 vaccine_20210407.pdf. 4. Rosenberger LH, Smith PW, Sawyer RG, et al. Bilateral adrenal hemorrhage: the unrecognized cause of hemodynamic collapse associated with heparin-induced thrombocytopenia. Crit Care Med. 2011;39(4):833-838. 5. Scully M, Singh D, Lown R, et al. Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med.

3252

2021;384(23):2202-2211. 6. Bussel JB, Connors J, Cines DB, et al. Thrombosis with thrombocytopenia syndrome (also termed vaccine-induced thrombotic thrombocytopenia). version 1.5. Accessed on Jul 26, 2021. Available from: https://www.hematology.org/covid-19/vaccine-induced-immunethrombotic-thrombocytopenia. 7. Vayne C, Rollin J, Gruel Y, et al. PF4 Immunoassays in vaccineinduced thrombotic thrombocytopenia. N Engl J Med. 2021; 385(4):376-378. 8. Pouplard C, Leroux D, Rollin J, et al. Incidence of antibodies to protamine sulfate/heparin complexes incardiac surgery patients and impact on platelet activation and clinical outcome. Thromb Haemost. 2013;109(6):1141-1147. 9. Nguyen T-H, Medvedev N, Delcea M, Greinacher A. Anti-platelet factor 4/polyanion antibodies mediate a new mechanism of autoimmunity. Nat Commun. 2017;8:14945. 10.Platton S, Bartlett A, MacCallum P, et al. Evaluation of laboratory assays for anti-platelet factor 4 antibodies after ChAdOx1 nCOV-19 vaccination. J Thromb Haemost. 2021;19(8):2007-2013. 11.Bourguignon A, Arnold DM, Warkentin TE, et al. Adjunct immune globulin for vaccine-induced thrombotic thrombocytopenia. N Engl J Med. 2021;385(8):720-728. 12.Rollin J, Pouplard C, Sung HC, et al. Increased risk of thrombosis in FcgammaRIIA 131RR patients with HIT due to defective control of platelet activation by plasma IgG2. Blood. 2015;125(15):2397-2404. 13.Warkentin TE. High-dose intravenous immunoglobulin for the treatment and prevention of heparin-induced thrombocytopenia: a review. Expert Rev Hematol. 2019;12(8):685-698. 14.Svendsen PJ, Ward AM, Dati F, et al. New international reference preparation for proteins in human serum (RPPHS). Clin Chem. 1994;40(6):934-938. 15.Koleba T, Ensom MH. Pharmacokinetics of intravenous immunoglobulin: a systematic review. Pharmacotherapy. 2006; 26(6):813-827.

haematologica | 2021; 106(12)

haematologica — Vol. 106 n. 12 — December 2021 —3029-3252